This is a copy of regalloc_code2_BRANCH from Netscape's private repository,

as it existed in January of 1998.


git-svn-id: svn://10.0.0.236/branches/regalloc_code2_BRANCH@22571 18797224-902f-48f8-a5cc-f745e15eee43

mozilla/ef/Compiler/RegisterAllocator/BitSet.cpp

@@ -0,0 +1,134 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "BitSet.h"
+
+// Return the next bit after index set to true or -1 if none.
+//
+Int32 BitSet::nextOne(Int32 pos) const
+{
+	++pos;
+
+	if (pos < 0 || Uint32(pos) >= universeSize)
+		return -1;
+
+	Uint32 offset = getWordOffset(pos);
+	Uint8 index = getBitOffset(pos);
+	Word* ptr = &word[offset];
+	Word currentWord = *ptr++ >> index;
+	
+	if (currentWord != Word(0)) {
+		while ((currentWord & Word(1)) == 0) {
+			++index;
+			currentWord >>= 1;
+		}
+		return (offset << nBitsInWordLog2) + index;
+	}
+
+	Word* limit = &word[getSizeInWords(universeSize)];
+	while (ptr < limit) {
+		++offset;
+		currentWord = *ptr++;
+		if (currentWord != Word(0)) {
+			index = 0;
+			while ((currentWord & Word(1)) == 0) {
+				++index;
+				currentWord >>= 1;
+			}
+			return (offset << nBitsInWordLog2) + index;
+		}
+	}
+	return -1;
+}
+
+// Return the next bit after index set to false or -1 if none.
+//
+Int32 BitSet::nextZero(Int32 pos) const
+{
+	++pos;
+
+	if (pos < 0 || Uint32(pos) >= universeSize)
+		return -1;
+
+	Uint32 offset = getWordOffset(pos);
+	Uint8 index = getBitOffset(pos);
+	Word* ptr = &word[offset];
+	Word currentWord = *ptr++ >> index;
+	
+	if (currentWord != Word(~0)) {
+		for (; index < nBitsInWord; ++index) {
+			if ((currentWord & Word(1)) == 0) {
+				Int32 ret = (offset << nBitsInWordLog2) + index;
+				return (Uint32(ret) < universeSize) ? ret : -1;
+			}
+			currentWord >>= 1;
+		}
+	}
+
+	Word* limit = &word[getSizeInWords(universeSize)];
+	while (ptr < limit) {
+		++offset;
+		currentWord = *ptr++;
+		if (currentWord != Word(~0)) {
+			for (index = 0; index < nBitsInWord; ++index) {
+				if ((currentWord & Word(1)) == 0) {
+					Int32 ret = (offset << nBitsInWordLog2) + index;
+					return (Uint32(ret) < universeSize) ? ret : -1;
+				}
+				currentWord >>= 1;
+			}
+		}
+	}
+	return -1;
+}
+
+#ifdef DEBUG_LOG
+
+// Print the set.
+//
+void BitSet::printPretty(LogModuleObject log)
+{
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("[ "));
+
+	for (Int32 i = firstOne(); i != -1; i = nextOne(i)) {
+		Int32 currentBit = i;
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("%d", currentBit));
+
+		Int32 nextBit = nextOne(currentBit);
+		if (nextBit != currentBit + 1) {
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, (" "));
+			continue;
+		}
+		  
+		while ((nextBit != -1) && (nextBit == (currentBit + 1))) {
+			currentBit = nextBit;
+			nextBit = nextOne(nextBit);
+		}
+
+		if (currentBit > (i+1))
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("-%d ", currentBit));
+		else
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, (" %d ", currentBit));
+
+		i = currentBit;
+	}
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("]\n"));
+}
+
+#endif // DEBUG_LOG

mozilla/ef/Compiler/RegisterAllocator/BitSet.h

@@ -0,0 +1,195 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _BITSET_H_
+#define _BITSET_H_
+
+#include "Fundamentals.h"
+#include "LogModule.h"
+#include "Pool.h"
+#include <string.h>
+
+//------------------------------------------------------------------------------
+// BitSet -
+
+class BitSet
+{
+private:
+
+#if (PR_BITS_PER_WORD == 64)
+	typedef Uint64 Word;
+#elif (PR_BITS_PER_WORD == 32)
+	typedef Uint32 Word;
+#endif
+
+	static const nBitsInWord = 		PR_BITS_PER_WORD;
+	static const nBytesInWord = 	PR_BYTES_PER_WORD;
+	static const nBitsInWordLog2 =	PR_BITS_PER_WORD_LOG2;
+	static const nBytesInWordLog2 = PR_BYTES_PER_WORD_LOG2;
+
+	// Return the number of Word need to store the universe.
+	static Uint32 getSizeInWords(Uint32 sizeOfUniverse) {return (sizeOfUniverse + (nBitsInWord - 1)) >> nBitsInWordLog2;}
+	// Return the given element offset in its containing Word.
+	static Uint32 getBitOffset(Uint32 element) {return element & (nBitsInWord - 1);}
+	// Return the Word offset for the given element int the universe.
+	static Uint32 getWordOffset(Uint32 element) {return element >> nBitsInWordLog2;}
+	// Return the mask for the given bit index.
+	static Word getMask(Uint8 index) {return Word(1) << index;}
+	
+private:
+
+	Uint32							universeSize;			// Size of the universe
+	Word*							word;					// universe memory.
+
+private:
+
+	// No copy constructor.
+	BitSet(const BitSet&);
+
+	// Check if the given set's universe is of the same size than this universe.
+	void checkUniverseCompatibility(const BitSet& set) const {assert(set.universeSize == universeSize);}
+	// Check if pos is valid for this set's universe.
+	void checkMember(Int32 pos) const {assert(pos >=0 && Uint32(pos) < universeSize);}
+
+public:
+
+	// Create a bitset of universeSize bits.
+	BitSet(Pool& pool, Uint32 universeSize) : universeSize(universeSize) {word = new(pool) Word[getSizeInWords(universeSize)]; clear();}
+
+	// Return the size of this bitset.
+	Uint32 getSize() const {return universeSize;}
+
+	// Clear the bitset.
+	void clear() {memset(word, 0x00, getSizeInWords(universeSize) << nBytesInWordLog2);}
+	// Clear the bit at index.
+	void clear(Uint32 index) {checkMember(index); word[getWordOffset(index)] &= ~getMask(index);}
+	// Set the bitset.
+	void set() {memset(word, 0xFF, getSizeInWords(universeSize) << nBytesInWordLog2);}
+	// Set the bit at index.
+	void set(Uint32 index) {checkMember(index); word[getWordOffset(index)] |= getMask(index);}
+	// Return true if the bit at index is set.
+	bool test(Uint32 index) const {checkMember(index); return (word[getWordOffset(index)] & getMask(index)) != 0;}
+	// Union with the given bitset.
+	inline void or(const BitSet& set);
+	// Intersection with the given bitset.
+	inline void and(const BitSet& set);
+	// Difference with the given bitset.
+	inline void difference(const BitSet& set);
+	// Copy set.
+	inline BitSet& operator = (const BitSet& set);
+	// Return true if the bitset are identical.
+	friend bool operator == (const BitSet& set1, const BitSet& set2);
+	// Return true if the bitset are different.
+	friend bool operator != (const BitSet& set1, const BitSet& set2);
+
+	// Logical operators.
+	BitSet& operator |= (const BitSet& set) {or(set); return *this;}
+	BitSet& operator &= (const BitSet& set) {and(set); return *this;}
+	BitSet& operator -= (const BitSet& set) {difference(set); return *this;}
+
+	// Return the first bit at set to true or -1 if none.
+	Int32 firstOne() const {return nextOne(-1);}
+	// Return the next bit after index set to true or -1 if none.
+	Int32 nextOne(Int32 pos) const;
+	// Return the first bit at set to false or -1 if none.
+	Int32 firstZero() const {return nextZero(-1);}
+	// Return the next bit after index set to false or -1 if none.
+	Int32 nextZero(Int32 pos) const;
+
+	// Iterator to conform with the set API.
+	typedef Int32 iterator;
+	// Return true if the walk is ordered.
+	static bool isOrdered() {return true;}
+	// Return the iterator for the first element of this set.
+	iterator begin() const {return firstOne();}
+	// Return the next iterator.
+	iterator advance(iterator pos) const {return nextOne(pos);}
+	// Return true if the iterator is at the end of the set.
+	bool done(iterator pos) const {return pos == -1;}
+	// Return the element corresponding to the given iterator.
+	Uint32 get(iterator pos) const {return pos;}
+
+#ifdef DEBUG_LOG
+	// Print the set.
+	void printPretty(LogModuleObject log);
+#endif // DEBUG_LOG
+};
+
+// Union with the given bitset.
+//
+inline void  BitSet::or(const BitSet& set)
+{
+	checkUniverseCompatibility(set);
+	Word* src = set.word;
+	Word* dst = word;
+	Word* limit = &src[getSizeInWords(universeSize)];
+
+	while (src < limit)
+		*dst++ |= *src++;
+}
+
+// Intersection with the given bitset.
+//
+inline void  BitSet::and(const BitSet& set)
+{
+	checkUniverseCompatibility(set);
+	Word* src = set.word;
+	Word* dst = word;
+	Word* limit = &src[getSizeInWords(universeSize)];
+
+	while (src < limit)
+		*dst++ &= *src++;
+}
+
+// Difference with the given bitset.
+//
+inline void BitSet::difference(const BitSet& set)
+{
+	checkUniverseCompatibility(set);
+	Word* src = set.word;
+	Word* dst = word;
+	Word* limit = &src[getSizeInWords(universeSize)];
+
+	while (src < limit)
+		*dst++ &= ~*src++;
+}
+
+// Copy the given set into this set.
+//
+inline BitSet& BitSet::operator = (const BitSet& set)
+{
+	checkUniverseCompatibility(set);
+	if (this != &set)
+		memcpy(word, set.word, getSizeInWords(universeSize) << nBytesInWordLog2);
+	return *this;
+}
+
+// Return true if the given set is identical to this set.
+inline bool operator == (const BitSet& set1, const BitSet& set2)
+{
+	set1.checkUniverseCompatibility(set2);
+
+	if (&set1 == &set2)
+		return true;
+
+	return memcmp(set1.word, set2.word, BitSet::getSizeInWords(set1.universeSize) << BitSet::nBytesInWordLog2) == 0;
+}
+
+inline bool operator != (const BitSet& set1, const BitSet& set2) {return !(set1 == set2);}
+
+#endif // _BITSET_H

mozilla/ef/Compiler/RegisterAllocator/Coalescing.h

@@ -0,0 +1,159 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _COALESCING_H_
+#define _COALESCING_H_
+
+#include "Fundamentals.h"
+#include "Pool.h"
+#include "RegisterPressure.h"
+#include "InterferenceGraph.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "SparseSet.h"
+#include "RegisterAllocator.h"
+#include "RegisterAllocatorTools.h"
+
+#if 1
+// Performing an ultra conservative coalescing meens that when we look at
+// candidates (source,destination) for coalescing we need to make sure
+// that the combined interference of the source and destination register
+// will not exceed the total number of register available for the register
+// class. 
+#define ULTRA_CONSERVATIVE_COALESCING
+#else
+// If we are not doing an ultra conservative coalescing we have to make sure 
+// that the total number of neighbor whose degree is greater than the total 
+// number of register is not greater than the total number of register.
+#undef ULTRA_CONSERVATIVE_COALESCING
+#endif
+
+template <class RegisterPressure>
+struct Coalescing
+{
+	static bool coalesce(RegisterAllocator& registerAllocator);
+};
+
+template <class RegisterPressure>
+bool Coalescing<RegisterPressure>::coalesce(RegisterAllocator& registerAllocator)
+{
+	Pool& pool = registerAllocator.pool;
+
+   	// Initialize the lookup table
+	//
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	RegisterName* newRange = new RegisterName[2 * rangeCount];
+	RegisterName* coalescedRange = &newRange[rangeCount];
+	RegisterName* name2range = registerAllocator.name2range;
+
+	init(coalescedRange, rangeCount);
+
+	SparseSet interferences(pool, rangeCount);
+	InterferenceGraph<RegisterPressure>& iGraph = registerAllocator.iGraph;
+	bool removedInstructions = false;
+
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.lndList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+
+	// Walk the nodes in the loop nesting depth list.
+	for (Int32 n = nNodes - 1; n >= 0; n--) {
+		InstructionList& instructions = nodes[n]->getInstructions();
+
+		InstructionList::iterator it = instructions.begin();
+		while (!instructions.done(it)) {
+			Instruction& instruction = instructions.get(it);
+			it = instructions.advance(it);
+
+			if ((instruction.getFlags() & ifCopy) != 0) {
+				assert(instruction.getInstructionUseBegin() != instruction.getInstructionUseEnd() && instruction.getInstructionUseBegin()[0].isRegister());
+				assert(instruction.getInstructionDefineBegin() != instruction.getInstructionDefineEnd() && instruction.getInstructionDefineBegin()[0].isRegister());
+
+				RegisterName source = findRoot(name2range[instruction.getInstructionUseBegin()[0].getRegisterName()], coalescedRange);
+				RegisterName destination = findRoot(name2range[instruction.getInstructionDefineBegin()[0].getRegisterName()], coalescedRange);
+
+				if (source == destination) {
+					instruction.remove();
+				} else if (!iGraph.interfere(source, destination)) {
+					InterferenceVector* sourceVector = iGraph.getInterferenceVector(source);
+					InterferenceVector* destinationVector = iGraph.getInterferenceVector(destination);
+
+#ifdef ULTRA_CONSERVATIVE_COALESCING
+					interferences.clear();
+
+					InterferenceVector* vector;
+					for (vector = sourceVector; vector != NULL; vector = vector->next) {
+						RegisterName* neighbors = vector->neighbors;
+						for (Uint32 i = 0; i < vector->count; i++)
+							interferences.set(findRoot(neighbors[i], coalescedRange));
+					}
+					for (vector = destinationVector; vector != NULL; vector = vector->next) {
+						RegisterName* neighbors = vector->neighbors;
+						for (Uint32 i = 0; i < vector->count; i++)
+							interferences.set(findRoot(neighbors[i], coalescedRange));
+					}
+
+					Uint32 count = interferences.getSize();
+#else // ULTRA_CONSERVATIVE_COALESCING
+					trespass("not implemented");
+					Uint32 count = 0;
+#endif // ULTRA_CONSERVATIVE_COALESCING
+
+					if (count < 6 /* FIX: should get the number from the class */) {
+						// Update the interferences vector.
+						if (sourceVector == NULL) {
+							iGraph.setInterferenceVector(source, destinationVector);
+							sourceVector = destinationVector;
+						} else if (destinationVector == NULL)
+							iGraph.setInterferenceVector(destination, sourceVector);
+						else {
+							InterferenceVector* last = NULL;
+							for (InterferenceVector* v = sourceVector; v != NULL; v = v->next)
+								last = v;
+							assert(last);
+							last->next = destinationVector;
+							iGraph.setInterferenceVector(destination, sourceVector);
+						}
+						// Update the interference matrix.
+						for (InterferenceVector* v = sourceVector; v != NULL; v = v->next) {
+							RegisterName* neighbors = v->neighbors;
+							for (Uint32 i = 0; i < v->count; i++) {
+								RegisterName neighbor = findRoot(neighbors[i], coalescedRange);
+								iGraph.setInterference(neighbor, source);
+								iGraph.setInterference(neighbor, destination);
+							}
+						}
+
+						instruction.remove();
+						coalescedRange[source] = destination;
+						removedInstructions = true;
+					}
+				}
+			}
+		}
+	}
+
+	registerAllocator.rangeCount = compress(registerAllocator.name2range, coalescedRange, registerAllocator.nameCount, rangeCount);
+	delete newRange;
+
+	return removedInstructions;
+}
+
+#endif // _COALESCING_H_

mozilla/ef/Compiler/RegisterAllocator/Coloring.cpp

@@ -0,0 +1,283 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef NEW_LAURENTM_CODE
+
+#include "Coloring.h"
+#include "VirtualRegister.h"
+#include "FastBitSet.h"
+#include "FastBitMatrix.h"
+#include "CpuInfo.h"
+
+bool Coloring::
+assignRegisters(FastBitMatrix& interferenceMatrix)
+{
+  PRUint32 *stackPtr = new(pool) PRUint32[vRegManager.count()];
+
+  return select(interferenceMatrix, stackPtr, simplify(interferenceMatrix, stackPtr));
+}
+
+PRInt32 Coloring::
+getLowestSpillCostRegister(FastBitSet& bitset)
+{
+  PRInt32 lowest = bitset.firstOne();
+  if (lowest != -1)
+	{
+	  Flt32 cost = vRegManager.getVirtualRegister(lowest).spillInfo.spillCost;
+	  for (PRInt32 r = bitset.nextOne(lowest); r != -1; r = bitset.nextOne(r))
+		{
+		  VirtualRegister& vReg = vRegManager.getVirtualRegister(r);
+		  if (!vReg.spillInfo.infiniteSpillCost && (vReg.spillInfo.spillCost < cost))
+			{
+			  cost = vReg.spillInfo.spillCost;
+			  lowest = r;
+			}
+		}
+	}
+  return lowest;
+}
+
+PRUint32* Coloring::
+simplify(FastBitMatrix interferenceMatrix, PRUint32* stackPtr)
+{
+  // first we construct the sets low and high. low contains all nodes of degree
+  // inferior to the number of register available on the processor. All the
+  // nodes with an high degree and a finite spill cost are placed in high.
+  // Nodes of high degree and infinite spill cost are not included in either sets.
+
+  PRUint32 nRegisters = vRegManager.count();
+  FastBitSet low(pool, nRegisters);
+  FastBitSet high(pool, nRegisters);
+  FastBitSet stack(pool, nRegisters);
+
+  for (VirtualRegisterManager::iterator i = vRegManager.begin(); !vRegManager.done(i); i = vRegManager.advance(i))
+	{
+	  VirtualRegister& vReg = vRegManager.getVirtualRegister(i);
+	  
+	  if (vReg.getClass() == vrcStackSlot)
+		{
+		  stack.set(i);
+		  vReg.colorRegister(nRegisters);
+		}
+	  else
+		{
+		  if (vReg.colorInfo.interferenceDegree < NUMBER_OF_REGISTERS)
+			low.set(i);
+		  else // if (!vReg.spillInfo.infiniteSpillCost)
+			high.set(i);
+
+		  // Set coloring info.
+		  vReg.spillInfo.willSpill = false;
+
+		  switch(vReg.getClass())
+			{
+			case vrcInteger:
+			  vReg.colorRegister(LAST_GREGISTER + 1);
+			  break;
+			case vrcFloatingPoint:
+			case vrcFixedPoint:
+			  vReg.colorRegister(LAST_FPREGISTER + 1);
+			  break;
+			default:
+			  PR_ASSERT(false); // Cannot happen.
+			}
+		}
+	}
+
+  // push the stack registers
+  PRInt32 j;
+  for (j = stack.firstOne(); j != -1; j = stack.nextOne(j))
+	*stackPtr++ = j;
+
+  // simplify
+  while (true)
+	{
+	  PRInt32 r;
+	  while ((r = getLowestSpillCostRegister(low)) != -1)
+		{
+		  VirtualRegister& vReg = vRegManager.getVirtualRegister(r);
+
+		  /* update low and high */
+		  FastBitSet inter(interferenceMatrix.getRow(r), nRegisters);
+		  for (j = inter.firstOne(); j != -1; j = inter.nextOne(j))
+			{
+			  VirtualRegister& neighbor = vRegManager.getVirtualRegister(j);
+			  // if the new interference degree of one of his neighbor becomes
+			  // NUMBER_OF_REGISTERS - 1 then it is added to the set 'low'.
+			  
+			  PRUint32 maxInterference = 0;
+			  switch (neighbor.getClass())
+				{
+				case vrcInteger:
+				  maxInterference = NUMBER_OF_GREGISTERS;
+				  break;
+				case vrcFloatingPoint:
+				case vrcFixedPoint:
+				  maxInterference = NUMBER_OF_FPREGISTERS;
+				  break;
+				default:
+				  PR_ASSERT(false);
+				}
+			  if ((vRegManager.getVirtualRegister(j).colorInfo.interferenceDegree-- == maxInterference))
+				{
+				  high.clear(j);
+				  low.set(j);
+				}
+			  vReg.colorInfo.interferenceDegree--;
+			  interferenceMatrix.clear(r, j);
+			  interferenceMatrix.clear(j, r);
+			}
+		  low.clear(r);
+
+		  // Push this register.
+		  *stackPtr++ = r;
+		}
+	  if ((r = getLowestSpillCostRegister(high)) != -1)
+		{
+		  high.clear(r);
+		  low.set(r);
+		}
+	  else
+		break;
+	}
+
+  return stackPtr;
+}
+
+bool Coloring::
+select(FastBitMatrix& interferenceMatrix, PRUint32* stackBase, PRUint32* stackPtr)
+{
+  PRUint32 nRegisters = vRegManager.count();
+  FastBitSet usedRegisters(NUMBER_OF_REGISTERS + 1); // usedRegisters if used for both GR & FPR.
+  FastBitSet preColoredRegisters(NUMBER_OF_REGISTERS + 1);
+  FastBitSet usedStack(nRegisters + 1);
+  bool success = true;
+  Int32 lastUsedSSR = -1;
+
+  // select
+  while (stackPtr != stackBase)
+	{
+	  // Pop one register.
+	  PRUint32 r = *--stackPtr;
+	  VirtualRegister& vReg = vRegManager.getVirtualRegister(r);
+
+	  FastBitSet neighbors(interferenceMatrix.getRow(r), nRegisters);
+		  
+	  if (vReg.getClass() == vrcStackSlot)
+		// Stack slots coloring.
+		{
+		  usedStack.clear();
+
+		  for (PRInt32 i = neighbors.firstOne(); i != -1; i = neighbors.nextOne(i))
+			usedStack.set(vRegManager.getVirtualRegister(i).getColor());
+
+		  Int32 color = usedStack.firstZero();
+		  vReg.colorRegister(color);
+		  if (color > lastUsedSSR)
+			  lastUsedSSR = color;
+		}
+	  else
+		// Integer & Floating point register coloring.
+		{
+		  usedRegisters.clear();
+		  preColoredRegisters.clear();
+		  
+		  for (PRInt32 i = neighbors.firstOne(); i != -1; i = neighbors.nextOne(i))
+			{
+			  VirtualRegister& nvReg = vRegManager.getVirtualRegister(i);
+			  usedRegisters.set(nvReg.getColor());
+			  if (nvReg.isPreColored())
+				preColoredRegisters.set(nvReg.getPreColor());
+			}
+		  if (vReg.hasSpecialInterference)
+			usedRegisters |= vReg.specialInterference;
+
+		  PRInt8 c = -1;
+		  PRInt8 maxColor = 0;
+		  PRInt8 firstColor = 0;
+		  switch (vReg.getClass())
+			{
+			case vrcInteger:
+			  firstColor = FIRST_GREGISTER;
+			  maxColor = LAST_GREGISTER;
+			  break;
+			case vrcFloatingPoint:
+			case vrcFixedPoint:
+			  firstColor = FIRST_FPREGISTER;
+			  maxColor = LAST_FPREGISTER;
+			  break;
+			default:
+			  PR_ASSERT(false);
+			}
+
+		  if (vReg.isPreColored())
+			{
+			  c = vReg.getPreColor();
+			  if (usedRegisters.test(c))
+				c = -1;
+			}
+		  else
+			{
+			  for (c = usedRegisters.nextZero(firstColor - 1); (c >= 0) && (c <= maxColor) && (preColoredRegisters.test(c)); 
+				   c = usedRegisters.nextZero(c)) {}
+			}
+
+		  if ((c >= 0) && (c <= maxColor))
+			{
+			  vReg.colorRegister(c);
+			}
+		  else
+			{
+			  VirtualRegister& stackRegister = vRegManager.newVirtualRegister(vrcStackSlot);
+			  vReg.equivalentRegister[vrcStackSlot] = &stackRegister;	  
+			  vReg.spillInfo.willSpill = true;
+			  success = false;
+			}
+		}
+	}
+  
+#ifdef DEBUG
+  if (success)
+	{
+	  for (VirtualRegisterManager::iterator i = vRegManager.begin(); !vRegManager.done(i); i = vRegManager.advance(i))
+		{
+		  VirtualRegister& vReg = vRegManager.getVirtualRegister(i);
+		  switch (vReg.getClass())
+			{
+			case vrcInteger:
+			  if (vReg.getColor() > LAST_GREGISTER)
+				PR_ASSERT(false);
+			  break;
+			case vrcFloatingPoint:
+			case vrcFixedPoint:
+#if NUMBER_OF_FPREGISTERS != 0
+			  if (vReg.getColor() > LAST_FPREGISTER)
+				PR_ASSERT(false);
+#endif
+			  break;
+			default:
+			  break;
+			}
+		}
+	}
+#endif
+
+  vRegManager.nUsedStackSlots = lastUsedSSR + 1;
+  return success;
+}
+#endif // NEW_LAURENTM_CODE

mozilla/ef/Compiler/RegisterAllocator/Coloring.h

@@ -0,0 +1,284 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "RegisterAllocator.h"
+#include "VirtualRegister.h"
+#include "InterferenceGraph.h"
+#include "SparseSet.h"
+#include "Spilling.h"
+#include "Splits.h"
+
+UT_EXTERN_LOG_MODULE(RegAlloc);
+
+template <class RegisterPressure>
+class Coloring
+{
+private:
+	static RegisterName* simplify(RegisterAllocator& registerAllocator, RegisterName* coloringStack);
+	static bool select(RegisterAllocator& registerAllocator, RegisterName* coloringStack, RegisterName* coloringStackPtr);
+
+public:
+	static bool color(RegisterAllocator& registerAllocator);
+	static void finalColoring(RegisterAllocator& registerAllocator);
+};
+
+
+template <class RegisterPressure>
+void Coloring<RegisterPressure>::finalColoring(RegisterAllocator& registerAllocator)
+{
+	RegisterName* color = registerAllocator.color;
+	RegisterName* name2range = registerAllocator.name2range;
+
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		InstructionList& instructions = nodes[n]->getInstructions();
+
+		for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					usePtr->setRegisterName(color[name2range[usePtr->getRegisterName()]]);
+#ifdef DEBUG
+					RegisterID rid = usePtr->getRegisterID();
+					setColoredRegister(rid);
+					usePtr->setRegisterID(rid);
+#endif // DEBUG
+				}
+
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister()) {
+					definePtr->setRegisterName(color[name2range[definePtr->getRegisterName()]]);
+#ifdef DEBUG
+					RegisterID rid = definePtr->getRegisterID();
+					setColoredRegister(rid);
+					definePtr->setRegisterID(rid);
+#endif // DEBUG
+				}
+		}
+	}
+}
+
+template <class RegisterPressure>
+bool Coloring<RegisterPressure>::select(RegisterAllocator& registerAllocator, RegisterName* coloringStack, RegisterName* coloringStackPtr)
+{
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	RegisterName* color = new RegisterName[rangeCount];
+	registerAllocator.color = color;
+
+	for (Uint32 r = 1; r < rangeCount; r++)
+		color[r] = RegisterName(6); // FIX;
+
+	// Color the preColored registers.
+	//
+	VirtualRegisterManager& vrManager = registerAllocator.vrManager;
+	RegisterName* name2range = registerAllocator.name2range;
+	PreColoredRegister* machineEnd = vrManager.getMachineRegistersEnd();
+	for (PreColoredRegister* machinePtr = vrManager.getMachineRegistersBegin(); machinePtr < machineEnd; machinePtr++)
+		if (machinePtr->id != invalidID) {
+			color[name2range[getName(machinePtr->id)]] = machinePtr->color;
+			UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\twill preColor range %d as %d\n", name2range[getName(machinePtr->id)], machinePtr->color));
+		}
+
+	SpillCost* cost = registerAllocator.spillCost;
+	Pool& pool = registerAllocator.pool;
+	SparseSet& spill = *new(pool) SparseSet(pool, rangeCount);
+	registerAllocator.willSpill = &spill;
+	SparseSet neighborColors(pool, 6); // FIX
+	InterferenceGraph<RegisterPressure>& iGraph = registerAllocator.iGraph;
+
+	bool coloringFailed = false;
+	while (coloringStackPtr > coloringStack) {
+		RegisterName range = *--coloringStackPtr;
+
+		if (!cost[range].infinite && cost[range].cost < 0) {
+			coloringFailed = true;
+			spill.set(range);
+			UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\tfailed to color %d, will spill.\n", range));
+		} else {
+			neighborColors.clear();
+
+			for (InterferenceVector* vector = iGraph.getInterferenceVector(range); vector != NULL; vector = vector->next)
+				for (Int32 i = vector->count - 1; i >= 0; --i) {
+					RegisterName neighborColor = color[vector->neighbors[i]];
+					if (neighborColor < 6) // FIX
+						neighborColors.set(neighborColor);
+				}
+
+			if (neighborColors.getSize() == 6) { // FIX
+				coloringFailed = true;
+				UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\tfailed to color %d, ", range));
+
+				if (!Splits<RegisterPressure>::findSplit(registerAllocator, color, range)) {
+					UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("will spill.\n"));
+					spill.set(range);
+				} else
+					UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("will split.\n"));
+			} else {
+				for (Uint32 i = 0; i < 6; i++) // FIX
+					if (!neighborColors.test(i)) {
+						fprintf(stdout, "\twill color %d as %d\n", range, i);
+						color[range] = RegisterName(i);
+						break;
+					}
+			}
+		}
+	}
+
+#ifdef DEBUG_LOG
+	if (coloringFailed) {
+		UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("Coloring failed:\n"));
+		UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\twill spill: "));
+		spill.printPretty(UT_LOG_MODULE(RegAlloc));
+	} else {
+		UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("Coloring succeeded:\n"));
+		for (Uint32 i = 1; i < rangeCount; i++)
+			UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\trange %d colored as %d\n", i, color[i]));
+	}
+#endif
+
+	return !coloringFailed;
+}
+
+template <class RegisterPressure>
+RegisterName* Coloring<RegisterPressure>::simplify(RegisterAllocator& registerAllocator, RegisterName* coloringStack)
+{
+	InterferenceGraph<RegisterPressure>& iGraph = registerAllocator.iGraph;
+	SpillCost* spillCost = registerAllocator.spillCost;
+	Uint32 rangeCount = registerAllocator.rangeCount;
+
+	Uint32* degree = new Uint32[rangeCount];
+	for (RegisterName i = RegisterName(1); i < rangeCount; i = RegisterName(i + 1)) {
+		InterferenceVector* vector = iGraph.getInterferenceVector(i);
+		degree[i] = (vector != NULL) ? vector->count : 0;
+	}
+
+	Pool& pool = registerAllocator.pool;
+	SparseSet low(pool, rangeCount);
+	SparseSet high(pool, rangeCount);
+	SparseSet highInfinite(pool, rangeCount);
+	SparseSet preColored(pool, rangeCount);
+
+	// Get the precolored registers.
+	//
+	VirtualRegisterManager& vrManager = registerAllocator.vrManager;
+	RegisterName* name2range = registerAllocator.name2range;
+	PreColoredRegister* machineEnd = vrManager.getMachineRegistersEnd();
+	for (PreColoredRegister* machinePtr = vrManager.getMachineRegistersBegin(); machinePtr < machineEnd; machinePtr++)
+		if (machinePtr->id != invalidID)
+			preColored.set(name2range[getName(machinePtr->id)]);
+
+	// Insert the live ranges in the sets.
+	//
+	for (Uint32 range = 1; range < rangeCount; range++)
+		if (!preColored.test(range))
+			if (degree[range] < 6) // FIX
+				low.set(range);
+			else if (!spillCost[range].infinite)
+				high.set(range);
+			else
+				highInfinite.set(range);
+
+#ifdef DEBUG_LOG
+	UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("Coloring sets:\n\tlow = "));
+	low.printPretty(UT_LOG_MODULE(RegAlloc));
+	UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\thigh = "));
+	high.printPretty(UT_LOG_MODULE(RegAlloc));
+	UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\thighInfinite = "));
+	highInfinite.printPretty(UT_LOG_MODULE(RegAlloc));
+	UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\tpreColored = "));
+	preColored.printPretty(UT_LOG_MODULE(RegAlloc));
+#endif // DEBUG_LOG
+
+	RegisterName* coloringStackPtr = coloringStack;
+
+	while (low.getSize() != 0 || high.getSize() != 0) {
+		while (low.getSize() != 0) {
+			RegisterName range = RegisterName(low.getOne());
+			low.clear(range);
+			*coloringStackPtr++ = range;
+
+			for (InterferenceVector* vector = iGraph.getInterferenceVector(range); vector != NULL; vector = vector->next)
+				for (Int32 i = (vector->count - 1); i >= 0; --i) {
+					RegisterName neighbor = vector->neighbors[i];
+					degree[neighbor]--;
+
+					if (degree[neighbor] < 6) // FIX
+						if (high.test(neighbor)) {
+							high.clear(neighbor);
+							low.set(neighbor);
+						} else if (highInfinite.test(neighbor)) {
+							highInfinite.clear(neighbor);
+							low.set(neighbor);
+						}
+				}
+		}
+
+		if (high.getSize() != 0) {
+			RegisterName best = RegisterName(high.getOne());
+			double bestCost = spillCost[best].cost;
+			double bestDegree = degree[best];
+
+			// Choose the next best candidate.
+			//
+			for (SparseSet::iterator i = high.begin(); !high.done(i); i = high.advance(i)) {
+				RegisterName range = RegisterName(high.get(i));
+				double thisCost = spillCost[range].cost;
+				double thisDegree = degree[range];
+
+				if (thisCost * bestDegree < bestCost * thisDegree) {
+					best = range;
+					bestCost = thisCost;
+					bestDegree = thisDegree;
+				}
+			}
+			
+			high.clear(best);
+			low.set(best);
+		}
+ 	}
+	assert(highInfinite.getSize() == 0);
+
+	delete degree;
+
+#ifdef DEBUG_LOG
+	UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("Coloring stack:\n\t"));
+	for (RegisterName* sp = coloringStack; sp < coloringStackPtr; ++sp)
+		UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("%d ", *sp));
+	UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\n"));
+#endif // DEBUG_LOG
+
+	return coloringStackPtr;
+}
+
+
+template <class RegisterPressure>
+bool Coloring<RegisterPressure>::color(RegisterAllocator& registerAllocator)
+{
+	RegisterName* coloringStack = new RegisterName[registerAllocator.rangeCount];
+	return select(registerAllocator, coloringStack, simplify(registerAllocator, coloringStack));
+}

mozilla/ef/Compiler/RegisterAllocator/DominatorGraph.cpp

@@ -0,0 +1,212 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include <string.h>
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+
+#include "DominatorGraph.h"
+
+DominatorGraph::DominatorGraph(ControlGraph& controlGraph) : controlGraph(controlGraph)
+{
+	Uint32 nNodes = controlGraph.nNodes;
+
+	GtoV = new Uint32[nNodes + 1];
+	VtoG = new Uint32[nNodes + 1];
+
+	Uint32 v = 1;
+	for (Uint32 n = 0; n < nNodes; n++) {
+		VtoG[v] = n;
+		GtoV[n] = v++;
+	}
+
+	// Initialize all the 1-based arrays.
+	//
+	parent =	new Uint32[v];
+	semi =		new Uint32[v];
+	vertex =	new Uint32[v];
+	label =		new Uint32[v];
+	size =		new Uint32[v];
+	ancestor =	new Uint32[v];
+	child =		new Uint32[v];
+	dom =		new Uint32[v];
+	bucket =	new DGLinkedList*[v];
+
+	memset(semi, '\0', v * sizeof(Uint32));
+	memset(bucket, '\0', v * sizeof(DGLinkedList*));
+
+	vCount = v;
+
+	build();
+
+	delete parent;
+	delete semi;
+	delete vertex;
+	delete label;
+	delete size;
+	delete ancestor;
+	delete child;
+	delete dom;
+	delete bucket;
+}
+
+Uint32 DominatorGraph::DFS(Uint32 vx, Uint32 n)
+{
+	semi[vx] = ++n;
+	vertex[n] = label[vx] = vx;
+	ancestor[vx] = child[vx] = 0;
+	size[vx] = 1;
+
+
+	ControlNode& node = *controlGraph.dfsList[VtoG[vx]];
+	ControlEdge* successorEnd = node.getSuccessorsEnd();
+	for (ControlEdge* successorPtr = node.getSuccessorsBegin(); successorPtr < successorEnd; successorPtr++) {
+		Uint32 w = GtoV[successorPtr->getTarget().dfsNum];
+		if (semi[w] == 0) {
+			parent[w] = vx;
+			n = DFS(w, n);
+		}
+	}
+	return n;
+}
+
+void DominatorGraph::LINK(Uint32 vx, Uint32 w)
+{
+    Uint32 s = w;
+
+	while (semi[label[w]] < semi[label[child[s]]]) {
+		if (size[s] + size[child[child[s]]] >= (size[child[s]] << 1)) {
+			ancestor[child[s]] = s;
+			child[s] = child[child[s]];
+		} else {
+			size[child[s]] = size[s];
+			s = ancestor[s] = child[s];
+		}
+	}
+	label[s] = label[w];
+	size[vx] += size[w];
+	if(size[vx] < (size[w] << 1)) {
+		Uint32 t = s;
+		s = child[vx]; 
+		child[vx] = t;
+	}
+	while( s != 0 ) {
+		ancestor[s] = vx;
+		s = child[s];
+	}
+}
+
+
+void DominatorGraph::COMPRESS(Uint32 vx)
+{
+	if(ancestor[ancestor[vx]] != 0) {
+		COMPRESS(ancestor[vx]);
+		if(semi[label[ancestor[vx]]] < semi[label[vx]])
+			label[vx] = label[ancestor[vx]];
+		ancestor[vx] = ancestor[ancestor[vx]];
+	}
+}
+
+Uint32 DominatorGraph::EVAL(Uint32 vx)
+{
+	if(ancestor[vx] == 0) 
+		return label[vx];
+	COMPRESS(vx);
+	return (semi[label[ancestor[vx]]] >= semi[label[vx]]) ? label[vx] : label[ancestor[vx]];
+}
+
+void DominatorGraph::build()
+{
+	Uint32 n = DFS(GtoV[0], 0);
+	size[0] = label[0] = semi[0];
+
+	for (Uint32 i = n; i >= 2; i--) {
+		Uint32 w = vertex[i];
+
+		ControlNode& node = *controlGraph.dfsList[VtoG[w]];
+		const DoublyLinkedList<ControlEdge>& predecessors = node.getPredecessors();
+		for (DoublyLinkedList<ControlEdge>::iterator p = predecessors.begin(); !predecessors.done(p); p = predecessors.advance(p)) {
+			Uint32 vx = GtoV[predecessors.get(p).getSource().dfsNum];
+			Uint32 u = EVAL(vx);
+
+			if(semi[u] < semi[w])
+				semi[w] = semi[u];
+		}
+
+		DGLinkedList* elem = new DGLinkedList();
+		elem->next = bucket[vertex[semi[w]]];
+		elem->index = w;
+		bucket[vertex[semi[w]]] = elem;
+
+		LINK(parent[w], w);
+
+		elem = bucket[parent[w]];
+		while(elem != NULL) {
+			Uint32 vx = elem->index;
+			Uint32 u = EVAL(vx);
+			dom[vx] = (semi[u] < semi[vx]) ? u : parent[w];
+			elem = elem->next;
+		}
+	}
+
+	memset(size, '\0', n * sizeof(Uint32));
+	Pool& pool = controlGraph.pool;
+	nodes = new(pool) DGNode[n];
+
+	for(Uint32 j = 2; j <= n; j++) {
+		Uint32 w = vertex[j];
+		Uint32 d = dom[w];
+		if(d != vertex[semi[w]]) {
+			d = dom[d];
+			dom[w] = d;
+		}
+		size[d]++;
+	}
+	dom[GtoV[0]] = 0;
+
+	for (Uint32 k = 1; k <= n; k++) {
+		DGNode& node = nodes[VtoG[k]];
+		Uint32 count = size[k];
+		node.successorsEnd = node.successorsBegin = (count) ? new(pool) Uint32[count] : (Uint32*) 0;
+	}
+
+	for (Uint32 l = 2; l <= n; l++)
+		*(nodes[VtoG[dom[l]]].successorsEnd)++ = VtoG[l];
+}
+
+#ifdef DEBUG_LOG
+void DominatorGraph::printPretty(LogModuleObject log)
+{
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Dominator Graph:\n"));
+	Uint32 nNodes = controlGraph.nNodes;
+	for (Uint32 i = 0; i < nNodes; i++) {
+		DGNode& node = nodes[i];
+		if (node.successorsBegin !=  node.successorsEnd) {
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\tN%d dominates ", i));
+			for (Uint32* successorsPtr = node.successorsBegin; successorsPtr < node.successorsEnd; successorsPtr++)
+				UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("N%d ", *successorsPtr));
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\n"));
+		}
+	}
+}
+
+#endif // DEBUG_LOG
+
+
+

mozilla/ef/Compiler/RegisterAllocator/DominatorGraph.h

@@ -0,0 +1,80 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _DOMINATOR_GRAPH_H_
+#define _DOMINATOR_GRAPH_H_
+
+#include "LogModule.h"
+
+class ControlGraph;
+
+struct DGNode
+{
+	Uint32* successorsBegin;
+	Uint32* successorsEnd;
+};
+
+struct DGLinkedList
+{
+	DGLinkedList* next;
+	Uint32 index;
+};
+
+class DominatorGraph
+{
+private:
+
+	ControlGraph&	controlGraph;
+
+	Uint32			vCount;
+
+	Uint32*			VtoG;
+	Uint32*			GtoV;
+	Uint32*			parent;
+	Uint32*			semi;
+	Uint32*			vertex;
+	Uint32*			label;
+	Uint32*			size;
+	Uint32*			ancestor;
+	Uint32*			child;
+	Uint32*			dom;
+	DGLinkedList**	bucket;
+	DGNode*			nodes;
+	
+private:
+
+	void build();
+
+	Uint32 DFS(Uint32 vx, Uint32 n);
+	void LINK(Uint32 vx, Uint32 w);
+	void COMPRESS(Uint32 vx);
+	Uint32 EVAL(Uint32 vx);
+
+public:
+
+	DominatorGraph(ControlGraph& controlGraph);
+
+	Uint32* getSuccessorsBegin(Uint32 n) const {return nodes[n].successorsBegin;}
+	Uint32* getSuccessorsEnd(Uint32 n) const {return nodes[n].successorsEnd;}
+
+#ifdef DEBUG_LOG
+	void printPretty(LogModuleObject log);
+#endif // DEBUG_LOG
+};
+
+#endif // _DOMINATOR_GRAPH_H_

mozilla/ef/Compiler/RegisterAllocator/HashSet.cpp

@@ -0,0 +1,20 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "HashSet.h"

mozilla/ef/Compiler/RegisterAllocator/HashSet.h

@@ -0,0 +1,97 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _HASH_SET_H_
+#define _HASH_SET_H_
+
+#include "Fundamentals.h"
+#include "Pool.h"
+#include <string.h>
+
+struct HashSetElement
+{
+	Uint32 index;
+	HashSetElement* next;
+};
+
+class HashSet
+{
+private:
+
+	static const hashSize = 64;
+
+	// Return the hash code for the given element index.
+	static Uint32 getHashCode(Uint32 index) {return index & (hashSize - 1);} // Could be better !
+
+private:
+
+	Pool&					allocationPool;
+	HashSetElement**		bucket;
+	HashSetElement*			free;
+
+private:
+
+	// No copy constructor.
+	HashSet(const HashSet&);
+	// No copy operator.
+	void operator = (const HashSet&);
+
+public:
+
+	// Create a new HashSet.
+	inline HashSet(Pool& pool, Uint32 universeSize);
+
+	// Clear the hashset.
+	void clear();
+	// Clear the element for the given index.
+	void clear(Uint32 index);
+	// Set the element for the given index.
+	void set(Uint32 index);
+	// Return true if the element at index is a member.
+	bool test(Uint32 index) const;
+	// Union with the given hashset.
+	inline void or(const HashSet& set);
+	// Intersection with the given hashset.
+	inline void and(const HashSet& set);
+	// Difference with the given hashset.
+	inline void difference(const HashSet& set);
+
+	// Logical operators.
+	HashSet& operator |= (const HashSet& set) {or(set); return *this;}
+	HashSet& operator &= (const HashSet& set) {and(set); return *this;}
+	HashSet& operator -= (const HashSet& set) {difference(set); return *this;}
+
+	// Iterator to conform with the set API.
+	typedef HashSetElement* iterator;
+	// Return the iterator for the first element of this set.
+	iterator begin() const;
+	// Return the next iterator.
+	iterator advance(iterator pos) const;
+	// Return true if the iterator is at the end of the set.
+	bool done(iterator pos) const {return pos == NULL;}
+};
+
+
+inline HashSet::HashSet(Pool& pool, Uint32 /*universeSize*/)
+	: allocationPool(pool), free(NULL)
+{
+	bucket = new(pool) HashSetElement*[hashSize];
+	memset(bucket, '\0', sizeof(HashSetElement*));
+}
+
+#endif // _HASH_SET_H_

mozilla/ef/Compiler/RegisterAllocator/IndexedPool.h

@@ -0,0 +1,213 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _INDEXED_POOL_H_
+#define _INDEXED_POOL_H_
+
+#include "Fundamentals.h"
+#include <string.h>
+#include <stdlib.h>
+
+//------------------------------------------------------------------------------
+// IndexedPool<IndexedObjectSubclass> is an indexed pool of objects. The
+// template parameter 'IndexedObjectSubclass' must be a subclass of the struct
+// IndexedObject.
+// 
+// When the indexed pool is ask to allocate and initialize a new object (using
+// the operator new(anIndexedPool) it will zero the memory used to store the
+// object and initialize the field 'index' of this object to its position in
+// the pool.
+//
+// An object allocated by the indexed pool can be freed by calling the method
+// IndexedPool::release(IndexedElement& objectIndex).
+//
+// example:
+//
+//   IndexedPool<IndexedElement> elementPool;
+//
+//   IndexedElement& element1 = *new(elementPool) IndexedElement();
+//   IndexedElement& element2 = *new(elementPool) IndexedElement();
+//
+//   indexedPool.release(element1);
+//   IndexedElement& element3 = *new(elementPool) IndexedElement();
+//
+//  At this point element1 is no longer a valid object, element2 is at
+//  index 2 and element3 is at index 1.
+//
+
+//------------------------------------------------------------------------------
+// IndexedObject -
+//
+
+template<class Object>
+struct IndexedObject
+{
+	Uint32			index;				// Index in the pool.
+	Object*			next;				// Used to link IndexedObject together.
+
+	Uint32 getIndex() {return index;}
+};
+
+//------------------------------------------------------------------------------
+// IndexedPool<IndexedObject> -
+//
+
+template <class IndexedObject>
+class IndexedPool
+{
+private:
+
+	static const	blockSize = 4;		// Size of one block.
+
+	Uint32			nBlocks;			// Number of blocks in the pool.
+	IndexedObject**	block;				// Array of block pointers.
+	IndexedObject*  freeObjects;		// Chained list of free IndexedObjects.
+	Uint32			nextIndex;			// Index of the next free object in the last block. 
+
+private:
+
+	void allocateAnotherBlock();
+	IndexedObject& newObject();
+
+public:
+
+	IndexedPool() : nBlocks(0), block(NULL), freeObjects(NULL), nextIndex(1) {}
+	~IndexedPool();
+
+	IndexedObject& get(Uint32 index) const;
+	void release(IndexedObject& object);
+
+	void setSize(Uint32 size) {assert(size < nextIndex); nextIndex = size;}
+
+	// Return the universe size.
+	Uint32 getSize() {return nextIndex;}
+
+	friend void* operator new(size_t, IndexedPool<IndexedObject>& pool); // Needs to call newObject().
+};
+
+// Free all the memory allocated for this object.
+//
+template <class IndexedObject>
+IndexedPool<IndexedObject>::~IndexedPool()
+{
+	for (Uint32 n = 0; n < nBlocks; n++)
+		free(&((IndexedObject **) &block[n][n*blockSize])[-(n + 1)]);
+}
+
+// Release the given. This object will be iserted in the chained
+// list of free IndexedObjects. To minimize the fragmentation the chained list 
+// is ordered by ascending indexes.
+//
+template <class IndexedObject>
+void IndexedPool<IndexedObject>::release(IndexedObject& object)
+{
+	Uint32 index = object.index;
+	IndexedObject* list = freeObjects;
+
+	assert(&object == &get(index)); // Make sure that object is owned by this pool.
+
+	if (list == NULL) { // The list is empty.
+		freeObjects = &object;
+		object.next = NULL;
+	} else { // The list contains at least 1 element.
+		if (index < list->index) { // insert as first element.
+			freeObjects = &object;
+			object.next = list;
+		} else { // Find this object's place.
+			while ((list->next) != NULL && (list->next->index < index))
+				list = list->next;
+
+			object.next = list->next;
+			list->next = &object;
+		}
+	}
+
+#ifdef DEBUG
+	// Sanity check to be sure that the list is correctly ordered.
+	for (IndexedObject* obj = freeObjects; obj != NULL; obj = obj->next)
+		if (obj->next != NULL)
+			assert(obj->index < obj->next->index);
+#endif
+}
+
+// Create a new block of IndexedObjects. We will allocate the memory to
+// store IndexedPool::blockSize IndexedObject and the new Array of block
+// pointers.
+// The newly created IndexedObjects will not be initialized.
+//
+template <class IndexedObject>
+void IndexedPool<IndexedObject>::allocateAnotherBlock()
+{
+	void* memory = (void *) malloc((nBlocks + 1) * sizeof(Uint32) + blockSize * sizeof(IndexedObject));
+
+	memcpy(memory, block, nBlocks * sizeof(Uint32));
+
+	block = (IndexedObject **) memory;
+	IndexedObject* objects = (IndexedObject *) &block[nBlocks + 1];
+
+	block[nBlocks] = &objects[-(nBlocks * blockSize)];
+	nBlocks++;
+}
+
+// Return the IndexedObject at the position 'index' in the pool.
+//
+template <class IndexedObject>
+IndexedObject& IndexedPool<IndexedObject>::get(Uint32 index) const
+{
+	Uint32 blockIndex = index / blockSize;
+	assert(blockIndex < nBlocks);
+
+	return block[blockIndex][index];
+}
+
+// Return the reference of an unused object in the pool.
+//
+template <class IndexedObject>
+IndexedObject& IndexedPool<IndexedObject>::newObject()
+{
+	if (freeObjects != NULL) {
+		IndexedObject& newObject = *freeObjects;
+		freeObjects = newObject.next;
+		return newObject;
+	}
+	
+	Uint32 nextIndex = this->nextIndex++;
+	Uint32 blockIndex = nextIndex / blockSize;
+
+	while (blockIndex >= nBlocks)
+		allocateAnotherBlock();
+	
+	IndexedObject& newObject = block[blockIndex][nextIndex];
+	newObject.index = nextIndex;
+
+	return newObject;
+}
+
+// Return the address of the next unsused object in the given
+// indexed pool. The field index of the newly allocated object
+// will be initialized to the corresponding index of this object
+// in the pool.
+//
+template <class IndexedObject>
+void* operator new(size_t size, IndexedPool<IndexedObject>& pool)
+{
+	assert(size == sizeof(IndexedObject));
+	return (void *) &pool.newObject();
+}
+
+#endif // _INDEXED_POOL_H_

mozilla/ef/Compiler/RegisterAllocator/InterferenceGraph.h

@@ -0,0 +1,258 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _INTERFERENCE_GRAPH_H_
+#define _INTERFERENCE_GRAPH_H_
+
+#include "Fundamentals.h"
+#include "ControlGraph.h"
+#include "Primitives.h"
+#include "Instruction.h"
+#include "VirtualRegister.h"
+#include "RegisterPressure.h"
+#include "SparseSet.h"
+#include <string.h>
+
+struct InterferenceVector
+{
+	Uint32				count;
+	InterferenceVector*	next;
+	RegisterName*		neighbors;
+
+	InterferenceVector() : count(0), next(NULL) {}
+};
+
+class RegisterAllocator;
+
+template <class RegisterPressure>
+class InterferenceGraph
+{
+private:
+
+	RegisterAllocator&			registerAllocator;
+
+	RegisterPressure::Set*		interferences;
+	InterferenceVector**		vector;
+	Uint32*						offset;
+	Uint32 						rangeCount;
+
+private:
+
+	// No copy constructor.
+	InterferenceGraph(const InterferenceGraph&);
+	// No copy operator.
+	void operator = (const InterferenceGraph&);
+
+	// Check if reg is a member of the universe.
+	void checkMember(RegisterName name) {assert(name < rangeCount);}
+	// Return the edge index for the interference between name1 and name2.
+	Uint32 getEdgeIndex(RegisterName name1, RegisterName name2);
+
+public:
+	InterferenceGraph(RegisterAllocator& registerAllocator) : registerAllocator(registerAllocator) {}
+
+	// Calculate the interferences.
+	void build();
+	// Return true if reg1 and reg2 interfere.
+	bool interfere(RegisterName name1, RegisterName name2);
+	// Return the interference vector for the given register or NULL if there is none.
+	InterferenceVector* getInterferenceVector(RegisterName name) {return vector[name];}
+	// Set the interference between name1 and name2.
+	void setInterference(RegisterName name1, RegisterName name2);
+	// Set the interference vector for the given register.
+	void setInterferenceVector(RegisterName name, InterferenceVector* v) {vector[name] = v;}
+
+
+#ifdef DEBUG_LOG
+	// Print the interferences.
+	void printPretty(LogModuleObject log);
+#endif // DEBUG_LOG
+};
+
+template <class RegisterPressure>
+void InterferenceGraph<RegisterPressure>::build()
+{
+	Pool& pool = registerAllocator.pool;
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	this->rangeCount = rangeCount;
+
+	// Initialize the structures.
+	//
+	offset = new(pool) Uint32[rangeCount + 1];
+	vector = new(pool) InterferenceVector*[rangeCount];
+	memset(vector, '\0', sizeof(InterferenceVector*) * rangeCount);
+
+	Uint32 o = 0;
+	offset[0] = 0;
+	for (Uint32 i = 1; i <= rangeCount; ++i) {
+		offset[i] = o;
+		o += i;
+	}
+	
+	interferences = new(pool) RegisterPressure::Set(pool, (rangeCount * rangeCount) / 2);
+
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	RegisterName* name2range = registerAllocator.name2range;
+	LivenessInfo<RegisterPressure> liveness = Liveness<RegisterPressure>::analysis(controlGraph, rangeCount, name2range);
+	registerAllocator.liveness = liveness;
+	SparseSet currentLive(pool, rangeCount);
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		ControlNode& node = *nodes[n];
+		currentLive = liveness.liveOut[n];
+
+		InstructionList& instructions = node.getInstructions();
+		for (InstructionList::iterator i = instructions.end(); !instructions.done(i); i = instructions.retreat(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useBegin = instruction.getInstructionUseBegin();
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			InstructionUse* usePtr;
+			InstructionDefine* defineBegin = instruction.getInstructionDefineBegin();
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			InstructionDefine* definePtr;
+
+			// Handle the copy instruction to avoid unnecessary interference between the 2 registers.
+			if ((instruction.getFlags() & ifCopy) != 0) {
+				assert(useBegin != useEnd && useBegin[0].isRegister());
+				currentLive.clear(name2range[useBegin[0].getRegisterName()]);
+			}
+
+			// Create the interferences.
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister()) {
+					RegisterName define = name2range[definePtr->getRegisterName()];
+
+					for (SparseSet::iterator e = currentLive.begin(); !currentLive.done(e); e = currentLive.advance(e)) {
+						RegisterName live = RegisterName(currentLive.get(e));
+
+						if ((live != define) && !interfere(live, define) && registerAllocator.canInterfere(live, define)) {
+
+							if (vector[define] == NULL)
+								vector[define] = new(pool) InterferenceVector();
+							vector[define]->count++;
+
+							if (vector[live] == NULL)
+								vector[live] = new(pool) InterferenceVector();
+							vector[live]->count++;
+
+							setInterference(live, define);
+						}
+					}
+				}
+
+			// Now update the liveness.
+			//
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					currentLive.clear(name2range[definePtr->getRegisterName()]);
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					currentLive.set(name2range[usePtr->getRegisterName()]);
+		}
+	}
+
+	// Allocate the memory to store the interferences.
+	//
+	for (Uint32 e = 0; e < rangeCount; e++)
+		if (vector[e] != NULL) {
+			InterferenceVector& v = *vector[e];
+			v.neighbors = new(pool) RegisterName[v.count];
+			v.count = 0;
+		}
+
+	// Initialize the edges.
+	//
+	if (RegisterPressure::Set::isOrdered()) {
+		RegisterName name1 = RegisterName(0);
+
+		for (RegisterPressure::Set::iterator i = interferences->begin(); !interferences->done(i); i = interferences->advance(i)) {
+			Uint32 interferenceIndex = interferences->get(i);
+
+			while(interferenceIndex >= offset[name1 + 1])
+				name1 = RegisterName(name1 + 1);
+
+			assert((interferenceIndex >= offset[name1]) && (interferenceIndex < offset[name1 + 1]));
+
+			RegisterName name2 = RegisterName(interferenceIndex - offset[name1]);
+
+			assert(interfere(name1, name2));
+
+			InterferenceVector& vector1 = *vector[name1];
+			vector1.neighbors[vector1.count++] = name2;
+		
+			InterferenceVector& vector2 = *vector[name2];
+			vector2.neighbors[vector2.count++] = name1;
+		}
+	} else {
+		trespass("not Implemented"); // FIX: need one more pass to initialize the vectors.
+	}
+}
+
+template <class RegisterPressure>
+Uint32 InterferenceGraph<RegisterPressure>::getEdgeIndex(RegisterName name1, RegisterName name2)
+{
+	checkMember(name1); checkMember(name2);
+	assert(name1 != name2); // This is not possible.
+	return (name1 < name2) ? offset[name2] + name1 : offset[name1] + name2;
+}
+
+template <class RegisterPressure>
+void InterferenceGraph<RegisterPressure>::setInterference(RegisterName name1, RegisterName name2)
+{
+	interferences->set(getEdgeIndex(name1, name2));
+}
+
+template <class RegisterPressure>
+bool InterferenceGraph<RegisterPressure>::interfere(RegisterName name1, RegisterName name2)
+{
+	return interferences->test(getEdgeIndex(name1, name2));
+}
+
+#ifdef DEBUG_LOG
+template <class RegisterPressure>
+void InterferenceGraph<RegisterPressure>::printPretty(LogModuleObject log)
+{
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Interference Vectors:\n"));
+	for (Uint32 i = 1; i < rangeCount; i++) {
+		if (vector[i] != NULL) {
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\tvr%d: (", i));
+			for (InterferenceVector* v = vector[i]; v != NULL; v = v->next)
+				for (Uint32 j = 0; j < v->count; j++) {
+					UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("%d", v->neighbors[j]));
+					if (v->next != NULL || j != (v->count - 1))
+						UT_OBJECTLOG(log, PR_LOG_ALWAYS, (","));
+				}
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, (")\n"));
+		}
+	}
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Interference Matrix:\n"));
+	for (RegisterName name1 = RegisterName(1); name1 < rangeCount; name1 = RegisterName(name1 + 1)) {
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\t%d:\t", name1));
+		for (RegisterName name2 = RegisterName(1); name2 < rangeCount; name2 = RegisterName(name2 + 1))
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("%c", ((name1 != name2) && interfere(name1, name2)) ? '1' : '0'));
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\n"));
+	}
+}
+#endif // DEBUG_LOG
+
+#endif // _INTERFERENCE_GRAPH_H_

mozilla/ef/Compiler/RegisterAllocator/LiveRange.h

@@ -0,0 +1,87 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _LIVE_RANGE_H_
+#define _LIVE_RANGE_H_
+
+#include "Fundamentals.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Primitives.h"
+#include "Instruction.h"
+#include "RegisterAllocator.h"
+#include "RegisterAllocatorTools.h"
+
+template <class RegisterPressure>
+struct LiveRange
+{
+	static void build(RegisterAllocator& registerAllocator);
+};
+
+template <class RegisterPressure>
+void LiveRange<RegisterPressure>::build(RegisterAllocator& registerAllocator)
+{
+	// Intialize the lookup table.
+	//
+	Uint32 nameCount = registerAllocator.nameCount;
+	RegisterName* nameTable = new(registerAllocator.pool) RegisterName[2*nameCount];
+	RegisterName* rangeName = &nameTable[nameCount];
+
+	init(rangeName, nameCount);
+
+	// Walk the graph.
+	//
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	SparseSet destination(registerAllocator.pool, nameCount);
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		InstructionList& phiNodes = nodes[n]->getPhiNodeInstructions();
+
+		destination.clear();
+		for (InstructionList::iterator i = phiNodes.begin(); !phiNodes.done(i); i = phiNodes.advance(i)) {
+			Instruction& phiNode = phiNodes.get(i);
+			assert(phiNode.getInstructionDefineBegin() != phiNode.getInstructionDefineEnd() && phiNode.getInstructionDefineBegin()[0].isRegister());
+			destination.set(findRoot(phiNode.getInstructionDefineBegin()[0].getRegisterName(), rangeName));
+		}
+
+		for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+			Instruction& phiNode = phiNodes.get(p);
+
+			assert(phiNode.getInstructionDefineBegin() != phiNode.getInstructionDefineEnd() && phiNode.getInstructionDefineBegin()[0].isRegister());
+			RegisterName destinationName = phiNode.getInstructionDefineBegin()[0].getRegisterName();
+			RegisterName destinationRoot = findRoot(destinationName, rangeName);
+
+			InstructionUse* useEnd = phiNode.getInstructionUseEnd();
+			for (InstructionUse* usePtr = phiNode.getInstructionUseBegin(); usePtr < useEnd; usePtr++) {
+				assert(usePtr->isRegister());
+				RegisterName sourceName = usePtr->getRegisterName();
+				RegisterName sourceRoot = findRoot(sourceName, rangeName);
+
+				if (sourceRoot != destinationRoot && !destination.test(sourceRoot))
+					rangeName[sourceRoot] = destinationRoot;
+			}
+		}
+	}
+
+	registerAllocator.rangeCount = compress(registerAllocator.name2range, rangeName, nameCount, nameCount);
+}
+
+#endif // _LIVE_RANGE_H_

mozilla/ef/Compiler/RegisterAllocator/LiveRangeGraph.h

@@ -0,0 +1,163 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _LIVE_RANGE_GRAPH_
+#define _LIVE_RANGE_GRAPH_
+
+#include "Fundamentals.h"
+#include "Pool.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "RegisterTypes.h"
+
+class RegisterAllocator;
+
+template <class RegisterPressure>
+class LiveRangeGraph
+{
+private:
+
+	RegisterAllocator&		registerAllocator;
+
+	RegisterPressure::Set*	edges;
+	Uint32					rangeCount;
+
+public:
+	//
+	//
+	LiveRangeGraph(RegisterAllocator& registerAllocator) : registerAllocator(registerAllocator) {}
+
+	//
+	//
+	void build();
+
+	//
+	//
+	void addEdge(RegisterName name1, RegisterName name2);
+
+	//
+	//
+	bool haveEdge(RegisterName name1, RegisterName name2);
+
+#ifdef DEBUG_LOG
+	//
+	//
+	void printPretty(LogModuleObject log);
+#endif // DEBUG_LOG
+};
+
+template <class RegisterPressure>
+void LiveRangeGraph<RegisterPressure>::build()
+{
+	Pool& pool = registerAllocator.pool;
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	this->rangeCount = rangeCount;
+
+	edges = new(pool) RegisterPressure::Set(pool, rangeCount * rangeCount);
+
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	RegisterName* name2range = registerAllocator.name2range;
+	LivenessInfo<RegisterPressure>& liveness = registerAllocator.liveness;
+	SparseSet currentLive(pool, rangeCount);
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		ControlNode& node = *nodes[n];
+		currentLive = liveness.liveOut[n];
+
+		InstructionList& instructions = node.getInstructions();
+		for (InstructionList::iterator i = instructions.end(); !instructions.done(i); i = instructions.retreat(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useBegin = instruction.getInstructionUseBegin();
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			InstructionUse* usePtr;
+			InstructionDefine* defineBegin = instruction.getInstructionDefineBegin();
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			InstructionDefine* definePtr;
+
+			if ((instruction.getFlags() & ifCopy) != 0) {
+				assert(useBegin != useEnd && useBegin[0].isRegister());
+				currentLive.clear(name2range[useBegin[0].getRegisterName()]);
+			}
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister()) {
+					RegisterName define = name2range[definePtr->getRegisterName()];
+
+					for (SparseSet::iterator l = currentLive.begin(); !currentLive.done(l); l = currentLive.advance(l)) {
+						RegisterName live = RegisterName(currentLive.get(l));
+						if (define != live && registerAllocator.canInterfere(define, live))
+							addEdge(define, live);
+					}
+				}
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					currentLive.clear(name2range[definePtr->getRegisterName()]);
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					currentLive.set(name2range[usePtr->getRegisterName()]);
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					RegisterName use = name2range[usePtr->getRegisterName()];
+
+					for (SparseSet::iterator l = currentLive.begin(); !currentLive.done(l); l = currentLive.advance(l)) {
+						RegisterName live = RegisterName(currentLive.get(l));
+						if (use != live && registerAllocator.canInterfere(use, live))
+							addEdge(use, live);
+					}
+				}
+		}
+	}
+}
+
+template <class RegisterPressure>
+void LiveRangeGraph<RegisterPressure>::addEdge(RegisterName name1, RegisterName name2)
+{
+	assert(name1 != name2);
+	edges->set(name1 * rangeCount + name2);
+}
+
+template <class RegisterPressure>
+bool LiveRangeGraph<RegisterPressure>::haveEdge(RegisterName name1, RegisterName name2)
+{
+	assert(name1 != name2);
+	return edges->test(name1 * rangeCount + name2);
+}
+
+#ifdef DEBUG_LOG
+template <class RegisterPressure>
+void LiveRangeGraph<RegisterPressure>::printPretty(LogModuleObject log)
+{
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Live ranges graph:\n"));
+	for (RegisterName name1 = RegisterName(1); name1 < rangeCount; name1 = RegisterName(name1 + 1)) {
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\t%d:\t", name1));
+		for (RegisterName name2 = RegisterName(1); name2 < rangeCount; name2 = RegisterName(name2 + 1))
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("%c", ((name1 != name2) && haveEdge(name1, name2)) ? '1' : '0'));
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\n"));
+	}
+}
+#endif // DEBUG_LOG
+
+#endif // _LIVE_RANGE_GRAPH_

mozilla/ef/Compiler/RegisterAllocator/Liveness.cpp

@@ -0,0 +1,21 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "Liveness.h"
+

mozilla/ef/Compiler/RegisterAllocator/Liveness.h

@@ -0,0 +1,301 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _LIVENESS_H_
+#define _LIVENESS_H_
+
+#include "Fundamentals.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "RegisterTypes.h"
+
+// ----------------------------------------------------------------------------
+// LivenessInfo -
+
+template <class RegisterPressure>
+struct LivenessInfo
+{
+	RegisterPressure::Set*	liveIn;
+	RegisterPressure::Set*	liveOut;
+	DEBUG_LOG_ONLY(Uint32	size);
+
+#ifdef DEBUG_LOG
+	void printPretty(LogModuleObject log);
+#endif // DEBUG_LOG
+};
+
+// ----------------------------------------------------------------------------
+// Liveness
+//
+// The liveness is defined by the following data-flow equations:
+//
+//		LiveIn(n) = LocalLive(n) U (LiveOut(n) - Killed(n)).
+//		LiveOut(n) = U LiveIn(s) (s a successor of n).
+//
+//	where LocalLive(n) is the set of used registers in the block n, Killed(n)
+//  is the set of defined registers in the block n, LiveIn(n) is the set of
+//  live registers at the begining of the block n and LiveOut(n) is the set
+//  of live registers at the end of the block n.
+//
+//
+// We will compute the liveness analysis in two stages:
+//
+//	1- Build LocalLive(n) (wich is an approximation of LiveIn(n)) and Killed(n) 
+//     for each block n.
+//  2- Perform a backward data-flow analysis to propagate the liveness information
+//     through the entire control-flow graph.
+//
+
+template <class RegisterPressure>
+struct Liveness
+{
+	static LivenessInfo<RegisterPressure> analysis(ControlGraph& controlGraph, Uint32 rangeCount, const RegisterName* name2range);
+	static LivenessInfo<RegisterPressure> analysis(ControlGraph& controlGraph, Uint32 nameCount);
+};
+
+template <class RegisterPressure>
+LivenessInfo<RegisterPressure> Liveness<RegisterPressure>::analysis(ControlGraph& controlGraph, Uint32 rangeCount, const RegisterName* name2range)
+{
+	Pool& pool = controlGraph.pool;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	// Allocate the temporary sets.
+	RegisterPressure::Set* killed = new(pool) RegisterPressure::Set[nNodes](pool, rangeCount);
+	
+	// Allocate the globals sets.
+	RegisterPressure::Set* liveIn = new(pool) RegisterPressure::Set[nNodes](pool, rangeCount);
+	RegisterPressure::Set* liveOut = new(pool) RegisterPressure::Set[nNodes](pool, rangeCount);
+
+	// First stage of the liveness analysis: Compute the sets LocalLive(stored in LiveIn) and Killed.
+	//
+	for (Uint32 n = 0; n < (nNodes - 1); n++) {
+		ControlNode& node = *nodes[n];
+
+		RegisterPressure::Set& currentLocalLive = liveIn[n];
+		RegisterPressure::Set& currentKilled = killed[n];
+		
+		// Find the instructions contributions to the sets LocalLive and Killed.
+		//
+		InstructionList& instructions = node.getInstructions();
+		for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			// If a VirtualRegister is 'used' before being 'defined' then we add it to set LocalLive.
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					Uint32 index = name2range[usePtr->getRegisterName()];
+
+					if (!currentKilled.test(index))
+						currentLocalLive.set(index);
+				}
+
+			// If a Virtualregister is 'defined' then we add it to the set Killed.
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					currentKilled.set(name2range[definePtr->getRegisterName()]);
+		}
+	}
+
+	// Second stage of the liveness analysis: We propagate the LiveIn & LiveOut through the entire 
+	// control-flow graph.
+	//
+	RegisterPressure::Set temp(pool, rangeCount);
+
+	bool changed;
+	do {
+		changed = false;
+
+		// For all nodes is this graph except the endNode.
+		for (Int32 n = (nNodes - 2); n >= 0; n--) {
+			ControlNode& node = *nodes[n];
+
+			RegisterPressure::Set& currentLiveIn = liveIn[n];
+			RegisterPressure::Set& currentLiveOut = liveOut[n];
+
+			// Compute temp = Union of LiveIn(s) (s a successor of this node) | usedByPhiNodes(n).
+			// temp will be the new LiveOut(n).
+			Uint32 nSuccessors = node.nSuccessors();
+			if (nSuccessors != 0) {
+				temp = liveIn[node.nthSuccessor(0).getTarget().dfsNum];
+				for (Uint32 s = 1; s < nSuccessors; s++)
+					temp |= liveIn[node.nthSuccessor(s).getTarget().dfsNum];
+			} else 
+				temp.clear();
+
+			// If temp and LiveOut(n) differ then set LiveOut(n) = temp and recalculate the
+			// new LiveIn(n).
+			if (currentLiveOut != temp) {
+				currentLiveOut = temp;
+				temp -= killed[n]; // FIX: could be optimized with one call to unionDiff !
+				temp |= currentLiveIn;
+			
+				if (currentLiveIn != temp) {
+					currentLiveIn = temp;
+					changed = true;
+				}
+			}
+		}
+	} while(changed);
+
+	LivenessInfo<RegisterPressure> liveness;
+	liveness.liveIn = liveIn;
+	liveness.liveOut = liveOut;
+	DEBUG_LOG_ONLY(liveness.size = nNodes);
+	return liveness;
+}
+
+template <class RegisterPressure>
+LivenessInfo<RegisterPressure> Liveness<RegisterPressure>::analysis(ControlGraph& controlGraph, Uint32 nameCount)
+{
+	Pool& pool = controlGraph.pool;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	// Allocate the temporary sets.
+	RegisterPressure::Set* killed = new(pool) RegisterPressure::Set[nNodes](pool, nameCount);
+	RegisterPressure::Set* usedByPhiNodes = NULL;
+	
+	// Allocate the globals sets.
+	RegisterPressure::Set* liveIn = new(pool) RegisterPressure::Set[nNodes](pool, nameCount);
+	RegisterPressure::Set* liveOut = new(pool) RegisterPressure::Set[nNodes](pool, nameCount);
+
+	// First stage of the liveness analysis: Compute the sets LocalLive(stored in LiveIn) and Killed.
+	//
+	for (Uint32 n = 0; n < (nNodes - 1); n++) {
+		ControlNode& node = *nodes[n];
+
+		RegisterPressure::Set& currentLocalLive = liveIn[n];
+		RegisterPressure::Set& currentKilled = killed[n];
+		
+		InstructionList& phiNodes = node.getPhiNodeInstructions();
+
+		if ((usedByPhiNodes == NULL) && !phiNodes.empty())
+			usedByPhiNodes = new(pool) RegisterPressure::Set[nNodes](pool, nameCount);
+
+		for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+			Instruction& phiNode = phiNodes.get(p);
+
+			InstructionDefine& define = phiNode.getInstructionDefineBegin()[0];
+			currentKilled.set(define.getRegisterName());
+
+			typedef DoublyLinkedList<ControlEdge> ControlEdgeList;
+			const ControlEdgeList& predecessors = node.getPredecessors();
+			ControlEdgeList::iterator p = predecessors.begin();
+			InstructionUse* useEnd = phiNode.getInstructionUseEnd();
+			for (InstructionUse* usePtr = phiNode.getInstructionUseBegin(); usePtr < useEnd; usePtr++, p = predecessors.advance(p))
+				if (usePtr->isRegister())
+					usedByPhiNodes[predecessors.get(p).getSource().dfsNum].set(usePtr->getRegisterName());
+		}
+
+		// Find the instructions contributions to the sets LocalLive and Killed.
+		//
+		InstructionList& instructions = node.getInstructions();
+		for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			// If a VirtualRegister is 'used' before being 'defined' then we add it to set LocalLive.
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					Uint32 index = usePtr->getRegisterName();
+
+					if (!currentKilled.test(index))
+						currentLocalLive.set(index);
+				}
+
+			// If a Virtualregister is 'defined' then we add it to the set Killed.
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					currentKilled.set(definePtr->getRegisterName());
+		}
+	}
+
+	// Second stage of the liveness analysis: We propagate the LiveIn & LiveOut through the entire 
+	// control-flow graph.
+	//
+	RegisterPressure::Set temp(pool, nameCount);
+
+	bool changed;
+	do {
+		changed = false;
+
+		// For all nodes is this graph except the endNode.
+		for (Int32 n = (nNodes - 2); n >= 0; n--) {
+			ControlNode& node = *nodes[n];
+
+			RegisterPressure::Set& currentLiveIn = liveIn[n];
+			RegisterPressure::Set& currentLiveOut = liveOut[n];
+
+			// Compute temp = Union of LiveIn(s) (s a successor of this node) | usedByPhiNodes(n).
+			// temp will be the new LiveOut(n).
+			Uint32 nSuccessors = node.nSuccessors();
+			if (nSuccessors != 0) {
+				temp = liveIn[node.nthSuccessor(0).getTarget().dfsNum];
+				for (Uint32 s = 1; s < nSuccessors; s++)
+					temp |= liveIn[node.nthSuccessor(s).getTarget().dfsNum];
+			} else 
+				temp.clear();
+
+			// Insert the phiNodes contribution.
+			if (usedByPhiNodes != NULL)
+				temp |= usedByPhiNodes[n];
+
+			// If temp and LiveOut(n) differ then set LiveOut(n) = temp and recalculate the
+			// new LiveIn(n).
+			if (currentLiveOut != temp) {
+				currentLiveOut = temp;
+				temp -= killed[n]; // FIX: could be optimized with one call to unionDiff !
+				temp |= currentLiveIn;
+			
+				if (currentLiveIn != temp) {
+					currentLiveIn = temp;
+					changed = true;
+				}
+			}
+		}
+	} while(changed);
+
+	LivenessInfo<RegisterPressure> liveness;
+	liveness.liveIn = liveIn;
+	liveness.liveOut = liveOut;
+	DEBUG_LOG_ONLY(liveness.size = nNodes);
+	return liveness;
+}
+
+#ifdef DEBUG_LOG
+template <class RegisterPressure>
+void LivenessInfo<RegisterPressure>::printPretty(LogModuleObject log)
+{
+	for (Uint32 n = 0; n < size; n++) {
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Node N%d:\n\tliveIn = ", n));
+		liveIn[n].printPretty(log);
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\tliveOut = "));
+		liveOut[n].printPretty(log);
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\n"));
+	}
+}
+#endif // DEBUG_LOG
+
+#endif // _LIVENESS_H_

mozilla/ef/Compiler/RegisterAllocator/Makefile

@@ -0,0 +1,40 @@
+#! gmake
+
+DEPTH			= ../..
+
+MODULE_NAME		= RegisterAllocator
+
+include $(DEPTH)/config/config.mk
+
+INCLUDES		+=										\
+	-I$(DEPTH)/Utilities/General						\
+	-I$(DEPTH)/Utilities/zlib							\
+	-I$(DEPTH)/Runtime/ClassReader						\
+	-I$(DEPTH)/Runtime/NativeMethods					\
+	-I$(DEPTH)/Runtime/System							\
+	-I$(DEPTH)/Runtime/ClassInfo						\
+	-I$(DEPTH)/Runtime/FileReader						\
+	-I$(DEPTH)/Compiler/PrimitiveGraph					\
+	-I$(DEPTH)/Compiler/FrontEnd						\
+	-I$(DEPTH)/Compiler/Optimizer						\
+	-I$(DEPTH)/Compiler/CodeGenerator					\
+	-I$(DEPTH)/Compiler/CodeGenerator/md				\
+	-I$(DEPTH)/Compiler/CodeGenerator/md/$(CPU_ARCH)	\
+	-I$(DEPTH)/Compiler/RegisterAllocator				\
+	-I$(DEPTH)/Driver/StandAloneJava					\
+	-I$(DEPTH)/Debugger									\
+	$(NULL)
+
+CXXSRCS 		=										\
+	RegisterAllocator.cpp								\
+	RegisterAllocatorTools.cpp							\
+	DominatorGraph.cpp									\
+	VirtualRegister.cpp									\
+	BitSet.cpp											\
+	SparseSet.cpp										\
+	$(NULL)
+
+
+include $(DEPTH)/config/rules.mk
+
+libs:: $(MODULE)

mozilla/ef/Compiler/RegisterAllocator/PhiNodeRemover.h

@@ -0,0 +1,392 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _PHI_NODE_REMOVER_H_
+#define _PHI_NODE_REMOVER_H_
+
+#include "Fundamentals.h"
+#include "Pool.h"
+#include "ControlGraph.h"
+#include "DominatorGraph.h"
+#include "VirtualRegister.h"
+#include "RegisterPressure.h"
+#include "Liveness.h"
+#include "Instruction.h"
+#include "InstructionEmitter.h"
+#include "SparseSet.h"
+#include <string.h>
+
+//------------------------------------------------------------------------------
+// RegisterNameNode -
+
+struct RegisterNameNode 
+{
+	RegisterNameNode*	next;
+	RegisterName		newName;
+	Uint32				nextPushed;
+};
+
+//------------------------------------------------------------------------------
+// CopyData -
+
+struct CopyData
+{
+	RegisterName		source;
+	RegisterClassKind	classKind;
+	Uint32				useCount;
+	bool				isLiveOut;
+	RegisterName		sourceNameToUse;
+	RegisterName		temporaryName;
+	RegisterNameNode*	newName;
+};
+
+//------------------------------------------------------------------------------
+// PhiNodeRemover<RegisterPressure> -
+
+template <class RegisterPressure>
+struct PhiNodeRemover
+{
+	// Replace the phi nodes by copy instructions.
+	static void replacePhiNodes(ControlGraph& controlGraph, VirtualRegisterManager& vrManager, InstructionEmitter& emitter);
+};
+
+// Split some of the critical edges and return true if there are still some
+// in the graph after that.
+//
+static bool splitCriticalEdges(ControlGraph& /*cg*/)
+{
+	// FIX: not implemented.
+	return true;
+}
+
+inline void pushName(Pool& pool, RegisterNameNode** stack, SparseSet& pushed, Uint32* nodeListPointer, RegisterName oldName, RegisterName newName)
+{
+	RegisterNameNode& newNode = *new(pool) RegisterNameNode();
+
+	if (pushed.test(oldName))
+		(*stack)->newName = newName;
+	else {
+		newNode.newName = newName;
+		newNode.nextPushed = *nodeListPointer;
+		*nodeListPointer = oldName;
+		newNode.next = *stack;
+		*stack = &newNode;
+		pushed.set(oldName);
+	}
+}
+
+template <class RegisterPressure>
+void PhiNodeRemover<RegisterPressure>::replacePhiNodes(ControlGraph& controlGraph, VirtualRegisterManager& vrManager, InstructionEmitter& emitter)
+{
+	Pool& pool = controlGraph.pool;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	// Initialize the local variables.
+	//
+
+	// When we insert the copies we will also need to create new VirtualRegisters for
+	// the insertion of temporaries. The maximum number of temporary register will not
+	// exceed the number of phiNodes in the primitive graph.
+	Uint32 nameCount = vrManager.getSize();
+	Uint32 maxNameCount = nameCount;
+	for (Uint32 n = 0; n < nNodes; n++)
+		maxNameCount += nodes[n]->getPhiNodes().length();
+
+	// If the CFG contains some critical edges (backward edge which source has more than one
+	// outgoing edge and destination has more than one incomimg edge) then we need the liveness
+	// information to be able to insert temporary copies.
+	RegisterPressure::Set* liveOut = NULL;
+	if (splitCriticalEdges(controlGraph))
+		liveOut = Liveness<LowRegisterPressure>::analysis(controlGraph, nameCount).liveOut;
+
+	DominatorGraph dGraph(controlGraph);
+
+	SparseSet pushed(pool, maxNameCount);
+	SparseSet destinationList(pool, maxNameCount);
+	SparseSet workList(pool, maxNameCount);
+
+	CopyData* copyStats = new(pool) CopyData[maxNameCount];
+	memset(copyStats, '\0', maxNameCount*sizeof(CopyData));
+
+	struct NodeStack {
+		Uint32* next;
+		Uint32* limit;
+		Uint32 pushedList;
+	};
+
+	// Allocate the node stack and initialize the node stack pointer.
+	NodeStack* nodeStack = new(pool) NodeStack[nNodes + 1];
+	NodeStack* nodeStackPtr = nodeStack;
+
+	// We start by the begin node. 
+	Uint32 startNode = 0;
+	Uint32* next = &startNode;
+	Uint32* limit = &startNode + 1;
+
+	while (true) {
+
+		if (next == limit) {
+			// If there are no more node in the sibling, we have to pop the current
+			// frame from the stack and update the copyStats of the pushed nodes.
+			//
+			if (nodeStackPtr == nodeStack)
+				// We are at the bottom of the stack and there are no more nodes
+				// to look at. We are done !
+				break;
+
+			--nodeStackPtr;
+			// We are done with all the children of this node in the dominator tree.
+			// We need to update the copy information of all the new names pushed
+			// during the walk over this node.
+			Uint32 pushedList = nodeStackPtr->pushedList;
+			while (pushedList != 0) {
+				Uint32 nextName = copyStats[pushedList].newName->nextPushed;
+				copyStats[pushedList].newName = copyStats[pushedList].newName->next;
+				pushedList = nextName;
+			}
+
+			// restore the previous frame.
+			next = nodeStackPtr->next;
+			limit = nodeStackPtr->limit;
+		} else {
+			Uint32 currentNode = *next++;
+			Uint32 pushedList = 0;
+
+
+			// Initialize the sets.
+			pushed.clear();
+			destinationList.clear();
+
+			// STEP1:
+			//  Walk the instruction list and to replace all the instruction uses with their new name. 
+			//  If the instruction is a phi node and its defined register is alive at the end of this 
+			//  block then we push the defined register into the stack.
+			//
+			ControlNode& node = *nodes[currentNode];
+			RegisterPressure::Set* currentLiveOut = (liveOut != NULL) ? &liveOut[currentNode] : (RegisterPressure::Set*) 0;
+
+			InstructionList& phiNodes = node.getPhiNodeInstructions();
+			for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+				Instruction& phiNode = phiNodes.get(p);
+
+				InstructionUse* useEnd = phiNode.getInstructionUseEnd();
+				for (InstructionUse* usePtr = phiNode.getInstructionUseBegin(); usePtr < useEnd; usePtr++) {
+					assert(usePtr->isRegister());
+					RegisterName name = usePtr->getRegisterName();
+
+					if (copyStats[name].newName != NULL && copyStats[name].newName->newName != name)
+						usePtr->setRegisterName(copyStats[name].newName->newName);
+				}
+
+				if (currentLiveOut != NULL) {
+					// This is a phi node and we have to push its defined name if it is live
+					// at the end of the node. We only need to do this if the CFG has critical edges.
+					assert(phiNode.getInstructionDefineBegin() != phiNode.getInstructionDefineEnd() && phiNode.getInstructionDefineBegin()[0].isRegister());
+					RegisterName name = phiNode.getInstructionDefineBegin()[0].getRegisterName();
+
+					if (currentLiveOut->test(name))
+						pushName(pool, &(copyStats[name].newName), pushed, &pushedList, name, name);
+				}
+				
+			}
+
+			InstructionList& instructions = node.getInstructions();
+			for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+				Instruction& instruction = instructions.get(i);
+
+				InstructionUse* useEnd = instruction.getInstructionUseEnd();
+				for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+					if (usePtr->isRegister()) {
+						RegisterName name = usePtr->getRegisterName();
+
+						if (copyStats[name].newName != NULL && copyStats[name].newName->newName != name)
+							usePtr->setRegisterName(copyStats[name].newName->newName);
+					}
+			}
+
+			// STEP2:
+			//  Look at this node's successors' phiNodes. We keep track of the number of time
+			//  a VR will be used by another copy instruction and insert each definition into the 
+			//  destinationList. This is the only pass over this node's successors as we will
+			//  get all the information we need in the CopyData structures.
+			//
+			ControlEdge* successorEdgeEnd = node.getSuccessorsEnd();
+			for (ControlEdge* successorEdgePtr = node.getSuccessorsBegin(); successorEdgePtr < successorEdgeEnd; successorEdgePtr++) {
+				Uint32 useIndex = successorEdgePtr->getIndex();
+				ControlNode& successor = successorEdgePtr->getTarget();
+
+				// Look at its phi nodes. The phi nodes are at the top of the instruction list. We exit
+				// as soon as we find an instruction which is not a phi node
+				InstructionList& phiNodes = successor.getPhiNodeInstructions();
+				for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+					Instruction& phiNode = phiNodes.get(p);
+					
+					assert((phiNode.getInstructionUseBegin() + useIndex) < phiNode.getInstructionUseEnd());
+					assert(phiNode.getInstructionDefineBegin() != phiNode.getInstructionDefineEnd());
+
+					InstructionUse& source = phiNode.getInstructionUseBegin()[useIndex];
+					InstructionDefine& destination = phiNode.getInstructionDefineBegin()[0];
+
+					assert(source.isRegister() && destination.isRegister());
+
+					RegisterName sourceName = source.getRegisterName();
+					RegisterName destinationName = destination.getRegisterName();
+
+					// Get the correct name for the source.
+					if (copyStats[sourceName].newName != NULL)
+						sourceName = copyStats[sourceName].newName->newName;
+
+					// Update the CopyData structures.
+					if ((sourceName != rnInvalid) && (sourceName != destinationName)) {
+						copyStats[destinationName].source = sourceName;
+						copyStats[destinationName].classKind = destination.getRegisterClass();
+						copyStats[destinationName].isLiveOut = (currentLiveOut != NULL) ? currentLiveOut->test(destinationName) : false;
+						copyStats[destinationName].sourceNameToUse = destinationName;
+						copyStats[sourceName].sourceNameToUse = sourceName;
+						copyStats[sourceName].useCount++;
+						destinationList.set(destinationName);
+					}
+				}
+			}
+
+			// STEP3:
+			//  Insert into the worklist only the destination registers that will be not used in 
+			//  another copy instruction in this block.
+			//
+			assert(workList.getSize() == 0);
+			for (SparseSet::iterator d = destinationList.begin(); !destinationList.done(d); d = destinationList.advance(d)) {
+				Uint32 dest = destinationList.get(d);
+				if (copyStats[dest].useCount == 0)
+					workList.set(dest);
+			}
+
+			// STEP4:
+			//  Insert the copy instructions.
+			//
+			Uint32 destinationListSize = destinationList.getSize();
+			InstructionList::iterator endOfTheNode = instructions.end();
+
+			// Find the right place to insert the copy instructions.
+			if (destinationListSize != 0)
+				while (instructions.get(endOfTheNode).getFlags() & ifControl)
+					endOfTheNode = instructions.retreat(endOfTheNode);
+
+			while (destinationListSize != 0) {
+				while(workList.getSize()) {
+					RegisterName destinationName = RegisterName(workList.getOne());
+					RegisterName sourceName = copyStats[destinationName].source;
+
+					workList.clear(destinationName);
+					if (copyStats[destinationName].isLiveOut && !copyStats[destinationName].temporaryName) {
+						// Lost copy problem.
+						copyStats[destinationName].isLiveOut = false;
+
+						RegisterName sourceName = destinationName;
+						RegisterClassKind classKind = copyStats[sourceName].classKind;
+						RegisterName destinationName = getName(vrManager.newVirtualRegister(classKind));
+						assert(destinationName < maxNameCount);
+
+						copyStats[destinationName].classKind = classKind;
+						copyStats[sourceName].useCount = 0;
+				
+						// We need to insert a copy to a temporary register to keep the
+						// source register valid at the end of the node defining it.
+						// This copy will be inserted right after the phi node defining it.
+						RegisterName from = copyStats[sourceName].sourceNameToUse;
+						Instruction* definingPhiNode = vrManager.getVirtualRegister(from).getDefiningInstruction();
+						assert(definingPhiNode && (definingPhiNode->getFlags() & ifPhiNode) != 0);
+
+						RegisterID fromID = buildRegisterID(from, classKind);
+						RegisterID toID = buildRegisterID(destinationName, classKind);
+						Instruction& copy = emitter.newCopy(*definingPhiNode->getPrimitive(), fromID, toID);
+						vrManager.getVirtualRegister(destinationName).setDefiningInstruction(copy);
+						definingPhiNode->getPrimitive()->getContainer()->getInstructions().addFirst(copy);
+
+						copyStats[sourceName].temporaryName = destinationName;
+						copyStats[sourceName].sourceNameToUse = destinationName;
+						pushName(pool, &(copyStats[sourceName].newName), pushed, &pushedList, sourceName, destinationName);
+					}
+
+					// Insert the copy instruction at the end of the current node.
+					RegisterName from = copyStats[sourceName].sourceNameToUse;
+
+					RegisterClassKind classKind = copyStats[destinationName].classKind;
+					RegisterID fromID = buildRegisterID(from, classKind);
+					RegisterID toID = buildRegisterID(destinationName, classKind);
+					Instruction& copy = emitter.newCopy(*vrManager.getVirtualRegister(from).getDefiningInstruction()->getPrimitive(), fromID, toID);
+					instructions.insertAfter(copy, endOfTheNode);
+					endOfTheNode = instructions.advance(endOfTheNode);
+
+					copyStats[sourceName].useCount = 0;
+					if (destinationList.test(sourceName) && copyStats[sourceName].isLiveOut)
+						pushName(pool, &(copyStats[sourceName].newName), pushed, &pushedList, sourceName, destinationName);
+					copyStats[sourceName].isLiveOut = false;
+					copyStats[sourceName].sourceNameToUse = destinationName;
+
+					if (destinationList.test(sourceName))
+						workList.set(sourceName);
+					destinationList.clear(destinationName);
+				}
+
+				destinationListSize = destinationList.getSize();
+				if (destinationListSize != 0) {
+					RegisterName sourceName = RegisterName(destinationList.getOne());
+					RegisterName destinationName;
+
+					if (!copyStats[sourceName].temporaryName) {
+						// Cycle problem.
+						RegisterClassKind classKind = copyStats[sourceName].classKind;
+						destinationName = getName(vrManager.newVirtualRegister(classKind));
+						assert(destinationName < maxNameCount);
+				
+						copyStats[destinationName].classKind = classKind;
+						copyStats[sourceName].temporaryName = destinationName;
+
+						// Insert the copy instruction at the end of the current node.
+						RegisterName from = copyStats[sourceName].sourceNameToUse;
+
+						RegisterID fromID = buildRegisterID(from, classKind);
+						RegisterID toID = buildRegisterID(destinationName, classKind);
+						Instruction& copy = emitter.newCopy(*vrManager.getVirtualRegister(from).getDefiningInstruction()->getPrimitive(), fromID, toID);
+						vrManager.getVirtualRegister(destinationName).setDefiningInstruction(copy);
+						instructions.insertAfter(copy, endOfTheNode);
+						endOfTheNode = instructions.advance(endOfTheNode);
+					} else 
+						destinationName = copyStats[sourceName].temporaryName;
+
+					copyStats[sourceName].useCount = 0;
+					copyStats[sourceName].isLiveOut = false;
+					copyStats[sourceName].sourceNameToUse = destinationName;
+					pushName(pool, &(copyStats[sourceName].newName), pushed, &pushedList, sourceName, destinationName);
+
+					workList.set(sourceName);
+				}
+			}
+
+			nodeStackPtr->pushedList = pushedList;
+			nodeStackPtr->next = next;
+			nodeStackPtr->limit = limit;
+			++nodeStackPtr;
+			next = dGraph.getSuccessorsBegin(currentNode);
+			limit = dGraph.getSuccessorsEnd(currentNode);
+		}
+	}
+}
+
+#endif // _PHI_NODE_REMOVER_H_

mozilla/ef/Compiler/RegisterAllocator/RegisterAllocator.cpp

@@ -0,0 +1,155 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "LogModule.h"
+#include "RegisterAllocator.h"
+#include "RegisterPressure.h"
+#include "RegisterAllocatorTools.h"
+#include "PhiNodeRemover.h"
+#include "LiveRange.h"
+#include "Liveness.h"
+#include "InterferenceGraph.h"
+#include "LiveRangeGraph.h"
+#include "Coalescing.h"
+#include "Spilling.h"
+#include "Coloring.h"
+#include "Splits.h"
+
+class Pool;
+class ControlGraph;
+class VirtualRegisterManager;
+class InstructionEmitter;
+
+UT_DEFINE_LOG_MODULE(RegAlloc);
+
+void RegisterAllocator::allocateRegisters(Pool& pool, ControlGraph& controlGraph, VirtualRegisterManager& vrManager, InstructionEmitter& emitter)
+{
+	// Insert the phi node instructions. We want to do this to have a single defined register per instruction.
+	// If we keep the PhiNode (as a DataNode) and a PhiNode is of DoubleWordKind then we have to execute
+	// some special code for the high word annotation.
+	//
+	RegisterAllocatorTools::insertPhiNodeInstructions(controlGraph, emitter);
+
+	// Perform some tests on the instruction graph.
+	//
+	DEBUG_ONLY(RegisterAllocatorTools::testTheInstructionGraph(controlGraph, vrManager));
+
+	// Replace the phi node instructions by their equivalent copy instructions.
+	//
+	PhiNodeRemover<LowRegisterPressure>::replacePhiNodes(controlGraph, vrManager, emitter);
+
+	// Do the register allocation.
+	//
+	RegisterAllocator registerAllocator(pool, controlGraph, vrManager, emitter);
+	registerAllocator.doGraphColoring();
+}
+
+void RegisterAllocator::doGraphColoring()
+{
+	// Initialize the liverange map.
+	//
+	initLiveRanges();
+
+	// Build the live ranges. We do this to compress the number of RegisterNames 
+	// used in the insterference graph.
+	//
+	LiveRange<LowRegisterPressure>::build(*this);
+
+	// Remove unnecessary copies.
+	//
+	RegisterAllocatorTools::removeUnnecessaryCopies(*this);
+
+	for (Uint8 loop = 0; loop < 10; loop++) {
+
+		UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("********* RegisterAllocator loop %d *********\n", loop));
+
+		while(true) {
+			// Build the interference graph.
+			//
+			iGraph.build();
+
+			// Coalesce the copy instructions.
+			//
+			if (!Coalescing<LowRegisterPressure>::coalesce(*this))
+				break;
+		}
+
+		// Print the interference graph.
+		//
+		DEBUG_LOG_ONLY(iGraph.printPretty(UT_LOG_MODULE(RegAlloc)));
+
+		// Calculate the spill costs.
+		//
+		Spilling<LowRegisterPressure>::calculateSpillCosts(*this);
+		DEBUG_LOG_ONLY(RegisterAllocatorTools::printSpillCosts(*this));
+
+		// Calculate the split costs.
+		//
+		Splits<LowRegisterPressure>::calculateSplitCosts(*this);
+		DEBUG_LOG_ONLY(RegisterAllocatorTools::printSplitCosts(*this));
+
+		// Build the live range graph.
+		//
+		lGraph.build();
+		DEBUG_LOG_ONLY(lGraph.printPretty(UT_LOG_MODULE(RegAlloc)));
+
+		// Color the graph. If it succeeds then we're done with the
+		// register allocation.
+		//
+		if (Coloring<LowRegisterPressure>::color(*this)) {
+			// Write the final colors in the instruction graph.
+			//
+			Coloring<LowRegisterPressure>::finalColoring(*this);
+
+			UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("********** RegisterAllocator  done **********\n"));
+			DEBUG_LOG_ONLY(RegisterAllocatorTools::printInstructions(*this));
+
+			return;
+		}
+
+		// We need to spill some registers.
+		//
+		Spilling<LowRegisterPressure>::insertSpillCode(*this);
+
+		// Insert the split instructions.
+		//
+		Splits<LowRegisterPressure>::insertSplitCode(*this);
+
+		// Update the live ranges.
+		// 
+		// FIX
+	}
+
+#ifdef DEBUG_LOG
+	RegisterAllocatorTools::updateInstructionGraph(*this);
+	RegisterAllocatorTools::printInstructions(*this);
+#endif
+	fprintf(stderr, "!!! Coloring failed after 10 loops !!!\n");
+	abort();
+}
+
+void RegisterAllocator::initLiveRanges()
+{
+	Uint32 count = this->nameCount;
+	RegisterName* name2range = new(pool) RegisterName[nameCount];
+	for (RegisterName r = RegisterName(1); r < count; r = RegisterName(r + 1))
+		name2range[r] = r;
+	this->name2range = name2range;
+	rangeCount = count;
+}

mozilla/ef/Compiler/RegisterAllocator/RegisterAllocator.h

@@ -0,0 +1,88 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _REGISTER_ALLOCATOR_H_
+#define _REGISTER_ALLOCATOR_H_
+
+class Pool;
+class ControlGraph;
+class InstructionEmitter;
+struct SpillCost;
+struct SplitCost;
+
+#include "Liveness.h"
+#include "VirtualRegister.h"
+#include "RegisterPressure.h" // This should included by Backend.cpp
+#include "InterferenceGraph.h"
+#include "LiveRangeGraph.h"
+
+//template <class RegisterPressure>
+class RegisterAllocator
+{
+public:
+
+	Pool&									pool;			//
+	ControlGraph&							controlGraph;	//
+	VirtualRegisterManager&					vrManager;		//
+	InstructionEmitter&						emitter;		//
+
+	RegisterName*							name2range;		//
+	RegisterName*							color;			//
+	SpillCost*								spillCost;		//
+	SparseSet*								willSpill;		//
+	SplitCost*								splitCost;		//
+	NameLinkedList**						splitAround;	//
+	InterferenceGraph<LowRegisterPressure>	iGraph;			//
+	LiveRangeGraph<LowRegisterPressure>		lGraph;			//
+	LivenessInfo<LowRegisterPressure>		liveness;		//
+	Uint32									nameCount;		//
+	Uint32									rangeCount;		//
+	bool									splitFound;		//
+
+private:
+
+	//
+	//
+	void doGraphColoring();
+
+public:
+
+	//
+	//
+	inline RegisterAllocator(Pool& pool, ControlGraph& controlGraph, VirtualRegisterManager& vrManager, InstructionEmitter& emitter);
+
+	//
+	//
+	bool canInterfere(RegisterName /*name1*/, RegisterName /*name2*/) const {return true;}
+
+	//
+	//
+	void initLiveRanges();
+
+	//
+	//
+	static void allocateRegisters(Pool& pool, ControlGraph& controlGraph, VirtualRegisterManager& vrManager, InstructionEmitter& emitter);
+};
+
+//
+//
+inline RegisterAllocator::RegisterAllocator(Pool& pool, ControlGraph& controlGraph, VirtualRegisterManager& vrManager, InstructionEmitter& emitter)
+	: pool(pool), controlGraph(controlGraph), vrManager(vrManager), emitter(emitter), iGraph(*this), lGraph(*this), nameCount(vrManager.getSize()) {}
+
+#endif // _REGISTER_ALLOCATOR_H_
+

mozilla/ef/Compiler/RegisterAllocator/RegisterAllocatorTools.cpp

@@ -0,0 +1,355 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "LogModule.h"
+#include "RegisterAllocatorTools.h"
+#include "Pool.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Primitives.h"
+#include "InstructionEmitter.h"
+#include "Instruction.h"
+#include "RegisterAllocator.h"
+#include "Spilling.h"
+#include "Splits.h"
+#include "BitSet.h"
+
+UT_EXTERN_LOG_MODULE(RegAlloc);
+
+#ifdef DEBUG
+void RegisterAllocatorTools::testTheInstructionGraph(ControlGraph& controlGraph, VirtualRegisterManager& vrManager)
+{
+	// Test the declared VirtualRegisters. The register allocator tries to condense the register universe.
+	// Any gap in the VirtualRegister names will be a loss of efficiency !!!!
+
+	Uint32 nameCount = vrManager.getSize();
+	BitSet registerSeen(controlGraph.pool, nameCount);
+
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+
+		InstructionList& instructions = nodes[n]->getInstructions();
+		for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					registerSeen.set(usePtr->getRegisterName());
+
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					registerSeen.set(definePtr->getRegisterName());
+		}
+
+		InstructionList& phiNodes = nodes[n]->getPhiNodeInstructions();
+		for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+			Instruction& instruction = phiNodes.get(p);
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					registerSeen.set(usePtr->getRegisterName());
+
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					registerSeen.set(definePtr->getRegisterName());
+		}
+	}
+
+	bool renameRegisters = false;
+	for (BitSet::iterator i = registerSeen.nextZero(0); !registerSeen.done(i); i = registerSeen.nextZero(i)) {
+		renameRegisters = true;
+		fprintf(stderr, 
+				"WARNING: The VirtualRegister vr%d has been allocated during CodeGeneration but\n"
+				"         is never used nor defined by any instruction in the instruction graph\n"
+				"                                  PLEASE FIX                                  \n",
+				i);
+	}
+	if (renameRegisters) {
+		Instruction** definingInstruction = new Instruction*[nameCount];
+		memset(definingInstruction, '\0', nameCount * sizeof(Instruction*));
+		RegisterName* newName = new RegisterName[nameCount];
+		memset(newName, '\0', nameCount * sizeof(RegisterName));
+		RegisterName nextName = RegisterName(1);
+
+		for (Uint32 n = 0; n < nNodes; n++) {
+
+			InstructionList& instructions = nodes[n]->getInstructions();
+			for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+				Instruction& instruction = instructions.get(i);
+
+				InstructionUse* useEnd = instruction.getInstructionUseEnd();
+				for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+					if (usePtr->isRegister()) {
+						RegisterName name = usePtr->getRegisterName();
+						if (newName[name] == rnInvalid) {
+							newName[name] = nextName;
+							definingInstruction[nextName] = vrManager.getVirtualRegister(name).getDefiningInstruction();
+							nextName = RegisterName(nextName + 1);
+						}
+						usePtr->setRegisterName(newName[name]);
+					}
+
+				InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+				for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+					if (definePtr->isRegister()) {
+						RegisterName name = definePtr->getRegisterName();
+						if (newName[name] == rnInvalid) {
+							newName[name] = nextName;
+							definingInstruction[nextName] = vrManager.getVirtualRegister(name).getDefiningInstruction();
+							nextName = RegisterName(nextName + 1);
+						}
+						definePtr->setRegisterName(newName[name]);
+					}
+			}
+
+			InstructionList& phiNodes = nodes[n]->getPhiNodeInstructions();
+			for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+				Instruction& instruction = phiNodes.get(p);
+
+				InstructionUse* useEnd = instruction.getInstructionUseEnd();
+				for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+					if (usePtr->isRegister()) {
+						RegisterName name = usePtr->getRegisterName();
+						if (newName[name] == rnInvalid) {
+							newName[name] = nextName;
+							definingInstruction[nextName] = vrManager.getVirtualRegister(name).getDefiningInstruction();
+							nextName = RegisterName(nextName + 1);
+						}
+						usePtr->setRegisterName(newName[name]);
+					}
+
+				InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+				for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+					if (definePtr->isRegister()) {
+						RegisterName name = definePtr->getRegisterName();
+						if (newName[name] == rnInvalid) {
+							newName[name] = nextName;
+							definingInstruction[nextName] = vrManager.getVirtualRegister(name).getDefiningInstruction();
+							nextName = RegisterName(nextName + 1);
+						}
+						definePtr->setRegisterName(newName[name]);
+					}
+			}
+		}
+		
+		vrManager.setSize(nextName);
+
+		for (RegisterName r = RegisterName(1); r < nextName; r = RegisterName(r + 1))
+			vrManager.getVirtualRegister(r).definingInstruction = definingInstruction[r];
+
+		UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("RegisterMap:\n"));
+		for (Uint32 i = 1; i < nameCount; i++)
+			if (newName[i] != 0)
+				UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\tvr%d becomes vr%d.\n", i, newName[i]));
+			else
+				UT_OBJECTLOG(UT_LOG_MODULE(RegAlloc), PR_LOG_ALWAYS, ("\tvr%d is dead.\n", i));
+
+
+		delete newName;
+		delete definingInstruction;
+	}
+
+}
+#endif // DEBUG
+
+void RegisterAllocatorTools::removeUnnecessaryCopies(RegisterAllocator& registerAllocator)
+{
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+	RegisterName* name2range = registerAllocator.name2range;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		InstructionList& instructions = nodes[n]->getInstructions();
+		for (InstructionList::iterator i = instructions.begin(); !instructions.done(i);) {
+			Instruction& instruction = instructions.get(i);
+			i = instructions.advance(i);
+
+			if (instruction.getFlags() & ifCopy) {
+				assert(instruction.getInstructionUseBegin() != instruction.getInstructionUseEnd() && instruction.getInstructionUseBegin()[0].isRegister());
+				assert(instruction.getInstructionDefineBegin() != instruction.getInstructionDefineEnd() && instruction.getInstructionDefineBegin()[0].isRegister());
+				
+				RegisterName source = name2range[instruction.getInstructionUseBegin()[0].getRegisterName()];
+				RegisterName destination = name2range[instruction.getInstructionDefineBegin()[0].getRegisterName()];
+
+				if (source == destination)
+					instruction.remove();
+			}
+		}
+	}
+}
+
+
+void RegisterAllocatorTools::updateInstructionGraph(RegisterAllocator& registerAllocator)
+{
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+	RegisterName* name2range = registerAllocator.name2range;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		InstructionList& instructions = nodes[n]->getInstructions();
+		for (InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					usePtr->setRegisterName(name2range[usePtr->getRegisterName()]);
+
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					definePtr->setRegisterName(name2range[definePtr->getRegisterName()]);
+		}
+
+		InstructionList& phiNodes = nodes[n]->getPhiNodeInstructions();
+		for (InstructionList::iterator p = phiNodes.begin(); !phiNodes.done(p); p = phiNodes.advance(p)) {
+			Instruction& instruction = phiNodes.get(p);
+
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			for (InstructionUse* usePtr = instruction.getInstructionUseBegin(); usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					usePtr->setRegisterName(name2range[usePtr->getRegisterName()]);
+
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			for (InstructionDefine* definePtr = instruction.getInstructionDefineBegin(); definePtr < defineEnd; definePtr++)
+				if (definePtr->isRegister())
+					definePtr->setRegisterName(name2range[definePtr->getRegisterName()]);
+		}
+	}
+}
+
+
+void RegisterAllocatorTools::insertPhiNodeInstructions(ControlGraph& controlGraph, InstructionEmitter& emitter)
+{
+	Pool& pool = controlGraph.pool;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+	
+	for (Uint32 n = 0; n < nNodes; n++) {
+		ControlNode& node = *nodes[n];
+		DoublyLinkedList<PhiNode>& phiNodes = node.getPhiNodes();
+
+		if (!phiNodes.empty()) {
+
+			// Set the index of the incoming edges.
+			Uint32 index = 0;
+			const DoublyLinkedList<ControlEdge>& predecessors = node.getPredecessors();
+			for (DoublyLinkedList<ControlEdge>::iterator p = predecessors.begin(); !predecessors.done(p); p = predecessors.advance(p))
+				predecessors.get(p).setIndex(index++);
+
+			// Insert the phi node instruction in the instruction list.
+			for (DoublyLinkedList<PhiNode>::iterator i = phiNodes.begin(); !phiNodes.done(i); i = phiNodes.advance(i)) {
+				PhiNode& phiNode = phiNodes.get(i);
+				ValueKind kind = phiNode.getKind();
+
+				if (!isStorableKind(kind))
+					continue;
+
+				RegisterClassKind classKind = rckGeneral; // FIX: get class kind from phi node kind.
+				Uint32 nInputs = phiNode.nInputs();
+
+				PhiNodeInstruction& phiNodeInstruction = *new(pool) PhiNodeInstruction(&phiNode, pool, nInputs);
+
+				emitter.defineProducer(phiNode, phiNodeInstruction, 0, classKind, drLow);
+				for (Uint32 whichInput = 0; whichInput < nInputs; whichInput++)
+					emitter.useProducer(phiNode.nthInputVariable(whichInput), phiNodeInstruction, whichInput, classKind, drLow);
+
+				node.addPhiNodeInstruction(phiNodeInstruction);
+
+				if (isDoublewordKind(kind)) {
+					PhiNodeInstruction& phiNodeInstruction = *new(pool) PhiNodeInstruction(&phiNode, pool, nInputs);
+
+					emitter.defineProducer(phiNode, phiNodeInstruction, 0, classKind, drHigh);
+					for (Uint32 whichInput = 0; whichInput < nInputs; whichInput++)
+						emitter.useProducer(phiNode.nthInputVariable(whichInput), phiNodeInstruction, whichInput, classKind, drHigh);
+
+					node.addPhiNodeInstruction(phiNodeInstruction);
+				}
+			}
+		}
+	}
+}
+
+#ifdef DEBUG_LOG
+
+void RegisterAllocatorTools::printSpillCosts(RegisterAllocator& registerAllocator)
+{
+	LogModuleObject log = UT_LOG_MODULE(RegAlloc);
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	SpillCost* cost = registerAllocator.spillCost;
+
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Spill costs:\n"));
+	for (Uint32 i = 1; i < rangeCount; i++) {
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\trange %d : ", i));
+		if (cost[i].infinite)
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("infinite\n"));
+		else
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("%f\n", cost[i].cost));
+	}
+}
+
+void RegisterAllocatorTools::printSplitCosts(RegisterAllocator& registerAllocator)
+{
+	LogModuleObject log = UT_LOG_MODULE(RegAlloc);
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	SplitCost* cost = registerAllocator.splitCost;
+
+	UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("Split costs:\n"));
+	for (Uint32 i = 1; i < rangeCount; i++) {
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\trange %d : loads = %f stores = %f\n", i, cost[i].loads, cost[i].stores));
+	}
+}
+
+void RegisterAllocatorTools::printInstructions(RegisterAllocator& registerAllocator)
+{
+	LogModuleObject log = UT_LOG_MODULE(RegAlloc);
+	ControlNode** nodes = registerAllocator.controlGraph.dfsList;
+	Uint32 nNodes = registerAllocator.controlGraph.nNodes;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("N%d:\n", n));
+
+		InstructionList& phiNodes = nodes[n]->getPhiNodeInstructions();
+		InstructionList& instructions = nodes[n]->getInstructions();
+
+		if (!phiNodes.empty()) {
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, (" PhiNodes:\n", n));
+			for(InstructionList::iterator i = phiNodes.begin(); !phiNodes.done(i); i = phiNodes.advance(i)) {
+				phiNodes.get(i).printPretty(log);
+				UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\n"));
+			}
+			if (!instructions.empty())
+				UT_OBJECTLOG(log, PR_LOG_ALWAYS, (" Instructions:\n", n));
+		}
+
+		for(InstructionList::iterator i = instructions.begin(); !instructions.done(i); i = instructions.advance(i)) {
+			instructions.get(i).printPretty(log);
+			UT_OBJECTLOG(log, PR_LOG_ALWAYS, ("\n"));
+		}
+	}
+}
+#endif // DEBUG_LOG

mozilla/ef/Compiler/RegisterAllocator/RegisterAllocatorTools.h

@@ -0,0 +1,117 @@
+//	-*- mode:C++; tab-width:4; truncate-lines:t -*-
+//
+//           CONFIDENTIAL AND PROPRIETARY SOURCE CODE OF
+//              NETSCAPE COMMUNICATIONS CORPORATION
+// Copyright © 1996, 1997 Netscape Communications Corporation.  All Rights
+// Reserved.  Use of this Source Code is subject to the terms of the
+// applicable license agreement from Netscape Communications Corporation.
+// The copyright notice(s) in this Source Code does not indicate actual or
+// intended publication of this Source Code.
+//
+// $Id: RegisterAllocatorTools.h,v 1.1.2.1 1999-03-02 16:12:05 fur%netscape.com Exp $
+//
+
+#ifndef _REGISTER_ALLOCATOR_TOOLS_H_
+#define _REGISTER_ALLOCATOR_TOOLS_H_
+
+#include "LogModule.h"
+#include "RegisterTypes.h"
+#include <string.h>
+
+class RegisterAllocator;
+class ControlGraph;
+class InstructionEmitter;
+class VirtualRegisterManager;
+
+struct RegisterAllocatorTools
+{
+	//
+	//
+	static void insertPhiNodeInstructions(ControlGraph& controlGraph, InstructionEmitter& emitter);
+
+	//
+	//
+	static void updateInstructionGraph(RegisterAllocator& registerAllocator);
+
+	//
+	//
+	static void removeUnnecessaryCopies(RegisterAllocator& registerAllocator);
+
+#ifdef DEBUG
+	//
+	//
+	static void testTheInstructionGraph(ControlGraph& controlGraph, VirtualRegisterManager& vrManager);
+#endif // DEBUG
+
+#ifdef DEBUG_LOG
+	//
+	//
+	static void printInstructions(RegisterAllocator& registerAllocator);
+
+	//
+	//
+	static void printSpillCosts(RegisterAllocator& registerAllocator);
+
+	//
+	//
+	static void printSplitCosts(RegisterAllocator& registerAllocator);
+#endif // DEBUG_LOG
+};
+
+//
+// FIX: this should go in a class (LookupTable ?)
+//
+
+inline RegisterName findRoot(RegisterName name, RegisterName* table)
+{
+	RegisterName* stack = table;
+	RegisterName* stackPtr = stack;
+
+	RegisterName newName;
+	while((newName = table[name]) != name) {
+		*--stackPtr = name;
+		name = newName;
+	}
+
+	while (stackPtr != stack)
+		table[*stackPtr++] = name;
+
+	return name;
+}
+
+inline void init(RegisterName* table, Uint32 nameCount)
+{
+	for (RegisterName r = RegisterName(0); r < nameCount; r = RegisterName(r + 1))
+		table[r] = r;
+}
+
+inline Uint32 compress(RegisterName* name2range, RegisterName* table, Uint32 nameCount, Uint32 tableSize)
+{
+	RegisterName* liveRange = new RegisterName[tableSize];
+	memset(liveRange, '\0', tableSize * sizeof(RegisterName));
+
+	// Update the lookup table.
+	for (RegisterName r = RegisterName(1); r < tableSize; r = RegisterName(r + 1))
+		findRoot(r, table);
+
+	// Count the liveranges.
+	Uint32 liveRangeCount = 1;
+	for (RegisterName s = RegisterName(1); s < tableSize; s = RegisterName(s + 1))
+		if (table[s] == s)
+			liveRange[s] = RegisterName(liveRangeCount++);
+
+	for (RegisterName t = RegisterName(1); t < nameCount; t = RegisterName(t + 1))
+		name2range[t] = liveRange[table[name2range[t]]];
+
+	return liveRangeCount;
+}
+
+inline double doLog10(Uint32 power)
+{
+	double log = 1.0;
+	while (power--)
+		log *= 10.0;
+	return log;
+}
+
+#endif // _REGISTER_ALLOCATOR_TOOLS_H_

mozilla/ef/Compiler/RegisterAllocator/RegisterAssigner.h

@@ -0,0 +1,38 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _REGISTER_ASSIGNER_H_
+#define _REGISTER_ASSIGNER_H_
+
+#include "Fundamentals.h"
+#include "VirtualRegister.h"
+
+class FastBitMatrix;
+
+class RegisterAssigner
+{
+protected:
+  VirtualRegisterManager& vRegManager;
+  
+public:
+  RegisterAssigner(VirtualRegisterManager& vrMan) : vRegManager(vrMan) {}
+  
+  virtual bool assignRegisters(FastBitMatrix& interferenceMatrix) = 0;
+};
+
+#endif /* _REGISTER_ASSIGNER_H_ */

mozilla/ef/Compiler/RegisterAllocator/RegisterClass.h

@@ -0,0 +1,25 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _REGISTER_CLASS_H_
+#define _REGISTER_CLASS_H_
+
+#include "Fundamentals.h"
+#include "RegisterTypes.h"
+
+#endif // _REGISTER_CLASS_H_

mozilla/ef/Compiler/RegisterAllocator/RegisterPressure.h

@@ -0,0 +1,37 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _REGISTER_PRESSURE_H_
+#define _REGISTER_PRESSURE_H_
+
+#include "BitSet.h"
+#include "HashSet.h"
+
+struct LowRegisterPressure
+{
+	typedef BitSet Set;
+	static const bool setIsOrdered = true;
+};
+
+struct HighRegisterPressure
+{
+	typedef HashSet Set;
+	static const bool setIsOrdered = false;
+};
+
+#endif // _REGISTER_PRESSURE_H_

mozilla/ef/Compiler/RegisterAllocator/RegisterTypes.h

@@ -0,0 +1,104 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _REGISTER_TYPES_H_
+#define _REGISTER_TYPES_H_
+
+#include "Fundamentals.h"
+
+//------------------------------------------------------------------------------
+// RegisterName -
+//
+
+enum RegisterName {
+	rnInvalid = 0,
+};
+
+//------------------------------------------------------------------------------
+// RegisterClassKind -
+//
+
+enum RegisterClassKind {
+	rckInvalid = 0,
+	rckGeneral,
+	rckStackSlot,
+
+	nRegisterClassKind
+};
+
+//------------------------------------------------------------------------------
+// RegisterID -
+//
+
+enum RegisterID {
+	invalidID = 0
+};
+
+//------------------------------------------------------------------------------
+// RegisterKind -
+//
+
+enum RegisterKind {
+	rkCallerSave = 0,
+	rkCalleeSave,
+};
+
+struct NameLinkedList {
+	RegisterName name;
+	NameLinkedList* next;
+};
+
+#ifdef DEBUG
+
+const registerNameMask = 		0x03ffffff;
+const coloredRegisterMask = 	0x04000000;
+const machineRegisterMask = 	0x08000000;
+const registerClassMask = 		0xf0000000;
+
+const registerNameShift = 		0;
+const coloredRegisterShift = 	26;
+const machineRegisterShift = 	27;
+const registerClassShift = 		28;
+
+#else // DEBUG
+
+const registerNameMask = 		0x0fffffff;
+const registerClassMask = 		0xf0000000;
+
+const registerNameShift = 		0;
+const registerClassShift = 		28;
+
+#endif // DEBUG
+
+
+inline RegisterClassKind getClass(RegisterID registerID) {return RegisterClassKind((registerID & registerClassMask) >> registerClassShift);}
+inline RegisterName getName(RegisterID registerID) {return RegisterName((registerID & registerNameMask) >> registerNameShift);}
+inline void setClass(RegisterID& registerID, RegisterClassKind classKind) {registerID = RegisterID((registerID & ~registerClassMask) | ((classKind << registerClassShift) & registerClassMask));}
+inline void setName(RegisterID& registerID, RegisterName name) {assert((name & ~registerNameMask) == 0); registerID = RegisterID((registerID & ~registerNameMask) | ((name << registerNameShift) & registerNameMask));}
+inline RegisterID buildRegisterID(RegisterName name, RegisterClassKind classKind) {return RegisterID(((classKind << registerClassShift) & registerClassMask) | ((name << registerNameShift) & registerNameMask));}
+
+#ifdef DEBUG
+
+inline bool isMachineRegister(RegisterID rid) {return (rid & machineRegisterMask) != 0;}
+inline void setMachineRegister(RegisterID& rid) {rid = RegisterID(rid | machineRegisterMask);}
+inline bool isColoredRegister(RegisterID rid) {return (rid & coloredRegisterMask) != 0;}
+inline void setColoredRegister(RegisterID& rid) {rid = RegisterID(rid | coloredRegisterMask);}
+
+#endif // DEBUG
+
+#endif // _REGISTER_TYPES_H_

mozilla/ef/Compiler/RegisterAllocator/SSATools.cpp

@@ -0,0 +1,32 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "SSATools.h"
+#include "ControlGraph.h"
+#include "VirtualRegister.h"
+#include "Liveness.h"
+
+void replacePhiNodes(ControlGraph& controlGraph, VirtualRegisterManager& vrManager)
+{
+	if (!controlGraph.hasBackEdges)
+		return;
+
+	Liveness liveness(controlGraph.pool);
+	liveness.buildLivenessAnalysis(controlGraph, vrManager);
+}

mozilla/ef/Compiler/RegisterAllocator/SSATools.h

@@ -0,0 +1,29 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _SSA_TOOLS_H_
+#define _SSA_TOOLS_H_
+
+#include "Fundamentals.h"
+
+class ControlGraph;
+class VirtualRegisterManager;
+
+extern void replacePhiNodes(ControlGraph& controlGraph, VirtualRegisterManager& vrManager);
+
+#endif // _SSA_TOOLS_H_

mozilla/ef/Compiler/RegisterAllocator/SparseSet.cpp

@@ -0,0 +1,37 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "SparseSet.h"
+#include "BitSet.h"
+#include "Pool.h"
+
+#ifdef DEBUG_LOG
+// Print the set.
+//
+void SparseSet::printPretty(LogModuleObject log)
+{
+	Pool pool;
+	BitSet set(pool, universeSize);
+
+	for (Uint32 i = 0; i < count; i++)
+	   set.set(node[i].element);
+
+	set.printPretty(log);
+}
+#endif // DEBUG_LOG

mozilla/ef/Compiler/RegisterAllocator/SparseSet.h

@@ -0,0 +1,168 @@
+//	-*- mode:C++; tab-width:4; truncate-lines:t -*-
+//
+//           CONFIDENTIAL AND PROPRIETARY SOURCE CODE OF
+//              NETSCAPE COMMUNICATIONS CORPORATION
+// Copyright © 1996, 1997 Netscape Communications Corporation.  All Rights
+// Reserved.  Use of this Source Code is subject to the terms of the
+// applicable license agreement from Netscape Communications Corporation.
+// The copyright notice(s) in this Source Code does not indicate actual or
+// intended publication of this Source Code.
+//
+// $Id: SparseSet.h,v 1.1.2.1 1999-03-02 16:12:07 fur%netscape.com Exp $
+//
+
+#ifndef _SPARSE_SET_H_
+#define _SPARSE_SET_H_
+
+#include "Fundamentals.h"
+#include "Pool.h"
+#include "LogModule.h"
+#include "BitSet.h"
+
+class SparseSet
+{
+private:
+
+	struct Node {
+		Uint32		element;
+		Uint32		stackIndex;
+	};
+
+	Node*			node;
+	Uint32			count;
+	Uint32			universeSize;
+
+private:
+
+	// No copy constructor.
+	SparseSet(const SparseSet&);
+
+	// Check if the given set's universe is of the same size than this universe.
+	void checkUniverseCompatibility(const SparseSet& set) const {assert(set.universeSize == universeSize);}
+	// Check if pos is valid for this set's universe.
+	void checkMember(Int32 pos) const {assert(pos >=0 && Uint32(pos) < universeSize);}
+
+public:
+
+	SparseSet(Pool& pool, Uint32 universeSize) : universeSize(universeSize) {node = new(pool) Node[universeSize]; clear();}
+
+	// Clear the sparse set.
+	void clear() {count = 0;}
+	// Clear the element at index.
+	inline void clear(Uint32 index);
+	// Set the element at index.
+	inline void set(Uint32 index);
+	// Return true if the element at index is set.
+	inline bool test(Uint32 index) const;
+	// Union with the given sparse set.
+	inline void or(const SparseSet& set);
+	// Intersection with the given sparse set.
+	inline void and(const SparseSet& set);
+	// Difference with the given sparse set.
+	inline void difference(const SparseSet& set);
+	// Copy set.
+	inline SparseSet& operator = (const SparseSet& set);
+	inline SparseSet& operator = (const BitSet& set);
+	// Return true if the sparse sets are identical.
+	friend bool operator == (const SparseSet& set1, const SparseSet& set2);
+	// Return true if the sparse sets are different.
+	friend bool operator != (const SparseSet& set1, const SparseSet& set2);
+
+	// Logical operators.
+	SparseSet& operator |= (const SparseSet& set) {or(set); return *this;}
+	SparseSet& operator &= (const SparseSet& set) {and(set); return *this;}
+	SparseSet& operator -= (const SparseSet& set) {difference(set); return *this;}
+
+	// Iterator to conform with the set API.
+	typedef Int32 iterator;
+	// Return the iterator for the first element of this set.
+	iterator begin() const {return count - 1;}
+	// Return the next iterator.
+	iterator advance(iterator pos) const {return --pos;}
+	// Return true if the iterator is at the end of the set.
+	bool done(iterator pos) const {return pos < 0;}
+	// Return the element for the given iterator;
+	Uint32 get(iterator pos) const {return node[pos].element;}
+	// Return one element of this set.
+	Uint32 getOne() const {assert(count > 0); return node[0].element;}
+	// Return the size of this set.
+	Uint32 getSize() const {return count;}
+
+#ifdef DEBUG_LOG
+	// Print the set.
+	void printPretty(LogModuleObject log);
+#endif // DEBUG_LOG
+};
+
+inline void SparseSet::clear(Uint32 element)
+{
+	checkMember(element);
+	Uint32 count = this->count;
+	Node* node = this->node;
+
+	Uint32 stackIndex = node[element].stackIndex;
+
+	if ((stackIndex < count) && (node[stackIndex].element == element)) {
+		Uint32 stackTop = node[count - 1].element;
+
+		node[stackIndex].element = stackTop;
+		node[stackTop].stackIndex = stackIndex;
+		this->count = count - 1;
+	}
+}
+
+inline void SparseSet::set(Uint32 element)
+{
+	checkMember(element);
+  	Uint32 count = this->count;
+	Node* node = this->node;
+
+	Uint32 stackIndex = node[element].stackIndex;
+
+	if ((stackIndex >= count) || (node[stackIndex].element != element)) {
+		node[count].element = element;
+		node[element].stackIndex = count;
+		this->count = count + 1;
+	}
+}
+
+inline bool SparseSet::test(Uint32 element) const
+{
+	checkMember(element);
+	Node* node = this->node;
+
+	Uint32 stackIndex = node[element].stackIndex;
+	return ((stackIndex < count) && (node[stackIndex].element == element));
+}
+
+inline SparseSet& SparseSet::operator = (const SparseSet& set)
+{
+	checkUniverseCompatibility(set);
+	Uint32 sourceCount = set.getSize();
+	Node* node = this->node;
+
+	memcpy(node, set.node, sourceCount * sizeof(Node));
+
+	for (Uint32 i = 0; i < sourceCount; i++) {
+		Uint32 element = node[i].element;
+		node[element].stackIndex = i;
+	}
+
+	count = sourceCount;
+
+	return *this;
+}
+
+
+inline SparseSet& SparseSet::operator = (const BitSet& set)
+{
+	// FIX: there's room for optimization here.
+	assert(universeSize == set.getSize());
+
+	clear();
+	for (Int32 i = set.firstOne(); i != -1; i = set.nextOne(i))
+		this->set(i);
+	return *this;
+}
+
+#endif // _SPARSE_SET_H_

mozilla/ef/Compiler/RegisterAllocator/Spilling.cpp

@@ -0,0 +1,270 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef NEW_LAURENTM_CODE
+#define INCLUDE_EMITTER
+#include "CpuInfo.h"
+#include "Fundamentals.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "InstructionEmitter.h"
+#include "Spilling.h"
+
+
+void Spilling::
+insertSpillCode(ControlNode** dfsList, Uint32 nNodes)
+{
+  PRUint32 nVirtualRegisters = vRegManager.count();
+  FastBitSet currentLive(vRegManager.pool, nVirtualRegisters);
+  FastBitSet usedInThisInstruction(vRegManager.pool, nVirtualRegisters);
+  RegisterFifo grNeedLoad(nVirtualRegisters);
+  RegisterFifo fpNeedLoad(nVirtualRegisters);
+
+  for (PRInt32 n = nNodes - 1; n >= 0; n--)
+	{
+	  PR_ASSERT(grNeedLoad.empty() & fpNeedLoad.empty());
+	  ControlNode& node = *dfsList[n];
+
+	  currentLive = node.liveAtEnd;
+
+	  PRUint32 nGeneralAlive = 0;
+	  PRUint32 nFloatingPointAlive = 0;
+
+	  // Get the number of registers alive at the end of this node.
+	  for (PRInt32 j = currentLive.firstOne(); j != -1; j = currentLive.nextOne(j))
+		{
+		  VirtualRegister& vReg = vRegManager.getVirtualRegister(j);
+		  if (vReg.spillInfo.willSpill)
+			{
+			  currentLive.clear(j);
+			}
+		  else
+			{
+			  switch (vReg.getClass())
+				{
+				case vrcInteger:
+				  nGeneralAlive++;
+				  break;
+				case vrcFloatingPoint:
+				case vrcFixedPoint:
+				  nFloatingPointAlive++;
+				  break;
+				default:
+				  break;
+				}
+			}
+		}
+	  
+//		  if(node.dfsNum == 8) printf("\n________Begin Node %d________\n", node.dfsNum);
+
+		InstructionList& instructions = node.getInstructions();
+	  for (InstructionList::iterator i = instructions.end(); !instructions.done(i); i = instructions.retreat(i))
+		{
+		  Instruction&       instruction = instructions.get(i);
+		  InstructionUse*    useBegin    = instruction.getInstructionUseBegin();
+		  InstructionUse*    useEnd      = instruction.getInstructionUseEnd();
+		  InstructionUse*    usePtr;
+		  InstructionDefine* defBegin    = instruction.getInstructionDefineBegin();
+		  InstructionDefine* defEnd      = instruction.getInstructionDefineEnd();
+		  InstructionDefine* defPtr;
+
+//		if(node.dfsNum == 8)  { printf("\n");
+//		instruction.printPretty(stdout);
+//		printf("\n"); }
+
+		  // Handle definitions
+		  for (defPtr = defBegin; defPtr < defEnd; defPtr++)
+			if (defPtr->isVirtualRegister())
+			  {
+				VirtualRegister& vReg = defPtr->getVirtualRegister();
+				currentLive.clear(vReg.getRegisterIndex());
+				switch (vReg.getClass())
+				  {
+				  case vrcInteger:
+					nGeneralAlive--;
+					break;
+				  case vrcFloatingPoint:
+				  case vrcFixedPoint:
+					nFloatingPointAlive--;
+					break;
+				  default:
+					break;
+				  }
+			  }
+
+		  // Check for deaths
+		  for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+			if (usePtr->isVirtualRegister())
+			  {
+				VirtualRegister& vReg = usePtr->getVirtualRegister();
+				if (!currentLive.test(vReg.getRegisterIndex()))
+				  // This is the last use of this register.
+				  {
+					currentLive.set(vReg.getRegisterIndex());
+					switch (vReg.getClass())
+					  {
+					  case vrcInteger:
+						nGeneralAlive++;
+						while (/*(nGeneralAlive > NUMBER_OF_GREGISTERS) &&*/ !grNeedLoad.empty())
+						  {
+							PRUint32 toLoad = grNeedLoad.get();
+							currentLive.clear(toLoad);
+							nGeneralAlive--;
+							
+							VirtualRegister& nReg = vRegManager.getVirtualRegister(toLoad);
+							Instruction& lastUsingInstruction = *nReg.spillInfo.lastUsingInstruction;
+							emitter.emitLoadAfter(*lastUsingInstruction.getPrimitive(), lastUsingInstruction.getLinks().prev,
+												  nReg.getAlias(), *nReg.equivalentRegister[vrcStackSlot]);
+							nReg.releaseSelf();
+						  }
+						break;
+					  case vrcFloatingPoint:
+					  case vrcFixedPoint:
+						nFloatingPointAlive++;
+						while (/*(nFloatingPointAlive > NUMBER_OF_FPREGISTERS) &&*/ !fpNeedLoad.empty())
+						  {
+							PRUint32 toLoad = fpNeedLoad.get();
+							currentLive.clear(toLoad);
+							nFloatingPointAlive--;
+							
+							VirtualRegister& nReg = vRegManager.getVirtualRegister(toLoad);
+							Instruction& lastUsingInstruction = *nReg.spillInfo.lastUsingInstruction;
+							emitter.emitLoadAfter(*lastUsingInstruction.getPrimitive(), lastUsingInstruction.getLinks().prev,
+												  nReg.getAlias(), *nReg.equivalentRegister[vrcStackSlot]);
+							nReg.releaseSelf();
+						  }
+						break;
+					  default:
+						break;
+					  }
+				  }
+			  }
+
+		  // Handle uses
+		  for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+			if (usePtr->isVirtualRegister())
+			  {
+				VirtualRegister& vReg = usePtr->getVirtualRegister();
+				PRUint32 registerIndex = vReg.getRegisterIndex();
+
+				if (vReg.spillInfo.willSpill) {
+#if defined(GENERATE_FOR_X86)
+				  if (!instruction.switchUseToSpill((usePtr - useBegin), *vReg.equivalentRegister[vrcStackSlot]))
+#endif
+					{
+					  switch (vReg.getClass())
+						{
+						case vrcInteger:
+						  if (!grNeedLoad.test(registerIndex))
+							{
+							  grNeedLoad.put(registerIndex);
+							  VirtualRegister& alias = vRegManager.newVirtualRegister(vrcInteger);
+							  if (vReg.isPreColored())
+								  alias.preColorRegister(vReg.getPreColor());
+							  /* if (vReg.hasSpecialInterference) {
+								  alias.specialInterference.sizeTo(NUMBER_OF_REGISTERS);
+								  alias.specialInterference = vReg.specialInterference;
+								  alias.hasSpecialInterference = true;
+							  } */
+							  vReg.setAlias(alias);
+							  vReg.retainSelf();
+							}
+						  break;
+						case vrcFloatingPoint:
+						case vrcFixedPoint:
+						  if (!fpNeedLoad.test(registerIndex))
+							{
+							  fpNeedLoad.put(registerIndex);
+							  VirtualRegister& alias = vRegManager.newVirtualRegister(vReg.getClass());
+							  if (vReg.isPreColored())
+								  alias.preColorRegister(vReg.getPreColor());
+							  /*if (vReg.hasSpecialInterference) {
+								  alias.specialInterference.sizeTo(NUMBER_OF_REGISTERS);
+								  alias.specialInterference = vReg.specialInterference;
+								  alias.hasSpecialInterference = true;
+							  } */
+							  vReg.setAlias(alias);
+							  vReg.retainSelf();
+							}
+						  break;
+						default:
+						  break;
+						}
+					  usePtr->getVirtualRegisterPtr().initialize(vReg.getAlias());
+					  usedInThisInstruction.set(registerIndex);
+					  vReg.spillInfo.lastUsingInstruction = &instruction;
+					}
+				  currentLive.clear(registerIndex);
+				} else { // will not spill
+				  currentLive.set(registerIndex);
+				}
+			  }
+		  
+		  // Handle definitions
+		  for (defPtr = defBegin; defPtr < defEnd; defPtr++)
+			if (defPtr->isVirtualRegister())
+			  {
+				VirtualRegister& vReg = defPtr->getVirtualRegister();
+				
+				if (vReg.spillInfo.willSpill)
+#if defined(GENERATE_FOR_X86)
+				  if (!instruction.switchDefineToSpill((defPtr - defBegin), *vReg.equivalentRegister[vrcStackSlot]))
+#endif
+					{
+					  if (usedInThisInstruction.test(vReg.getRegisterIndex()))
+						// this virtualRegister was used in this instruction and is also defined. We need to move 
+						// this virtual register to its alias first and then save it to memory.
+						{
+						  emitter.emitStoreAfter(*instruction.getPrimitive(), &instruction.getLinks(), 
+												 vReg.getAlias(), *vReg.equivalentRegister[vrcStackSlot]);
+						  defPtr->getVirtualRegisterPtr().initialize(vReg.getAlias());
+						}
+					  else
+						{
+						  emitter.emitStoreAfter(*instruction.getPrimitive(), &instruction.getLinks(), 
+												 vReg, *vReg.equivalentRegister[vrcStackSlot]);
+						}
+					}
+			  }
+		}
+	  while (!grNeedLoad.empty())
+		{
+		  PRUint32 nl = grNeedLoad.get();
+		  VirtualRegister& nlReg = vRegManager.getVirtualRegister(nl);
+		  Instruction& lastUse = *nlReg.spillInfo.lastUsingInstruction;
+		  
+		  emitter.emitLoadAfter(*lastUse.getPrimitive(), lastUse.getLinks().prev, 
+								nlReg.getAlias(), *nlReg.equivalentRegister[vrcStackSlot]);
+		  nlReg.releaseSelf();
+		}
+	  while (!fpNeedLoad.empty())
+		{
+		  PRUint32 nl = fpNeedLoad.get();
+		  VirtualRegister& nlReg = vRegManager.getVirtualRegister(nl);
+		  Instruction& lastUse = *nlReg.spillInfo.lastUsingInstruction;
+
+		  emitter.emitLoadAfter(*lastUse.getPrimitive(), lastUse.getLinks().prev, 
+								nlReg.getAlias(), *nlReg.equivalentRegister[vrcStackSlot]);
+		  nlReg.releaseSelf();
+		}
+
+//	  if(node.dfsNum == 8) printf("\n________End Node %d________\n", node.dfsNum);
+
+	}
+}
+#endif

mozilla/ef/Compiler/RegisterAllocator/Spilling.h

@@ -0,0 +1,269 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _SPILLING_H_
+#define _SPILLING_H_
+
+#include "Fundamentals.h"
+#include <string.h>
+#include "RegisterAllocator.h"
+#include "RegisterAllocatorTools.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "SparseSet.h"
+
+template <class RegisterPressure>
+class Spilling
+{
+private:
+	static void insertStoreAfter(Instruction& instruction, RegisterName name);
+	static void insertLoadBefore(Instruction& instruction, RegisterName name);
+
+public:
+	static void calculateSpillCosts(RegisterAllocator& registerAllocator);
+	static void insertSpillCode(RegisterAllocator& registerAllocator);
+};
+
+struct SpillCost
+{
+	double loads;
+	double stores;
+	double copies;
+	double cost;
+	bool infinite;
+};
+
+template <class RegisterPressure>
+void Spilling<RegisterPressure>::insertSpillCode(RegisterAllocator& registerAllocator)
+{
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	RegisterName* name2range = registerAllocator.name2range;
+
+	Pool& pool = registerAllocator.pool;
+	SparseSet currentLive(pool, rangeCount);
+	SparseSet needLoad(pool, rangeCount);
+	SparseSet mustSpill(pool, rangeCount);
+	SparseSet& willSpill = *registerAllocator.willSpill;
+	
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	RegisterPressure::Set* liveOut = registerAllocator.liveness.liveOut;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+
+		needLoad.clear();
+		currentLive = liveOut[n];
+		mustSpill = currentLive;
+
+		InstructionList& instructions = nodes[n]->getInstructions();
+		for (InstructionList::iterator i = instructions.end(); !instructions.done(i);) {
+			Instruction& instruction = instructions.get(i);
+			i = instructions.retreat(i);
+
+			InstructionUse* useBegin = instruction.getInstructionUseBegin();
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			InstructionUse* usePtr;
+			InstructionDefine* defineBegin = instruction.getInstructionDefineBegin();
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			InstructionDefine* definePtr;
+
+			bool foundLiveDefine = false;
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister()) {
+					if (currentLive.test(name2range[definePtr->getRegisterName()])) {
+						foundLiveDefine = true;
+						break;
+					}
+				} else {
+					foundLiveDefine = true;
+					break;
+				}
+			if (defineBegin != defineEnd && !foundLiveDefine) {
+				fprintf(stderr, "!!! Removed instruction because it was only defining unused registers !!!\n");
+				instruction.remove();
+			}
+
+ 			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister()) {
+					RegisterName range = name2range[definePtr->getRegisterName()];
+#ifdef DEBUG
+					if (needLoad.test(range))
+						if (!mustSpill.test(range) && registerAllocator.spillCost[range].infinite && willSpill.test(range)) {
+							fprintf(stderr, "Tried to spill a register with infinite spill cost\n");
+							abort();
+						}
+#endif // DEBUG
+					if (willSpill.test(range))
+						insertStoreAfter(instruction, range);
+
+					needLoad.clear(range);
+				}
+
+			if (instruction.getFlags() & ifCopy)
+				for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+					if (usePtr->isRegister()) {
+						RegisterName range = name2range[usePtr->getRegisterName()];
+						if (!currentLive.test(range))
+							for (SparseSet::iterator r = needLoad.begin(); !needLoad.done(r); r = needLoad.advance(r)) {
+								RegisterName load = RegisterName(needLoad.get(r));
+								if (willSpill.test(load))
+									insertLoadBefore(instruction, load);
+								mustSpill.set(load);
+							}
+						needLoad.clear();
+					}
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister())
+					currentLive.clear(name2range[definePtr->getRegisterName()]);
+
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					RegisterName range = name2range[usePtr->getRegisterName()];
+					currentLive.set(range);
+					needLoad.set(range);
+				}
+		}
+
+		for (SparseSet::iterator l = needLoad.begin(); !needLoad.done(l); l = needLoad.advance(l)) {
+			RegisterName load = RegisterName(needLoad.get(l));
+			if (willSpill.test(load))
+				insertLoadBefore(instructions.first(), load);
+		}		
+	}
+}
+
+template <class RegisterPressure>
+void Spilling<RegisterPressure>::insertLoadBefore(Instruction& /*instruction*/, RegisterName name)
+{
+	fprintf(stdout, "will insert load for range %d\n", name);
+}
+
+template <class RegisterPressure>
+void Spilling<RegisterPressure>::insertStoreAfter(Instruction& /*instruction*/, RegisterName name)
+{
+	fprintf(stdout, "will insert store for range %d\n", name);
+}
+
+template <class RegisterPressure>
+void Spilling<RegisterPressure>::calculateSpillCosts(RegisterAllocator& registerAllocator)
+{
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	RegisterName* name2range = registerAllocator.name2range;
+
+	Pool& pool = registerAllocator.pool;
+	SparseSet live(pool, rangeCount);
+	SparseSet needLoad(pool, rangeCount);
+	SparseSet mustSpill(pool, rangeCount);
+
+	SparseSet alreadyStored(pool, rangeCount); // FIX: should get this from previous spilling.
+
+	SpillCost* cost = new SpillCost[rangeCount];
+	memset(cost, '\0', rangeCount * sizeof(SpillCost));
+
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	RegisterPressure::Set* liveOut = registerAllocator.liveness.liveOut;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		ControlNode& node = *nodes[n];
+
+		double weight = doLog10(node.loopDepth);
+
+		needLoad.clear();
+		live = liveOut[n];
+		mustSpill = live;
+
+		InstructionList& instructions = nodes[n]->getInstructions();
+		for (InstructionList::iterator i = instructions.end(); !instructions.done(i); i = instructions.retreat(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useBegin = instruction.getInstructionUseBegin();
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			InstructionUse* usePtr;
+			InstructionDefine* defineBegin = instruction.getInstructionDefineBegin();
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			InstructionDefine* definePtr;
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister()) {
+					RegisterName range = name2range[definePtr->getRegisterName()];
+
+					if (needLoad.test(range))
+						if (!mustSpill.test(range))
+							cost[range].infinite = true;
+
+					if ((false /* !rematerializable(range) */ || !needLoad.test(range)) && !alreadyStored.test(range))
+						cost[range].stores += weight;
+
+					needLoad.clear(range);
+				}
+
+			if (instruction.getFlags() & ifCopy)
+				for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+					if (usePtr->isRegister())
+						if (!live.test(name2range[usePtr->getRegisterName()])) {
+							for (SparseSet::iterator l = needLoad.begin(); !needLoad.done(l); l = needLoad.advance(l)) {
+								Uint32 range = needLoad.get(l);
+								cost[range].loads += weight;
+								mustSpill.set(range);
+							}
+							needLoad.clear();
+						}
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister())
+					live.clear(name2range[definePtr->getRegisterName()]);
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					RegisterName range = name2range[usePtr->getRegisterName()];
+
+					live.set(range);
+					needLoad.set(range);
+				}
+
+			if (instruction.getFlags() & ifCopy) {
+				assert(useBegin != useEnd && useBegin[0].isRegister());
+				assert(defineBegin != defineEnd && defineBegin[0].isRegister());
+
+				RegisterName source = name2range[useBegin[0].getRegisterName()];
+				RegisterName destination = name2range[defineBegin[0].getRegisterName()];
+
+				cost[source].copies += weight;
+				cost[destination].copies += weight;
+			}			
+		}
+
+		for (SparseSet::iterator s = needLoad.begin(); !needLoad.done(s); s = needLoad.advance(s))
+			cost[needLoad.get(s)].loads += weight;
+	}
+	
+	for (Uint32 r = 0; r < rangeCount; r++) {
+		SpillCost& c = cost[r];
+		c.cost = 2 * (c.loads + c.stores) - c.copies;
+	}
+
+	registerAllocator.spillCost = cost;
+}
+
+#endif // _SPILLING_H_

mozilla/ef/Compiler/RegisterAllocator/Splits.h

@@ -0,0 +1,239 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _SPLITS_H_
+#define _SPLITS_H_
+
+#include "Fundamentals.h"
+#include <string.h>
+#include "Pool.h"
+#include "ControlGraph.h"
+#include "ControlNodes.h"
+#include "Instruction.h"
+#include "RegisterAllocator.h"
+#include "RegisterAllocatorTools.h"
+
+UT_EXTERN_LOG_MODULE(RegAlloc);
+
+template <class RegisterPressure>
+struct Splits
+{
+	static void calculateSplitCosts(RegisterAllocator& registerAllocator);
+	static bool findSplit(RegisterAllocator& registerAllocator, RegisterName* color, RegisterName range);
+	static void insertSplitCode(RegisterAllocator& registerAllocator);
+};
+
+struct SplitCost
+{
+	double loads;
+	double stores;
+};
+
+template <class RegisterPressure>
+void Splits<RegisterPressure>::insertSplitCode(RegisterAllocator& /*registerAllocator*/)
+{
+	// FIX
+}
+
+template <class RegisterPressure>
+bool Splits<RegisterPressure>::findSplit(RegisterAllocator& registerAllocator, RegisterName* color, RegisterName range)
+{
+	Pool& pool = registerAllocator.pool;
+	NameLinkedList** neighborsWithColor = new(pool) NameLinkedList*[6]; // FIX
+	memset(neighborsWithColor, '\0', 6 * sizeof(NameLinkedList*));
+
+	InterferenceGraph<RegisterPressure>& iGraph = registerAllocator.iGraph;
+
+	for (InterferenceVector* vector = iGraph.getInterferenceVector(range); vector != NULL; vector = vector->next)
+		for (Int32 i = vector->count - 1; i >=0; --i) {
+			RegisterName neighbor = vector->neighbors[i];
+			RegisterName c = color[neighbor];
+
+			if (c < 6) { // FIX
+				NameLinkedList* node = new(pool) NameLinkedList();
+				node->name = neighbor;
+				node->next = neighborsWithColor[c];
+				neighborsWithColor[c] = node;
+			}
+		}
+
+	bool splitAroundName = true;
+
+	LiveRangeGraph<RegisterPressure>& lGraph = registerAllocator.lGraph;
+	RegisterName bestColor = RegisterName(6); // FIX
+	double bestCost = registerAllocator.spillCost[range].cost;
+	SplitCost* splitCost = registerAllocator.splitCost;
+
+	for (RegisterName i = RegisterName(0); i < 6; i = RegisterName(i + 1)) { // FIX
+
+		double splitAroundNameCost = 0.0;
+		bool canSplitAroundName = true;
+
+		SplitCost& sCost = splitCost[range];
+		double addedCost = 2.0 * (sCost.stores + sCost.loads);
+
+		for (NameLinkedList* node = neighborsWithColor[i]; node != NULL; node = node->next) {
+			RegisterName neighbor = node->name;
+			if (lGraph.haveEdge(neighbor, range)) {
+				canSplitAroundName = false;
+				break;
+			} else
+				splitAroundNameCost += addedCost;
+		}
+		if (canSplitAroundName && splitAroundNameCost < bestCost) {
+			bestCost = splitAroundNameCost;
+			bestColor = i;
+			splitAroundName = true;
+		}
+
+		double splitAroundColorCost = 0.0;
+		bool canSplitAroundColor = true;
+
+		for (NameLinkedList* node = neighborsWithColor[i]; node != NULL; node = node->next) {
+			RegisterName neighbor = node->name;
+			if (lGraph.haveEdge(range, neighbor)) {
+				canSplitAroundColor = false;
+				break;
+			} else {
+				SplitCost& sCost = splitCost[neighbor];
+				double addedCost = 2.0 * (sCost.stores + sCost.loads);
+				splitAroundColorCost += addedCost;
+			}
+		}
+		if (canSplitAroundColor && splitAroundColorCost < bestCost) {
+			bestCost = splitAroundColorCost;
+			bestColor = i;
+			splitAroundName = false;
+		}
+ 	}
+	if (bestColor < RegisterName(6)) {
+		color[range] = bestColor;
+		registerAllocator.splitFound = true;
+
+		NameLinkedList** splitAround = registerAllocator.splitAround;
+
+		if (splitAroundName)
+			for (NameLinkedList* node = neighborsWithColor[bestColor]; node != NULL; node = node->next) {
+				NameLinkedList* newNode = new(pool) NameLinkedList();
+				newNode->name = node->name;
+				newNode->next = splitAround[range];
+				splitAround[range] = newNode;
+			}
+		else
+			for (NameLinkedList* node = neighborsWithColor[bestColor]; node != NULL; node = node->next) {
+				NameLinkedList* newNode = new(pool) NameLinkedList();
+				RegisterName neighbor = node->name;
+				newNode->name = range;
+				newNode->next = splitAround[neighbor];
+				splitAround[neighbor] = newNode;
+			}
+
+		trespass("Found a split");
+		return true;
+	}
+
+	return false;
+}
+
+template <class RegisterPressure>
+void Splits<RegisterPressure>::calculateSplitCosts(RegisterAllocator& registerAllocator)
+{
+	Pool& pool = registerAllocator.pool;
+	Uint32 rangeCount = registerAllocator.rangeCount;
+	RegisterName* name2range = registerAllocator.name2range;
+
+	SplitCost* splitCost = new(pool) SplitCost[rangeCount];
+	memset(splitCost, '\0', rangeCount * sizeof(SplitCost));
+
+	SparseSet live(pool, rangeCount);
+	RegisterPressure::Set* liveIn = registerAllocator.liveness.liveIn;
+	RegisterPressure::Set* liveOut = registerAllocator.liveness.liveOut;
+
+	ControlGraph& controlGraph = registerAllocator.controlGraph;
+	ControlNode** nodes = controlGraph.dfsList;
+	Uint32 nNodes = controlGraph.nNodes;
+
+	for (Uint32 n = 0; n < nNodes; n++) {
+		ControlNode& node = *nodes[n];
+		double weight = doLog10(node.loopDepth);
+
+		live = liveOut[n];
+
+		ControlEdge* successorsEnd = node.getSuccessorsEnd();
+		for (ControlEdge* successorsPtr = node.getSuccessorsBegin(); successorsPtr < successorsEnd; successorsPtr++) {
+			ControlNode& successor = successorsPtr->getTarget();
+
+			if (successor.getControlKind() != ckEnd) {
+				RegisterPressure::Set& successorLiveIn = liveIn[successor.dfsNum];
+
+				for (SparseSet::iterator i = live.begin(); !live.done(i); i = live.advance(i)) {
+					RegisterName name = RegisterName(live.get(i));
+					if (!successorLiveIn.test(name))
+						splitCost[name].loads += doLog10(successor.loopDepth);
+				}
+			}
+		}
+
+		InstructionList& instructions = node.getInstructions();
+		for (InstructionList::iterator i = instructions.end(); !instructions.done(i); i = instructions.retreat(i)) {
+			Instruction& instruction = instructions.get(i);
+
+			InstructionUse* useBegin = instruction.getInstructionUseBegin();
+			InstructionUse* useEnd = instruction.getInstructionUseEnd();
+			InstructionUse* usePtr;
+			InstructionDefine* defineBegin = instruction.getInstructionDefineBegin();
+			InstructionDefine* defineEnd = instruction.getInstructionDefineEnd();
+			InstructionDefine* definePtr;
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister())
+					splitCost[name2range[definePtr->getRegisterName()]].stores += weight;
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister()) {
+					RegisterName range = name2range[usePtr->getRegisterName()];
+					if (!live.test(range)) {
+						if (&instruction != &instructions.last())
+							splitCost[range].loads += weight;
+						else {
+							ControlEdge* successorsEnd = node.getSuccessorsEnd();
+							for (ControlEdge* successorsPtr = node.getSuccessorsBegin(); successorsPtr < successorsEnd; successorsPtr++)
+								splitCost[range].loads += doLog10(successorsPtr->getTarget().loopDepth);
+						}
+					}
+				}
+
+			for (definePtr = defineBegin; definePtr < defineEnd; definePtr++) 
+				if (definePtr->isRegister())
+					live.clear(name2range[definePtr->getRegisterName()]);
+
+			for (usePtr = useBegin; usePtr < useEnd; usePtr++)
+				if (usePtr->isRegister())
+					live.set(name2range[usePtr->getRegisterName()]);
+		}
+	}
+
+	NameLinkedList** splitAround = new(pool) NameLinkedList*[rangeCount];
+	memset(splitAround, '\0', rangeCount * sizeof(NameLinkedList*));
+	registerAllocator.splitAround = splitAround;
+
+	registerAllocator.splitCost = splitCost;
+	registerAllocator.splitFound = false;
+}
+
+#endif // _SPLITS_H_

mozilla/ef/Compiler/RegisterAllocator/Timer.cpp

@@ -0,0 +1,186 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "HashTable.h"
+#include "Timer.h"
+#include "Pool.h"
+
+static Pool pool;											// Pool for the Timer class.
+static HashTable<TimerEntry*> timerEntries(pool);			// Timers hashtable.
+
+const nTimersInABlock = 128;								// Number of timers in a block.
+static PRTime *timers = new(pool) PRTime[nTimersInABlock];	// A block of timers.
+static Uint8 nextTimer = 0;									// nextAvailableTimer.
+
+//
+// Calibrate the call to PR_Now().
+//
+static PRTime calibrate()
+{
+	PRTime t = PR_Now();
+	PRTime& a = *new(pool) PRTime();
+
+	// Call 10 times the PR_Now() function.
+	a = PR_Now(); a = PR_Now(); a = PR_Now(); a = PR_Now(); a = PR_Now(); a = PR_Now(); 
+	a = PR_Now(); a = PR_Now(); a = PR_Now(); a = PR_Now(); a = PR_Now(); a = PR_Now(); 
+	t = (PR_Now() - t + 9) / 10;
+
+	return t;
+}
+
+static PRTime adjust = calibrate();
+
+//
+// Return the named timer..
+//
+TimerEntry& Timer::getTimerEntry(const char* name)
+{
+	if (!timerEntries.exists(name)) {
+		TimerEntry* newEntry = new(pool) TimerEntry();
+		newEntry->accumulator = 0;
+		newEntry->running = false;
+		timerEntries.add(name, newEntry);
+	}
+
+	return *timerEntries[name];
+}
+
+//
+// Return a reference to a new timer.
+//
+PRTime& Timer::getNewTimer()
+{
+	if (nextTimer >= nTimersInABlock) {
+		timers = new(pool) PRTime[nTimersInABlock];
+		nextTimer = 0;
+	}
+	return timers[nextTimer++];
+}
+
+static Uint32 timersAreFrozen = 0;
+
+//
+// Start the named timer.
+//
+void Timer::start(const char* name)
+{
+	if (timersAreFrozen)
+		return;
+
+	freezeTimers();
+
+	TimerEntry& timer = getTimerEntry(name);
+	PR_ASSERT(!timer.running);
+
+	timer.accumulator = 0;
+	timer.running = true;
+	timer.done = false;
+
+	unfreezeTimers();
+}
+
+//
+// Stop the named timer.
+//
+void Timer::stop(const char* name)
+{
+	if (timersAreFrozen)
+		return;
+
+	freezeTimers();
+
+	TimerEntry& timer = getTimerEntry(name);
+	PR_ASSERT(timer.running);
+	timer.running = false;
+	timer.done = true;
+
+	unfreezeTimers();
+}
+
+//
+// Freeze all the running timers.
+//
+void Timer::freezeTimers()
+{
+	PRTime when = PR_Now() - adjust;
+
+	if (timersAreFrozen == 0) {
+		Vector<TimerEntry*> entries = timerEntries;
+		Uint32 count = entries.size();
+
+		for (Uint32 i = 0; i < count; i++) {
+			TimerEntry& entry = *entries[i];
+			if (entry.running) {
+				entry.accumulator += (when - *entry.startTime);
+			}
+		}
+	}
+	timersAreFrozen++;
+}
+ 
+//
+// Unfreeze all the running timers.
+//
+void Timer::unfreezeTimers()
+{
+	PR_ASSERT(timersAreFrozen != 0);
+	timersAreFrozen--;
+
+	if (timersAreFrozen == 0) {
+		Vector<TimerEntry *> entries = timerEntries;
+		Uint32 count = entries.size();
+		
+		PRTime& newStart = getNewTimer();
+		
+		for (Uint32 i = 0; i < count; i++) {
+			TimerEntry& entry = *entries[i];
+			if (entry.running) {
+				entry.startTime = &newStart;
+			}
+		}
+		
+		newStart = PR_Now();
+	}
+}
+
+//
+// Print the named timer in the file f.
+//
+void Timer::print(FILE* f, const char *name)
+{
+	if (timersAreFrozen)
+		return;
+
+	freezeTimers();
+
+	TimerEntry& timer = getTimerEntry(name);
+  
+	PR_ASSERT(timer.done);
+	PRTime elapsed = timer.accumulator;
+
+	if (elapsed >> 32) {
+		fprintf(f, "[timer %s out of range]\n", name);
+	} else {
+		fprintf(f, "[%dus in %s]\n", Uint32(elapsed), name);
+	}
+	fflush(f);
+
+	unfreezeTimers();
+}
+

mozilla/ef/Compiler/RegisterAllocator/Timer.h

@@ -0,0 +1,80 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _TIMER_H_
+#define _TIMER_H_
+
+#include "Fundamentals.h"
+#include "HashTable.h"
+#include "prtime.h"
+
+//
+// Naming convention:
+//  As the class Timer contains only static methods, the timer's name should start with the 
+//  module name. Otherwise starting 2 timers with the same name will assert.
+//
+
+#ifndef NO_TIMER
+
+struct TimerEntry
+{
+	PRTime *startTime;					// Current time when we start the timer.
+	PRTime accumulator;					// Time spent in this timer.
+	bool running;						// True if the timer is running.
+	bool done;							// True if the timer was running and was stopped. 
+};
+
+class Timer
+{
+private:
+
+	// Return the named timer.
+	static TimerEntry& getTimerEntry(const char* name);
+	// Return a reference to a new Timer.
+	static PRTime& getNewTimer();
+
+public:
+
+	// Start the timer.
+	static void start(const char* name);
+	// Stop the timer.
+	static void stop(const char* name);
+	// Freeze all the running timers.
+	static void freezeTimers();
+	// Unfreeze all the running timers.
+	static void unfreezeTimers();
+	// Print the timer.
+	static void print(FILE* f, const char *name);
+};
+
+inline void startTimer(const char* name) 		{Timer::start(name);}
+inline void stopTimer(const char* name) 		{Timer::stop(name); Timer::print(stdout, name);}
+#define START_TIMER_SAFE						Timer::freezeTimers();
+#define END_TIMER_SAFE							Timer::unfreezeTimers();
+#define TIMER_SAFE(x)							START_TIMER_SAFE x; END_TIMER_SAFE
+
+#else /* NO_TIMER */
+
+inline void startTimer(const char* /*name*/) 	{} 
+inline void stopTimer(const char* /*name*/) 	{} 
+#define START_TIMER_SAFE
+#define END_TIMER_SAFE
+#define TIMER_SAFE(x)							x;
+
+#endif /* NO_TIMER */
+#endif /* _TIMER_H_ */

mozilla/ef/Compiler/RegisterAllocator/VirtualRegister.cpp

@@ -0,0 +1,40 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#include "Fundamentals.h"
+#include "VirtualRegister.h"
+#include "Instruction.h"
+
+//------------------------------------------------------------------------------
+// VirtualRegister -
+
+#ifdef MANUAL_TEMPLATES
+template class IndexedPool<VirtualRegister>;
+#endif
+
+// Set the defining instruction.
+//
+void VirtualRegister::setDefiningInstruction(Instruction& instruction)
+{
+	if (definingInstruction != NULL) {
+		if ((instruction.getFlags() & ifCopy) && (definingInstruction->getFlags() & ifPhiNode))
+			return;
+	}
+	definingInstruction = &instruction;
+}
+

mozilla/ef/Compiler/RegisterAllocator/VirtualRegister.h

@@ -0,0 +1,116 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ *
+ * The contents of this file are subject to the Netscape Public License
+ * Version 1.0 (the "NPL"); you may not use this file except in
+ * compliance with the NPL.  You may obtain a copy of the NPL at
+ * http://www.mozilla.org/NPL/
+ *
+ * Software distributed under the NPL is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
+ * for the specific language governing rights and limitations under the
+ * NPL.
+ *
+ * The Initial Developer of this code under the NPL is Netscape
+ * Communications Corporation.  Portions created by Netscape are
+ * Copyright (C) 1998 Netscape Communications Corporation.  All Rights
+ * Reserved.
+ */
+
+#ifndef _VIRTUAL_REGISTER_H_
+#define _VIRTUAL_REGISTER_H_
+
+#include "Fundamentals.h"
+#include "IndexedPool.h"
+#include <string.h>
+
+#include "RegisterTypes.h"
+#include "RegisterClass.h"
+
+
+//------------------------------------------------------------------------------
+// VirtualRegister - 24b
+
+class Instruction;
+
+class VirtualRegister : public IndexedObject<VirtualRegister>
+{
+public:
+
+	Instruction*					definingInstruction;	// Instruction defining this VR.
+
+	// Initialize a VR of the given classKind.
+	VirtualRegister(RegisterClassKind /*classKind*/) : definingInstruction(NULL) {}
+
+	// Return the defining instruction for this VR.
+	Instruction* getDefiningInstruction() const {return definingInstruction;}
+	// Set the defining instruction.
+	void setDefiningInstruction(Instruction& insn);
+};
+
+// Return true if the VirtualRegisters are equals. The only way 2 VRs can be equal is if
+// they have the same index. If they have the same index then they are at the same
+// address in the indexed pool.
+//
+inline bool operator == (const VirtualRegister& regA, const VirtualRegister& regB) {return &regA == &regB;}
+
+//------------------------------------------------------------------------------
+// VirtualRegisterManager -
+
+struct PreColoredRegister
+{
+	RegisterID id;
+	RegisterName color;
+};
+
+class VirtualRegisterManager
+{
+private:
+
+	IndexedPool<VirtualRegister> registerPool;
+	PreColoredRegister machineRegister[6];
+
+public:
+	VirtualRegisterManager()
+		{
+			for (Uint32 i = 0; i < 6; i++)
+				machineRegister[i].id = invalidID;
+		}
+
+	// Return the VirtualRegister at the given index.
+	VirtualRegister& getVirtualRegister(RegisterName name) const {return registerPool.get(name);}
+
+	// Return a new VirtualRegister.
+	RegisterID newVirtualRegister(RegisterClassKind classKind)
+		{
+			VirtualRegister& vReg = *new(registerPool) VirtualRegister(classKind);
+			RegisterID rid;
+
+			setName(rid, RegisterName(vReg.getIndex()));
+			setClass(rid, classKind);
+			return rid;
+		}
+
+	RegisterID newMachineRegister(RegisterName name, RegisterClassKind classKind)
+		{
+			RegisterID rid = machineRegister[name].id;
+
+			if (rid == invalidID) {
+				rid = newVirtualRegister(classKind);
+				DEBUG_ONLY(setMachineRegister(rid));
+				machineRegister[name].id = rid;
+				machineRegister[name].color = name;
+			}
+
+			return rid;
+		}
+
+	PreColoredRegister* getMachineRegistersBegin() const {return (PreColoredRegister*) machineRegister;} // FIX
+	PreColoredRegister* getMachineRegistersEnd() const {return (PreColoredRegister*) &machineRegister[6];} // FIX
+
+	// Return the VirtualRegister universe size.
+	Uint32 getSize() {return registerPool.getSize();}
+
+	void setSize(Uint32 size) {registerPool.setSize(size);}
+};
+
+#endif // _VIRTUAL_REGISTER_H_

mozilla/security/nss/lib/freebl/manifest.mn

@@ -1,152 +0,0 @@
-# 
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the Netscape security libraries.
-#
-# The Initial Developer of the Original Code is
-# Netscape Communications Corporation.
-# Portions created by the Initial Developer are Copyright (C) 1994-2000
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#   Dr Vipul Gupta <vipul.gupta@sun.com> and
-#   Douglas Stebila <douglas@stebila.ca>, Sun Microsystems
-#   Laboratories
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-CORE_DEPTH = ../../..
-
-MODULE = nss
-
-ifndef FREEBL_RECURSIVE_BUILD
-  LIBRARY_NAME = freebl
-else
-  ifdef USE_PURE_32
-    CORE_DEPTH = ../../../..
-    LIBRARY_NAME = freebl_pure32
-  else
-    LIBRARY_NAME = freebl_hybrid
-  endif
-endif
-
-# same version as rest of freebl
-LIBRARY_VERSION = _3
-
-DEFINES += -DSHLIB_SUFFIX=\"$(DLL_SUFFIX)\" -DSHLIB_PREFIX=\"$(DLL_PREFIX)\"
-
-REQUIRES = 
-
-EXPORTS = \
-	blapit.h \
-	shsign.h \
-	ecl-exp.h \
-	$(NULL)
-
-PRIVATE_EXPORTS = \
-	blapi.h \
-	secmpi.h \
-	secrng.h \
-	ec.h \
-	ecl.h \
-	ecl-curve.h \
-	$(NULL)
-
-MPI_HDRS = mpi-config.h mpi.h mpi-priv.h mplogic.h mpprime.h logtab.h mp_gf2m.h
-MPI_SRCS = mpprime.c mpmontg.c mplogic.c mpi.c mp_gf2m.c mpcpucache.c
-ECL_HDRS = ecl-exp.h ecl.h ec2.h ecp.h ecl-priv.h
-ifdef NSS_ENABLE_ECC
-ECL_SRCS = ecl.c ecl_curve.c ecl_mult.c ecl_gf.c \
-	ec2_aff.c ec2_mont.c ec2_proj.c \
-	ec2_163.c ec2_193.c ec2_233.c \
-	ecp_aff.c ecp_jac.c ecp_mont.c \
-	ecp_192.c ecp_224.c \
-	ec_naf.c ecp_jm.c
-else
-ECL_SRCS = $(NULL)
-endif
-
-CSRCS = \
-	ldvector.c \
-	prng_fips1861.c \
-	sysrand.c \
-	sha_fast.c \
-	md2.c \
-	md5.c \
-	sha512.c \
-	alg2268.c \
-	arcfour.c \
-	arcfive.c \
-	desblapi.c \
-	des.c \
-	rijndael.c \
-	aeskeywrap.c \
-	dh.c \
-	ec.c \
-	pqg.c \
-	dsa.c \
-	rsa.c \
-	shvfy.c \
-	$(MPI_SRCS) \
-	$(ECL_SRCS) \
-	$(NULL)
-
-ALL_CSRCS := $(CSRCS)
-
-ALL_HDRS =  \
-	blapi.h \
-	blapit.h \
-	des.h \
-	ec.h \
-	loader.h \
-	rijndael.h \
-	secmpi.h \
-	sha.h \
-	sha_fast.h \
-	shsign.h \
-	vis_proto.h \
-	$(NULL)
-
-ifdef NSS_ENABLE_ECC
-DEFINES += -DNSS_ENABLE_ECC
-endif
-
-ifdef AES_GEN_TBL
-DEFINES += -DRIJNDAEL_GENERATE_TABLES
-else 
-ifdef AES_GEN_TBL_M
-DEFINES += -DRIJNDAEL_GENERATE_TABLES_MACRO
-else
-ifdef AES_GEN_VAL
-DEFINES += -DRIJNDAEL_GENERATE_VALUES
-else
-ifdef AES_GEN_VAL_M
-DEFINES += -DRIJNDAEL_GENERATE_VALUES_MACRO
-else
-DEFINES += -DRIJNDAEL_INCLUDE_TABLES
-endif
-endif
-endif
-endif

mozilla/security/nss/lib/freebl/mpi/Makefile

@@ -1,278 +0,0 @@
-#
-# Makefile for MPI library
-
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>.
-# Portions created by the Initial Developer are Copyright (C) 1998
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#   Netscape Communications Corporation
-#   Richard C. Swift        (swift@netscape.com)
-#   Douglas Stebila <douglas@stebila.ca>, Sun Microsystems Laboratories
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-#
-# $Id: Makefile,v 1.21 2005-02-02 22:28:22 gerv%gerv.net Exp $
-#
-
-## Define CC to be the C compiler you wish to use.  The GNU cc
-## compiler (gcc) should work, at the very least
-#CC=cc
-#CC=gcc
-
-## 
-## Define PERL to point to your local Perl interpreter.  It
-## should be Perl 5.x, although it's conceivable that Perl 4
-## might work ... I haven't tested it.
-##
-#PERL=/usr/bin/perl
-PERL=perl
-
-include target.mk
-
-CFLAGS+= $(XCFLAGS)
-
-##
-## Define LIBS to include any libraries you need to link against.
-## If NO_TABLE is define, LIBS should include '-lm' or whatever is
-## necessary to bring in the math library.  Otherwise, it can be
-## left alone, unless your system has other peculiar requirements.
-##
-LIBS=#-lmalloc#-lefence#-lm
-
-## 
-## Define RANLIB to be the library header randomizer; you might not
-## need this on some systems (just set it to 'echo' on these systems,
-## such as IRIX)
-##
-RANLIB=echo
-
-##
-## This is the version string used for the documentation and 
-## building the distribution tarball.  Don't mess with it unless
-## you are releasing a new version
-VERS=1.7p6
-
-## ----------------------------------------------------------------------
-## You probably don't need to change anything below this line...
-##
-
-##
-## This is the list of source files that need to be packed into
-## the distribution file
-SRCS=   mpi.c mpprime.c mplogic.c mp_gf2m.c mpmontg.c mpi-test.c primes.c tests/ \
-	utils/gcd.c utils/invmod.c utils/lap.c \
-	utils/ptab.pl utils/sieve.c utils/isprime.c\
-	utils/dec2hex.c utils/hex2dec.c utils/bbs_rand.c \
-	utils/bbsrand.c utils/prng.c utils/primegen.c \
-	utils/basecvt.c utils/makeprime.c\
-	utils/fact.c utils/exptmod.c utils/pi.c utils/metime.c \
-	utils/mpi.h utils/mpprime.h mulsqr.c \
-	make-test-arrays test-arrays.txt all-tests make-logtab \
-	types.pl stats timetest multest
-
-## These are the header files that go into the distribution file
-HDRS=mpi.h mpi-config.h utils/mpi.h utils/mpi-config.h mpprime.h mplogic.h mp_gf2m.h \
-     mp_gf2m-priv.h utils/bbs_rand.h tests/mpi.h tests/mpprime.h
-
-## These are the documentation files that go into the distribution file
-DOCS=README doc utils/README utils/PRIMES 
-
-## This is the list of tools built by 'make tools'
-TOOLS=gcd invmod isprime lap dec2hex hex2dec primegen prng \
-	basecvt fact exptmod pi makeprime identest
-
-LIBOBJS = mpprime.o mpmontg.o mplogic.o mp_gf2m.o mpi.o $(AS_OBJS)
-LIBHDRS = mpi-config.h mpi-priv.h mpi.h
-APPHDRS = mpi-config.h mpi.h mplogic.h mp_gf2m.h mpprime.h
-
-help:
-	@ echo ""
-	@ echo "The following targets can be built with this Makefile:"
-	@ echo ""
-	@ echo "libmpi.a     - arithmetic and prime testing library"
-	@ echo "mpi-test     - test driver (requires MP_IOFUNC)"
-	@ echo "tools        - command line tools"
-	@ echo "doc          - manual pages for tools"
-	@ echo "clean        - clean up objects and such"
-	@ echo "distclean    - get ready for distribution"
-	@ echo "dist         - distribution tarball"
-	@ echo ""
-
-.SUFFIXES: .c .o .i
-
-.c.i:
-	$(CC) $(CFLAGS) -E $< > $@
-
-#.c.o: $*.h $*.c
-#	$(CC) $(CFLAGS) -c $<
-
-#---------------------------------------
-
-$(LIBOBJS): $(LIBHDRS)
-
-logtab.h: make-logtab
-	$(PERL) make-logtab > logtab.h
-
-mpi.o: mpi.c logtab.h $(LIBHDRS)
-
-mplogic.o: mplogic.c mpi-priv.h mplogic.h $(LIBHDRS)
-
-mp_gf2m.o: mp_gf2m.c mpi-priv.h mp_gf2m.h mp_gf2m-priv.h $(LIBHDRS)
-
-mpmontg.o: mpmontg.c mpi-priv.h mplogic.h mpprime.h $(LIBHDRS)
-
-mpprime.o: mpprime.c mpi-priv.h mpprime.h mplogic.h primes.c $(LIBHDRS)
-
-mpi_mips.o: mpi_mips.s
-	$(CC) -o $@ $(ASFLAGS) -c mpi_mips.s
-
-mpi_sparc.o : montmulf.h
-
-mpv_sparcv9.s: vis_64.il mpv_sparc.c
-	$(CC) -o $@ $(SOLARIS_FPU_FLAGS) -S vis_64.il mpv_sparc.c
-
-mpv_sparcv8.s: vis_64.il mpv_sparc.c
-	$(CC) -o $@ $(SOLARIS_FPU_FLAGS) -S vis_32.il mpv_sparc.c
-
-montmulfv8.o montmulfv9.o mpv_sparcv8.o mpv_sparcv9.o : %.o : %.s 
-	$(CC) -o $@ $(SOLARIS_ASM_FLAGS) -c $<
-
-# This rule is used to build the .s sources, which are then hand optimized.
-#montmulfv8.s montmulfv9.s : montmulf%.s : montmulf%.il montmulf.c montmulf.h 
-#	$(CC) -o $@ $(SOLARIS_ASM_FLAGS) -S montmulf$*.il montmulf.c
-
-
-libmpi.a: $(LIBOBJS)
-	ar -cvr libmpi.a $(LIBOBJS)
-	$(RANLIB) libmpi.a
-
-lib libs: libmpi.a
-
-mpi.i: mpi.h
-
-#---------------------------------------
-
-MPTESTOBJS = mptest1.o mptest2.o mptest3.o mptest3a.o mptest4.o mptest4a.o \
-	mptest4b.o mptest6.o mptest7.o mptest8.o mptest9.o mptestb.o
-MPTESTS = $(MPTESTOBJS:.o=)
-
-$(MPTESTOBJS): mptest%.o: tests/mptest-%.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -o $@ -c $<
-
-$(MPTESTS): mptest%: mptest%.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^  $(LIBS)
-
-tests: mptest1 mptest2 mptest3 mptest3a mptest4 mptest4a mptest4b mptest6 \
-	mptestb bbsrand
-
-utests: mptest7 mptest8 mptest9
-
-#---------------------------------------
-
-EXTRAOBJS = bbsrand.o bbs_rand.o prng.o
-UTILOBJS = primegen.o metime.o identest.o basecvt.o fact.o exptmod.o pi.o \
-	makeprime.o gcd.o invmod.o lap.o isprime.o \
-	dec2hex.o hex2dec.o
-UTILS = $(UTILOBJS:.o=) 
-
-$(UTILS): % : %.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
-
-$(UTILOBJS) $(EXTRAOBJS): %.o : utils/%.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -o $@ -c $<
-
-prng: prng.o bbs_rand.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
-
-bbsrand: bbsrand.o bbs_rand.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
-
-utils: $(UTILS) prng bbsrand
-
-#---------------------------------------
-
-test-info.c: test-arrays.txt
-	$(PERL) make-test-arrays test-arrays.txt > test-info.c
-
-mpi-test.o: mpi-test.c test-info.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -o $@ -c $<
-
-mpi-test: mpi-test.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
-
-mdxptest.o: mdxptest.c $(LIBHDRS) mpi-priv.h
-
-mdxptest: mdxptest.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
-
-mulsqr.o: mulsqr.c logtab.h mpi.h mpi-config.h mpprime.h 
-	$(CC) $(CFLAGS) -DMP_SQUARE=1 -o $@ -c mulsqr.c 
-
-mulsqr: mulsqr.o libmpi.a
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
-
-#---------------------------------------
-
-alltests: tests utests mpi-test
-
-tools: $(TOOLS)
-
-doc:
-	(cd doc; ./build)
-
-clean:
-	rm -f *.o *.a *.i
-	rm -f core
-	rm -f *~ .*~
-	rm -f utils/*.o
-	rm -f utils/core
-	rm -f utils/*~ utils/.*~
-
-clobber: clean
-	rm -f $(TOOLS) $(UTILS)
-
-distclean: clean
-	rm -f mptest? mpi-test metime mulsqr karatsuba
-	rm -f mptest?a mptest?b
-	rm -f utils/mptest?
-	rm -f test-info.c logtab.h
-	rm -f libmpi.a
-	rm -f $(TOOLS)
-
-dist: Makefile $(HDRS) $(SRCS) $(DOCS)
-	tar -cvf mpi-$(VERS).tar Makefile $(HDRS) $(SRCS) $(DOCS)
-	pgps -ab mpi-$(VERS).tar
-	chmod +r mpi-$(VERS).tar.asc
-	gzip -9 mpi-$(VERS).tar
-
-# END

mozilla/security/nss/lib/freebl/mpi/Makefile.os2

@@ -1,280 +0,0 @@
-#
-# Makefile.win - gmake Makefile for building MPI with VACPP on OS/2
-#
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>.
-# Portions created by the Initial Developer are Copyright (C) 1998
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#   Netscape Communications Corporation
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-#
-# $Id: Makefile.os2,v 1.3 2005-02-02 22:28:22 gerv%gerv.net Exp $
-#
-
-## Define CC to be the C compiler you wish to use.  The GNU cc
-## compiler (gcc) should work, at the very least
-#CC=cc
-#CC=gcc
-CC=icc.exe
-AS=alp.exe
-
-## 
-## Define PERL to point to your local Perl interpreter.  It
-## should be Perl 5.x, although it's conceivable that Perl 4
-## might work ... I haven't tested it.
-##
-#PERL=/usr/bin/perl
-PERL=perl
-
-##
-## Define CFLAGS to contain any local options your compiler
-## setup requires.
-##
-## Conditional compilation options are no longer here; see
-## the file 'mpi-config.h' instead.
-##
-MPICMN = -I. -DMP_API_COMPATIBLE -DMP_IOFUNC -DMP_USE_UINT_DIGIT -DMP_NO_MP_WORD
-
-#OS/2
-AS_SRCS = mpi_x86.asm
-MPICMN += -DMP_ASSEMBLY_MULTIPLY -DMP_ASSEMBLY_SQUARE -DMP_ASSEMBLY_DIV_2DX1D
-#CFLAGS= -Od -Z7 -MD -W3 -nologo -D_X86_ -DXP_PC \
- -DDEBUG -D_DEBUG -UNDEBUG -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-#CFLAGS = -O2 -MD -W3 -nologo -D_X86_ -DXP_PC -UDEBUG -U_DEBUG -DNDEBUG \
- -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-#CFLAGS = -Od -Z7 -MD -W3 -nologo -D_X86_ -DXP_PC -UDEBUG -U_DEBUG -DNDEBUG \
- -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-CFLAGS = /Ti+ -D_X86_ -DXP_PC -UDEBUG -U_DEBUG -DNDEBUG \
- $(MPICMN)
-ASFLAGS =
-
-##
-## Define LIBS to include any libraries you need to link against.
-## If NO_TABLE is define, LIBS should include '-lm' or whatever is
-## necessary to bring in the math library.  Otherwise, it can be
-## left alone, unless your system has other peculiar requirements.
-##
-LIBS=#-lmalloc#-lefence#-lm
-
-## 
-## Define RANLIB to be the library header randomizer; you might not
-## need this on some systems (just set it to 'echo' on these systems,
-## such as IRIX)
-##
-RANLIB=echo
-
-##
-## This is the version string used for the documentation and 
-## building the distribution tarball.  Don't mess with it unless
-## you are releasing a new version
-VERS=1.7p6
-
-## ----------------------------------------------------------------------
-## You probably don't need to change anything below this line...
-##
-
-##
-## This is the list of source files that need to be packed into
-## the distribution file
-SRCS=   mpi.c mpprime.c mplogic.c mpmontg.c mpi-test.c primes.c tests/ \
-	utils/gcd.c utils/invmod.c utils/lap.c \
-	utils/ptab.pl utils/sieve.c utils/isprime.c\
-	utils/dec2hex.c utils/hex2dec.c utils/bbs_rand.c \
-	utils/bbsrand.c utils/prng.c utils/primegen.c \
-	utils/basecvt.c utils/makeprime.c\
-	utils/fact.c utils/exptmod.c utils/pi.c utils/metime.c \
-	utils/mpi.h utils/mpprime.h mulsqr.c \
-	make-test-arrays test-arrays.txt all-tests make-logtab \
-	types.pl stats timetest multest
-
-## These are the header files that go into the distribution file
-HDRS=mpi.h mpi-config.h utils/mpi.h utils/mpi-config.h mpprime.h mplogic.h \
-     utils/bbs_rand.h tests/mpi.h tests/mpprime.h
-
-## These are the documentation files that go into the distribution file
-DOCS=README doc utils/README utils/PRIMES 
-
-## This is the list of tools built by 'make tools'
-TOOLS=gcd.exe invmod.exe isprime.exe lap.exe dec2hex.exe hex2dec.exe \
- primegen.exe prng.exe basecvt.exe fact.exe exptmod.exe pi.exe makeprime.exe
-
-AS_OBJS = $(AS_SRCS:.asm=.obj)
-LIBOBJS = mpprime.obj mpmontg.obj mplogic.obj mpi.obj $(AS_OBJS)
-LIBHDRS = mpi-config.h mpi-priv.h mpi.h
-APPHDRS = mpi-config.h mpi.h mplogic.h mpprime.h
-
-
-help:
-	@ echo ""
-	@ echo "The following targets can be built with this Makefile:"
-	@ echo ""
-	@ echo "mpi.lib      - arithmetic and prime testing library"
-	@ echo "mpi-test.exe - test driver (requires MP_IOFUNC)"
-	@ echo "tools        - command line tools"
-	@ echo "doc          - manual pages for tools"
-	@ echo "clean        - clean up objects and such"
-	@ echo "distclean    - get ready for distribution"
-	@ echo "dist         - distribution tarball"
-	@ echo ""
-
-.SUFFIXES: .c .obj .i .lib .exe .asm
-
-.c.i:
-	$(CC) $(CFLAGS) -E $< > $@
-
-.c.obj: 
-	$(CC) $(CFLAGS) -c $<
-
-.asm.obj:
-	$(AS) $(ASFLAGS) $<
-
-.obj.exe:
-	$(CC) $(CFLAGS) -Fo$@ $<
-
-#---------------------------------------
-
-$(LIBOBJS): $(LIBHDRS)
-
-logtab.h: make-logtab
-	$(PERL) make-logtab > logtab.h
-
-mpi.obj: mpi.c logtab.h $(LIBHDRS)
-
-mplogic.obj: mplogic.c mpi-priv.h mplogic.h $(LIBHDRS)
-
-mpmontg.obj: mpmontg.c mpi-priv.h mplogic.h mpprime.h $(LIBHDRS)
-
-mpprime.obj: mpprime.c mpi-priv.h mpprime.h mplogic.h primes.c $(LIBHDRS)
-
-mpi_mips.obj: mpi_mips.s
-	$(CC) -Fo$@ $(ASFLAGS) -c mpi_mips.s
-
-mpi.lib: $(LIBOBJS)
-	ilib /out:mpi.lib $(LIBOBJS)
-	$(RANLIB) mpi.lib
-
-lib libs: mpi.lib
-
-#---------------------------------------
-
-MPTESTOBJS = mptest1.obj mptest2.obj mptest3.obj mptest3a.obj mptest4.obj \
- mptest4a.obj mptest4b.obj mptest6.obj mptest7.obj mptest8.obj mptest9.obj
-MPTESTS = $(MPTESTOBJS:.obj=.exe)
-
-$(MPTESTOBJS): mptest%.obj: tests/mptest-%.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -Fo$@ -c $<
-
-$(MPTESTS): mptest%.exe: mptest%.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^ 
-
-tests: mptest1.exe mptest2.exe mptest3.exe mptest3a.exe mptest4.exe \
- mptest4a.exe mptest4b.exe mptest6.exe bbsrand.exe
-
-utests: mptest7.exe mptest8.exe mptest9.exe
-
-#---------------------------------------
-
-EXTRAOBJS = bbsrand.obj bbs_rand.obj prng.obj
-UTILOBJS = primegen.obj metime.obj identest.obj basecvt.obj fact.obj \
- exptmod.obj pi.obj makeprime.obj karatsuba.obj gcd.obj invmod.obj lap.obj \
- isprime.obj dec2hex.obj hex2dec.obj
-UTILS = $(UTILOBJS:.obj=.exe) 
-
-$(UTILS): %.exe : %.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^ 
-
-$(UTILOBJS) $(EXTRAOBJS): %.obj : utils/%.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -Fo$@ -c $<
-
-prng.exe: prng.obj bbs_rand.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-bbsrand.exe: bbsrand.obj bbs_rand.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-utils: $(UTILS) prng.exe bbsrand.exe
-
-#---------------------------------------
-
-test-info.c: test-arrays.txt
-	$(PERL) make-test-arrays test-arrays.txt > test-info.c
-
-mpi-test.obj: mpi-test.c test-info.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -Fo$@ -c $<
-
-mpi-test.exe: mpi-test.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-mdxptest.obj: mdxptest.c $(LIBHDRS) mpi-priv.h
-
-mdxptest.exe: mdxptest.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-mulsqr.obj: mulsqr.c logtab.h mpi.h mpi-config.h mpprime.h 
-	$(CC) $(CFLAGS) -DMP_SQUARE=1 -Fo$@ -c mulsqr.c 
-
-mulsqr.exe: mulsqr.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-#---------------------------------------
-
-alltests: tests utests mpi-test.exe
-
-tools: $(TOOLS)
-
-doc:
-	(cd doc; ./build)
-
-clean:
-	rm -f *.obj *.lib *.pdb *.ilk
-	cd utils; rm -f *.obj *.lib *.pdb *.ilk
-
-distclean: clean
-	rm -f mptest? mpi-test metime mulsqr karatsuba
-	rm -f mptest?a mptest?b
-	rm -f utils/mptest?
-	rm -f test-info.c logtab.h
-	rm -f mpi.lib
-	rm -f $(TOOLS)
-
-dist: Makefile $(HDRS) $(SRCS) $(DOCS)
-	tar -cvf mpi-$(VERS).tar Makefile $(HDRS) $(SRCS) $(DOCS)
-	pgps -ab mpi-$(VERS).tar
-	chmod +r mpi-$(VERS).tar.asc
-	gzip -9 mpi-$(VERS).tar
-
-
-print: 
-	@echo LIBOBJS = $(LIBOBJS)
-# END

mozilla/security/nss/lib/freebl/mpi/Makefile.win

@@ -1,280 +0,0 @@
-#
-# Makefile.win - gmake Makefile for building MPI with MSVC on NT
-
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>.
-# Portions created by the Initial Developer are Copyright (C) 1998
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#   Netscape Communications Corporation
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-#
-# $Id: Makefile.win,v 1.3 2005-02-02 22:28:22 gerv%gerv.net Exp $
-#
-
-## Define CC to be the C compiler you wish to use.  The GNU cc
-## compiler (gcc) should work, at the very least
-#CC=cc
-#CC=gcc
-CC=cl.exe
-AS=ml.exe
-
-## 
-## Define PERL to point to your local Perl interpreter.  It
-## should be Perl 5.x, although it's conceivable that Perl 4
-## might work ... I haven't tested it.
-##
-#PERL=/usr/bin/perl
-PERL=perl
-
-##
-## Define CFLAGS to contain any local options your compiler
-## setup requires.
-##
-## Conditional compilation options are no longer here; see
-## the file 'mpi-config.h' instead.
-##
-MPICMN = -I. -DMP_API_COMPATIBLE -DMP_IOFUNC
-
-#NT
-AS_SRCS = mpi_x86.asm
-MPICMN += -DMP_ASSEMBLY_MULTIPLY -DMP_ASSEMBLY_SQUARE -DMP_ASSEMBLY_DIV_2DX1D
-#CFLAGS= -Od -Z7 -MD -W3 -nologo -D_X86_ -DXP_PC \
- -DDEBUG -D_DEBUG -UNDEBUG -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-#CFLAGS = -O2 -MD -W3 -nologo -D_X86_ -DXP_PC -UDEBUG -U_DEBUG -DNDEBUG \
- -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-#CFLAGS = -Od -Z7 -MD -W3 -nologo -D_X86_ -DXP_PC -UDEBUG -U_DEBUG -DNDEBUG \
- -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-CFLAGS = -O2 -Z7 -MD -W3 -nologo -D_X86_ -DXP_PC -UDEBUG -U_DEBUG -DNDEBUG \
- -DWIN32 -D_WINDOWS -DWIN95 $(MPICMN)
-ASFLAGS = -Cp -Sn -Zi -coff -I. 
-
-##
-## Define LIBS to include any libraries you need to link against.
-## If NO_TABLE is define, LIBS should include '-lm' or whatever is
-## necessary to bring in the math library.  Otherwise, it can be
-## left alone, unless your system has other peculiar requirements.
-##
-LIBS=#-lmalloc#-lefence#-lm
-
-## 
-## Define RANLIB to be the library header randomizer; you might not
-## need this on some systems (just set it to 'echo' on these systems,
-## such as IRIX)
-##
-RANLIB=echo
-
-##
-## This is the version string used for the documentation and 
-## building the distribution tarball.  Don't mess with it unless
-## you are releasing a new version
-VERS=1.7p6
-
-## ----------------------------------------------------------------------
-## You probably don't need to change anything below this line...
-##
-
-##
-## This is the list of source files that need to be packed into
-## the distribution file
-SRCS=   mpi.c mpprime.c mplogic.c mpmontg.c mpi-test.c primes.c tests/ \
-	utils/gcd.c utils/invmod.c utils/lap.c \
-	utils/ptab.pl utils/sieve.c utils/isprime.c\
-	utils/dec2hex.c utils/hex2dec.c utils/bbs_rand.c \
-	utils/bbsrand.c utils/prng.c utils/primegen.c \
-	utils/basecvt.c utils/makeprime.c\
-	utils/fact.c utils/exptmod.c utils/pi.c utils/metime.c \
-	utils/mpi.h utils/mpprime.h mulsqr.c \
-	make-test-arrays test-arrays.txt all-tests make-logtab \
-	types.pl stats timetest multest
-
-## These are the header files that go into the distribution file
-HDRS=mpi.h mpi-config.h utils/mpi.h utils/mpi-config.h mpprime.h mplogic.h \
-     utils/bbs_rand.h tests/mpi.h tests/mpprime.h
-
-## These are the documentation files that go into the distribution file
-DOCS=README doc utils/README utils/PRIMES 
-
-## This is the list of tools built by 'make tools'
-TOOLS=gcd.exe invmod.exe isprime.exe lap.exe dec2hex.exe hex2dec.exe \
- primegen.exe prng.exe basecvt.exe fact.exe exptmod.exe pi.exe makeprime.exe
-
-AS_OBJS = $(AS_SRCS:.asm=.obj)
-LIBOBJS = mpprime.obj mpmontg.obj mplogic.obj mpi.obj $(AS_OBJS)
-LIBHDRS = mpi-config.h mpi-priv.h mpi.h
-APPHDRS = mpi-config.h mpi.h mplogic.h mpprime.h
-
-
-help:
-	@ echo ""
-	@ echo "The following targets can be built with this Makefile:"
-	@ echo ""
-	@ echo "mpi.lib     - arithmetic and prime testing library"
-	@ echo "mpi-test     - test driver (requires MP_IOFUNC)"
-	@ echo "tools        - command line tools"
-	@ echo "doc          - manual pages for tools"
-	@ echo "clean        - clean up objects and such"
-	@ echo "distclean    - get ready for distribution"
-	@ echo "dist         - distribution tarball"
-	@ echo ""
-
-.SUFFIXES: .c .obj .i .lib .exe .asm
-
-.c.i:
-	$(CC) $(CFLAGS) -E $< > $@
-
-.c.obj: 
-	$(CC) $(CFLAGS) -c $<
-
-.asm.obj:
-	$(AS) $(ASFLAGS) -c $<
-
-.obj.exe:
-	$(CC) $(CFLAGS) -Fo$@ $<
-
-#---------------------------------------
-
-$(LIBOBJS): $(LIBHDRS)
-
-logtab.h: make-logtab
-	$(PERL) make-logtab > logtab.h
-
-mpi.obj: mpi.c logtab.h $(LIBHDRS)
-
-mplogic.obj: mplogic.c mpi-priv.h mplogic.h $(LIBHDRS)
-
-mpmontg.obj: mpmontg.c mpi-priv.h mplogic.h mpprime.h $(LIBHDRS)
-
-mpprime.obj: mpprime.c mpi-priv.h mpprime.h mplogic.h primes.c $(LIBHDRS)
-
-mpi_mips.obj: mpi_mips.s
-	$(CC) -Fo$@ $(ASFLAGS) -c mpi_mips.s
-
-mpi.lib: $(LIBOBJS)
-	ar -cvr mpi.lib $(LIBOBJS)
-	$(RANLIB) mpi.lib
-
-lib libs: mpi.lib
-
-#---------------------------------------
-
-MPTESTOBJS = mptest1.obj mptest2.obj mptest3.obj mptest3a.obj mptest4.obj \
- mptest4a.obj mptest4b.obj mptest6.obj mptest7.obj mptest8.obj mptest9.obj
-MPTESTS = $(MPTESTOBJS:.obj=.exe)
-
-$(MPTESTOBJS): mptest%.obj: tests/mptest-%.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -Fo$@ -c $<
-
-$(MPTESTS): mptest%.exe: mptest%.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^ 
-
-tests: mptest1.exe mptest2.exe mptest3.exe mptest3a.exe mptest4.exe \
- mptest4a.exe mptest4b.exe mptest6.exe bbsrand.exe
-
-utests: mptest7.exe mptest8.exe mptest9.exe
-
-#---------------------------------------
-
-EXTRAOBJS = bbsrand.obj bbs_rand.obj prng.obj
-UTILOBJS = primegen.obj metime.obj identest.obj basecvt.obj fact.obj \
- exptmod.obj pi.obj makeprime.obj karatsuba.obj gcd.obj invmod.obj lap.obj \
- isprime.obj dec2hex.obj hex2dec.obj
-UTILS = $(UTILOBJS:.obj=.exe) 
-
-$(UTILS): %.exe : %.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^ 
-
-$(UTILOBJS) $(EXTRAOBJS): %.obj : utils/%.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -Fo$@ -c $<
-
-prng.exe: prng.obj bbs_rand.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-bbsrand.exe: bbsrand.obj bbs_rand.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-utils: $(UTILS) prng.exe bbsrand.exe
-
-#---------------------------------------
-
-test-info.c: test-arrays.txt
-	$(PERL) make-test-arrays test-arrays.txt > test-info.c
-
-mpi-test.obj: mpi-test.c test-info.c $(LIBHDRS)
-	$(CC) $(CFLAGS) -Fo$@ -c $<
-
-mpi-test.exe: mpi-test.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-mdxptest.obj: mdxptest.c $(LIBHDRS) mpi-priv.h
-
-mdxptest.exe: mdxptest.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-mulsqr.obj: mulsqr.c logtab.h mpi.h mpi-config.h mpprime.h 
-	$(CC) $(CFLAGS) -DMP_SQUARE=1 -Fo$@ -c mulsqr.c 
-
-mulsqr.exe: mulsqr.obj mpi.lib $(LIBS)
-	$(CC) $(CFLAGS) -Fo$@ $^
-
-#---------------------------------------
-
-alltests: tests utests mpi-test.exe
-
-tools: $(TOOLS)
-
-doc:
-	(cd doc; ./build)
-
-clean:
-	rm -f *.obj *.lib *.pdb *.ilk
-	cd utils; rm -f *.obj *.lib *.pdb *.ilk
-
-distclean: clean
-	rm -f mptest? mpi-test metime mulsqr karatsuba
-	rm -f mptest?a mptest?b
-	rm -f utils/mptest?
-	rm -f test-info.c logtab.h
-	rm -f mpi.lib
-	rm -f $(TOOLS)
-
-dist: Makefile $(HDRS) $(SRCS) $(DOCS)
-	tar -cvf mpi-$(VERS).tar Makefile $(HDRS) $(SRCS) $(DOCS)
-	pgps -ab mpi-$(VERS).tar
-	chmod +r mpi-$(VERS).tar.asc
-	gzip -9 mpi-$(VERS).tar
-
-
-print: 
-	@echo LIBOBJS = $(LIBOBJS)
-# END

mozilla/security/nss/lib/freebl/mpi/README

@@ -1,799 +0,0 @@
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version
-1.1 (the "License"); you may not use this file except in compliance with
-the License. You may obtain a copy of the License at
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1997-2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-About the MPI Library
----------------------
-
-The files 'mpi.h' and 'mpi.c' define a simple, arbitrary precision
-signed integer arithmetic package.  The implementation is not the most
-efficient possible, but the code is small and should be fairly easily
-portable to just about any machine that supports an ANSI C compiler,
-as long as it is capable of at least 16-bit arithmetic (but also see
-below for more on this).
-
-This library was written with an eye to cryptographic applications;
-thus, some care is taken to make sure that temporary values are not
-left lying around in memory when they are no longer in use.  This adds
-some overhead for zeroing buffers before they are released back into
-the free pool; however, it gives you the assurance that there is only
-one copy of your important values residing in your process's address
-space at a time.  Obviously, it is difficult to guarantee anything, in
-a pre-emptive multitasking environment, but this at least helps you
-keep a lid on the more obvious ways your data can get spread around in
-memory.
-
-
-Using the Library
------------------
-
-To use the MPI library in your program, you must include the header:
-
-#include "mpi.h"
-
-This header provides all the type and function declarations you'll
-need to use the library.  Almost all the names defined by the library
-begin with the prefix 'mp_', so it should be easy to keep them from
-clashing with your program's namespace (he says, glibly, knowing full
-well there are always pathological cases).
-
-There are a few things you may want to configure about the library.
-By default, the MPI library uses an unsigned short for its digit type,
-and an unsigned int for its word type.  The word type must be big
-enough to contain at least two digits, for the primitive arithmetic to
-work out.  On my machine, a short is 2 bytes and an int is 4 bytes --
-but if you have 64-bit ints, you might want to use a 4-byte digit and
-an 8-byte word.  I have tested the library using 1-byte digits and
-2-byte words, as well.  Whatever you choose to do, the things you need
-to change are:
-
-(1) The type definitions for mp_digit and mp_word.
-
-(2) The macro DIGIT_FMT which tells mp_print() how to display a
-    single digit.  This is just a printf() format string, so you
-    can adjust it appropriately.
-
-(3) The macros DIGIT_MAX and MP_WORD_MAX, which specify the 
-    largest value expressible in an mp_digit and an mp_word,
-    respectively.
-
-Both the mp_digit and mp_word should be UNSIGNED integer types.  The
-code relies on having the full positive precision of the type used for
-digits and words.
-
-The remaining type definitions should be left alone, for the most
-part.  The code in the library does not make any significant
-assumptions about the sizes of things, but there is little if any
-reason to change the other parameters, so I would recommend you leave
-them as you found them.
-
-The library comes with a Perl script, 'types.pl', which will scan your
-current Makefile settings, and attempt to find good definitions for
-these types.  It relies on a Unix sort of build environment, so it
-probably won't work under MacOS or Windows, but it can be convenient
-if you're porting to a new flavour of Unix.  Just run 'types.pl' at
-the command line, and it will spit out its results to the standard
-output.
-
-
-Conventions
------------
-
-Most functions in the library return a value of type mp_err.  This
-permits the library to communicate success or various kinds of failure
-to the calling program.  The return values currently defined are:
-
-        MP_OKAY         - okay, operation succeeded, all's well
-        MP_YES          - okay, the answer is yes (same as MP_OKAY)
-        MP_NO           - okay, but answer is no (not MP_OKAY)
-        MP_MEM          - operation ran out of memory
-        MP_RANGE        - input parameter was out of range
-        MP_BADARG       - an invalid input parameter was provided
-        MP_UNDEF        - no output value is defined for this input
-
-The only function which currently uses MP_UNDEF is mp_invmod().
-Division by zero is undefined, but the division functions will return
-MP_RANGE for a zero divisor.  MP_BADARG usually means you passed a
-bogus mp_int structure to the function.  MP_YES and MP_NO are not used
-by the library itself; they're defined so you can use them in your own
-extensions.
-
-If you need a readable interpretation of these error codes in your
-program, you may also use the mp_strerror() function.  This function
-takes an mp_err as input, and returns a pointer to a human-readable
-string describing the meaning of the error.  These strings are stored
-as constants within the library, so the caller should not attempt to
-modify or free the memory associated with these strings.
-
-The library represents values in signed-magnitude format.  Values
-strictly less than zero are negative, all others are considered
-positive (zero is positive by fiat).  You can access the 'sign' member
-of the mp_int structure directly, but better is to use the mp_cmp_z()
-function, to find out which side of zero the value lies on.
-
-Most arithmetic functions have a single-digit variant, as well as the
-full arbitrary-precision.  An mp_digit is an unsigned value between 0
-and DIGIT_MAX inclusive.  The radix is available as RADIX.  The number
-of bits in a given digit is given as DIGIT_BIT.
-
-Generally, input parameters are given before output parameters.
-Unless otherwise specified, any input parameter can be re-used as an
-output parameter, without confusing anything.
-
-The basic numeric type defined by the library is an mp_int.  Virtually
-all the functions in the library take a pointer to an mp_int as one of
-their parameters.  An explanation of how to create and use these
-<HR>
-<A NAME="p23">
-<H3>Problem 23:</H3>
-
-structures follows.  And so, without further ado...
-
-
-Initialization and Cleanup
---------------------------
-
-The basic numeric type defined by the library is an 'mp_int'.
-However, it is not sufficient to simply declare a variable of type
-mp_int in your program.  These variables also need to be initialized
-before they can be used, to allocate the internal storage they require
-for computation.
-
-This is done using one of the following functions:
-
-        mp_init(mp_int *mp);
-        mp_init_copy(mp_int *mp, mp_int *from);
-        mp_init_size(mp_int *mp, mp_size p);
-
-Each of these requires a pointer to a structure of type mp_int.  The
-basic mp_init() simply initializes the mp_int to a default size, and
-sets its value to zero.  If you would like to initialize a copy of an
-existing mp_int, use mp_init_copy(), where the 'from' parameter is the
-mp_int you'd like to make a copy of.  The third function,
-mp_init_size(), permits you to specify how many digits of precision
-should be preallocated for your mp_int.  This can help the library
-avoid unnecessary re-allocations later on.
-
-The default precision used by mp_init() can be retrieved using:
-
-        precision = mp_get_prec();
-
-This returns the number of digits that will be allocated.  You can
-change this value by using:
-
-        mp_set_prec(unsigned int prec);
-
-Any positive value is acceptable -- if you pass zero, the default
-precision will be re-set to the compiled-in library default (this is
-specified in the header file 'mpi-config.h', and typically defaults to
-8 or 16).
-
-Just as you must allocate an mp_int before you can use it, you must
-clean up the structure when you are done with it.  This is performed
-using the mp_clear() function.  Remember that any mp_int that you
-create as a local variable in a function must be mp_clear()'d before
-that function exits, or else the memory allocated to that mp_int will
-be orphaned and unrecoverable.
-
-To set an mp_int to a given value, the following functions are given:
-
-        mp_set(mp_int *mp, mp_digit d);
-        mp_set_int(mp_int *mp, long z);
-
-The mp_set() function sets the mp_int to a single digit value, while
-mp_set_int() sets the mp_int to a signed long integer value.
-
-To set an mp_int to zero, use:
-
-        mp_zero(mp_int *mp);
-
-
-Copying and Moving
-------------------
-
-If you have two initialized mp_int's, and you want to copy the value
-of one into the other, use:
-
-        mp_copy(from, to)
-
-This takes care of clearing the old value of 'to', and copies the new
-value into it.  If 'to' is not yet initialized, use mp_init_copy()
-instead (see above).
-
-Note:   The library tries, whenever possible, to avoid allocating
-----    new memory.  Thus, mp_copy() tries first to satisfy the needs
-        of the copy by re-using the memory already allocated to 'to'.
-        Only if this proves insufficient will mp_copy() actually
-        allocate new memory.
-
-        For this reason, if you know a priori that 'to' has enough
-        available space to hold 'from', you don't need to check the
-        return value of mp_copy() for memory failure.  The USED()
-        macro tells you how many digits are used by an mp_int, and
-        the ALLOC() macro tells you how many are allocated.
-
-If you have two initialized mp_int's, and you want to exchange their
-values, use:
-
-        mp_exch(a, b)
-
-This is better than using mp_copy() with a temporary, since it will
-not (ever) touch the memory allocator -- it just swaps the exact
-contents of the two structures.  The mp_exch() function cannot fail;
-if you pass it an invalid structure, it just ignores it, and does
-nothing.
-
-
-Basic Arithmetic
-----------------
-
-Once you have initialized your integers, you can operate on them.  The
-basic arithmetic functions on full mp_int values are:
-
-mp_add(a, b, c)         - computes c = a + b
-mp_sub(a, b, c)         - computes c = a - b
-mp_mul(a, b, c)         - computes c = a * b
-mp_sqr(a, b)            - computes b = a * a
-mp_div(a, b, q, r)      - computes q, r such that a = bq + r
-mp_div_2d(a, d, q, r)   - computes q = a / 2^d, r = a % 2^d
-mp_expt(a, b, c)        - computes c = a ** b
-mp_2expt(a, k)          - computes a = 2^k
-mp_sqrt(a, c)           - computes c = floor(sqrt(a))
-
-The mp_div_2d() function efficiently computes division by powers of
-two.  Either the q or r parameter may be NULL, in which case that
-portion of the computation will be discarded.
-
-The algorithms used for some of the computations here are described in
-the following files which are included with this distribution:
-
-mul.txt         Describes the multiplication algorithm
-div.txt         Describes the division algorithm
-expt.txt        Describes the exponentiation algorithm
-sqrt.txt        Describes the square-root algorithm
-square.txt      Describes the squaring algorithm
-
-There are single-digit versions of most of these routines, as well.
-In the following prototypes, 'd' is a single mp_digit:
-
-mp_add_d(a, d, c)       - computes c = a + d
-mp_sub_d(a, d, c)       - computes c = a - d
-mp_mul_d(a, d, c)       - computes c = a * d
-mp_mul_2(a, c)          - computes c = a * 2
-mp_div_d(a, d, q, r)    - computes q, r such that a = bq + r
-mp_div_2(a, c)          - computes c = a / 2
-mp_expt_d(a, d, c)      - computes c = a ** d
-
-The mp_mul_2() and mp_div_2() functions take advantage of the internal
-representation of an mp_int to do multiplication by two more quickly
-than mp_mul_d() would.  Other basic functions of an arithmetic variety
-include:
-
-mp_zero(a)              - assign 0 to a
-mp_neg(a, c)            - negate a: c = -a
-mp_abs(a, c)            - absolute value: c = |a|
-
-
-Comparisons
------------
-
-Several comparison functions are provided.  Each of these, unless
-otherwise specified, returns zero if the comparands are equal, < 0 if
-the first is less than the second, and > 0 if the first is greater
-than the second:
-
-mp_cmp_z(a)             - compare a <=> 0
-mp_cmp_d(a, d)          - compare a <=> d, d is a single digit
-mp_cmp(a, b)            - compare a <=> b
-mp_cmp_mag(a, b)        - compare |a| <=> |b|
-mp_cmp_int(a, z)        - compare a <=> z, z is a signed long integer
-mp_isodd(a)             - return nonzero if odd, zero otherwise
-mp_iseven(a)            - return nonzero if even, zero otherwise
-
-
-Modular Arithmetic
-------------------
-
-Modular variations of the basic arithmetic functions are also
-supported.  These are available if the MP_MODARITH parameter in
-mpi-config.h is turned on (it is by default).  The modular arithmetic
-functions are:
-
-mp_mod(a, m, c)         - compute c = a (mod m), 0 <= c < m
-mp_mod_d(a, d, c)       - compute c = a (mod d), 0 <= c < d (see below)
-mp_addmod(a, b, m, c)   - compute c = (a + b) mod m
-mp_submod(a, b, m, c)   - compute c = (a - b) mod m
-mp_mulmod(a, b, m, c)   - compute c = (a * b) mod m
-mp_sqrmod(a, m, c)      - compute c = (a * a) mod m
-mp_exptmod(a, b, m, c)  - compute c = (a ** b) mod m
-mp_exptmod_d(a, d, m, c)- compute c = (a ** d) mod m
-
-The mp_sqr() function squares its input argument.  A call to mp_sqr(a,
-c) is identical in meaning to mp_mul(a, a, c); however, if the
-MP_SQUARE variable is set true in mpi-config.h (see below), then it
-will be implemented with a different algorithm, that is supposed to
-take advantage of the redundant computation that takes place during
-squaring.  Unfortunately, some compilers result in worse performance
-on this code, so you can change the behaviour at will.  There is a
-utility program "mulsqr.c" that lets you test which does better on
-your system.
-
-The mp_sqrmod() function is analogous to the mp_sqr() function; it
-uses the mp_sqr() function rather than mp_mul(), and then performs the
-modular reduction.  This probably won't help much unless you are doing
-a lot of them.
-
-See the file 'square.txt' for a synopsis of the algorithm used.
-
-Note:   The mp_mod_d() function computes a modular reduction around
-----    a single digit d.  The result is a single digit c.
-
-Because an inverse is defined for a (mod m) if and only if (a, m) = 1
-(that is, if a and m are relatively prime), mp_invmod() may not be
-able to compute an inverse for the arguments.  In this case, it
-returns the value MP_UNDEF, and does not modify c.  If an inverse is
-defined, however, it returns MP_OKAY, and sets c to the value of the
-inverse (mod m).
-
-See the file 'redux.txt' for a description of the modular reduction
-algorithm used by mp_exptmod().
-
-
-Greatest Common Divisor
------------------------
-
-If The greates common divisor of two values can be found using one of the
-following functions:
-
-mp_gcd(a, b, c)         - compute c = (a, b) using binary algorithm
-mp_lcm(a, b, c)         - compute c = [a, b] = ab / (a, b)
-mp_xgcd(a, b, g, x, y)  - compute g, x, y so that ax + by = g = (a, b)
-
-Also provided is a function to compute modular inverses, if they
-exist:
-
-mp_invmod(a, m, c)      - compute c = a^-1 (mod m), if it exists
-
-The function mp_xgcd() computes the greatest common divisor, and also
-returns values of x and y satisfying Bezout's identity.  This is used
-by mp_invmod() to find modular inverses.  However, if you do not need
-these values, you will find that mp_gcd() is MUCH more efficient,
-since it doesn't need all the intermediate values that mp_xgcd()
-requires in order to compute x and y. 
-
-The mp_gcd() (and mp_xgcd()) functions use the binary (extended) GCD
-algorithm due to Josef Stein.
-
-
-Input & Output Functions
-------------------------
-
-The following basic I/O routines are provided.  These are present at
-all times:
-
-mp_read_radix(mp, str, r)  - convert a string in radix r to an mp_int
-mp_read_raw(mp, s, len)    - convert a string of bytes to an mp_int
-mp_radix_size(mp, r)       - return length of buffer needed by mp_toradix()
-mp_raw_size(mp)            - return length of buffer needed by mp_toraw()
-mp_toradix(mp, str, r)     - convert an mp_int to a string of radix r 
-                             digits
-mp_toraw(mp, str)          - convert an mp_int to a string of bytes
-mp_tovalue(ch, r)          - convert ch to its value when taken as
-                             a radix r digit, or -1 if invalid
-mp_strerror(err)           - get a string describing mp_err value 'err'
-
-If you compile the MPI library with MP_IOFUNC defined, you will also
-have access to the following additional I/O function:
-
-mp_print(mp, ofp)       - print an mp_int as text to output stream ofp
-
-Note that mp_radix_size() returns a size in bytes guaranteed to be AT
-LEAST big enough for the digits output by mp_toradix().  Because it
-uses an approximation technique to figure out how many digits will be
-needed, it may return a figure which is larger than necessary.  Thus,
-the caller should not rely on the value to determine how many bytes
-will actually be written by mp_toradix().  The string mp_toradix()
-creates will be NUL terminated, so the standard C library function
-strlen() should be able to ascertain this for you, if you need it.
-
-The mp_read_radix() and mp_toradix() functions support bases from 2 to
-64 inclusive.  If you require more general radix conversion facilities
-than this, you will need to write them yourself (that's why mp_div_d()
-is provided, after all).
-
-Note:   mp_read_radix() will accept as digits either capital or 
-----    lower-case letters.  However, the current implementation of
-        mp_toradix() only outputs upper-case letters, when writing
-        bases betwee 10 and 36.  The underlying code supports using
-        lower-case letters, but the interface stub does not have a
-        selector for it.  You can add one yourself if you think it
-        is worthwhile -- I do not.  Bases from 36 to 64 use lower-
-        case letters as distinct from upper-case.  Bases 63 and
-        64 use the characters '+' and '/' as digits.
-
-        Note also that compiling with MP_IOFUNC defined will cause
-        inclusion of <stdio.h>, so if you are trying to write code
-        which does not depend on the standard C library, you will
-        probably want to avoid this option.  This is needed because
-        the mp_print() function takes a standard library FILE * as
-        one of its parameters, and uses the fprintf() function.
-
-The mp_toraw() function converts the integer to a sequence of bytes,
-in big-endian ordering (most-significant byte first).  Assuming your
-bytes are 8 bits wide, this corresponds to base 256.  The sign is
-encoded as a single leading byte, whose value is 0 for zero or
-positive values, or 1 for negative values.  The mp_read_raw() function
-reverses this process -- it takes a buffer of bytes, interprets the
-first as a sign indicator (0 = zero/positive, nonzero = negative), and
-the rest as a sequence of 1-byte digits in big-endian ordering.
-
-The mp_raw_size() function returns the exact number of bytes required
-to store the given integer in "raw" format (as described in the
-previous paragraph).  Zero is returned in case of error; a valid
-integer will require at least three bytes of storage.
-
-In previous versions of the MPI library, an "external representation
-format" was supported.  This was removed, however, because I found I
-was never using it, it was not as portable as I would have liked, and
-I decided it was a waste of space.
-
-
-Other Functions
----------------
-
-The files 'mpprime.h' and 'mpprime.c' define some routines which are
-useful for divisibility testing and probabilistic primality testing.
-The routines defined are:
-
-mpp_divis(a, b)          - is a divisible by b?
-mpp_divis_d(a, d)        - is a divisible by digit d?
-mpp_random(a)            - set a to random value at current precision
-mpp_random_size(a, prec) - set a to random value at given precision
-
-Note:  The mpp_random() and mpp_random_size() functions use the C
-----   library's rand() function to generate random values.  It is
-       up to the caller to seed this generator before it is called.
-       These functions are not suitable for generating quantities
-       requiring cryptographic-quality randomness; they are intended
-       primarily for use in primality testing.
-
-       Note too that the MPI library does not call srand(), so your
-       application should do this, if you ever want the sequence
-       to change.
-
-mpp_divis_vector(a, v, s, w)  - is a divisible by any of the s digits
-                                in v?  If so, let w be the index of 
-                                that digit
-
-mpp_divis_primes(a, np)       - is a divisible by any of the first np
-                                primes?  If so, set np to the prime 
-                                which divided a.
-
-mpp_fermat(a, d)              - test if w^a = w (mod a).  If so, 
-                                returns MP_YES, otherwise MP_NO.
-
-mpp_pprime(a, nt)             - perform nt iterations of the Rabin-
-                                Miller probabilistic primality test
-                                on a.  Returns MP_YES if all tests
-                                passed, or MP_NO if any test fails.
-
-The mpp_fermat() function works based on Fermat's little theorem, a
-consequence of which is that if p is a prime, and (w, p) = 1, then:
-
-        w^p = w (mod p)
-
-Put another way, if w^p != w (mod p), then p is not prime.  The test
-is expensive to compute, but it helps to quickly eliminate an enormous
-class of composite numbers prior to Rabin-Miller testing.
-
-Building the Library
---------------------
-
-The MPI library is designed to be as self-contained as possible.  You
-should be able to compile it with your favourite ANSI C compiler, and
-link it into your program directly.  If you are on a Unix system using
-the GNU C compiler (gcc), the following should work:
-
-% gcc -ansi -pedantic -Wall -O2 -c mpi.c
-
-The file 'mpi-config.h' defines several configurable parameters for
-the library, which you can adjust to suit your application.  At the
-time of this writing, the available options are:
-
-MP_IOFUNC       - Define true to include the mp_print() function, 
-                  which is moderately useful for debugging.  This
-                  implicitly includes <stdio.h>.
-
-MP_MODARITH     - Define true to include the modular arithmetic
-                  functions.  If you don't need modular arithmetic
-                  in your application, you can set this to zero to
-                  leave out all the modular routines.
-
-MP_NUMTH        - Define true to include number theoretic functions
-                  such as mp_gcd(), mp_lcm(), and mp_invmod().
-
-MP_LOGTAB       - If true, the file "logtab.h" is included, which
-                  is basically a static table of base 2 logarithms.
-                  These are used to compute how big the buffers for
-                  radix conversion need to be.  If you set this false,
-                  the library includes <math.h> and uses log().  This
-                  typically forces you to link against math libraries.
-
-MP_MEMSET       - If true, use memset() to zero buffers.  If you run
-                  into weird alignment related bugs, set this to zero
-                  and an explicit loop will be used.
-
-MP_MEMCPY       - If true, use memcpy() to copy buffers.  If you run
-                  into weird alignment bugs, set this to zero and an
-                  explicit loop will be used.
-
-MP_CRYPTO       - If true, whenever arrays of digits are free'd, they
-                  are zeroed first.  This is useful if you're using
-                  the library in a cryptographic environment; however,
-                  it does add overhead to each free operation.  For
-                  performance, if you don't care about zeroing your
-                  buffers, set this to false.
-
-MP_ARGCHK       - Set to 0, 1, or 2.  This defines how the argument
-                  checking macro, ARGCHK(), gets expanded.  If this 
-                  is set to zero, ARGCHK() expands to nothing; no 
-                  argument checks are performed.  If this is 1, the
-                  ARGCHK() macro expands to code that returns MP_BADARG
-                  or similar at runtime.  If it is 2, ARGCHK() expands 
-                  to an assert() call that aborts the program on a 
-                  bad input.
-
-MP_DEBUG        - Turns on debugging output.  This is probably not at
-                  all useful unless you are debugging the library.  It
-                  tends to spit out a LOT of output.
-
-MP_DEFPREC      - The default precision of a newly-created mp_int, in
-                  digits.  The precision can be changed at runtime by
-                  the mp_set_prec() function, but this is its initial
-                  value.
-
-MP_SQUARE       - If this is set to a nonzero value, the mp_sqr() 
-                  function will use an alternate algorithm that takes
-                  advantage of the redundant inner product computation
-                  when both multiplicands are identical.  Unfortunately,
-                  with some compilers this is actually SLOWER than just
-                  calling mp_mul() with the same argument twice.  So
-                  if you set MP_SQUARE to zero, mp_sqr() will be expan-
-                  ded into a call to mp_mul().  This applies to all 
-                  the uses of mp_sqr(), including mp_sqrmod() and the
-                  internal calls to s_mp_sqr() inside mpi.c
-
-                  The program 'mulsqr' (mulsqr.c) can be used to test
-                  which works best for your configuration.  Set up the
-                  CC and CFLAGS variables in the Makefile, then type:
-
-                        make mulsqr
-
-                  Invoke it with arguments similar to the following:
-
-                        mulsqr 25000 1024
-
-                  That is, 25000 products computed on 1024-bit values.
-                  The output will compare the two timings, and recommend
-                  a setting for MP_SQUARE.  It is off by default.
-
-If you would like to use the mp_print() function (see above), be sure
-to define MP_IOFUNC in mpi-config.h.  Many of the test drivers in the
-'tests' subdirectory expect this to be defined (although the test
-driver 'mpi-test' doesn't need it)
-
-The Makefile which comes with the library should take care of building
-the library for you, if you have set the CC and CFLAGS variables at
-the top of the file appropriately.  By default, they are set up to
-use the GNU C compiler:
-
-CC=gcc
-CFLAGS=-ansi -pedantic -Wall -O2
-
-If all goes well, the library should compile without warnings using
-this combination.  You should, of course, make whatever adjustments
-you find necessary.  
-
-The MPI library distribution comes with several additional programs
-which are intended to demonstrate the use of the library, and provide
-a framework for testing it.  There are a handful of test driver
-programs, in the files named 'mptest-X.c', where X is a digit.  Also,
-there are some simple command-line utilities (in the 'utils'
-directory) for manipulating large numbers.  These include:
-
-basecvt.c       A radix-conversion program, supporting bases from
-                2 to 64 inclusive.
-
-bbsrand.c       A BBS (quadratic residue) pseudo-random number 
-                generator.  The file 'bbsrand.c' is just the driver
-                for the program; the real code lives in the files
-                'bbs_rand.h' and 'bbs_rand.c'
-
-dec2hex.c       Converts decimal to hexadecimal
-
-gcd.c           Computes the greatest common divisor of two values.
-                If invoked as 'xgcd', also computes constants x and
-                y such that (a, b) = ax + by, in accordance with
-                Bezout's identity.
-
-hex2dec.c       Converts hexadecimal to decimal
-
-invmod.c        Computes modular inverses
-
-isprime.c       Performs the Rabin-Miller probabilistic primality
-                test on a number.  Values which fail this test are
-                definitely composite, and those which pass are very
-                likely to be prime (although there are no guarantees)
-
-lap.c           Computes the order (least annihilating power) of
-                a value v modulo m.  Very dumb algorithm.
-
-primegen.c      Generates large (probable) primes.
-
-prng.c          A pseudo-random number generator based on the
-                BBS generator code in 'bbs_rand.c'
-
-sieve.c         Implements the Sieve of Eratosthenes, using a big
-                bitmap, to generate a list of prime numbers.
-
-fact.c          Computes the factorial of an arbitrary precision
-                integer (iterative).
-
-exptmod.c       Computes arbitrary precision modular exponentiation
-                from the command line (exptmod a b m -> a^b (mod m))
-
-Most of these can be built from the Makefile that comes with the
-library.  Try 'make tools', if your environment supports it.  (If you
-are compiling on a Macintosh, I'm afraid you'll have to build them by
-hand -- fortunately, this is not difficult -- the library itself
-should compile just fine under Metrowerks CodeWarrior).
-
-
-Testing the Library
--------------------
-
-Automatic test vectors are included, in the form of a program called
-'mpi-test'.  To build this program and run all the tests, simply
-invoke the shell script 'all-tests'.  If all the tests pass, you
-should see a message:
-
-        All tests passed
-
-If something went wrong, you'll get:
-
-        One or more tests failed.
-
-If this happens, scan back through the preceding lines, to see which
-test failed.  Any failure indicates a bug in the library, which needs
-to be fixed before it will give accurate results.  If you get any such
-thing, please let me know, and I'll try to fix it.  Please let me know
-what platform and compiler you were using, as well as which test
-failed.  If a reason for failure was given, please send me that text
-as well.
-
-If you're on a system such as the Macintosh, where the standard Unix
-build tools don't work, you can build the 'mpi-test' program manually,
-and run it by hand.  This is tedious and obnoxious, sorry.
-
-Further manual testing can be performed by building the manual testing
-programs, whose source is found in the 'tests' subdirectory.  Each
-test is in a source file called 'mptest-X.c'.  The Makefile contains a
-target to build all of them at once:
-
-        make tests
-
-Read the comments at the top of each source file to see what the
-driver is supposed to test.  You probably don't need to do this; these
-programs were only written to help me as I was developing the library.
-
-The relevant files are:
-
-mpi-test.c              The source for the test driver
-
-make-test-arrays        A Perl script to generate some of the internal
-                        data structures used by mpi-test.c
-
-test-arrays.txt         The source file for make-test-arrays
-
-all-tests               A Bourne shell script which runs all the
-                        tests in the mpi-test suite
-
-Running 'make mpi-test' should build the mpi-test program.  If you
-cannot use make, here is what needs to be done:
-
-(1) Use 'make-test-arrays' to generate the file 'test-info.c' from
-    the 'test-arrays.txt' file.  Since Perl can be found everywhere,
-    even on the Macintosh, this should be no trouble.  Under Unix, 
-    this looks like:
-
-        make-test-arrays test-arrays.txt > test-info.c
-
-(2) Build the MPI library:
-
-        gcc -ansi -pedantic -Wall -c mpi.c
-
-(3) Build the mpi-test program:
-
-        gcc -ansi -pedantic -Wall -o mpi-test mpi.o mpi-test.c
-
-When you've got mpi-test, you can use 'all-tests' to run all the tests
-made available by mpi-test.  If any of them fail, there should be a
-diagnostic indicating what went wrong.  These are fairly high-level
-diagnostics, and won't really help you debug the problem; they're
-simply intended to help you isolate which function caused the problem.
-If you encounter a problem of this sort, feel free to e-mail me, and I
-will certainly attempt to help you debug it.
-
-Note:   Several of the tests hard-wired into 'mpi-test' operate under
-----    the assumption that you are using at least a 16-bit mp_digit 
-        type.  If that is not true, several tests might fail, because 
-        of range problems with the maximum digit value.
-
-        If you are using an 8-bit digit, you will also need to 
-        modify the code for mp_read_raw(), which assumes that
-        multiplication by 256 can be done with mp_mul_d(), a
-        fact that fails when DIGIT_MAX is 255.  You can replace
-        the call with s_mp_lshd(), which will give you the same
-        effect, and without doing as much work. :)
-
-Acknowledgements:
-----------------
-
-The algorithms used in this library were drawn primarily from Volume
-2 of Donald Knuth's magnum opus, _The Art of Computer Programming_, 
-"Semi-Numerical Methods".  Barrett's algorithm for modular reduction
-came from Menezes, Oorschot, and Vanstone's _Handbook of Applied
-Cryptography_, Chapter 14.
-
-Thanks are due to Tom St. Denis, for finding an obnoxious sign-related
-bug in mp_read_raw() that made things break on platforms which use
-signed chars.
-
-About the Author
-----------------
-
-This software was written by Michael J. Fromberger.  You can contact
-the author as follows:
-
-E-mail:   <sting@linguist.dartmouth.edu>
-
-Postal:   8000 Cummings Hall, Thayer School of Engineering
-          Dartmouth College, Hanover, New Hampshire, USA
-
-PGP key:  http://linguist.dartmouth.edu/~sting/keys/mjf.html
-          9736 188B 5AFA 23D6 D6AA  BE0D 5856 4525 289D 9907
-
-Last updated:  16-Jan-2000

mozilla/security/nss/lib/freebl/mpi/all-tests

@@ -1,115 +0,0 @@
-#!/bin/sh
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>.
-# Portions created by the Initial Developer are Copyright (C) 1997
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-ECHO=/bin/echo
-MAKE=gmake
-
-$ECHO "\n** Running unit tests for MPI library\n"
-
-# Build the mpi-test program, which comprises all the unit tests for
-# the MPI library...
-
-$ECHO "Bringing mpi-test up to date ... "
-if $MAKE mpi-test ; then
-  :
-else
-  $ECHO " "
-  $ECHO "Make failed to build mpi-test."
-  $ECHO " "
-  exit 1
-fi
-
-if [ ! -x mpi-test ] ; then
-  $ECHO " "
-  $ECHO "Cannot find 'mpi-test' program, testing cannot continue."
-  $ECHO " "
-  exit 1
-fi
-
-# Get the list of available test suites...
-tests=`mpi-test list | awk '{print $1}'`
-errs=0
-
-# Run each test suite and check the result code of mpi-test
-for test in $tests ; do
-  $ECHO "$test ... \c"
-  if mpi-test $test ; then
-    $ECHO "passed"
-  else
-    $ECHO "FAILED"
-    errs=1
-  fi
-done
-
-# If any tests failed, we'll stop at this point
-if [ "$errs" = "0" ] ; then
-  $ECHO "All unit tests passed"
-else
-  $ECHO "One or more tests failed"
-  exit 1
-fi
-
-# Now try to build the 'pi' program, and see if it can compute the
-# first thousand digits of pi correctly
-$ECHO "\n** Running other tests\n"
-
-$ECHO "Bringing 'pi' up to date ... "
-if $MAKE pi ; then
-    :
-else
-    $ECHO "\nMake failed to build pi.\n"
-    exit 1
-fi
-
-if [ ! -x pi ] ; then
-    $ECHO "\nCannot find 'pi' program; testing cannot continue.\n"
-    exit 1
-fi
-
-./pi 2000 > /tmp/pi.tmp.$$
-if cmp tests/pi2k.txt /tmp/pi.tmp.$$ ; then
-    $ECHO "Okay!  The pi test passes."
-else
-    $ECHO "Oops!  The pi test failed. :("
-    exit 1
-fi
-
-rm -f /tmp/pi.tmp.$$
-
-exit 0
-
-# Here there be dragons

mozilla/security/nss/lib/freebl/mpi/doc/LICENSE

@@ -1,11 +0,0 @@
-Within this directory, each of the file listed below is licensed under 
-the terms given in the file LICENSE-MPL, also in this directory.
-
-basecvt.pod
-gcd.pod
-invmod.pod
-isprime.pod
-lap.pod
-mpi-test.pod
-prime.txt
-prng.pod

mozilla/security/nss/lib/freebl/mpi/doc/LICENSE-MPL

@@ -1,32 +0,0 @@
-The contents of this file are subject to the Mozilla Public
-License Version 1.1 (the "License"); you may not use this file
-except in compliance with the License. You may obtain a copy of
-the License at http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS
-IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
-implied. See the License for the specific language governing
-rights and limitations under the License.
-
-The Original Code is the Netscape security libraries.
-
-The Initial Developer of the Original Code is Netscape
-Communications Corporation.  Portions created by Netscape are 
-Copyright (C) 1994-2000 Netscape Communications Corporation.  All
-Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the
-terms of the GNU General Public License Version 2 or later (the
-"GPL"), in which case the provisions of the GPL are applicable 
-instead of those above.  If you wish to allow use of your 
-version of this file only under the terms of the GPL and not to
-allow others to use your version of this file under the MPL,
-indicate your decision by deleting the provisions above and
-replace them with the notice and other provisions required by
-the GPL.  If you do not delete the provisions above, a recipient
-may use your version of this file under either the MPL or the
-GPL.
-
-

mozilla/security/nss/lib/freebl/mpi/doc/basecvt.pod

@@ -1,63 +0,0 @@
-=head1 NAME
-
- basecvt - radix conversion for arbitrary precision integers
-
-=head1 SYNOPSIS
-
- basecvt <ibase> <obase> [values]
-
-=head1 DESCRIPTION
-
-The B<basecvt> program is a command-line tool for converting integers
-of arbitrary precision from one radix to another.  The current version
-supports radix values from 2 (binary) to 64, inclusive.  The first two
-command line arguments specify the input and output radix, in base 10.
-Any further arguments are taken to be integers notated in the input
-radix, and these are converted to the output radix.  The output is
-written, one integer per line, to standard output.
-
-When reading integers, only digits considered "valid" for the input
-radix are considered.  Processing of an integer terminates when an
-invalid input digit is encountered.  So, for example, if you set the
-input radix to 10 and enter '10ACF', B<basecvt> would assume that you
-had entered '10' and ignore the rest of the string.
-
-If no values are provided, no output is written, but the program
-simply terminates with a zero exit status.  Error diagnostics are
-written to standard error in the event of out-of-range radix
-specifications.  Regardless of the actual values of the input and
-output radix, the radix arguments are taken to be in base 10 (decimal)
-notation.
-
-=head1 DIGITS
-
-For radices from 2-10, standard ASCII decimal digits 0-9 are used for
-both input and output.  For radices from 11-36, the ASCII letters A-Z
-are also included, following the convention used in hexadecimal.  In
-this range, input is accepted in either upper or lower case, although
-on output only lower-case letters are used.
-
-For radices from 37-62, the output includes both upper- and lower-case
-ASCII letters, and case matters.  In this range, case is distinguished
-both for input and for output values.
-
-For radices 63 and 64, the characters '+' (plus) and '/' (forward
-solidus) are also used.  These are derived from the MIME base64
-encoding scheme.  The overall encoding is not the same as base64,
-because the ASCII digits are used for the bottom of the range, and the
-letters are shifted upward; however, the output will consist of the
-same character set.
-
-This input and output behaviour is inherited from the MPI library used
-by B<basecvt>, and so is not configurable at runtime.
-
-=head1 SEE ALSO
-
- dec2hex(1), hex2dec(1)
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Thayer School of Engineering, Hanover, New Hampshire, USA
-
- $Date: 2000-07-14 00:44:31 $

mozilla/security/nss/lib/freebl/mpi/doc/build

@@ -1,66 +0,0 @@
-#!/bin/sh
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>.
-# Portions created by the Initial Developer are Copyright (C) 1998
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#   Netscape Communications Corporation
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-# $Id: build,v 1.3 2005-02-02 22:28:22 gerv%gerv.net Exp $
-#
-
-VERS="1.7p6"
-SECT="1"
-NAME="MPI Tools"
-
-echo "Building manual pages ..."
-case $# in
-  0)
-    files=`ls *.pod`
-    ;;
-  *)
-    files=$*
-    ;;
-esac
-
-for name in $files
-do
-   echo -n "$name ... "
-#  sname=`noext $name`
-   sname=`basename $name .pod`
-   pod2man --section="$SECT" --center="$NAME" --release="$VERS" $name > $sname.$SECT
-   echo "(done)"
-done
-
-echo "Finished building."
-

mozilla/security/nss/lib/freebl/mpi/doc/div.txt

@@ -1,101 +0,0 @@
-Division
-
-This describes the division algorithm used by the MPI library.
-
-Input:    a, b; a > b
-Compute:  Q, R; a = Qb + R
-
-The input numbers are normalized so that the high-order digit of b is
-at least half the radix.  This guarantees that we have a reasonable
-way to guess at the digits of the quotient (this method was taken from
-Knuth, vol. 2, with adaptations).
-
-To normalize, test the high-order digit of b.  If it is less than half
-the radix, multiply both a and b by d, where:
-
-             radix - 1
-	d = -----------
-              bmax + 1
-
-...where bmax is the high-order digit of b.  Otherwise, set d = 1.
-
-Given normalize values for a and b, let the notation a[n] denote the
-nth digit of a.  Let #a be the number of significant figures of a (not
-including any leading zeroes).
-
-	Let R = 0
-	Let p = #a - 1
-
-	while(p >= 0)
-	  do
-	    R = (R * radix) + a[p]
-	    p = p - 1
-	  while(R < b and p >= 0)
-
-	  if(R < b)
-	    break
-
-	  q = (R[#R - 1] * radix) + R[#R - 2]
-	  q = q / b[#b - 1]
-
-	  T = b * q
-
-	  while(T > L)
-	    q = q - 1
-	    T = T - b
-	  endwhile
-
-	  L = L - T
-
-	  Q = (Q * radix) + q
-
-	endwhile
-
-At this point, Q is the quotient, and R is the normalized remainder.
-To denormalize R, compute:
-
-	R = (R / d)
-
-At this point, you are finished.
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: div.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-

mozilla/security/nss/lib/freebl/mpi/doc/expt.txt

@@ -1,132 +0,0 @@
-Exponentiation
-
-For exponentiation, the MPI library uses a simple and fairly standard
-square-and-multiply method.  The algorithm is this:
-
-Input:	a, b
-Output: a ** b
-
-	s = 1
-
-	while(b != 0)
-	  if(b is odd)
-	    s = s * a
-	  endif
-
-	  b = b / 2
-
-	  x = x * x
-	endwhile
-
-	return s
-
-The modular exponentiation is done the same way, except replacing:
-
-	s = s * a
-
-with
-	s = (s * a) mod m
-
-and replacing
-
-	x = x * x
-
-with
-
-	x = (x * x) mod m
-
-Here is a sample exponentiation using the MPI library, as compared to
-the same problem solved by the Unix 'bc' program on my system:
-
-Computation of 2,381,283 ** 235
-
-'bc' says:
-
-4385CA4A804D199FBEAD95FAD0796FAD0D0B51FC9C16743C45568C789666985DB719\
-4D90E393522F74C9601262C0514145A49F3B53D00983F95FDFCEA3D0043ECEF6227E\
-6FB59C924C3EE74447B359B5BF12A555D46CB819809EF423F004B55C587D6F0E8A55\
-4988036A42ACEF9F71459F97CEF6E574BD7373657111648626B1FF8EE15F663B2C0E\
-6BBE5082D4CDE8E14F263635AE8F35DB2C280819517BE388B5573B84C5A19C871685\
-FD408A6471F9D6AFAF5129A7548EAE926B40874B340285F44765BF5468CE20A13267\
-CD88CE6BC786ACED36EC7EA50F67FF27622575319068A332C3C0CB23E26FB55E26F4\
-5F732753A52B8E2FB4D4F42D894242613CA912A25486C3DEC9C66E5DB6182F6C1761\
-CF8CD0D255BE64B93836B27D452AE38F950EB98B517D4CF50D48F0165EF0CCCE1F5C\
-49BF18219FDBA0EEDD1A7E8B187B70C2BAED5EC5C6821EF27FAFB1CFF70111C52235\
-5E948B93A015AA1AE152B110BB5658CB14D3E45A48BFE7F082C1182672A455A695CD\
-A1855E8781E625F25B41B516E77F589FA420C3B058861EA138CF7A2C58DB3C7504FD\
-D29554D78237834CC5AE710D403CC4F6973D5012B7E117A8976B14A0B5AFA889BD47\
-92C461F0F96116F00A97AE9E83DC5203680CAF9A18A062566C145650AB86BE4F907F\
-A9F7AB4A700B29E1E5BACCD6DCBFA513E10832815F710807EED2E279081FEC61D619\
-AB270BEB3D3A1787B35A9DD41A8766CF21F3B5C693B3BAB1C2FA14A4ED202BC35743\
-E5CBE2391624D4F8C9BFBBC78D69764E7C6C5B11BF005677BFAD17D9278FFC1F158F\
-1B3683FF7960FA0608103792C4163DC0AF3E06287BB8624F8FE3A0FFBDF82ACECA2F\
-CFFF2E1AC93F3CA264A1B
-
-MPI says:
-
-4385CA4A804D199FBEAD95FAD0796FAD0D0B51FC9C16743C45568C789666985DB719\
-4D90E393522F74C9601262C0514145A49F3B53D00983F95FDFCEA3D0043ECEF6227E\
-6FB59C924C3EE74447B359B5BF12A555D46CB819809EF423F004B55C587D6F0E8A55\
-4988036A42ACEF9F71459F97CEF6E574BD7373657111648626B1FF8EE15F663B2C0E\
-6BBE5082D4CDE8E14F263635AE8F35DB2C280819517BE388B5573B84C5A19C871685\
-FD408A6471F9D6AFAF5129A7548EAE926B40874B340285F44765BF5468CE20A13267\
-CD88CE6BC786ACED36EC7EA50F67FF27622575319068A332C3C0CB23E26FB55E26F4\
-5F732753A52B8E2FB4D4F42D894242613CA912A25486C3DEC9C66E5DB6182F6C1761\
-CF8CD0D255BE64B93836B27D452AE38F950EB98B517D4CF50D48F0165EF0CCCE1F5C\
-49BF18219FDBA0EEDD1A7E8B187B70C2BAED5EC5C6821EF27FAFB1CFF70111C52235\
-5E948B93A015AA1AE152B110BB5658CB14D3E45A48BFE7F082C1182672A455A695CD\
-A1855E8781E625F25B41B516E77F589FA420C3B058861EA138CF7A2C58DB3C7504FD\
-D29554D78237834CC5AE710D403CC4F6973D5012B7E117A8976B14A0B5AFA889BD47\
-92C461F0F96116F00A97AE9E83DC5203680CAF9A18A062566C145650AB86BE4F907F\
-A9F7AB4A700B29E1E5BACCD6DCBFA513E10832815F710807EED2E279081FEC61D619\
-AB270BEB3D3A1787B35A9DD41A8766CF21F3B5C693B3BAB1C2FA14A4ED202BC35743\
-E5CBE2391624D4F8C9BFBBC78D69764E7C6C5B11BF005677BFAD17D9278FFC1F158F\
-1B3683FF7960FA0608103792C4163DC0AF3E06287BB8624F8FE3A0FFBDF82ACECA2F\
-CFFF2E1AC93F3CA264A1B
-
-Diff says:
-% diff bc.txt mp.txt
-%
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: expt.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-
-

mozilla/security/nss/lib/freebl/mpi/doc/gcd.pod

@@ -1,27 +0,0 @@
-=head1 NAME
-
- gcd - compute greatest common divisor of two integers
-
-=head1 SYNOPSIS
-
- gcd <a> <b>
-
-=head1 DESCRIPTION
-
-The B<gcd> program computes the greatest common divisor of two
-arbitrary-precision integers I<a> and I<b>.  The result is written in
-standard decimal notation to the standard output.
-
-If I<b> is zero, B<gcd> will print an error message and exit.
-
-=head1 SEE ALSO
-
-invmod(1), isprime(1), lap(1)
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Thayer School of Engineering, Hanover, New Hampshire, USA
- 
- $Date: 2000-07-14 00:44:32 $
-

mozilla/security/nss/lib/freebl/mpi/doc/invmod.pod

@@ -1,33 +0,0 @@
-=head1 NAME
-
- invmod - compute modular inverse of an integer
-
-=head1 SYNOPSIS
-
- invmod <a> <m>
-
-=head1 DESCRIPTION
-
-The B<invmod> program computes the inverse of I<a>, modulo I<m>, if
-that inverse exists.  Both I<a> and I<m> are arbitrary-precision
-integers in decimal notation.  The result is written in standard
-decimal notation to the standard output.
-
-If there is no inverse, the message:
-
- No inverse
-
-...will be printed to the standard output (an inverse exists if and
-only if the greatest common divisor of I<a> and I<m> is 1).
-
-=head1 SEE ALSO
-
-gcd(1), isprime(1), lap(1)
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Thayer School of Engineering, Hanover, New Hampshire, USA
- 
- $Date: 2000-07-14 00:44:33 $
-

mozilla/security/nss/lib/freebl/mpi/doc/isprime.pod

@@ -1,62 +0,0 @@
-=head1 NAME
-
- isprime - probabilistic primality testing
-
-=head1 SYNOPSIS
-
- isprime <a>
-
-=head1 DESCRIPTION
-
-The B<isprime> program attempts to determine whether the arbitrary
-precision integer I<a> is prime.  It first tests I<a> for divisibility
-by the first 170 or so small primes, and assuming I<a> is not
-divisible by any of these, applies 15 iterations of the Rabin-Miller
-probabilistic primality test.
-
-If the program discovers that the number is composite, it will print:
-
- Not prime (reason)
-
-Where I<reason> is either:
-
-	divisible by small prime x
-
-Or:
-
-	failed nth pseudoprime test
-
-In the first case, I<x> indicates the first small prime factor that
-was found.  In the second case, I<n> indicates which of the
-pseudoprime tests failed (numbered from 1)
-
-If this happens, the number is definitely not prime.  However, if the
-number succeeds, this message results:
-
- Probably prime, 1 in 4^15 chance of false positive
-
-If this happens, the number is prime with very high probability, but
-its primality has not been absolutely proven, only demonstrated to a
-very convincing degree.
-
-The value I<a> can be input in standard decimal notation, or, if it is
-prefixed with I<Ox>, it will be read as hexadecimal.
-
-=head1 ENVIRONMENT
-
-You can control how many iterations of Rabin-Miller are performed on
-the candidate number by setting the I<RM_TESTS> environment variable
-to an integer value before starting up B<isprime>.  This will change
-the output slightly if the number passes all the tests.
-
-=head1 SEE ALSO
-
-gcd(1), invmod(1), lap(1)
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Thayer School of Engineering, Hanover, New Hampshire, USA
- 
- $Date: 2000-07-14 00:44:33 $
-

mozilla/security/nss/lib/freebl/mpi/doc/lap.pod

@@ -1,35 +0,0 @@
-=head1 NAME
-
- lap - compute least annihilating power of a number
-
-=head1 SYNOPSIS
-
- lap <a> <m>
-
-=head1 DESCRIPTION
-
-The B<lap> program computes the order of I<a> modulo I<m>, for
-arbitrary precision integers I<a> and I<m>.  The B<order> of I<a>
-modulo I<m> is defined as the smallest positive value I<n> for which
-I<a> raised to the I<n>th power, modulo I<m>, is equal to 1.  The
-order may not exist, if I<m> is composite.
-
-=head1 RESTRICTIONS
-
-This program is very slow, especially for large moduli.  It is
-intended as a way to help find primitive elements in a modular field,
-but it does not do so in a particularly inefficient manner.  It was
-written simply to help verify that a particular candidate does not
-have an obviously short cycle mod I<m>.
-
-=head1 SEE ALSO
-
-gcd(1), invmod(1), isprime(1)
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Thayer School of Engineering, Hanover, New Hampshire, USA
- 
- $Date: 2000-07-14 00:44:34 $
-

mozilla/security/nss/lib/freebl/mpi/doc/mpi-test.pod

@@ -1,49 +0,0 @@
-=head1 NAME
-
- mpi-test - automated test program for MPI library
-
-=head1 SYNOPSIS
-
- mpi-test <suite-name> [quiet]
- mpi-test list
- mpi-test help
-
-=head1 DESCRIPTION
-
-The B<mpi-test> program is a general unit test driver for the MPI
-library.  It is used to verify that the library works as it is
-supposed to on your architecture.  As with most such things, passing
-all the tests in B<mpi-test> does not guarantee the code is correct,
-but if any of them fail, there are certainly problems.
-
-Each major function of the library can be tested individually.  For a
-list of the test suites understood by B<mpi-test>, run it with the
-I<list> command line option:
-
- mpi-test list
-
-This will display a list of the available test suites and a brief
-synopsis of what each one does.  For a brief overview of this
-document, run B<mpi-test> I<help>.
-
-B<mpi-test> exits with a zero status if the selected test succeeds, or
-a nonzero status if it fails.  If a I<suite-name> which is not
-understood by B<mpi-test> is given, a diagnostic is printed to the
-standard error, and the program exits with a result code of 2.  If a
-test fails, the result code will be 1, and a diagnostic is ordinarily
-printed to the standard error.  However, if the I<quiet> option is
-provided, these diagnostics will be suppressed.
-
-=head1 RESTRICTIONS
-
-Only a few canned test cases are provided.  The solutions have been
-verified using the GNU bc(1) program, so bugs there may cause problems
-here; however, this is very unlikely, so if a test fails, it is almost
-certainly my fault, not bc(1)'s.
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Thayer School of Engineering, Hanover, New Hampshire, USA
- 
- $Date: 2000-07-14 00:44:34 $

mozilla/security/nss/lib/freebl/mpi/doc/mul.txt

@@ -1,114 +0,0 @@
-Multiplication
-
-This describes the multiplication algorithm used by the MPI library.
-
-This is basically a standard "schoolbook" algorithm.  It is slow --
-O(mn) for m = #a, n = #b -- but easy to implement and verify.
-Basically, we run two nested loops, as illustrated here (R is the
-radix):
-
-k = 0
-for j <- 0 to (#b - 1)
-  for i <- 0 to (#a - 1)
-    w = (a[j] * b[i]) + k + c[i+j]
-    c[i+j] = w mod R
-    k = w div R
-  endfor
-  c[i+j] = k;
-  k = 0;
-endfor
-
-It is necessary that 'w' have room for at least two radix R digits.
-The product of any two digits in radix R is at most:
-
-	(R - 1)(R - 1) = R^2 - 2R + 1
-
-Since a two-digit radix-R number can hold R^2 - 1 distinct values,
-this insures that the product will fit into the two-digit register.
-
-To insure that two digits is enough for w, we must also show that
-there is room for the carry-in from the previous multiplication, and
-the current value of the product digit that is being recomputed.
-Assuming each of these may be as big as R - 1 (and no larger,
-certainly), two digits will be enough if and only if:
-
-	(R^2 - 2R + 1) + 2(R - 1) <= R^2 - 1
-
-Solving this equation shows that, indeed, this is the case:
-
-	R^2 - 2R + 1 + 2R - 2 <= R^2 - 1
-
-	R^2 - 1 <= R^2 - 1
-
-This suggests that a good radix would be one more than the largest
-value that can be held in half a machine word -- so, for example, as
-in this implementation, where we used a radix of 65536 on a machine
-with 4-byte words.  Another advantage of a radix of this sort is that
-binary-level operations are easy on numbers in this representation.
-
-Here's an example multiplication worked out longhand in radix-10,
-using the above algorithm:
-
-   a =     999
-   b =   x 999
-  -------------
-   p =   98001
-
-w = (a[jx] * b[ix]) + kin + c[ix + jx]
-c[ix+jx] = w % RADIX
-k = w / RADIX
-                                                               product
-ix	jx	a[jx]	b[ix]	kin	w	c[i+j]	kout	000000
-0	0	9	9	0	81+0+0	1	8	000001
-0	1	9	9	8	81+8+0	9	8	000091
-0	2	9	9	8	81+8+0	9	8	000991
-				8			0	008991
-1	0	9	9	0	81+0+9	0	9	008901
-1	1	9	9	9	81+9+9	9	9	008901
-1	2	9	9	9	81+9+8	8	9	008901
-				9			0	098901
-2	0	9	9	0	81+0+9	0	9	098001
-2	1	9	9	9	81+9+8	8	9	098001
-2	2	9	9	9	81+9+9	9	9	098001
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: mul.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-

mozilla/security/nss/lib/freebl/mpi/doc/pi.txt

@@ -1,90 +0,0 @@
-This file describes how pi is computed by the program in 'pi.c' (see
-the utils subdirectory).
-
-Basically, we use Machin's formula, which is what everyone in the
-world uses as a simple method for computing approximations to pi.
-This works for up to a few thousand digits without too much effort.
-Beyond that, though, it gets too slow.
-
-Machin's formula states:
-
-	 pi := 16 * arctan(1/5) - 4 * arctan(1/239)
-
-We compute this in integer arithmetic by first multiplying everything
-through by 10^d, where 'd' is the number of digits of pi we wanted to
-compute.  It turns out, the last few digits will be wrong, but the
-number that are wrong is usually very small (ordinarly only 2-3).
-Having done this, we compute the arctan() function using the formula:
-
-                       1      1       1       1       1     
-       arctan(1/x) := --- - ----- + ----- - ----- + ----- - ...
-                       x    3 x^3   5 x^5   7 x^7   9 x^9
-
-This is done iteratively by computing the first term manually, and
-then iteratively dividing x^2 and k, where k = 3, 5, 7, ... out of the
-current figure.  This is then added to (or subtracted from) a running
-sum, as appropriate.  The iteration continues until we overflow our
-available precision and the current figure goes to zero under integer
-division.  At that point, we're finished.
-
-Actually, we get a couple extra bits of precision out of the fact that
-we know we're computing y * arctan(1/x), by setting up the multiplier
-as:
-
-      y * 10^d
-
-... instead of just 10^d.  There is also a bit of cleverness in how
-the loop is constructed, to avoid special-casing the first term.
-Check out the code for arctan() in 'pi.c', if you are interested in
-seeing how it is set up.
-
-Thanks to Jason P. for this algorithm, which I assembled from notes
-and programs found on his cool "Pile of Pi Programs" page, at:
-
-      http://www.isr.umd.edu/~jasonp/pipage.html
-
-Thanks also to Henrik Johansson <Henrik.Johansson@Nexus.Comm.SE>, from
-whose pi program I borrowed the clever idea of pre-multiplying by x in
-order to avoid a special case on the loop iteration.
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: pi.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-

mozilla/security/nss/lib/freebl/mpi/doc/prng.pod

@@ -1,41 +0,0 @@
-=head1 NAME
-
- prng - pseudo-random number generator
-
-=head1 SYNOPSIS
-
- prng [count]
-
-=head1 DESCRIPTION
-
-B<Prng> generates 32-bit pseudo-random integers using the
-Blum-Blum-Shub (BBS) quadratic residue generator.  It is seeded using
-the standard C library's rand() function, which itself seeded from the
-system clock and the process ID number.  Thus, the values generated
-are not particularly useful for cryptographic applications, but they
-are in general much better than the typical output of the usual
-multiplicative congruency generator used by most runtime libraries.
-
-You may optionally specify how many random values should be generated
-by giving a I<count> argument on the command line.  If you do not
-specify a count, only one random value will be generated.  The results
-are output to the standard output in decimal notation, one value per
-line.
-
-=head1 RESTRICTIONS
-
-As stated above, B<prng> uses the C library's rand() function to seed
-the generator, so it is not terribly suitable for cryptographic
-applications.  Also note that each time you run the program, a new
-seed is generated, so it is better to run it once with a I<count>
-parameter than it is to run it multiple times to generate several
-values.
-
-=head1 AUTHOR
-
- Michael J. Fromberger <sting@linguist.dartmouth.edu>
- Copyright (C) 1998 Michael J. Fromberger, All Rights Reserved
- Thayer School of Engineering, Dartmouth College, Hanover, NH  USA
-
- $Date: 2000-07-14 00:44:36 $
-

mozilla/security/nss/lib/freebl/mpi/doc/redux.txt

@@ -1,121 +0,0 @@
-Modular Reduction
-
-Usually, modular reduction is accomplished by long division, using the
-mp_div() or mp_mod() functions.  However, when performing modular
-exponentiation, you spend a lot of time reducing by the same modulus
-again and again.  For this purpose, doing a full division for each
-multiplication is quite inefficient.
-
-For this reason, the mp_exptmod() function does not perform modular
-reductions in the usual way, but instead takes advantage of an
-algorithm due to Barrett, as described by Menezes, Oorschot and
-VanStone in their book _Handbook of Applied Cryptography_, published
-by the CRC Press (see Chapter 14 for details).  This method reduces
-most of the computation of reduction to efficient shifting and masking
-operations, and avoids the multiple-precision division entirely.
-
-Here is a brief synopsis of Barrett reduction, as it is implemented in
-this library.
-
-Let b denote the radix of the computation (one more than the maximum
-value that can be denoted by an mp_digit).  Let m be the modulus, and
-let k be the number of significant digits of m.  Let x be the value to
-be reduced modulo m.  By the Division Theorem, there exist unique
-integers Q and R such that:
-
-	 x = Qm + R, 0 <= R < m
-
-Barrett reduction takes advantage of the fact that you can easily
-approximate Q to within two, given a value M such that:
-
-	                  2k
-	                 b
-	    M = floor( ----- )
-	                 m
-
-Computation of M requires a full-precision division step, so if you
-are only doing a single reduction by m, you gain no advantage.
-However, when multiple reductions by the same m are required, this
-division need only be done once, beforehand.  Using this, we can use
-the following equation to compute Q', an approximation of Q:
-
-                     x
-            floor( ------ ) M
-                      k-1
-                     b
-Q' = floor( ----------------- )
-                    k+1
-                   b
-
-The divisions by b^(k-1) and b^(k+1) and the floor() functions can be
-efficiently implemented with shifts and masks, leaving only a single
-multiplication to be performed to get this approximation.  It can be
-shown that Q - 2 <= Q' <= Q, so in the worst case, we can get out with
-two additional subtractions to bring the value into line with the
-actual value of Q.
-
-Once we've got Q', we basically multiply that by m and subtract from
-x, yielding:
-
-   x - Q'm = Qm + R - Q'm
-
-Since we know the constraint on Q', this is one of:
-
-      R
-      m + R
-      2m + R
-
-Since R < m by the Division Theorem, we can simply subtract off m
-until we get a value in the correct range, which will happen with no
-more than 2 subtractions:
-
-     v = x - Q'm
-
-     while(v >= m)
-       v = v - m
-     endwhile
-
-
-In random performance trials, modular exponentiation using this method
-of reduction gave around a 40% speedup over using the division for
-reduction.
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: redux.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $

mozilla/security/nss/lib/freebl/mpi/doc/sqrt.txt

@@ -1,87 +0,0 @@
-Square Root
-
-A simple iterative algorithm is used to compute the greatest integer
-less than or equal to the square root.  Essentially, this is Newton's
-linear approximation, computed by finding successive values of the
-equation:
-
-		    x[k]^2 - V
-x[k+1]	 =  x[k] - ------------
-	             2 x[k]
-
-...where V is the value for which the square root is being sought.  In
-essence, what is happening here is that we guess a value for the
-square root, then figure out how far off we were by squaring our guess
-and subtracting the target.  Using this value, we compute a linear
-approximation for the error, and adjust the "guess".  We keep doing
-this until the precision gets low enough that the above equation
-yields a quotient of zero.  At this point, our last guess is one
-greater than the square root we're seeking.
-
-The initial guess is computed by dividing V by 4, which is a heuristic
-I have found to be fairly good on average.  This also has the
-advantage of being very easy to compute efficiently, even for large
-values.
-
-So, the resulting algorithm works as follows:
-
-    x = V / 4   /* compute initial guess */
-    
-    loop
-	t = (x * x) - V   /* Compute absolute error  */
-	u = 2 * x         /* Adjust by tangent slope */
-	t = t / u
-
-	/* Loop is done if error is zero */
-	if(t == 0)
-	    break
-
-	/* Adjust guess by error term    */
-	x = x - t
-    end
-
-    x = x - 1
-
-The result of the computation is the value of x.
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: sqrt.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-

mozilla/security/nss/lib/freebl/mpi/doc/square.txt

@@ -1,109 +0,0 @@
-Squaring Algorithm
-
-When you are squaring a value, you can take advantage of the fact that
-half the multiplications performed by the more general multiplication
-algorithm (see 'mul.txt' for a description) are redundant when the
-multiplicand equals the multiplier.
-
-In particular, the modified algorithm is:
-
-k = 0
-for j <- 0 to (#a - 1)
-  w = c[2*j] + (a[j] ^ 2);
-  k = w div R
-
-  for i <- j+1 to (#a - 1)
-    w = (2 * a[j] * a[i]) + k + c[i+j]
-    c[i+j] = w mod R
-    k = w div R
-  endfor
-  c[i+j] = k;
-  k = 0;
-endfor
-
-On the surface, this looks identical to the multiplication algorithm;
-however, note the following differences:
-
-  - precomputation of the leading term in the outer loop
-
-  - i runs from j+1 instead of from zero
-
-  - doubling of a[i] * a[j] in the inner product
-
-Unfortunately, the construction of the inner product is such that we
-need more than two digits to represent the inner product, in some
-cases.  In a C implementation, this means that some gymnastics must be
-performed in order to handle overflow, for which C has no direct
-abstraction.  We do this by observing the following:
-
-If we have multiplied a[i] and a[j], and the product is more than half
-the maximum value expressible in two digits, then doubling this result
-will overflow into a third digit.  If this occurs, we take note of the
-overflow, and double it anyway -- C integer arithmetic ignores
-overflow, so the two digits we get back should still be valid, modulo
-the overflow.
-
-Having doubled this value, we now have to add in the remainders and
-the digits already computed by earlier steps.  If we did not overflow
-in the previous step, we might still cause an overflow here.  That
-will happen whenever the maximum value expressible in two digits, less
-the amount we have to add, is greater than the result of the previous
-step.  Thus, the overflow computation is:
-
-
-  u = 0
-  w = a[i] * a[j]
-
-  if(w > (R - 1)/ 2)
-    u = 1;
-
-  w = w * 2
-  v = c[i + j] + k
-
-  if(u == 0 && (R - 1 - v) < w)
-    u = 1
-
-If there is an overflow, u will be 1, otherwise u will be 0.  The rest
-of the parameters are the same as they are in the above description.
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: square.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-

mozilla/security/nss/lib/freebl/mpi/doc/timing.txt

@@ -1,250 +0,0 @@
-MPI Library Timing Tests
-
-Hardware/OS
-(A) SGI O2 1 x MIPS R10000 250MHz IRIX 6.5.3
-(B) IBM RS/6000 43P-240 1 x PowerPC 603e 223MHz AIX 4.3
-(C) Dell GX1/L+ 1 x Pentium III 550MHz Linux 2.2.12-20
-(D) PowerBook G3 1 x PowerPC 750 266MHz LinuxPPC 2.2.6-15apmac
-(E) PowerBook G3 1 x PowerPC 750 266MHz MacOS 8.5.1
-(F) PowerBook G3 1 x PowerPC 750 400MHz MacOS 9.0.2
-
-Compiler
-(1) MIPSpro C 7.2.1 -O3 optimizations
-(2) GCC 2.95.1 -O3 optimizations
-(3) IBM AIX xlc -O3 optimizations (version unknown)
-(4) EGCS 2.91.66 -O3 optimizations
-(5) Metrowerks CodeWarrior 5.0 C, all optimizations
-(6) MIPSpro C 7.30 -O3 optimizations
-(7) same as (6), with optimized libmalloc.so
-
-Timings are given in seconds, computed using the C library's clock()
-function.  The first column gives the hardware and compiler
-configuration used for the test. The second column indicates the
-number of tests that were aggregated to get the statistics for that
-size.  These were compiled using 16 bit digits.
-
-Source data were generated randomly using a fixed seed, so they should
-be internally consistent, but may vary on different systems depending
-on the C library.  Also, since the resolution of the timer accessed by
-clock() varies, there may be some variance in the precision of these
-measurements.
-
-Prime Generation (primegen)
-
-128 bits:
-A1      200     min=0.03, avg=0.19, max=0.72, sum=38.46
-A2      200     min=0.02, avg=0.16, max=0.62, sum=32.55
-B3      200     min=0.01, avg=0.07, max=0.22, sum=13.29
-C4      200     min=0.00, avg=0.03, max=0.20, sum=6.14
-D4      200     min=0.00, avg=0.05, max=0.33, sum=9.70
-A6      200     min=0.01, avg=0.09, max=0.36, sum=17.48
-A7      200     min=0.00, avg=0.05, max=0.24, sum=10.07
-
-192 bits:
-A1      200     min=0.05, avg=0.45, max=3.13, sum=89.96
-A2      200     min=0.04, avg=0.39, max=2.61, sum=77.55
-B3      200     min=0.02, avg=0.18, max=1.25, sum=36.97
-C4      200     min=0.01, avg=0.09, max=0.33, sum=18.24
-D4      200     min=0.02, avg=0.15, max=0.54, sum=29.63
-A6      200     min=0.02, avg=0.24, max=1.70, sum=47.84
-A7      200     min=0.01, avg=0.15, max=1.05, sum=30.88
-
-256 bits:
-A1      200     min=0.08, avg=0.92, max=6.13, sum=184.79
-A2      200     min=0.06, avg=0.76, max=5.03, sum=151.11
-B3      200     min=0.04, avg=0.41, max=2.68, sum=82.35
-C4      200     min=0.02, avg=0.19, max=0.69, sum=37.91
-D4      200     min=0.03, avg=0.31, max=1.15, sum=63.00
-A6      200     min=0.04, avg=0.48, max=3.13, sum=95.46
-A7      200     min=0.03, avg=0.37, max=2.36, sum=73.60
-
-320 bits:
-A1      200     min=0.11, avg=1.59, max=6.14, sum=318.81
-A2      200     min=0.09, avg=1.27, max=4.93, sum=254.03
-B3      200     min=0.07, avg=0.82, max=3.13, sum=163.80
-C4      200     min=0.04, avg=0.44, max=1.91, sum=87.59
-D4      200     min=0.06, avg=0.73, max=3.22, sum=146.73
-A6      200     min=0.07, avg=0.93, max=3.50, sum=185.01
-A7      200     min=0.05, avg=0.76, max=2.94, sum=151.78
-
-384 bits:
-A1      200     min=0.16, avg=2.69, max=11.41, sum=537.89
-A2      200     min=0.13, avg=2.15, max=9.03, sum=429.14
-B3      200     min=0.11, avg=1.54, max=6.49, sum=307.78
-C4      200     min=0.06, avg=0.81, max=4.84, sum=161.13
-D4      200     min=0.10, avg=1.38, max=8.31, sum=276.81
-A6      200     min=0.11, avg=1.73, max=7.36, sum=345.55
-A7      200     min=0.09, avg=1.46, max=6.12, sum=292.02
-
-448 bits:
-A1      200     min=0.23, avg=3.36, max=15.92, sum=672.63
-A2      200     min=0.17, avg=2.61, max=12.25, sum=522.86
-B3      200     min=0.16, avg=2.10, max=9.83, sum=420.86
-C4      200     min=0.09, avg=1.44, max=7.64, sum=288.36
-D4      200     min=0.16, avg=2.50, max=13.29, sum=500.17
-A6      200     min=0.15, avg=2.31, max=10.81, sum=461.58
-A7      200     min=0.14, avg=2.03, max=9.53, sum=405.16
-
-512 bits:
-A1      200     min=0.30, avg=6.12, max=22.18, sum=1223.35
-A2      200     min=0.25, avg=4.67, max=16.90, sum=933.18
-B3      200     min=0.23, avg=4.13, max=14.94, sum=825.45
-C4      200     min=0.13, avg=2.08, max=9.75, sum=415.22
-D4      200     min=0.24, avg=4.04, max=20.18, sum=808.11
-A6      200     min=0.22, avg=4.47, max=16.19, sum=893.83
-A7      200     min=0.20, avg=4.03, max=14.65, sum=806.02
-
-Modular Exponentation (metime)
-
-The following results are aggregated from 200 pseudo-randomly
-generated tests, based on a fixed seed. 
-
-                      base, exponent, and modulus size (bits)
-P/C       128   192   256   320   384   448   512   640   768   896  1024
-------- -----------------------------------------------------------------
-A1      0.015 0.027 0.047 0.069 0.098 0.133 0.176 0.294 0.458 0.680 1.040
-A2      0.013 0.024 0.037 0.053 0.077 0.102 0.133 0.214 0.326 0.476 0.668
-B3      0.005 0.011 0.021 0.036 0.056 0.084 0.121 0.222 0.370 0.573 0.840
-C4      0.002 0.006 0.011 0.020 0.032 0.048 0.069 0.129 0.223 0.344 0.507
-D4      0.004 0.010 0.019 0.034 0.056 0.085 0.123 0.232 0.390 0.609 0.899
-E5      0.007 0.015 0.031 0.055 0.088 0.133 0.183 0.342 0.574 0.893 1.317
-A6      0.008 0.016 0.038 0.042 0.064 0.093 0.133 0.239 0.393 0.604 0.880
-A7      0.005 0.011 0.020 0.036 0.056 0.083 0.121 0.223 0.374 0.583 0.855
-
-Multiplication and Squaring tests, (mulsqr)
-
-The following results are aggregated from 500000 pseudo-randomly
-generated tests, based on a per-run wall-clock seed.  Times are given
-in seconds, except where indicated in microseconds (us).
-
-(A1)
-
-bits    multiply    square  ad  percent time/mult   time/square
-64      9.33        9.15    >   1.9     18.7us      18.3us
-128     10.88       10.44   >   4.0     21.8us      20.9us
-192     13.30       11.89   >   10.6    26.7us      23.8us
-256     14.88       12.64   >   15.1    29.8us      25.3us
-320     18.64       15.01   >   19.5    37.3us      30.0us
-384     23.11       17.70   >   23.4    46.2us      35.4us
-448     28.28       20.88   >   26.2    56.6us      41.8us
-512     34.09       24.51   >   28.1    68.2us      49.0us
-640     47.86       33.25   >   30.5    95.7us      66.5us
-768     64.91       43.54   >   32.9    129.8us     87.1us
-896     84.49       55.48   >   34.3    169.0us     111.0us
-1024    107.25      69.21   >   35.5    214.5us     138.4us
-1536    227.97      141.91  >   37.8    456.0us     283.8us
-2048    394.05      242.15  >   38.5    788.1us     484.3us
-
-(A2)
-
-bits    multiply    square  ad  percent time/mult   time/square
-64      7.87        7.95    <   1.0     15.7us      15.9us
-128     9.40        9.19    >   2.2     18.8us      18.4us
-192     11.15       10.59   >   5.0     22.3us      21.2us
-256     12.02       11.16   >   7.2     24.0us      22.3us
-320     14.62       13.43   >   8.1     29.2us      26.9us
-384     17.72       15.80   >   10.8    35.4us      31.6us
-448     21.24       18.51   >   12.9    42.5us      37.0us
-512     25.36       21.78   >   14.1    50.7us      43.6us
-640     34.57       29.00   >   16.1    69.1us      58.0us
-768     46.10       37.60   >   18.4    92.2us      75.2us
-896     58.94       47.72   >   19.0    117.9us     95.4us
-1024    73.76       59.12   >   19.8    147.5us     118.2us
-1536    152.00      118.80  >   21.8    304.0us     237.6us
-2048    259.41      199.57  >   23.1    518.8us     399.1us
-
-(B3)
-
-bits    multiply    square  ad  percent time/mult   time/square
-64      2.60        2.47    >   5.0     5.20us      4.94us
-128     4.43        4.06    >   8.4     8.86us      8.12us
-192     7.03        6.10    >   13.2    14.1us      12.2us
-256     10.44       8.59    >   17.7    20.9us      17.2us
-320     14.44       11.64   >   19.4    28.9us      23.3us
-384     19.12       15.08   >   21.1    38.2us      30.2us
-448     24.55       19.09   >   22.2    49.1us      38.2us
-512     31.03       23.53   >   24.2    62.1us      47.1us
-640     45.05       33.80   >   25.0    90.1us      67.6us
-768     63.02       46.05   >   26.9    126.0us     92.1us
-896     83.74       60.29   >   28.0    167.5us     120.6us
-1024    106.73      76.65   >   28.2    213.5us     153.3us
-1536    228.94      160.98  >   29.7    457.9us     322.0us
-2048    398.08      275.93  >   30.7    796.2us     551.9us
-
-(C4)
-
-bits    multiply    square  ad  percent time/mult   time/square
-64      1.34        1.28    >   4.5     2.68us      2.56us
-128     2.76        2.59    >   6.2     5.52us      5.18us
-192     4.52        4.16    >   8.0     9.04us      8.32us
-256     6.64        5.99    >   9.8     13.3us      12.0us
-320     9.20        8.13    >   11.6    18.4us      16.3us
-384     12.01       10.58   >   11.9    24.0us      21.2us
-448     15.24       13.33   >   12.5    30.5us      26.7us
-512     19.02       16.46   >   13.5    38.0us      32.9us
-640     27.56       23.54   >   14.6    55.1us      47.1us
-768     37.89       31.78   >   16.1    75.8us      63.6us
-896     49.24       41.42   >   15.9    98.5us      82.8us
-1024    62.59       52.18   >   16.6    125.2us     104.3us
-1536    131.66      107.72  >   18.2    263.3us     215.4us
-2048    226.45      182.95  >   19.2    453.0us     365.9us
-
-(A7)
-
-bits    multiply    square  ad  percent time/mult   time/square
-64      1.74        1.71    >   1.7     3.48us      3.42us
-128     3.48        2.96    >   14.9    6.96us      5.92us
-192     5.74        4.60    >   19.9    11.5us      9.20us
-256     8.75        6.61    >   24.5    17.5us      13.2us
-320     12.5        8.99    >   28.1    25.0us      18.0us
-384     16.9        11.9    >   29.6    33.8us      23.8us
-448     22.2        15.2    >   31.7    44.4us      30.4us
-512     28.3        19.0    >   32.7    56.6us      38.0us
-640     42.4        28.0    >   34.0    84.8us      56.0us
-768     59.4        38.5    >   35.2    118.8us     77.0us
-896     79.5        51.2    >   35.6    159.0us     102.4us
-1024    102.6	    65.5    >	36.2	205.2us	    131.0us
-1536    224.3	    140.6   >	37.3	448.6us	    281.2us
-2048    393.4	    244.3   >	37.9	786.8us	    488.6us
-
-------------------------------------------------------------------
-***** BEGIN LICENSE BLOCK *****
-Version: MPL 1.1/GPL 2.0/LGPL 2.1
-
-The contents of this file are subject to the Mozilla Public License Version 
-1.1 (the "License"); you may not use this file except in compliance with 
-the License. You may obtain a copy of the License at 
-http://www.mozilla.org/MPL/
-
-Software distributed under the License is distributed on an "AS IS" basis,
-WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-for the specific language governing rights and limitations under the
-License.
-
-The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-library.
-
-The Initial Developer of the Original Code is
-Michael J. Fromberger <sting@linguist.dartmouth.edu>
-Portions created by the Initial Developer are Copyright (C) 1998, 2000
-the Initial Developer. All Rights Reserved.
-
-Contributor(s):
-
-Alternatively, the contents of this file may be used under the terms of
-either the GNU General Public License Version 2 or later (the "GPL"), or
-the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-in which case the provisions of the GPL or the LGPL are applicable instead
-of those above. If you wish to allow use of your version of this file only
-under the terms of either the GPL or the LGPL, and not to allow others to
-use your version of this file under the terms of the MPL, indicate your
-decision by deleting the provisions above and replace them with the notice
-and other provisions required by the GPL or the LGPL. If you do not delete
-the provisions above, a recipient may use your version of this file under
-the terms of any one of the MPL, the GPL or the LGPL.
-
-***** END LICENSE BLOCK *****
-
-$Id: timing.txt,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-

mozilla/security/nss/lib/freebl/mpi/hpma512.s

@@ -1,640 +0,0 @@
-/*
- * The contents of this file are subject to the Mozilla Public
- * License Version 1.1 (the "License"); you may not use this file
- * except in compliance with the License. You may obtain a copy of
- * the License at http://www.mozilla.org/MPL/
- * 
- * Software distributed under the License is distributed on an "AS
- * IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
- * implied. See the License for the specific language governing
- * rights and limitations under the License.
- * 
- * The Original Code is multacc512 multiple-precision integer arithmetic.
- * 
- * The Initial Developer of the Original Code is Hewlett-Packard Company.
- * Portions created by Hewlett-Packard Company are 
- * Copyright (C) March 1999, Hewlett-Packard Company. All Rights Reserved.
- * 
- * Contributor(s):
- *  coded by:   Bill Worley, Hewlett-Packard labs
- * 
- * Alternatively, the contents of this file may be used under the
- * terms of the GNU General Public License Version 2 or later (the
- * "GPL"), in which case the provisions of the GPL are applicable 
- * instead of those above.  If you wish to allow use of your 
- * version of this file only under the terms of the GPL and not to
- * allow others to use your version of this file under the MPL,
- * indicate your decision by deleting the provisions above and
- * replace them with the notice and other provisions required by
- * the GPL.  If you do not delete the provisions above, a recipient
- * may use your version of this file under either the MPL or the
- * GPL.
- *
- *  This PA-RISC 2.0 function computes the product of two unsigned integers,
- *  and adds the result to a previously computed integer.  The multiplicand
- *  is a 512-bit (64-byte, eight doubleword) unsigned integer, stored in
- *  memory in little-double-wordian order.  The multiplier is an unsigned
- *  64-bit integer.  The previously computed integer to which the product is
- *  added is located in the result ("res") area, and is assumed to be a
- *  576-bit (72-byte, nine doubleword) unsigned integer, stored in memory
- *  in little-double-wordian order.  This value normally will be the result
- *  of a previously computed nine doubleword result.  It is not necessary
- *  to pad the multiplicand with an additional 64-bit zero doubleword.
- *
- *  Multiplicand, multiplier, and addend ideally should be aligned at
- *  16-byte boundaries for best performance.  The code will function
- *  correctly for alignment at eight-byte boundaries which are not 16-byte
- *  boundaries, but the execution may be slightly slower due to even/odd
- *  bank conflicts on PA-RISC 8000 processors.
- *
- *  This function is designed to accept the same calling sequence as Bill
- *  Ackerman's "maxpy_little" function.  The carry from the ninth doubleword
- *  of the result is written to the tenth word of the result, as is done by
- *  Bill Ackerman's function.  The final carry also is returned as an
- *  integer, which may be ignored.  The function prototype may be either
- *  of the following:
- *
- *      void multacc512( int l, chunk* m, const chunk* a, chunk* res );
- *          or
- *      int multacc512( int l, chunk* m, const chunk* a, chunk* res );
- *
- *  where:  "l" originally denoted vector lengths.  This parameter is
- *      ignored.  This function always assumes a multiplicand length of
- *      512 bits (eight doublewords), and addend and result lengths of
- *      576 bits (nine doublewords).
- *
- *      "m" is a pointer to the doubleword multiplier, ideally aligned
- *      on a 16-byte boundary.
- *
- *      "a" is a pointer to the eight-doubleword multiplicand, stored
- *      in little-double-wordian order, and ideally aligned on a 16-byte
- *      boundary.
- *
- *      "res" is a pointer to the nine doubleword addend, and to the
- *      nine-doubleword product computed by this function.  The result
- *      also is stored in little-double-wordian order, and ideally is
- *      aligned on a 16-byte boundary. It is expected that the alignment
- *      of the "res" area may alternate between even/odd doubleword
- *      boundaries for successive calls for 512-bit x 512-bit
- *      multiplications.
- *
- *  The code for this function has been scheduled to use the parallelism
- *  of the PA-RISC 8000 series microprocessors as well as the author was
- *  able.  Comments and/or suggestions for improvement are welcomed.
- *
- *  The code is "64-bit safe".  This means it may be called in either
- *  the 32ILP context or the 64LP context.  All 64-bits of registers are
- *  saved and restored.
- *
- *  This code is self-contained.  It requires no other header files in order
- *  to compile and to be linkable on a PA-RISC 2.0 machine.  Symbolic
- *  definitions for registers and stack offsets are included within this
- *  one source file.
- *
- *  This is a leaf routine.  As such, minimal use is made of the stack area.
- *  Of the 192 bytes allocated, 64 bytes are used for saving/restoring eight
- *  general registers, and 128 bytes are used to move intermediate products
- *  from the floating-point registers to the general registers.  Stack
- *  protocols assure proper alignment of these areas.
- *
- */
-
-
-/*  ====================================================================*/
-/*      symbolic definitions for PA-RISC registers      */
-/*      in the MIPS style, avoids lots of case shifts       */
-/*      assigments (except t4) preserve register number parity  */
-/*  ====================================================================*/
-
-#define zero    %r0         /* permanent zero */
-#define t5      %r1         /* temp register, altered by addil */
-
-#define rp      %r2         /* return pointer */
-
-#define s1      %r3         /* callee saves register*/
-#define s0      %r4         /* callee saves register*/
-#define s3      %r5         /* callee saves register*/
-#define s2      %r6         /* callee saves register*/
-#define s5      %r7         /* callee saves register*/
-#define s4      %r8         /* callee saves register*/
-#define s7      %r9         /* callee saves register*/
-#define s6      %r10        /* callee saves register*/
-
-#define t1      %r19        /* caller saves register*/
-#define t0      %r20        /* caller saves register*/
-#define t3      %r21        /* caller saves register*/
-#define t2      %r22        /* caller saves register*/
-
-#define a3      %r23        /* fourth argument register, high word */
-#define a2      %r24        /* third argument register, low word*/
-#define a1      %r25        /* second argument register, high word*/
-#define a0      %r26        /* first argument register, low word*/
-
-#define v0      %r28        /* high order return value*/
-#define v1      %r29        /* low order return value*/
-
-#define sp      %r30        /* stack pointer*/
-#define t4      %r31        /* temporary register   */
-
-#define fa0     %fr4        /* first argument register*/
-#define fa1     %fr5        /* second argument register*/
-#define fa2     %fr6        /* third argument register*/
-#define fa3     %fr7        /* fourth argument register*/
-
-#define fa0r    %fr4R       /* first argument register*/
-#define fa1r    %fr5R       /* second argument register*/
-#define fa2r    %fr6R       /* third argument register*/
-#define fa3r    %fr7R       /* fourth argument register*/
-
-#define ft0     %fr8        /* caller saves register*/
-#define ft1     %fr9        /* caller saves register*/
-#define ft2     %fr10       /* caller saves register*/
-#define ft3     %fr11       /* caller saves register*/
-
-#define ft0r    %fr8R       /* caller saves register*/
-#define ft1r    %fr9R       /* caller saves register*/
-#define ft2r    %fr10R      /* caller saves register*/
-#define ft3r    %fr11R      /* caller saves register*/
-
-#define ft4     %fr22       /* caller saves register*/
-#define ft5     %fr23       /* caller saves register*/
-#define ft6     %fr24       /* caller saves register*/
-#define ft7     %fr25       /* caller saves register*/
-#define ft8     %fr26       /* caller saves register*/
-#define ft9     %fr27       /* caller saves register*/
-#define ft10    %fr28       /* caller saves register*/
-#define ft11    %fr29       /* caller saves register*/
-#define ft12    %fr30       /* caller saves register*/
-#define ft13    %fr31       /* caller saves register*/
-
-#define ft4r    %fr22R      /* caller saves register*/
-#define ft5r    %fr23R      /* caller saves register*/
-#define ft6r    %fr24R      /* caller saves register*/
-#define ft7r    %fr25R      /* caller saves register*/
-#define ft8r    %fr26R      /* caller saves register*/
-#define ft9r    %fr27R      /* caller saves register*/
-#define ft10r   %fr28R      /* caller saves register*/
-#define ft11r   %fr29R      /* caller saves register*/
-#define ft12r   %fr30R      /* caller saves register*/
-#define ft13r   %fr31R      /* caller saves register*/
-
-
-
-/*  ================================================================== */
-/*      functional definitions for PA-RISC registers           */
-/*  ================================================================== */
-
-/*              general registers           */
-
-#define T1      a0          /* temp, (length parameter ignored)             */
-
-#define pM      a1          /* -> 64-bit multiplier                         */
-#define T2      a1          /* temp, (after fetching multiplier)            */
-
-#define pA      a2          /* -> multiplicand vector (8 64-bit words)      */
-#define T3      a2          /* temp, (after fetching multiplicand)          */
-
-#define pR      a3          /* -> addend vector (8 64-bit doublewords,
-                                  result vector (9 64-bit words)            */
-
-#define S0      s0          /* callee saves summand registers               */
-#define S1      s1
-#define S2      s2
-#define S3      s3
-#define S4      s4
-#define S5      s5
-#define S6      s6
-#define S7      s7
-
-#define S8      v0          /* caller saves summand registers               */
-#define S9      v1
-#define S10     t0
-#define S11     t1
-#define S12     t2
-#define S13     t3
-#define S14     t4
-#define S15     t5
-
-
-
-/*              floating-point registers                                    */
-
-#define M       fa0         /* multiplier double word                       */
-#define MR      fa0r        /* low order half of multiplier double word     */
-#define ML      fa0         /* high order half of multiplier double word    */
-
-#define A0      fa2         /* multiplicand double word 0                   */
-#define A0R     fa2r        /* low order half of multiplicand double word   */
-#define A0L     fa2         /* high order half of multiplicand double word  */
-
-#define A1      fa3         /* multiplicand double word 1                   */
-#define A1R     fa3r        /* low order half of multiplicand double word   */
-#define A1L     fa3         /* high order half of multiplicand double word  */
-
-#define A2      ft0         /* multiplicand double word 2                   */
-#define A2R     ft0r        /* low order half of multiplicand double word   */
-#define A2L     ft0         /* high order half of multiplicand double word  */
-
-#define A3      ft1         /* multiplicand double word 3                   */
-#define A3R     ft1r        /* low order half of multiplicand double word   */
-#define A3L     ft1         /* high order half of multiplicand double word  */
-
-#define A4      ft2         /* multiplicand double word 4                   */
-#define A4R     ft2r        /* low order half of multiplicand double word   */
-#define A4L     ft2         /* high order half of multiplicand double word  */
-
-#define A5      ft3         /* multiplicand double word 5                   */
-#define A5R     ft3r        /* low order half of multiplicand double word   */
-#define A5L     ft3         /* high order half of multiplicand double word  */
-
-#define A6      ft4         /* multiplicand double word 6                   */
-#define A6R     ft4r        /* low order half of multiplicand double word   */
-#define A6L     ft4         /* high order half of multiplicand double word  */
-
-#define A7      ft5         /* multiplicand double word 7                   */
-#define A7R     ft5r        /* low order half of multiplicand double word   */
-#define A7L     ft5         /* high order half of multiplicand double word  */
-
-#define P0      ft6         /* product word 0                               */
-#define P1      ft7         /* product word 0                               */
-#define P2      ft8         /* product word 0                               */
-#define P3      ft9         /* product word 0                               */
-#define P4      ft10        /* product word 0                               */
-#define P5      ft11        /* product word 0                               */
-#define P6      ft12        /* product word 0                               */
-#define P7      ft13        /* product word 0                               */
-
-
-
-
-/*  ======================================================================  */
-/*      symbolic definitions for HP-UX stack offsets                        */
-/*      symbolic definitions for memory NOPs                                */
-/*  ======================================================================  */
-
-#define ST_SZ       192         /* stack area total size                    */
-
-#define SV0         -192(sp)    /* general register save area               */
-#define SV1         -184(sp)
-#define SV2         -176(sp)
-#define SV3         -168(sp)
-#define SV4         -160(sp)
-#define SV5         -152(sp)
-#define SV6         -144(sp)
-#define SV7         -136(sp)
-
-#define XF0         -128(sp)    /* data transfer area                       */
-#define XF1         -120(sp)    /* for floating-pt to integer regs          */
-#define XF2         -112(sp)
-#define XF3         -104(sp)
-#define XF4         -96(sp)
-#define XF5         -88(sp)
-#define XF6         -80(sp)
-#define XF7         -72(sp)
-#define XF8         -64(sp)
-#define XF9         -56(sp)
-#define XF10        -48(sp)
-#define XF11        -40(sp)
-#define XF12        -32(sp)
-#define XF13        -24(sp)
-#define XF14        -16(sp)
-#define XF15        -8(sp)
-
-#define mnop    proberi (sp),3,zero     /* memory NOP                       */
-
-
-
-
-/*  ======================================================================  */
-/*      assembler formalities                                               */
-/*  ======================================================================  */
-
-#ifdef __LP64__
-                .level  2.0W
-#else
-                .level  2.0
-#endif
-                .space    $TEXT$
-                .subspa   $CODE$
-                .align    16
-
-/*  ======================================================================  */
-/*      here to compute 64-bit x 512-bit product + 512-bit addend           */
-/*  ======================================================================  */
-
-multacc512
-        .PROC
-        .CALLINFO
-        .ENTER
-    fldd    0(pM),M                 ; multiplier double word
-    ldo     ST_SZ(sp),sp            ; push stack
-
-    fldd    0(pA),A0                ; multiplicand double word 0
-    std     S1,SV1                  ; save s1
-
-    fldd    16(pA),A2               ; multiplicand double word 2
-    std     S3,SV3                  ; save s3
-
-    fldd    32(pA),A4               ; multiplicand double word 4
-    std     S5,SV5                  ; save s5
-
-    fldd    48(pA),A6               ; multiplicand double word 6
-    std     S7,SV7                  ; save s7
-
-
-    std     S0,SV0                  ; save s0
-    fldd    8(pA),A1                ; multiplicand double word 1
-    xmpyu   MR,A0L,P0               ; A0 cross 32-bit word products
-    xmpyu   ML,A0R,P2
-
-    std     S2,SV2                  ; save s2
-    fldd    24(pA),A3               ; multiplicand double word 3
-    xmpyu   MR,A2L,P4               ; A2 cross 32-bit word products
-    xmpyu   ML,A2R,P6
-
-    std     S4,SV4                  ; save s4
-    fldd    40(pA),A5               ; multiplicand double word 5
-
-    std     S6,SV6                  ; save s6
-    fldd    56(pA),A7               ; multiplicand double word 7
-
-
-    fstd    P0,XF0                  ; MR * A0L
-    xmpyu   MR,A0R,P0               ; A0 right 32-bit word product
-    xmpyu   MR,A1L,P1               ; A1 cross 32-bit word product
-
-    fstd    P2,XF2                  ; ML * A0R
-    xmpyu   ML,A0L,P2               ; A0 left 32-bit word product
-    xmpyu   ML,A1R,P3               ; A1 cross 32-bit word product
-
-    fstd    P4,XF4                  ; MR * A2L
-    xmpyu   MR,A2R,P4               ; A2 right 32-bit word product
-    xmpyu   MR,A3L,P5               ; A3 cross 32-bit word product
-
-    fstd    P6,XF6                  ; ML * A2R
-    xmpyu   ML,A2L,P6               ; A2 parallel 32-bit word product
-    xmpyu   ML,A3R,P7               ; A3 cross 32-bit word product
-
-
-    ldd     XF0,S0                  ; MR * A0L
-    fstd    P1,XF1                  ; MR * A1L
-
-    ldd     XF2,S2                  ; ML * A0R
-    fstd    P3,XF3                  ; ML * A1R
-
-    ldd     XF4,S4                  ; MR * A2L
-    fstd    P5,XF5                  ; MR * A3L
-    xmpyu   MR,A1R,P1               ; A1 parallel 32-bit word products
-    xmpyu   ML,A1L,P3
-
-    ldd     XF6,S6                  ; ML * A2R
-    fstd    P7,XF7                  ; ML * A3R
-    xmpyu   MR,A3R,P5               ; A3 parallel 32-bit word products
-    xmpyu   ML,A3L,P7
-
-
-    fstd    P0,XF0                  ; MR * A0R
-    ldd     XF1,S1                  ; MR * A1L
-    nop
-    add     S0,S2,T1                ; A0 cross product sum
-
-    fstd    P2,XF2                  ; ML * A0L
-    ldd     XF3,S3                  ; ML * A1R
-    add,dc  zero,zero,S0            ; A0 cross product sum carry
-    depd,z  T1,31,32,S2             ; A0 cross product sum << 32
-
-    fstd    P4,XF4                  ; MR * A2R
-    ldd     XF5,S5                  ; MR * A3L
-    shrpd   S0,T1,32,S0             ; A0 carry | cross product sum >> 32
-    add     S4,S6,T3                ; A2 cross product sum
-
-    fstd    P6,XF6                  ; ML * A2L
-    ldd     XF7,S7                  ; ML * A3R
-    add,dc  zero,zero,S4            ; A2 cross product sum carry
-    depd,z  T3,31,32,S6             ; A2 cross product sum << 32
-
-
-    ldd     XF0,S8                  ; MR * A0R
-    fstd    P1,XF1                  ; MR * A1R
-    xmpyu   MR,A4L,P0               ; A4 cross 32-bit word product
-    xmpyu   MR,A5L,P1               ; A5 cross 32-bit word product
-
-    ldd     XF2,S10                 ; ML * A0L
-    fstd    P3,XF3                  ; ML * A1L
-    xmpyu   ML,A4R,P2               ; A4 cross 32-bit word product
-    xmpyu   ML,A5R,P3               ; A5 cross 32-bit word product
-
-    ldd     XF4,S12                 ; MR * A2R
-    fstd    P5,XF5                  ; MR * A3L
-    xmpyu   MR,A6L,P4               ; A6 cross 32-bit word product
-    xmpyu   MR,A7L,P5               ; A7 cross 32-bit word product
-
-    ldd     XF6,S14                 ; ML * A2L
-    fstd    P7,XF7                  ; ML * A3L
-    xmpyu   ML,A6R,P6               ; A6 cross 32-bit word product
-    xmpyu   ML,A7R,P7               ; A7 cross 32-bit word product
-
-
-    fstd    P0,XF0                  ; MR * A4L
-    ldd     XF1,S9                  ; MR * A1R
-    shrpd   S4,T3,32,S4             ; A2 carry | cross product sum >> 32
-    add     S1,S3,T1                ; A1 cross product sum
-
-    fstd    P2,XF2                  ; ML * A4R
-    ldd     XF3,S11                 ; ML * A1L
-    add,dc  zero,zero,S1            ; A1 cross product sum carry
-    depd,z  T1,31,32,S3             ; A1 cross product sum << 32
-
-    fstd    P4,XF4                  ; MR * A6L
-    ldd     XF5,S13                 ; MR * A3R
-    shrpd   S1,T1,32,S1             ; A1 carry | cross product sum >> 32
-    add     S5,S7,T3                ; A3 cross product sum
-
-    fstd    P6,XF6                  ; ML * A6R
-    ldd     XF7,S15                 ; ML * A3L
-    add,dc  zero,zero,S5            ; A3 cross product sum carry
-    depd,z  T3,31,32,S7             ; A3 cross product sum << 32
-
-
-    shrpd   S5,T3,32,S5             ; A3 carry | cross product sum >> 32
-    add     S2,S8,S8                ; M * A0 right doubleword, P0 doubleword
-
-    add,dc  S0,S10,S10              ; M * A0 left doubleword
-    add     S3,S9,S9                ; M * A1 right doubleword
-
-    add,dc  S1,S11,S11              ; M * A1 left doubleword
-    add     S6,S12,S12              ; M * A2 right doubleword
-
-
-    ldd     24(pR),S3               ; Addend word 3
-    fstd    P1,XF1                  ; MR * A5L
-    add,dc  S4,S14,S14              ; M * A2 left doubleword
-    xmpyu   MR,A5R,P1               ; A5 right 32-bit word product
-
-    ldd     8(pR),S1                ; Addend word 1
-    fstd    P3,XF3                  ; ML * A5R
-    add     S7,S13,S13              ; M * A3 right doubleword
-    xmpyu   ML,A5L,P3               ; A5 left 32-bit word product
-
-    ldd     0(pR),S7                ; Addend word 0
-    fstd    P5,XF5                  ; MR * A7L
-    add,dc  S5,S15,S15              ; M * A3 left doubleword
-    xmpyu   MR,A7R,P5               ; A7 right 32-bit word product
-
-    ldd     16(pR),S5               ; Addend word 2
-    fstd    P7,XF7                  ; ML * A7R
-    add     S10,S9,S9               ; P1 doubleword
-    xmpyu   ML,A7L,P7               ; A7 left 32-bit word products
-
-
-    ldd     XF0,S0                  ; MR * A4L
-    fstd    P1,XF9                  ; MR * A5R
-    add,dc  S11,S12,S12             ; P2 doubleword
-    xmpyu   MR,A4R,P0               ; A4 right 32-bit word product
-
-    ldd     XF2,S2                  ; ML * A4R
-    fstd    P3,XF11                 ; ML * A5L
-    add,dc  S14,S13,S13             ; P3 doubleword
-    xmpyu   ML,A4L,P2               ; A4 left 32-bit word product
-
-    ldd     XF6,S6                  ; ML * A6R
-    fstd    P5,XF13                 ; MR * A7R
-    add,dc  zero,S15,T2             ; P4 partial doubleword
-    xmpyu   MR,A6R,P4               ; A6 right 32-bit word product
-
-    ldd     XF4,S4                  ; MR * A6L
-    fstd    P7,XF15                 ; ML * A7L
-    add     S7,S8,S8                ; R0 + P0, new R0 doubleword
-    xmpyu   ML,A6L,P6               ; A6 left 32-bit word product
-
-
-    fstd    P0,XF0                  ; MR * A4R
-    ldd     XF7,S7                  ; ML * A7R
-    add,dc  S1,S9,S9                ; c + R1 + P1, new R1 doubleword
-
-    fstd    P2,XF2                  ; ML * A4L
-    ldd     XF1,S1                  ; MR * A5L
-    add,dc  S5,S12,S12              ; c + R2 + P2, new R2 doubleword
-
-    fstd    P4,XF4                  ; MR * A6R
-    ldd     XF5,S5                  ; MR * A7L
-    add,dc  S3,S13,S13              ; c + R3 + P3, new R3 doubleword
-
-    fstd    P6,XF6                  ; ML * A6L
-    ldd     XF3,S3                  ; ML * A5R
-    add,dc  zero,T2,T2              ; c + partial P4
-    add     S0,S2,T1                ; A4 cross product sum
-
-
-    std     S8,0(pR)                ; save R0
-    add,dc  zero,zero,S0            ; A4 cross product sum carry
-    depd,z  T1,31,32,S2             ; A4 cross product sum << 32
-
-    std     S9,8(pR)                ; save R1
-    shrpd   S0,T1,32,S0             ; A4 carry | cross product sum >> 32
-    add     S4,S6,T3                ; A6 cross product sum
-
-    std     S12,16(pR)              ; save R2
-    add,dc  zero,zero,S4            ; A6 cross product sum carry
-    depd,z  T3,31,32,S6             ; A6 cross product sum << 32
-
-
-    std     S13,24(pR)              ; save R3
-    shrpd   S4,T3,32,S4             ; A6 carry | cross product sum >> 32
-    add     S1,S3,T1                ; A5 cross product sum
-
-    ldd     XF0,S8                  ; MR * A4R
-    add,dc  zero,zero,S1            ; A5 cross product sum carry
-    depd,z  T1,31,32,S3             ; A5 cross product sum << 32
-
-    ldd     XF2,S10                 ; ML * A4L
-    ldd     XF9,S9                  ; MR * A5R
-    shrpd   S1,T1,32,S1             ; A5 carry | cross product sum >> 32
-    add     S5,S7,T3                ; A7 cross product sum
-
-    ldd     XF4,S12                 ; MR * A6R
-    ldd     XF11,S11                ; ML * A5L
-    add,dc  zero,zero,S5            ; A7 cross product sum carry
-    depd,z  T3,31,32,S7             ; A7 cross product sum << 32
-
-    ldd     XF6,S14                 ; ML * A6L
-    ldd     XF13,S13                ; MR * A7R
-    shrpd   S5,T3,32,S5             ; A7 carry | cross product sum >> 32
-    add     S2,S8,S8                ; M * A4 right doubleword
-
-
-    ldd     XF15,S15                ; ML * A7L
-    add,dc  S0,S10,S10              ; M * A4 left doubleword
-    add     S3,S9,S9                ; M * A5 right doubleword
-
-    add,dc  S1,S11,S11              ; M * A5 left doubleword
-    add     S6,S12,S12              ; M * A6 right doubleword
-
-    ldd     32(pR),S0               ; Addend word 4
-    ldd     40(pR),S1               ; Addend word 5
-    add,dc  S4,S14,S14              ; M * A6 left doubleword
-    add     S7,S13,S13              ; M * A7 right doubleword
-
-    ldd     48(pR),S2               ; Addend word 6
-    ldd     56(pR),S3               ; Addend word 7
-    add,dc  S5,S15,S15              ; M * A7 left doubleword
-    add     S8,T2,S8                ; P4 doubleword
-
-    ldd     64(pR),S4               ; Addend word 8
-    ldd     SV5,s5                  ; restore s5
-    add,dc  S10,S9,S9               ; P5 doubleword
-    add,dc  S11,S12,S12             ; P6 doubleword
-
-
-    ldd     SV6,s6                  ; restore s6
-    ldd     SV7,s7                  ; restore s7
-    add,dc  S14,S13,S13             ; P7 doubleword
-    add,dc  zero,S15,S15            ; P8 doubleword
-
-    add     S0,S8,S8                ; new R4 doubleword
-
-    ldd     SV0,s0                  ; restore s0
-    std     S8,32(pR)               ; save R4
-    add,dc  S1,S9,S9                ; new R5 doubleword
-
-    ldd     SV1,s1                  ; restore s1
-    std     S9,40(pR)               ; save R5
-    add,dc  S2,S12,S12              ; new R6 doubleword
-
-    ldd     SV2,s2                  ; restore s2
-    std     S12,48(pR)              ; save R6
-    add,dc  S3,S13,S13              ; new R7 doubleword
-
-    ldd     SV3,s3                  ; restore s3
-    std     S13,56(pR)              ; save R7
-    add,dc  S4,S15,S15              ; new R8 doubleword
-
-    ldd     SV4,s4                  ; restore s4
-    std     S15,64(pR)              ; save result[8]
-    add,dc  zero,zero,v0            ; return carry from R8
-
-    CMPIB,*= 0,v0,$L0               ; if no overflow, exit
-    LDO     8(pR),pR
-
-$FINAL1                             ; Final carry propagation
-    LDD     64(pR),v0
-    LDO     8(pR),pR
-    ADDI    1,v0,v0
-    CMPIB,*= 0,v0,$FINAL1           ; Keep looping if there is a carry.
-    STD     v0,56(pR)
-$L0
-    bv      zero(rp)                ; -> caller
-    ldo     -ST_SZ(sp),sp           ; pop stack
-
-/*  ======================================================================  */
-/*      end of module                                                       */
-/*  ======================================================================  */
-
-                .LEAVE
-
-                .PROCEND
-                .SPACE         $TEXT$
-                .SUBSPA        $CODE$
-                .EXPORT        multacc512,ENTRY
-
-        .end

mozilla/security/nss/lib/freebl/mpi/hppa20.s

@@ -1,929 +0,0 @@
-; The contents of this file are subject to the Mozilla Public
-; License Version 1.1 (the "License"); you may not use this file
-; except in compliance with the License. You may obtain a copy of
-; the License at http://www.mozilla.org/MPL/
-; 
-; Software distributed under the License is distributed on an "AS
-; IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
-; implied. See the License for the specific language governing
-; rights and limitations under the License.
-; 
-; The Original Code is MAXPY multiple-precision integer arithmetic.
-; 
-; The Initial Developer of the Original Code is the Hewlett-Packard Company.
-; Portions created by Hewlett-Packard Company are 
-; Copyright (C) 1997 Hewlett-Packard Company.  All Rights Reserved.
-; 
-; Contributor(s):
-;  coded by:   William B. Ackerman
-; 
-; Alternatively, the contents of this file may be used under the
-; terms of the GNU General Public License Version 2 or later (the
-; "GPL"), in which case the provisions of the GPL are applicable 
-; instead of those above.  If you wish to allow use of your 
-; version of this file only under the terms of the GPL and not to
-; allow others to use your version of this file under the MPL,
-; indicate your decision by deleting the provisions above and
-; replace them with the notice and other provisions required by
-; the GPL.  If you do not delete the provisions above, a recipient
-; may use your version of this file under either the MPL or the
-; GPL.
-
-#ifdef __LP64__
-        .LEVEL   2.0W
-#else
-;       .LEVEL   1.1
-;       .ALLOW   2.0N
-        .LEVEL   2.0N
-#endif
-        .SPACE   $TEXT$,SORT=8
-        .SUBSPA  $CODE$,QUAD=0,ALIGN=4,ACCESS=0x2c,CODE_ONLY,SORT=24
-
-; ***************************************************************
-;
-;                 maxpy_[little/big]
-;
-; ***************************************************************
-
-; There is no default -- you must specify one or the other.
-#define LITTLE_WORDIAN 1
-
-#ifdef LITTLE_WORDIAN
-#define EIGHT 8
-#define SIXTEEN 16
-#define THIRTY_TWO 32
-#define UN_EIGHT -8
-#define UN_SIXTEEN -16
-#define UN_TWENTY_FOUR -24
-#endif
-
-#ifdef BIG_WORDIAN
-#define EIGHT -8
-#define SIXTEEN -16
-#define THIRTY_TWO -32
-#define UN_EIGHT 8
-#define UN_SIXTEEN 16
-#define UN_TWENTY_FOUR 24
-#endif
-
-; This performs a multiple-precision integer version of "daxpy",
-; Using the selected addressing direction.  "Little-wordian" means that
-; the least significant word of a number is stored at the lowest address.
-; "Big-wordian" means that the most significant word is at the lowest
-; address.  Either way, the incoming address of the vector is that
-; of the least significant word.  That means that, for little-wordian
-; addressing, we move the address upward as we propagate carries
-; from the least significant word to the most significant.  For
-; big-wordian we move the address downward.
-
-; We use the following registers:
-;
-;     r2   return PC, of course
-;     r26 = arg1 =  length
-;     r25 = arg2 =  address of scalar
-;     r24 = arg3 =  multiplicand vector
-;     r23 = arg4 =  result vector
-;
-;     fr9 = scalar loaded once only from r25
-
-; The cycle counts shown in the bodies below are simply the result of a
-; scheduling by hand.  The actual PCX-U hardware does it differently.
-; The intention is that the overall speed is the same.
-
-; The pipeline startup and shutdown code is constructed in the usual way,
-; by taking the loop bodies and removing unnecessary instructions.
-; We have left the comments describing cycle numbers in the code.
-; These are intended for reference when comparing with the main loop,
-; and have no particular relationship to actual cycle numbers.
-
-#ifdef LITTLE_WORDIAN
-maxpy_little
-#else
-maxpy_big
-#endif
-        .PROC
-        .CALLINFO FRAME=120,ENTRY_GR=%r4
-        .ENTER
-
-; Of course, real men don't use the sissy "enter" and "leave" commands.
-; They write their own stack manipulation stuff.  Unfortunately,
-; that doesn't generate complete unwind info, whereas "enter" and
-; "leave" (if the documentation is to be believed) do so.  Therefore,
-; we use the sissy commands.  We have verified (by real-man methods)
-; that the above command generates what we want:
-;       STW,MA  %r3,128(%sp)
-;       STW     %r4,-124(%sp)
-
-        ADDIB,< -1,%r26,$L0         ; If N = 0, exit immediately.
-        FLDD    0(%r25),%fr9        ; fr9 = scalar
-
-; First startup
-
-        FLDD    0(%r24),%fr24       ; Cycle 1
-        XMPYU   %fr9R,%fr24R,%fr27  ; Cycle 3
-        XMPYU   %fr9R,%fr24L,%fr25  ; Cycle 4
-        XMPYU   %fr9L,%fr24L,%fr26  ; Cycle 5
-        CMPIB,> 3,%r26,$N_IS_SMALL  ; Pick out cases N = 1, 2, or 3
-        XMPYU   %fr9L,%fr24R,%fr24  ; Cycle 6
-        FLDD    EIGHT(%r24),%fr28   ; Cycle 8
-        XMPYU   %fr9L,%fr28R,%fr31  ; Cycle 10
-        FSTD    %fr24,-96(%sp)
-        XMPYU   %fr9R,%fr28L,%fr30  ; Cycle 11
-        FSTD    %fr25,-80(%sp)
-        LDO     SIXTEEN(%r24),%r24  ; Cycle 12
-        FSTD    %fr31,-64(%sp)
-        XMPYU   %fr9R,%fr28R,%fr29  ; Cycle 13
-        FSTD    %fr27,-48(%sp)
-
-; Second startup
-
-        XMPYU   %fr9L,%fr28L,%fr28  ; Cycle 1
-        FSTD    %fr30,-56(%sp)
-        FLDD    0(%r24),%fr24
-
-        FSTD    %fr26,-88(%sp)      ; Cycle 2
-
-        XMPYU   %fr9R,%fr24R,%fr27  ; Cycle 3
-        FSTD    %fr28,-104(%sp)
-
-        XMPYU   %fr9R,%fr24L,%fr25  ; Cycle 4
-        LDD     -96(%sp),%r3
-        FSTD    %fr29,-72(%sp)
-
-        XMPYU   %fr9L,%fr24L,%fr26  ; Cycle 5
-        LDD     -64(%sp),%r19
-        LDD     -80(%sp),%r21
-
-        XMPYU   %fr9L,%fr24R,%fr24  ; Cycle 6
-        LDD     -56(%sp),%r20
-        ADD     %r21,%r3,%r3
-
-        ADD,DC  %r20,%r19,%r19      ; Cycle 7
-        LDD     -88(%sp),%r4
-        SHRPD   %r3,%r0,32,%r21
-        LDD     -48(%sp),%r1
-
-        FLDD    EIGHT(%r24),%fr28   ; Cycle 8
-        LDD     -104(%sp),%r31
-        ADD,DC  %r0,%r0,%r20
-        SHRPD   %r19,%r3,32,%r3
-
-        LDD     -72(%sp),%r29       ; Cycle 9
-        SHRPD   %r20,%r19,32,%r20
-        ADD     %r21,%r1,%r1
-
-        XMPYU   %fr9L,%fr28R,%fr31  ; Cycle 10
-        ADD,DC  %r3,%r4,%r4
-        FSTD    %fr24,-96(%sp)
-
-        XMPYU   %fr9R,%fr28L,%fr30  ; Cycle 11
-        ADD,DC  %r0,%r20,%r20
-        LDD     0(%r23),%r3
-        FSTD    %fr25,-80(%sp)
-
-        LDO     SIXTEEN(%r24),%r24  ; Cycle 12
-        FSTD    %fr31,-64(%sp)
-
-        XMPYU   %fr9R,%fr28R,%fr29  ; Cycle 13
-        ADD     %r0,%r0,%r0         ; clear the carry bit
-        ADDIB,<= -4,%r26,$ENDLOOP   ; actually happens in cycle 12
-        FSTD    %fr27,-48(%sp)
-;        MFCTL   %cr16,%r21         ; for timing
-;        STD     %r21,-112(%sp)
-
-; Here is the loop.
-
-$LOOP   XMPYU   %fr9L,%fr28L,%fr28  ; Cycle 1
-        ADD,DC  %r29,%r4,%r4
-        FSTD    %fr30,-56(%sp)
-        FLDD    0(%r24),%fr24
-
-        LDO     SIXTEEN(%r23),%r23  ; Cycle 2
-        ADD,DC  %r0,%r20,%r20
-        FSTD    %fr26,-88(%sp)
-
-        XMPYU   %fr9R,%fr24R,%fr27  ; Cycle 3
-        ADD     %r3,%r1,%r1
-        FSTD    %fr28,-104(%sp)
-        LDD     UN_EIGHT(%r23),%r21
-
-        XMPYU   %fr9R,%fr24L,%fr25  ; Cycle 4
-        ADD,DC  %r21,%r4,%r28
-        FSTD    %fr29,-72(%sp)    
-        LDD     -96(%sp),%r3
-
-        XMPYU   %fr9L,%fr24L,%fr26  ; Cycle 5
-        ADD,DC  %r20,%r31,%r22
-        LDD     -64(%sp),%r19
-        LDD     -80(%sp),%r21
-
-        XMPYU   %fr9L,%fr24R,%fr24  ; Cycle 6
-        ADD     %r21,%r3,%r3
-        LDD     -56(%sp),%r20
-        STD     %r1,UN_SIXTEEN(%r23)
-
-        ADD,DC  %r20,%r19,%r19      ; Cycle 7
-        SHRPD   %r3,%r0,32,%r21
-        LDD     -88(%sp),%r4
-        LDD     -48(%sp),%r1
-
-        ADD,DC  %r0,%r0,%r20        ; Cycle 8
-        SHRPD   %r19,%r3,32,%r3
-        FLDD    EIGHT(%r24),%fr28
-        LDD     -104(%sp),%r31
-
-        SHRPD   %r20,%r19,32,%r20   ; Cycle 9
-        ADD     %r21,%r1,%r1
-        STD     %r28,UN_EIGHT(%r23)
-        LDD     -72(%sp),%r29
-
-        XMPYU   %fr9L,%fr28R,%fr31  ; Cycle 10
-        ADD,DC  %r3,%r4,%r4
-        FSTD    %fr24,-96(%sp)
-
-        XMPYU   %fr9R,%fr28L,%fr30  ; Cycle 11
-        ADD,DC  %r0,%r20,%r20
-        FSTD    %fr25,-80(%sp)
-        LDD     0(%r23),%r3
-
-        LDO     SIXTEEN(%r24),%r24  ; Cycle 12
-        FSTD    %fr31,-64(%sp)
-
-        XMPYU   %fr9R,%fr28R,%fr29  ; Cycle 13
-        ADD     %r22,%r1,%r1
-        ADDIB,> -2,%r26,$LOOP       ; actually happens in cycle 12
-        FSTD    %fr27,-48(%sp)
-
-$ENDLOOP
-
-; Shutdown code, first stage.
-
-;        MFCTL   %cr16,%r21         ; for timing
-;        STD     %r21,UN_SIXTEEN(%r23)
-;        LDD     -112(%sp),%r21
-;        STD     %r21,UN_EIGHT(%r23)
-
-        XMPYU   %fr9L,%fr28L,%fr28  ; Cycle 1
-        ADD,DC  %r29,%r4,%r4
-        CMPIB,= 0,%r26,$ONEMORE
-        FSTD    %fr30,-56(%sp)
-
-        LDO     SIXTEEN(%r23),%r23  ; Cycle 2
-        ADD,DC  %r0,%r20,%r20
-        FSTD    %fr26,-88(%sp)
-
-        ADD     %r3,%r1,%r1         ; Cycle 3
-        FSTD    %fr28,-104(%sp)
-        LDD     UN_EIGHT(%r23),%r21
-
-        ADD,DC  %r21,%r4,%r28       ; Cycle 4
-        FSTD    %fr29,-72(%sp)    
-        STD     %r28,UN_EIGHT(%r23) ; moved up from cycle 9
-        LDD     -96(%sp),%r3
-
-        ADD,DC  %r20,%r31,%r22      ; Cycle 5
-        STD     %r1,UN_SIXTEEN(%r23)
-$JOIN4
-        LDD     -64(%sp),%r19
-        LDD     -80(%sp),%r21
-
-        ADD     %r21,%r3,%r3        ; Cycle 6
-        LDD     -56(%sp),%r20
-
-        ADD,DC  %r20,%r19,%r19      ; Cycle 7
-        SHRPD   %r3,%r0,32,%r21
-        LDD     -88(%sp),%r4
-        LDD     -48(%sp),%r1
-
-        ADD,DC  %r0,%r0,%r20        ; Cycle 8
-        SHRPD   %r19,%r3,32,%r3
-        LDD     -104(%sp),%r31
-
-        SHRPD   %r20,%r19,32,%r20   ; Cycle 9
-        ADD     %r21,%r1,%r1
-        LDD     -72(%sp),%r29
-
-        ADD,DC  %r3,%r4,%r4         ; Cycle 10
-
-        ADD,DC  %r0,%r20,%r20       ; Cycle 11
-        LDD     0(%r23),%r3
-
-        ADD     %r22,%r1,%r1        ; Cycle 13
-
-; Shutdown code, second stage.
-
-        ADD,DC  %r29,%r4,%r4        ; Cycle 1
-
-        LDO     SIXTEEN(%r23),%r23  ; Cycle 2
-        ADD,DC  %r0,%r20,%r20
-
-        LDD     UN_EIGHT(%r23),%r21 ; Cycle 3
-        ADD     %r3,%r1,%r1
-
-        ADD,DC  %r21,%r4,%r28       ; Cycle 4
-
-        ADD,DC  %r20,%r31,%r22      ; Cycle 5
-
-        STD     %r1,UN_SIXTEEN(%r23); Cycle 6
-
-        STD     %r28,UN_EIGHT(%r23) ; Cycle 9
-
-        LDD     0(%r23),%r3         ; Cycle 11
-
-; Shutdown code, third stage.
-
-        LDO     SIXTEEN(%r23),%r23
-        ADD     %r3,%r22,%r1
-$JOIN1  ADD,DC  %r0,%r0,%r21
-        CMPIB,*= 0,%r21,$L0         ; if no overflow, exit
-        STD     %r1,UN_SIXTEEN(%r23)
-
-; Final carry propagation
-
-$FINAL1 LDO     EIGHT(%r23),%r23
-        LDD     UN_SIXTEEN(%r23),%r21
-        ADDI    1,%r21,%r21
-        CMPIB,*= 0,%r21,$FINAL1     ; Keep looping if there is a carry.
-        STD     %r21,UN_SIXTEEN(%r23)
-        B       $L0
-        NOP
-
-; Here is the code that handles the difficult cases N=1, N=2, and N=3.
-; We do the usual trick -- branch out of the startup code at appropriate
-; points, and branch into the shutdown code.
-
-$N_IS_SMALL
-        CMPIB,= 0,%r26,$N_IS_ONE
-        FSTD    %fr24,-96(%sp)      ; Cycle 10
-        FLDD    EIGHT(%r24),%fr28   ; Cycle 8
-        XMPYU   %fr9L,%fr28R,%fr31  ; Cycle 10
-        XMPYU   %fr9R,%fr28L,%fr30  ; Cycle 11
-        FSTD    %fr25,-80(%sp)
-        FSTD    %fr31,-64(%sp)      ; Cycle 12
-        XMPYU   %fr9R,%fr28R,%fr29  ; Cycle 13
-        FSTD    %fr27,-48(%sp)
-        XMPYU   %fr9L,%fr28L,%fr28  ; Cycle 1
-        CMPIB,= 2,%r26,$N_IS_THREE
-        FSTD    %fr30,-56(%sp)
-
-; N = 2
-        FSTD    %fr26,-88(%sp)      ; Cycle 2
-        FSTD    %fr28,-104(%sp)     ; Cycle 3
-        LDD     -96(%sp),%r3        ; Cycle 4
-        FSTD    %fr29,-72(%sp)
-        B       $JOIN4
-        ADD     %r0,%r0,%r22
-
-$N_IS_THREE
-        FLDD    SIXTEEN(%r24),%fr24
-        FSTD    %fr26,-88(%sp)      ; Cycle 2
-        XMPYU   %fr9R,%fr24R,%fr27  ; Cycle 3
-        FSTD    %fr28,-104(%sp)
-        XMPYU   %fr9R,%fr24L,%fr25  ; Cycle 4
-        LDD     -96(%sp),%r3
-        FSTD    %fr29,-72(%sp)
-        XMPYU   %fr9L,%fr24L,%fr26  ; Cycle 5
-        LDD     -64(%sp),%r19
-        LDD     -80(%sp),%r21
-        B       $JOIN3
-        ADD     %r0,%r0,%r22
-
-$N_IS_ONE
-        FSTD    %fr25,-80(%sp)
-        FSTD    %fr27,-48(%sp)
-        FSTD    %fr26,-88(%sp)      ; Cycle 2
-        B       $JOIN5
-        ADD     %r0,%r0,%r22
-
-; We came out of the unrolled loop with wrong parity.  Do one more
-; single cycle.  This is quite tricky, because of the way the
-; carry chains and SHRPD chains have been chopped up.
-
-$ONEMORE
-
-        FLDD    0(%r24),%fr24
-
-        LDO     SIXTEEN(%r23),%r23  ; Cycle 2
-        ADD,DC  %r0,%r20,%r20
-        FSTD    %fr26,-88(%sp)
-
-        XMPYU   %fr9R,%fr24R,%fr27  ; Cycle 3
-        FSTD    %fr28,-104(%sp)
-        LDD     UN_EIGHT(%r23),%r21
-        ADD     %r3,%r1,%r1
-
-        XMPYU   %fr9R,%fr24L,%fr25  ; Cycle 4
-        ADD,DC  %r21,%r4,%r28
-        STD     %r28,UN_EIGHT(%r23) ; moved from cycle 9
-        LDD     -96(%sp),%r3
-        FSTD    %fr29,-72(%sp)    
-
-        XMPYU   %fr9L,%fr24L,%fr26  ; Cycle 5
-        ADD,DC  %r20,%r31,%r22
-        LDD     -64(%sp),%r19
-        LDD     -80(%sp),%r21
-
-        STD     %r1,UN_SIXTEEN(%r23); Cycle 6
-$JOIN3
-        XMPYU   %fr9L,%fr24R,%fr24
-        LDD     -56(%sp),%r20
-        ADD     %r21,%r3,%r3
-
-        ADD,DC  %r20,%r19,%r19      ; Cycle 7
-        LDD     -88(%sp),%r4
-        SHRPD   %r3,%r0,32,%r21
-        LDD     -48(%sp),%r1
-
-        LDD     -104(%sp),%r31      ; Cycle 8
-        ADD,DC  %r0,%r0,%r20
-        SHRPD   %r19,%r3,32,%r3
-
-        LDD     -72(%sp),%r29       ; Cycle 9
-        SHRPD   %r20,%r19,32,%r20
-        ADD     %r21,%r1,%r1
-
-        ADD,DC  %r3,%r4,%r4         ; Cycle 10
-        FSTD    %fr24,-96(%sp)
-
-        ADD,DC  %r0,%r20,%r20       ; Cycle 11
-        LDD     0(%r23),%r3
-        FSTD    %fr25,-80(%sp)
-
-        ADD     %r22,%r1,%r1        ; Cycle 13
-        FSTD    %fr27,-48(%sp)
-
-; Shutdown code, stage 1-1/2.
-
-        ADD,DC  %r29,%r4,%r4        ; Cycle 1
-
-        LDO     SIXTEEN(%r23),%r23  ; Cycle 2
-        ADD,DC  %r0,%r20,%r20     
-        FSTD    %fr26,-88(%sp)
-
-        LDD     UN_EIGHT(%r23),%r21 ; Cycle 3
-        ADD     %r3,%r1,%r1
-
-        ADD,DC  %r21,%r4,%r28       ; Cycle 4
-        STD     %r28,UN_EIGHT(%r23) ; moved from cycle 9
-
-        ADD,DC  %r20,%r31,%r22      ; Cycle 5
-        STD     %r1,UN_SIXTEEN(%r23)
-$JOIN5
-        LDD     -96(%sp),%r3        ; moved from cycle 4
-        LDD     -80(%sp),%r21
-        ADD     %r21,%r3,%r3        ; Cycle 6
-        ADD,DC  %r0,%r0,%r19        ; Cycle 7
-        LDD     -88(%sp),%r4
-        SHRPD   %r3,%r0,32,%r21
-        LDD     -48(%sp),%r1
-        SHRPD   %r19,%r3,32,%r3     ; Cycle 8
-        ADD     %r21,%r1,%r1        ; Cycle 9
-        ADD,DC  %r3,%r4,%r4         ; Cycle 10
-        LDD     0(%r23),%r3         ; Cycle 11
-        ADD     %r22,%r1,%r1        ; Cycle 13
-
-; Shutdown code, stage 2-1/2.
-
-        ADD,DC  %r0,%r4,%r4         ; Cycle 1
-        LDO     SIXTEEN(%r23),%r23  ; Cycle 2
-        LDD     UN_EIGHT(%r23),%r21 ; Cycle 3
-        ADD     %r3,%r1,%r1
-        STD     %r1,UN_SIXTEEN(%r23)
-        ADD,DC  %r21,%r4,%r1
-        B       $JOIN1
-        LDO     EIGHT(%r23),%r23
-
-; exit
-
-$L0
-        .LEAVE
-
-; We have verified that the above command generates what we want:
-;       LDW     -124(%sp),%r4
-;       BVE     (%r2)
-;       LDW,MB  -128(%sp),%r3
-
-        .PROCEND
-
-; ***************************************************************
-;
-;                 add_diag_[little/big]
-;
-; ***************************************************************
-
-; The arguments are as follows:
-;     r2   return PC, of course
-;     r26 = arg1 =  length
-;     r25 = arg2 =  vector to square
-;     r24 = arg3 =  result vector
-
-#ifdef LITTLE_WORDIAN
-add_diag_little
-#else
-add_diag_big
-#endif
-        .PROC
-        .CALLINFO FRAME=120,ENTRY_GR=%r4
-        .ENTER
-
-        ADDIB,< -1,%r26,$Z0         ; If N=0, exit immediately.
-        NOP
-
-; Startup code
-
-        FLDD    0(%r25),%fr7        ; Cycle 2 (alternate body)
-        XMPYU   %fr7R,%fr7R,%fr29   ; Cycle 4
-        XMPYU   %fr7L,%fr7R,%fr27   ; Cycle 5
-        XMPYU   %fr7L,%fr7L,%fr30
-        LDO     SIXTEEN(%r25),%r25  ; Cycle 6
-        FSTD    %fr29,-88(%sp)
-        FSTD    %fr27,-72(%sp)      ; Cycle 7
-        CMPIB,= 0,%r26,$DIAG_N_IS_ONE ; Cycle 1 (main body)
-        FSTD    %fr30,-96(%sp)
-        FLDD    UN_EIGHT(%r25),%fr7 ; Cycle 2
-        LDD     -88(%sp),%r22       ; Cycle 3
-        LDD     -72(%sp),%r31       ; Cycle 4
-        XMPYU   %fr7R,%fr7R,%fr28
-        XMPYU   %fr7L,%fr7R,%fr24   ; Cycle 5
-        XMPYU   %fr7L,%fr7L,%fr31
-        LDD     -96(%sp),%r20       ; Cycle 6
-        FSTD    %fr28,-80(%sp)
-        ADD     %r0,%r0,%r0         ; clear the carry bit
-        ADDIB,<= -2,%r26,$ENDDIAGLOOP ; Cycle 7
-        FSTD    %fr24,-64(%sp)
-
-; Here is the loop.  It is unrolled twice, modelled after the "alternate body" and then the "main body".
-
-$DIAGLOOP
-        SHRPD   %r31,%r0,31,%r3     ; Cycle 1 (alternate body)
-        LDO     SIXTEEN(%r25),%r25
-        LDD     0(%r24),%r1
-        FSTD    %fr31,-104(%sp)
-        SHRPD   %r0,%r31,31,%r4     ; Cycle 2
-        ADD,DC  %r22,%r3,%r3
-        FLDD    UN_SIXTEEN(%r25),%fr7   
-        ADD,DC  %r0,%r20,%r20       ; Cycle 3
-        ADD     %r1,%r3,%r3
-        XMPYU   %fr7R,%fr7R,%fr29   ; Cycle 4
-        LDD     -80(%sp),%r21
-        STD     %r3,0(%r24)
-        XMPYU   %fr7L,%fr7R,%fr27   ; Cycle 5
-        XMPYU   %fr7L,%fr7L,%fr30
-        LDD     -64(%sp),%r29       
-        LDD     EIGHT(%r24),%r1  
-        ADD,DC  %r4,%r20,%r20       ; Cycle 6
-        LDD     -104(%sp),%r19
-        FSTD    %fr29,-88(%sp)
-        ADD     %r20,%r1,%r1        ; Cycle 7
-        FSTD    %fr27,-72(%sp)
-        SHRPD   %r29,%r0,31,%r4     ; Cycle 1 (main body)
-        LDO     THIRTY_TWO(%r24),%r24
-        LDD     UN_SIXTEEN(%r24),%r28
-        FSTD    %fr30,-96(%sp)
-        SHRPD   %r0,%r29,31,%r3     ; Cycle 2
-        ADD,DC  %r21,%r4,%r4
-        FLDD    UN_EIGHT(%r25),%fr7
-        STD     %r1,UN_TWENTY_FOUR(%r24)
-        ADD,DC  %r0,%r19,%r19       ; Cycle 3
-        ADD     %r28,%r4,%r4
-        XMPYU   %fr7R,%fr7R,%fr28   ; Cycle 4
-        LDD     -88(%sp),%r22
-        STD     %r4,UN_SIXTEEN(%r24)
-        XMPYU   %fr7L,%fr7R,%fr24   ; Cycle 5
-        XMPYU   %fr7L,%fr7L,%fr31
-        LDD     -72(%sp),%r31
-        LDD     UN_EIGHT(%r24),%r28
-        ADD,DC  %r3,%r19,%r19       ; Cycle 6
-        LDD     -96(%sp),%r20
-        FSTD    %fr28,-80(%sp)
-        ADD     %r19,%r28,%r28      ; Cycle 7
-        FSTD    %fr24,-64(%sp)
-        ADDIB,> -2,%r26,$DIAGLOOP   ; Cycle 8
-        STD     %r28,UN_EIGHT(%r24)
-
-$ENDDIAGLOOP
-
-        ADD,DC  %r0,%r22,%r22    
-        CMPIB,= 0,%r26,$ONEMOREDIAG
-        SHRPD   %r31,%r0,31,%r3
-
-; Shutdown code, first stage.
-
-        FSTD    %fr31,-104(%sp)     ; Cycle 1 (alternate body)
-        LDD     0(%r24),%r28
-        SHRPD   %r0,%r31,31,%r4     ; Cycle 2
-        ADD     %r3,%r22,%r3
-        ADD,DC  %r0,%r20,%r20       ; Cycle 3
-        LDD     -80(%sp),%r21
-        ADD     %r3,%r28,%r3
-        LDD     -64(%sp),%r29       ; Cycle 4
-        STD     %r3,0(%r24)
-        LDD     EIGHT(%r24),%r1     ; Cycle 5
-        LDO     SIXTEEN(%r25),%r25  ; Cycle 6
-        LDD     -104(%sp),%r19
-        ADD,DC  %r4,%r20,%r20
-        ADD     %r20,%r1,%r1        ; Cycle 7
-        ADD,DC  %r0,%r21,%r21       ; Cycle 8
-        STD     %r1,EIGHT(%r24)
-
-; Shutdown code, second stage.
-
-        SHRPD   %r29,%r0,31,%r4     ; Cycle 1 (main body)
-        LDO     THIRTY_TWO(%r24),%r24
-        LDD     UN_SIXTEEN(%r24),%r1
-        SHRPD   %r0,%r29,31,%r3      ; Cycle 2
-        ADD     %r4,%r21,%r4
-        ADD,DC  %r0,%r19,%r19       ; Cycle 3
-        ADD     %r4,%r1,%r4
-        STD     %r4,UN_SIXTEEN(%r24); Cycle 4
-        LDD     UN_EIGHT(%r24),%r28 ; Cycle 5
-        ADD,DC  %r3,%r19,%r19       ; Cycle 6       
-        ADD     %r19,%r28,%r28      ; Cycle 7
-        ADD,DC  %r0,%r0,%r22        ; Cycle 8
-        CMPIB,*= 0,%r22,$Z0         ; if no overflow, exit
-        STD     %r28,UN_EIGHT(%r24)
-
-; Final carry propagation
-
-$FDIAG2
-        LDO     EIGHT(%r24),%r24
-        LDD     UN_EIGHT(%r24),%r26
-        ADDI    1,%r26,%r26
-        CMPIB,*= 0,%r26,$FDIAG2     ; Keep looping if there is a carry.
-        STD     %r26,UN_EIGHT(%r24)
-
-        B   $Z0
-        NOP
-
-; Here is the code that handles the difficult case N=1.
-; We do the usual trick -- branch out of the startup code at appropriate
-; points, and branch into the shutdown code.
-
-$DIAG_N_IS_ONE
-
-        LDD     -88(%sp),%r22
-        LDD     -72(%sp),%r31
-        B       $JOINDIAG
-        LDD     -96(%sp),%r20
-
-; We came out of the unrolled loop with wrong parity.  Do one more
-; single cycle.  This is the "alternate body".  It will, of course,
-; give us opposite registers from the other case, so we need
-; completely different shutdown code.
-
-$ONEMOREDIAG
-        FSTD    %fr31,-104(%sp)     ; Cycle 1 (alternate body)
-        LDD     0(%r24),%r28
-        FLDD    0(%r25),%fr7        ; Cycle 2
-        SHRPD   %r0,%r31,31,%r4
-        ADD     %r3,%r22,%r3
-        ADD,DC  %r0,%r20,%r20       ; Cycle 3
-        LDD     -80(%sp),%r21
-        ADD     %r3,%r28,%r3
-        LDD     -64(%sp),%r29       ; Cycle 4
-        STD     %r3,0(%r24)
-        XMPYU   %fr7R,%fr7R,%fr29
-        LDD     EIGHT(%r24),%r1     ; Cycle 5
-        XMPYU   %fr7L,%fr7R,%fr27
-        XMPYU   %fr7L,%fr7L,%fr30
-        LDD     -104(%sp),%r19      ; Cycle 6
-        FSTD    %fr29,-88(%sp)
-        ADD,DC  %r4,%r20,%r20
-        FSTD    %fr27,-72(%sp)      ; Cycle 7
-        ADD     %r20,%r1,%r1
-        ADD,DC  %r0,%r21,%r21       ; Cycle 8
-        STD     %r1,EIGHT(%r24)
-
-; Shutdown code, first stage.
-
-        SHRPD   %r29,%r0,31,%r4     ; Cycle 1 (main body)
-        LDO     THIRTY_TWO(%r24),%r24
-        FSTD    %fr30,-96(%sp)
-        LDD     UN_SIXTEEN(%r24),%r1
-        SHRPD   %r0,%r29,31,%r3     ; Cycle 2
-        ADD     %r4,%r21,%r4
-        ADD,DC  %r0,%r19,%r19       ; Cycle 3
-        LDD     -88(%sp),%r22
-        ADD     %r4,%r1,%r4
-        LDD     -72(%sp),%r31       ; Cycle 4
-        STD     %r4,UN_SIXTEEN(%r24)
-        LDD     UN_EIGHT(%r24),%r28 ; Cycle 5
-        LDD     -96(%sp),%r20       ; Cycle 6
-        ADD,DC  %r3,%r19,%r19
-        ADD     %r19,%r28,%r28      ; Cycle 7
-        ADD,DC  %r0,%r22,%r22       ; Cycle 8
-        STD     %r28,UN_EIGHT(%r24)
-
-; Shutdown code, second stage.
-
-$JOINDIAG
-        SHRPD   %r31,%r0,31,%r3     ; Cycle 1 (alternate body)
-        LDD     0(%r24),%r28        
-        SHRPD   %r0,%r31,31,%r4     ; Cycle 2
-        ADD     %r3,%r22,%r3
-        ADD,DC  %r0,%r20,%r20       ; Cycle 3
-        ADD     %r3,%r28,%r3
-        STD     %r3,0(%r24)         ; Cycle 4
-        LDD     EIGHT(%r24),%r1     ; Cycle 5
-        ADD,DC  %r4,%r20,%r20
-        ADD     %r20,%r1,%r1        ; Cycle 7
-        ADD,DC  %r0,%r0,%r21        ; Cycle 8
-        CMPIB,*= 0,%r21,$Z0         ; if no overflow, exit
-        STD     %r1,EIGHT(%r24)
-
-; Final carry propagation
-
-$FDIAG1
-        LDO     EIGHT(%r24),%r24
-        LDD     EIGHT(%r24),%r26
-        ADDI    1,%r26,%r26
-        CMPIB,*= 0,%r26,$FDIAG1    ; Keep looping if there is a carry.
-        STD     %r26,EIGHT(%r24)
-
-$Z0
-        .LEAVE
-        .PROCEND
-;	.ALLOW
-
-        .SPACE         $TEXT$
-        .SUBSPA        $CODE$
-#ifdef LITTLE_WORDIAN
-        .EXPORT        maxpy_little,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,LONG_RETURN
-        .EXPORT        add_diag_little,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,LONG_RETURN
-#else
-        .EXPORT        maxpy_big,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,LONG_RETURN
-        .EXPORT        add_diag_big,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,LONG_RETURN
-#endif
-        .END
-
-
-; How to use "maxpy_PA20_little" and "maxpy_PA20_big"
-; 
-; The routine "maxpy_PA20_little" or "maxpy_PA20_big"
-; performs a 64-bit x any-size multiply, and adds the
-; result to an area of memory.  That is, it performs
-; something like
-; 
-;      A B C D
-;    *       Z
-;   __________
-;    P Q R S T
-; 
-; and then adds the "PQRST" vector into an area of memory,
-; handling all carries.
-; 
-; Digression on nomenclature and endian-ness:
-; 
-; Each of the capital letters in the above represents a 64-bit
-; quantity.  That is, you could think of the discussion as
-; being in terms of radix-16-quintillion arithmetic.  The data
-; type being manipulated is "unsigned long long int".  This
-; requires the 64-bit extension of the HP-UX C compiler,
-; available at release 10.  You need these compiler flags to
-; enable these extensions:
-; 
-;       -Aa +e +DA2.0 +DS2.0
-; 
-; (The first specifies ANSI C, the second enables the
-; extensions, which are beyond ANSI C, and the third and
-; fourth tell the compiler to use whatever features of the
-; PA2.0 architecture it wishes, in order to made the code more
-; efficient.  Since the presence of the assembly code will
-; make the program unable to run on anything less than PA2.0,
-; you might as well gain the performance enhancements in the C
-; code as well.)
-; 
-; Questions of "endian-ness" often come up, usually in the
-; context of byte ordering in a word.  These routines have a
-; similar issue, that could be called "wordian-ness".
-; Independent of byte ordering (PA is always big-endian), one
-; can make two choices when representing extremely large
-; numbers as arrays of 64-bit doublewords in memory.
-; 
-; "Little-wordian" layout means that the least significant
-; word of a number is stored at the lowest address.
-; 
-;   MSW     LSW
-;    |       |
-;    V       V
-; 
-;    A B C D E
-; 
-;    ^     ^ ^
-;    |     | |____ address 0
-;    |     |
-;    |     |_______address 8
-;    |
-;    address 32
-; 
-; "Big-wordian" means that the most significant word is at the
-; lowest address.
-; 
-;   MSW     LSW
-;    |       |
-;    V       V
-; 
-;    A B C D E
-; 
-;    ^     ^ ^
-;    |     | |____ address 32
-;    |     |
-;    |     |_______address 24
-;    |
-;    address 0
-; 
-; When you compile the file, you must specify one or the other, with
-; a switch "-DLITTLE_WORDIAN" or "-DBIG_WORDIAN".
-; 
-;     Incidentally, you assemble this file as part of your
-;     project with the same C compiler as the rest of the program.
-;     My "makefile" for a superprecision arithmetic package has
-;     the following stuff:
-; 
-;     # definitions:
-;     CC = cc -Aa +e -z +DA2.0 +DS2.0 +w1
-;     CFLAGS = +O3
-;     LDFLAGS = -L /usr/lib -Wl,-aarchive
-; 
-;     # general build rule for ".s" files:
-;     .s.o:
-;             $(CC) $(CFLAGS) -c $< -DBIG_WORDIAN
-; 
-;     # Now any bind step that calls for pa20.o will assemble pa20.s
-; 
-; End of digression, back to arithmetic:
-; 
-; The way we multiply two huge numbers is, of course, to multiply
-; the "ABCD" vector by each of the "WXYZ" doublewords, adding
-; the result vectors with increasing offsets, the way we learned
-; in school, back before we all used calculators:
-; 
-;            A B C D
-;          * W X Y Z
-;         __________
-;          P Q R S T
-;        E F G H I
-;      M N O P Q
-;  + R S T U V
-;    _______________
-;    F I N A L S U M
-; 
-; So we call maxpy_PA20_big (in my case; my package is
-; big-wordian) repeatedly, giving the W, X, Y, and Z arguments
-; in turn as the "scalar", and giving the "ABCD" vector each
-; time.  We direct it to add its result into an area of memory
-; that we have cleared at the start.  We skew the exact
-; location into that area with each call.
-; 
-; The prototype for the function is
-; 
-; extern void maxpy_PA20_big(
-;    int length,        /* Number of doublewords in the multiplicand vector. */
-;    const long long int *scalaraddr,    /* Address to fetch the scalar. */
-;    const long long int *multiplicand,  /* The multiplicand vector. */
-;    long long int *result);             /* Where to accumulate the result. */
-; 
-; (You should place a copy of this prototype in an include file
-; or in your C file.)
-; 
-; Now, IN ALL CASES, the given address for the multiplicand or
-; the result is that of the LEAST SIGNIFICANT DOUBLEWORD.
-; That word is, of course, the word at which the routine
-; starts processing.  "maxpy_PA20_little" then increases the
-; addresses as it computes.  "maxpy_PA20_big" decreases them.
-; 
-; In our example above, "length" would be 4 in each case.
-; "multiplicand" would be the "ABCD" vector.  Specifically,
-; the address of the element "D".  "scalaraddr" would be the
-; address of "W", "X", "Y", or "Z" on the four calls that we
-; would make.  (The order doesn't matter, of course.)
-; "result" would be the appropriate address in the result
-; area.  When multiplying by "Z", that would be the least
-; significant word.  When multiplying by "Y", it would be the
-; next higher word (8 bytes higher if little-wordian; 8 bytes
-; lower if big-wordian), and so on.  The size of the result
-; area must be the the sum of the sizes of the multiplicand
-; and multiplier vectors, and must be initialized to zero
-; before we start.
-; 
-; Whenever the routine adds its partial product into the result
-; vector, it follows carry chains as far as they need to go.
-; 
-; Here is the super-precision multiply routine that I use for
-; my package.  The package is big-wordian.  I have taken out
-; handling of exponents (it's a floating point package):
-; 
-; static void mul_PA20(
-;   int size,
-;   const long long int *arg1,
-;   const long long int *arg2,
-;   long long int *result)
-; {
-;    int i;
-; 
-;    for (i=0 ; i<2*size ; i++) result[i] = 0ULL;
-; 
-;    for (i=0 ; i<size ; i++) {
-;       maxpy_PA20_big(size, &arg2[i], &arg1[size-1], &result[size+i]);
-;    }
-; }

mozilla/security/nss/lib/freebl/mpi/hppatch.adb

@@ -1,54 +0,0 @@
-#/bin/sh
-#
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is script to change the system id in an object file from PA-RISC 2.0 to 1.1.
-#
-# The Initial Developer of the Original Code is
-# Hewlett-Packard Company.
-# Portions created by the Initial Developer are Copyright (C) 1999
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#   wrapped by Dennis Handly on Tue Mar 23 15:23:43 1999
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-# script to change the system id in an object file from PA-RISC 2.0 to 1.1
-
-adb -w $1 << EOF
-?m 0 -1 0
-0x0?X
-0x0?W (@0x0&~0x40000)|(~@0x0&0x40000)
-
-0?"change checksum"
-0x7c?X
-0x7c?W (@0x7c&~0x40000)|(~@0x7c&0x40000)
-$q
-EOF
-
-exit 0
-

mozilla/security/nss/lib/freebl/mpi/logtab.h

@@ -1,62 +0,0 @@
-/*
- *  logtab.h
- *
- *  Arbitrary precision integer arithmetic library
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1998
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: logtab.h,v 1.5 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-const float s_logv_2[] = {
-   0.000000000f, 0.000000000f, 1.000000000f, 0.630929754f,  /*  0  1  2  3 */
-   0.500000000f, 0.430676558f, 0.386852807f, 0.356207187f,  /*  4  5  6  7 */
-   0.333333333f, 0.315464877f, 0.301029996f, 0.289064826f,  /*  8  9 10 11 */
-   0.278942946f, 0.270238154f, 0.262649535f, 0.255958025f,  /* 12 13 14 15 */
-   0.250000000f, 0.244650542f, 0.239812467f, 0.235408913f,  /* 16 17 18 19 */
-   0.231378213f, 0.227670249f, 0.224243824f, 0.221064729f,  /* 20 21 22 23 */
-   0.218104292f, 0.215338279f, 0.212746054f, 0.210309918f,  /* 24 25 26 27 */
-   0.208014598f, 0.205846832f, 0.203795047f, 0.201849087f,  /* 28 29 30 31 */
-   0.200000000f, 0.198239863f, 0.196561632f, 0.194959022f,  /* 32 33 34 35 */
-   0.193426404f, 0.191958720f, 0.190551412f, 0.189200360f,  /* 36 37 38 39 */
-   0.187901825f, 0.186652411f, 0.185449023f, 0.184288833f,  /* 40 41 42 43 */
-   0.183169251f, 0.182087900f, 0.181042597f, 0.180031327f,  /* 44 45 46 47 */
-   0.179052232f, 0.178103594f, 0.177183820f, 0.176291434f,  /* 48 49 50 51 */
-   0.175425064f, 0.174583430f, 0.173765343f, 0.172969690f,  /* 52 53 54 55 */
-   0.172195434f, 0.171441601f, 0.170707280f, 0.169991616f,  /* 56 57 58 59 */
-   0.169293808f, 0.168613099f, 0.167948779f, 0.167300179f,  /* 60 61 62 63 */
-   0.166666667f
-};
-

mozilla/security/nss/lib/freebl/mpi/make-logtab

@@ -1,64 +0,0 @@
-#!/usr/linguist/bin/perl
-
-#
-# make-logtab
-#
-# Generate a table of logarithms of 2 in various bases, for use in
-# estimating the output sizes of various bases.
-
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic
-# library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>
-# Portions created by the Initial Developer are Copyright (C) 1998, 2000
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-# $Id: make-logtab,v 1.4 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-$ARRAYNAME = $ENV{'ARRAYNAME'} || "s_logv_2";
-$ARRAYTYPE = $ENV{'ARRAYTYPE'} || "float";
-
-printf("const %s %s[] = {\n   %0.9ff, %0.9ff, ", 
-       $ARRAYTYPE, $ARRAYNAME, 0, 0);
-$brk = 2;
-for($ix = 2; $ix < 64; $ix++) {
-    printf("%0.9ff, ", (log(2)/log($ix)));
-    $brk = ($brk + 1) & 3;
-    if(!$brk) {
-	printf(" /* %2d %2d %2d %2d */\n   ",
-	       $ix - 3, $ix - 2, $ix - 1, $ix);
-    }
-}
-printf("%0.9ff\n};\n\n", (log(2)/log($ix)));
-
-exit 0;

mozilla/security/nss/lib/freebl/mpi/make-test-arrays

@@ -1,133 +0,0 @@
-#!/usr/linguist/bin/perl
-
-#
-# make-test-arrays
-#
-# Given a test-arrays file, which specifies the test suite names, the
-# names of the functions which perform those test suites, and
-# descriptive comments, this script generates C structures for the
-# mpi-test program.  The input consists of lines of the form:
-#
-# suite-name:function-name:comment
-#
-# The output is written to the standard output.  Blank lines are
-# ignored, and comments beginning with '#' are stripped.
-
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-#
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-#
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-#
-# The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
-#
-# The Initial Developer of the Original Code is
-# Michael J. Fromberger <sting@linguist.dartmouth.edu>.
-# Portions created by the Initial Developer are Copyright (C) 1998
-# the Initial Developer. All Rights Reserved.
-#
-# Contributor(s):
-#
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-#
-# ***** END LICENSE BLOCK *****
-
-# $Id: make-test-arrays,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-#
-
-# Read parameters from the environment, if available
-$NAMEVAR = $ENV{'NAMEVAR'} || "g_names";
-$COUNTVAR = $ENV{'COUNTVAR'} || "g_count";
-$FUNCVAR = $ENV{'FUNCVAR'} || "g_tests";
-$DESCVAR = $ENV{'DESCVAR'} || "g_descs";
-$FUNCLEN = 13;
-$NAMELEN = 18;
-$DESCLEN = 45;
-
-#------------------------------------------------------------------------
-# Suck in input from the files on the command line, or standard input
-while(<>) {
-    chomp;
-    s/\#.*$//;
-    next if /^\s*$/;
-
-    ($suite, $func, $desc) = split(/:/, $_);
-
-    $tmp = { "suite" => $suite,
-	     "func"  => $func,
-	     "desc"  => $desc };
-
-    push(@item, $tmp);
-}
-$count = scalar(@item);
-$last = pop(@item);
-
-#------------------------------------------------------------------------
-# Output the table of names
-print "/* Table mapping test suite names to index numbers */\n";
-printf("const int   %s = %d;\n", $COUNTVAR, $count);
-printf("const char *%s[] = {\n", $NAMEVAR);
-
-foreach $elt (@item) {
-    printf("   \"%s\",%s/* %s%s */\n", $elt->{"suite"},
-	   " " x ($NAMELEN - length($elt->{"suite"})),
-	   $elt->{"desc"},
-	   " " x ($DESCLEN - length($elt->{"desc"})));
-}
-printf("   \"%s\" %s/* %s%s */\n", $last->{"suite"},
-       " " x ($NAMELEN - length($last->{"suite"})),
-       $last->{"desc"},
-       " " x ($DESCLEN - length($last->{"desc"})));
-print "};\n\n";
-
-#------------------------------------------------------------------------
-# Output the driver function prototypes
-print "/* Test function prototypes */\n";
-foreach $elt (@item, $last) {
-    printf("int  %s(void);\n", $elt->{"func"});
-}
-print "\n";
-
-#------------------------------------------------------------------------
-# Output the table of functions
-print "/* Table mapping index numbers to functions */\n";
-printf("int (*%s[])(void)  = {\n   ", $FUNCVAR);
-$brk = 0;
-
-foreach $elt (@item) {
-    print($elt->{"func"}, ", ", 
-	  " " x ($FUNCLEN - length($elt->{"func"})));
-    $brk = ($brk + 1) & 3;
-    print "\n   " unless($brk);
-}
-print $last->{"func"}, "\n};\n\n";
-
-#------------------------------------------------------------------------
-# Output the table of descriptions
-print "/* Table mapping index numbers to descriptions */\n";
-printf("const char *%s[] = {\n", $DESCVAR);
-
-foreach $elt (@item) {
-    printf("   \"%s\",\n", $elt->{"desc"});
-}
-printf("   \"%s\"\n};\n\n", $last->{"desc"});
-
-exit 0;
-

mozilla/security/nss/lib/freebl/mpi/mdxptest.c

@@ -1,342 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Netscape security libraries.
- *
- * The Initial Developer of the Original Code is
- * Netscape Communications Corporation.
- * Portions created by the Initial Developer are Copyright (C) 2000
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <malloc.h>
-#include <time.h>
-#include "mpi.h"
-#include "mpi-priv.h"
-
-/* #define OLD_WAY 1  */
-
-/* This key is the 1024-bit test key used for speed testing of RSA private 
-** key ops.
-*/
-
-#define CONST const
-
-static CONST unsigned char default_n[128] = {
-0xc2,0xae,0x96,0x89,0xaf,0xce,0xd0,0x7b,0x3b,0x35,0xfd,0x0f,0xb1,0xf4,0x7a,0xd1,
-0x3c,0x7d,0xb5,0x86,0xf2,0x68,0x36,0xc9,0x97,0xe6,0x82,0x94,0x86,0xaa,0x05,0x39,
-0xec,0x11,0x51,0xcc,0x5c,0xa1,0x59,0xba,0x29,0x18,0xf3,0x28,0xf1,0x9d,0xe3,0xae,
-0x96,0x5d,0x6d,0x87,0x73,0xf6,0xf6,0x1f,0xd0,0x2d,0xfb,0x2f,0x7a,0x13,0x7f,0xc8,
-0x0c,0x7a,0xe9,0x85,0xfb,0xce,0x74,0x86,0xf8,0xef,0x2f,0x85,0x37,0x73,0x0f,0x62,
-0x4e,0x93,0x17,0xb7,0x7e,0x84,0x9a,0x94,0x11,0x05,0xca,0x0d,0x31,0x4b,0x2a,0xc8,
-0xdf,0xfe,0xe9,0x0c,0x13,0xc7,0xf2,0xad,0x19,0x64,0x28,0x3c,0xb5,0x6a,0xc8,0x4b,
-0x79,0xea,0x7c,0xce,0x75,0x92,0x45,0x3e,0xa3,0x9d,0x64,0x6f,0x04,0x69,0x19,0x17
-};
-
-static CONST unsigned char default_d[128] = {
-0x13,0xcb,0xbc,0xf2,0xf3,0x35,0x8c,0x6d,0x7b,0x6f,0xd9,0xf3,0xa6,0x9c,0xbd,0x80,
-0x59,0x2e,0x4f,0x2f,0x11,0xa7,0x17,0x2b,0x18,0x8f,0x0f,0xe8,0x1a,0x69,0x5f,0x6e,
-0xac,0x5a,0x76,0x7e,0xd9,0x4c,0x6e,0xdb,0x47,0x22,0x8a,0x57,0x37,0x7a,0x5e,0x94,
-0x7a,0x25,0xb5,0xe5,0x78,0x1d,0x3c,0x99,0xaf,0x89,0x7d,0x69,0x2e,0x78,0x9d,0x1d,
-0x84,0xc8,0xc1,0xd7,0x1a,0xb2,0x6d,0x2d,0x8a,0xd9,0xab,0x6b,0xce,0xae,0xb0,0xa0,
-0x58,0x55,0xad,0x5c,0x40,0x8a,0xd6,0x96,0x08,0x8a,0xe8,0x63,0xe6,0x3d,0x6c,0x20,
-0x49,0xc7,0xaf,0x0f,0x25,0x73,0xd3,0x69,0x43,0x3b,0xf2,0x32,0xf8,0x3d,0x5e,0xee,
-0x7a,0xca,0xd6,0x94,0x55,0xe5,0xbd,0x25,0x34,0x8d,0x63,0x40,0xb5,0x8a,0xc3,0x01
-};
-
-
-#define DEFAULT_ITERS 50
-
-typedef clock_t timetype;
-#define gettime(x) *(x) = clock()
-#define subtime(a, b) a -= b
-#define msec(x) ((clock_t)((double)x * 1000.0 / CLOCKS_PER_SEC))
-#define sec(x) (x / CLOCKS_PER_SEC)
-
-struct TimingContextStr {
-    timetype start;
-    timetype end;
-    timetype interval;
-
-    int   minutes;  
-    int   seconds;  
-    int   millisecs;
-};
-
-typedef struct TimingContextStr TimingContext;
-
-TimingContext *CreateTimingContext(void) 
-{
-    return (TimingContext *)malloc(sizeof(TimingContext));
-}
-
-void DestroyTimingContext(TimingContext *ctx) 
-{
-    free(ctx);
-}
-
-void TimingBegin(TimingContext *ctx) 
-{
-    gettime(&ctx->start);
-}
-
-static void timingUpdate(TimingContext *ctx) 
-{
-
-    ctx->millisecs = msec(ctx->interval) % 1000;
-    ctx->seconds   = sec(ctx->interval);
-    ctx->minutes   = ctx->seconds / 60;
-    ctx->seconds  %= 60;
-
-}
-
-void TimingEnd(TimingContext *ctx) 
-{
-    gettime(&ctx->end);
-    ctx->interval = ctx->end;
-    subtime(ctx->interval, ctx->start);
-    timingUpdate(ctx);
-}
-
-char *TimingGenerateString(TimingContext *ctx) 
-{
-    static char sBuf[4096];
-
-    sprintf(sBuf, "%d minutes, %d.%03d seconds", ctx->minutes,
-				ctx->seconds, ctx->millisecs);
-    return sBuf;
-}
-
-static void
-dumpBytes( unsigned char * b, int l)
-{
-    int i;
-    if (l <= 0)
-        return;
-    for (i = 0; i < l; ++i) {
-        if (i % 16 == 0)
-            printf("\t");
-        printf(" %02x", b[i]);
-        if (i % 16 == 15)
-            printf("\n");
-    }
-    if ((i % 16) != 0)
-        printf("\n");
-    printf("\n");
-}
-
-static mp_err
-testNewFuncs(const unsigned char * modulusBytes, int modulus_len)
-{
-    mp_err mperr	= MP_OKAY;
-    mp_int modulus;
-    unsigned char buf[512];
-
-    mperr = mp_init(&modulus);
-    mperr = mp_read_unsigned_octets(&modulus,  modulusBytes,  modulus_len );
-    mperr = mp_to_fixlen_octets(&modulus, buf, modulus_len);
-    mperr = mp_to_fixlen_octets(&modulus, buf, modulus_len+1);
-    mperr = mp_to_fixlen_octets(&modulus, buf, modulus_len+4);
-    mperr = mp_to_unsigned_octets(&modulus, buf, modulus_len);
-    mperr = mp_to_signed_octets(&modulus, buf, modulus_len + 1);
-    mp_clear(&modulus);
-    return mperr;
-}
-
-int
-testModExp( const unsigned char *      modulusBytes, 
-	    const unsigned int         expo,
-	    const unsigned char *      input, 
-	          unsigned char *      output,
-	          int                  modulus_len)
-{
-    mp_err mperr	= MP_OKAY;
-    mp_int modulus;
-    mp_int base;
-    mp_int exponent;
-    mp_int result;
-
-    mperr  = mp_init(&modulus);
-    mperr += mp_init(&base);
-    mperr += mp_init(&exponent);
-    mperr += mp_init(&result);
-    /* we initialize all mp_ints unconditionally, even if some fail.
-    ** This guarantees that the DIGITS pointer is valid (even if null).
-    ** So, mp_clear will do the right thing below. 
-    */
-    if (mperr == MP_OKAY) {
-      mperr  = mp_read_unsigned_octets(&modulus,  
-		modulusBytes + (sizeof default_n - modulus_len),  modulus_len );
-      mperr += mp_read_unsigned_octets(&base,     input,         modulus_len );
-      mp_set(&exponent, expo);
-      if (mperr == MP_OKAY) {
-#if OLD_WAY
-	mperr = s_mp_exptmod(&base, &exponent, &modulus, &result);
-#else
-	mperr = mp_exptmod(&base, &exponent, &modulus, &result);
-#endif
-	if (mperr == MP_OKAY) {
-	  mperr = mp_to_fixlen_octets(&result, output, modulus_len);
-	}
-      }
-    }
-    mp_clear(&base);
-    mp_clear(&result);
-
-    mp_clear(&modulus);
-    mp_clear(&exponent);
-
-    return (int)mperr;
-}
-
-int
-doModExp(   const unsigned char *      modulusBytes, 
-	    const unsigned char *      exponentBytes, 
-	    const unsigned char *      input, 
-	          unsigned char *      output,
-	          int                  modulus_len)
-{
-    mp_err mperr	= MP_OKAY;
-    mp_int modulus;
-    mp_int base;
-    mp_int exponent;
-    mp_int result;
-
-    mperr  = mp_init(&modulus);
-    mperr += mp_init(&base);
-    mperr += mp_init(&exponent);
-    mperr += mp_init(&result);
-    /* we initialize all mp_ints unconditionally, even if some fail.
-    ** This guarantees that the DIGITS pointer is valid (even if null).
-    ** So, mp_clear will do the right thing below. 
-    */
-    if (mperr == MP_OKAY) {
-      mperr  = mp_read_unsigned_octets(&modulus,  
-		modulusBytes + (sizeof default_n - modulus_len),  modulus_len );
-      mperr += mp_read_unsigned_octets(&exponent, exponentBytes, modulus_len );
-      mperr += mp_read_unsigned_octets(&base,     input,         modulus_len );
-      if (mperr == MP_OKAY) {
-#if OLD_WAY
-	mperr = s_mp_exptmod(&base, &exponent, &modulus, &result);
-#else
-	mperr = mp_exptmod(&base, &exponent, &modulus, &result);
-#endif
-	if (mperr == MP_OKAY) {
-	  mperr = mp_to_fixlen_octets(&result, output, modulus_len);
-	}
-      }
-    }
-    mp_clear(&base);
-    mp_clear(&result);
-
-    mp_clear(&modulus);
-    mp_clear(&exponent);
-
-    return (int)mperr;
-}
-
-int
-main(int argc, char **argv)
-{
-    TimingContext *	  timeCtx;
-    char *		  progName;
-    long 		  iters 	= DEFAULT_ITERS;
-    unsigned int 	  modulus_len;
-    int		 	  i;
-    int 		  rv;
-    unsigned char 	  buf [1024];
-    unsigned char 	  buf2[1024];
-
-    progName = strrchr(argv[0], '/');
-    if (!progName)
-	progName = strrchr(argv[0], '\\');
-    progName = progName ? progName+1 : argv[0];
-
-    if (argc >= 2) {
-	iters = atol(argv[1]);
-    }
-
-    if (argc >= 3) {
-	modulus_len = atol(argv[2]);
-    } else 
-	modulus_len = sizeof default_n;
-
-    /* no library init function !? */
-
-    memset(buf, 0x41, sizeof buf);
-
-  if (iters < 2) {
-    testNewFuncs( default_n, modulus_len);
-    testNewFuncs( default_n+1, modulus_len - 1);
-    testNewFuncs( default_n+2, modulus_len - 2);
-    testNewFuncs( default_n+3, modulus_len - 3);
-
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-    rv = testModExp(default_n, 0, buf, buf2, modulus_len);
-    dumpBytes((unsigned char *)buf2, modulus_len);
-
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-    rv = testModExp(default_n, 1, buf, buf2, modulus_len);
-    dumpBytes((unsigned char *)buf2, modulus_len);
-
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-    rv = testModExp(default_n, 2, buf, buf2, modulus_len);
-    dumpBytes((unsigned char *)buf2, modulus_len);
-
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-    rv = testModExp(default_n, 3, buf, buf2, modulus_len);
-    dumpBytes((unsigned char *)buf2, modulus_len);
-  }
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-    rv = doModExp(default_n, default_d, buf, buf2, modulus_len);
-    if (rv != 0) {
-	fprintf(stderr, "Error in modexp operation:\n");
-	exit(1);
-    }
-    dumpBytes((unsigned char *)buf2, modulus_len);
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-
-    timeCtx = CreateTimingContext();
-    TimingBegin(timeCtx);
-    i = iters;
-    while (i--) {
-	rv = doModExp(default_n, default_d, buf, buf2, modulus_len);
-	if (rv != 0) {
-	    fprintf(stderr, "Error in modexp operation\n");
-	    exit(1);
-	}
-    }
-    TimingEnd(timeCtx);
-    printf("%ld iterations in %s\n", iters, TimingGenerateString(timeCtx));
-    printf("%lu allocations, %lu frees, %lu copies\n", mp_allocs, mp_frees, mp_copies);
-
-    return 0;
-}

mozilla/security/nss/lib/freebl/mpi/montmulf.c

@@ -1,329 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is SPARC optimized Montgomery multiply functions.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems Inc.
- * Portions created by the Initial Developer are Copyright (C) 1999-2000
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Netscape Communications Corporation
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: montmulf.c,v 1.7 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-#ifdef SOLARIS
-#define RF_INLINE_MACROS 1
-#endif
-
-static const double TwoTo16=65536.0;
-static const double TwoToMinus16=1.0/65536.0;
-static const double Zero=0.0;
-static const double TwoTo32=65536.0*65536.0;
-static const double TwoToMinus32=1.0/(65536.0*65536.0);
-
-#ifdef RF_INLINE_MACROS
-
-double upper32(double);
-double lower32(double, double);
-double mod(double, double, double);
-
-void i16_to_d16_and_d32x4(const double * /*1/(2^16)*/, 
-			  const double * /* 2^16*/,
-			  const double * /* 0 */,
-			  double *       /*result16*/, 
-			  double *       /* result32 */,
-			  float *  /*source - should be unsigned int*
-		          	       converted to float* */);
-
-#else
-#ifdef MP_USE_FLOOR
-#include <math.h>
-#else
-#define floor(d) ((double)((unsigned long long)(d)))
-#endif
-
-static double upper32(double x)
-{
-  return floor(x*TwoToMinus32);
-}
-
-static double lower32(double x, double y)
-{
-  return x-TwoTo32*floor(x*TwoToMinus32);
-}
-
-static double mod(double x, double oneoverm, double m)
-{
-  return x-m*floor(x*oneoverm);
-}
-
-#endif
-
-
-static void cleanup(double *dt, int from, int tlen)
-{
- int i;
- double tmp,tmp1,x,x1;
-
- tmp=tmp1=Zero;
- /* original code **
- for(i=2*from;i<2*tlen-2;i++)
-   {
-     x=dt[i];
-     dt[i]=lower32(x,Zero)+tmp1;
-     tmp1=tmp;
-     tmp=upper32(x);
-   }
- dt[tlen-2]+=tmp1;
- dt[tlen-1]+=tmp;
- **end original code ***/
- /* new code ***/
- for(i=2*from;i<2*tlen;i+=2)
-   {
-     x=dt[i];
-     x1=dt[i+1];
-     dt[i]=lower32(x,Zero)+tmp;
-     dt[i+1]=lower32(x1,Zero)+tmp1;
-     tmp=upper32(x);
-     tmp1=upper32(x1);
-   }
-  /** end new code **/
-}
-
-
-void conv_d16_to_i32(unsigned int *i32, double *d16, long long *tmp, int ilen)
-{
-int i;
-long long t, t1, a, b, c, d;
-
- t1=0;
- a=(long long)d16[0];
- b=(long long)d16[1];
- for(i=0; i<ilen-1; i++)
-   {
-     c=(long long)d16[2*i+2];
-     t1+=(unsigned int)a;
-     t=(a>>32);
-     d=(long long)d16[2*i+3];
-     t1+=(b&0xffff)<<16;
-     t+=(b>>16)+(t1>>32);
-     i32[i]=(unsigned int)t1;
-     t1=t;
-     a=c;
-     b=d;
-   }
- t1+=(unsigned int)a;
- t=(a>>32);
- t1+=(b&0xffff)<<16;
- i32[i]=(unsigned int)t1;
-}
-
-void conv_i32_to_d32(double *d32, unsigned int *i32, int len)
-{
-int i;
-
-#pragma pipeloop(0)
- for(i=0;i<len;i++) d32[i]=(double)(i32[i]);
-}
-
-
-void conv_i32_to_d16(double *d16, unsigned int *i32, int len)
-{
-int i;
-unsigned int a;
-
-#pragma pipeloop(0)
- for(i=0;i<len;i++)
-   {
-     a=i32[i];
-     d16[2*i]=(double)(a&0xffff);
-     d16[2*i+1]=(double)(a>>16);
-   }
-}
-
-
-void conv_i32_to_d32_and_d16(double *d32, double *d16, 
-			     unsigned int *i32, int len)
-{
-int i = 0;
-unsigned int a;
-
-#pragma pipeloop(0)
-#ifdef RF_INLINE_MACROS
- for(;i<len-3;i+=4)
-   {
-     i16_to_d16_and_d32x4(&TwoToMinus16, &TwoTo16, &Zero,
-			  &(d16[2*i]), &(d32[i]), (float *)(&(i32[i])));
-   }
-#endif
- for(;i<len;i++)
-   {
-     a=i32[i];
-     d32[i]=(double)(i32[i]);
-     d16[2*i]=(double)(a&0xffff);
-     d16[2*i+1]=(double)(a>>16);
-   }
-}
-
-
-void adjust_montf_result(unsigned int *i32, unsigned int *nint, int len)
-{
-long long acc;
-int i;
-
- if(i32[len]>0) i=-1;
- else
-   {
-     for(i=len-1; i>=0; i--)
-       {
-	 if(i32[i]!=nint[i]) break;
-       }
-   }
- if((i<0)||(i32[i]>nint[i]))
-   {
-     acc=0;
-     for(i=0;i<len;i++)
-       {
-	 acc=acc+(unsigned long long)(i32[i])-(unsigned long long)(nint[i]);
-	 i32[i]=(unsigned int)acc;
-	 acc=acc>>32;
-       }
-   }
-}
-
-
-
-
-/*
-** the lengths of the input arrays should be at least the following:
-** result[nlen+1], dm1[nlen], dm2[2*nlen+1], dt[4*nlen+2], dn[nlen], nint[nlen]
-** all of them should be different from one another
-**
-*/
-void mont_mulf_noconv(unsigned int *result,
-		     double *dm1, double *dm2, double *dt,
-		     double *dn, unsigned int *nint,
-		     int nlen, double dn0)
-{
- int i, j, jj;
- int tmp;
- double digit, m2j, nextm2j, a, b;
- double *dptmp, *pdm1, *pdm2, *pdn, *pdtj, pdn_0, pdm1_0;
-
- pdm1=&(dm1[0]);
- pdm2=&(dm2[0]);
- pdn=&(dn[0]);
- pdm2[2*nlen]=Zero;
-
- if (nlen!=16)
-   {
-     for(i=0;i<4*nlen+2;i++) dt[i]=Zero;
-
-     a=dt[0]=pdm1[0]*pdm2[0];
-     digit=mod(lower32(a,Zero)*dn0,TwoToMinus16,TwoTo16);
-
-     pdtj=&(dt[0]);
-     for(j=jj=0;j<2*nlen;j++,jj++,pdtj++)
-       {
-	 m2j=pdm2[j];
-	 a=pdtj[0]+pdn[0]*digit;
-	 b=pdtj[1]+pdm1[0]*pdm2[j+1]+a*TwoToMinus16;
-	 pdtj[1]=b;
-
-#pragma pipeloop(0)
-	 for(i=1;i<nlen;i++)
-	   {
-	     pdtj[2*i]+=pdm1[i]*m2j+pdn[i]*digit;
-	   }
- 	 if((jj==30)) {cleanup(dt,j/2+1,2*nlen+1); jj=0;}
-	 
-	 digit=mod(lower32(b,Zero)*dn0,TwoToMinus16,TwoTo16);
-       }
-   }
- else
-   {
-     a=dt[0]=pdm1[0]*pdm2[0];
-
-     dt[65]=     dt[64]=     dt[63]=     dt[62]=     dt[61]=     dt[60]=
-     dt[59]=     dt[58]=     dt[57]=     dt[56]=     dt[55]=     dt[54]=
-     dt[53]=     dt[52]=     dt[51]=     dt[50]=     dt[49]=     dt[48]=
-     dt[47]=     dt[46]=     dt[45]=     dt[44]=     dt[43]=     dt[42]=
-     dt[41]=     dt[40]=     dt[39]=     dt[38]=     dt[37]=     dt[36]=
-     dt[35]=     dt[34]=     dt[33]=     dt[32]=     dt[31]=     dt[30]=
-     dt[29]=     dt[28]=     dt[27]=     dt[26]=     dt[25]=     dt[24]=
-     dt[23]=     dt[22]=     dt[21]=     dt[20]=     dt[19]=     dt[18]=
-     dt[17]=     dt[16]=     dt[15]=     dt[14]=     dt[13]=     dt[12]=
-     dt[11]=     dt[10]=     dt[ 9]=     dt[ 8]=     dt[ 7]=     dt[ 6]=
-     dt[ 5]=     dt[ 4]=     dt[ 3]=     dt[ 2]=     dt[ 1]=Zero;
-
-     pdn_0=pdn[0];
-     pdm1_0=pdm1[0];
-
-     digit=mod(lower32(a,Zero)*dn0,TwoToMinus16,TwoTo16);
-     pdtj=&(dt[0]);
-
-     for(j=0;j<32;j++,pdtj++)
-       {
-
-	 m2j=pdm2[j];
-	 a=pdtj[0]+pdn_0*digit;
-	 b=pdtj[1]+pdm1_0*pdm2[j+1]+a*TwoToMinus16;
-	 pdtj[1]=b;
-
-	 /**** this loop will be fully unrolled:
-	 for(i=1;i<16;i++)
-	   {
-	     pdtj[2*i]+=pdm1[i]*m2j+pdn[i]*digit;
-	   }
-	 *************************************/
-	     pdtj[2]+=pdm1[1]*m2j+pdn[1]*digit;
-	     pdtj[4]+=pdm1[2]*m2j+pdn[2]*digit;
-	     pdtj[6]+=pdm1[3]*m2j+pdn[3]*digit;
-	     pdtj[8]+=pdm1[4]*m2j+pdn[4]*digit;
-	     pdtj[10]+=pdm1[5]*m2j+pdn[5]*digit;
-	     pdtj[12]+=pdm1[6]*m2j+pdn[6]*digit;
-	     pdtj[14]+=pdm1[7]*m2j+pdn[7]*digit;
-	     pdtj[16]+=pdm1[8]*m2j+pdn[8]*digit;
-	     pdtj[18]+=pdm1[9]*m2j+pdn[9]*digit;
-	     pdtj[20]+=pdm1[10]*m2j+pdn[10]*digit;
-	     pdtj[22]+=pdm1[11]*m2j+pdn[11]*digit;
-	     pdtj[24]+=pdm1[12]*m2j+pdn[12]*digit;
-	     pdtj[26]+=pdm1[13]*m2j+pdn[13]*digit;
-	     pdtj[28]+=pdm1[14]*m2j+pdn[14]*digit;
-	     pdtj[30]+=pdm1[15]*m2j+pdn[15]*digit;
-	 /* no need for cleenup, cannot overflow */
-	 digit=mod(lower32(b,Zero)*dn0,TwoToMinus16,TwoTo16);
-       }
-   }
-
- conv_d16_to_i32(result,dt+2*nlen,(long long *)dt,nlen+1);
-
- adjust_montf_result(result,nint,nlen); 
- 
-}
-

mozilla/security/nss/lib/freebl/mpi/montmulf.h

@@ -1,103 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is interface file for SPARC Montgomery multiply functions.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems Inc.
- * Portions created by the Initial Developer are Copyright (C) 1999-2000
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Netscape Communications Corporation
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: montmulf.h,v 1.4 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-/*  The functions that are to be called from outside of the .s file have the
- *  following interfaces and array size requirements:
- */
-
-
-void conv_i32_to_d32(double *d32, unsigned int *i32, int len);
-
-/*  Converts an array of int's to an array of doubles, so that each double
- *  corresponds to an int.  len is the number of items converted.
- *  Does not allocate the output array.
- *  The pointers d32 and i32 should point to arrays of size at least  len
- *  (doubles and unsigned ints, respectively)
- */
-
-
-void conv_i32_to_d16(double *d16, unsigned int *i32, int len);
-
-/*  Converts an array of int's to an array of doubles so that each element
- *  of the int array is converted to a pair of doubles, the first one
- *  corresponding to the lower (least significant) 16 bits of the int and
- *  the second one corresponding to the upper (most significant) 16 bits of
- *  the 32-bit int. len is the number of ints converted.
- *  Does not allocate the output array.
- *  The pointer d16 should point to an array of doubles of size at least
- *  2*len and i32 should point an array of ints of size at least  len
- */
-
-
-void conv_i32_to_d32_and_d16(double *d32, double *d16, 
-			     unsigned int *i32, int len);
-
-/*  Does the above two conversions together, it is much faster than doing
- *  both of those in succession
- */
-
-
-void mont_mulf_noconv(unsigned int *result,
-		     double *dm1, double *dm2, double *dt,
-		     double *dn, unsigned int *nint,
-		     int nlen, double dn0);
-
-/*  Does the Montgomery multiplication of the numbers stored in the arrays
- *  pointed to by dm1 and dm2, writing the result to the array pointed to by
- *  result. It uses the array pointed to by dt as a temporary work area.
- *  nint should point to the modulus in the array-of-integers representation, 
- *  dn should point to its array-of-doubles as obtained as a result of the
- *  function call   conv_i32_to_d32(dn, nint, nlen);
- *  nlen is the length of the array containing the modulus.
- *  The representation used for dm1 is the one that is a result of the function
- *  call   conv_i32_to_d32(dm1, m1, nlen), the representation for dm2 is the
- *  result of the function call   conv_i32_to_d16(dm2, m2, nlen).
- *  Note that m1 and m2 should both be of length nlen, so they should be
- *  padded with 0's if necessary before the conversion. The result comes in 
- *  this form (int representation, padded with 0's).
- *  dn0 is the value of the 16 least significant bits of n0'.
- *  The function does not allocate memory for any of the arrays, so the 
- *  pointers should point to arrays with the following minimal sizes:
- *  result - nlen+1
- *  dm1    - nlen
- *  dm2    - 2*nlen+1  ( the +1 is necessary for technical reasons )
- *  dt     - 4*nlen+2
- *  dn     - nlen
- *  nint   - nlen
- *  No two arrays should point to overlapping areas of memory.
- */  

mozilla/security/nss/lib/freebl/mpi/montmulf.il

@@ -1,141 +0,0 @@
-!  
-! ***** BEGIN LICENSE BLOCK *****
-! Version: MPL 1.1/GPL 2.0/LGPL 2.1
-!
-! The contents of this file are subject to the Mozilla Public License Version
-! 1.1 (the "License"); you may not use this file except in compliance with
-! the License. You may obtain a copy of the License at
-! http://www.mozilla.org/MPL/
-!
-! Software distributed under the License is distributed on an "AS IS" basis,
-! WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-! for the specific language governing rights and limitations under the
-! License.
-!
-! The Original Code is inline macros for SPARC Montgomery multiply functions.
-!
-! The Initial Developer of the Original Code is
-! Sun Microsystems Inc.
-! Portions created by the Initial Developer are Copyright (C) 1999-2000
-! the Initial Developer. All Rights Reserved.
-!
-! Contributor(s):
-!
-! Alternatively, the contents of this file may be used under the terms of
-! either the GNU General Public License Version 2 or later (the "GPL"), or
-! the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-! in which case the provisions of the GPL or the LGPL are applicable instead
-! of those above. If you wish to allow use of your version of this file only
-! under the terms of either the GPL or the LGPL, and not to allow others to
-! use your version of this file under the terms of the MPL, indicate your
-! decision by deleting the provisions above and replace them with the notice
-! and other provisions required by the GPL or the LGPL. If you do not delete
-! the provisions above, a recipient may use your version of this file under
-! the terms of any one of the MPL, the GPL or the LGPL.
-!
-! ***** END LICENSE BLOCK *****
-! $Id: montmulf.il,v 1.4 2004-04-27 23:04:36 gerv%gerv.net Exp $
-
-!
-! double upper32(double /*frs1*/);
-!
-        .inline upper32,8
-        std     %o0,[%sp+0x48]
-        ldd     [%sp+0x48],%f10
-
-	fdtox	%f10,%f10
-	fitod	%f10,%f0
-        .end
-
-!
-! double lower32(double /*frs1*/, double /* Zero */);
-!
-        .inline lower32,8
-        std     %o0,[%sp+0x48]
-        ldd     [%sp+0x48],%f10
-        std     %o2,[%sp+0x48]
-        ldd     [%sp+0x48],%f12
-
-	fdtox	%f10,%f10
-	fmovs	%f12,%f10
-	fxtod	%f10,%f0
-        .end
-
-!
-! double mod(double /*x*/, double /*1/m*/, double /*m*/);
-!
-        .inline mod,12
-        std     %o0,[%sp+0x48]
-        ldd     [%sp+0x48],%f2
-        std     %o2,[%sp+0x48]
-        ldd     [%sp+0x48],%f4
-        std     %o4,[%sp+0x48]
-        ldd     [%sp+0x48],%f6
-
-	fmuld	%f2,%f4,%f4
-	fdtox	%f4,%f4
-	fxtod	%f4,%f4
-	fmuld	%f4,%f6,%f4
-	fsubd	%f2,%f4,%f0
-        .end
-
-
-!
-! void i16_to_d16_and_d32x4(double * /*1/(2^16)*/, double * /* 2^16*/,
-!			    double * /* 0 */,
-!			    double * /*result16*/, double * /* result32 */
-!			    float *  /*source - should be unsigned int*
-!		            	       converted to float* */);
-!
-        .inline i16_to_d16_and_d32x4,24
-        ldd     [%o0],%f2  ! 1/(2^16)
-        ldd     [%o1],%f4  ! 2^16
-	ldd	[%o2],%f22
-
-	fmovd	%f22,%f6
-	ld	[%o5],%f7
-	fmovd	%f22,%f10
-	ld	[%o5+4],%f11
-	fmovd	%f22,%f14
-	ld	[%o5+8],%f15
-	fmovd	%f22,%f18
-	ld	[%o5+12],%f19
-	fxtod	%f6,%f6
-	std	%f6,[%o4]
-	fxtod	%f10,%f10
-	std	%f10,[%o4+8]
-	fxtod	%f14,%f14
-	std	%f14,[%o4+16]
-	fxtod	%f18,%f18
-	std	%f18,[%o4+24]
-	fmuld	%f2,%f6,%f8
-	fmuld	%f2,%f10,%f12
-	fmuld	%f2,%f14,%f16
-	fmuld	%f2,%f18,%f20
-	fdtox	%f8,%f8
-	fdtox	%f12,%f12
-	fdtox	%f16,%f16
-	fdtox	%f20,%f20
-	fxtod	%f8,%f8
-	std	%f8,[%o3+8]
-	fxtod	%f12,%f12
-	std	%f12,[%o3+24]
-	fxtod	%f16,%f16
-	std	%f16,[%o3+40]
-	fxtod	%f20,%f20
-	std	%f20,[%o3+56]
-	fmuld	%f8,%f4,%f8
-	fmuld	%f12,%f4,%f12
-	fmuld	%f16,%f4,%f16
-	fmuld	%f20,%f4,%f20
-	fsubd	%f6,%f8,%f8
-	std	%f8,[%o3]
-	fsubd	%f10,%f12,%f12
-	std	%f12,[%o3+16]
-	fsubd	%f14,%f16,%f16
-	std	%f16,[%o3+32]
-	fsubd	%f18,%f20,%f20
-	std	%f20,[%o3+48]
-        .end
-
-

mozilla/security/nss/lib/freebl/mpi/montmulfv8.il

@@ -1,141 +0,0 @@
-!  
-! ***** BEGIN LICENSE BLOCK *****
-! Version: MPL 1.1/GPL 2.0/LGPL 2.1
-!
-! The contents of this file are subject to the Mozilla Public License Version
-! 1.1 (the "License"); you may not use this file except in compliance with
-! the License. You may obtain a copy of the License at
-! http://www.mozilla.org/MPL/
-!
-! Software distributed under the License is distributed on an "AS IS" basis,
-! WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-! for the specific language governing rights and limitations under the
-! License.
-!
-! The Original Code is inline macros for SPARC Montgomery multiply functions.
-!
-! The Initial Developer of the Original Code is
-! Sun Microsystems Inc.
-! Portions created by the Initial Developer are Copyright (C) 1999-2000
-! the Initial Developer. All Rights Reserved.
-!
-! Contributor(s):
-!
-! Alternatively, the contents of this file may be used under the terms of
-! either the GNU General Public License Version 2 or later (the "GPL"), or
-! the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-! in which case the provisions of the GPL or the LGPL are applicable instead
-! of those above. If you wish to allow use of your version of this file only
-! under the terms of either the GPL or the LGPL, and not to allow others to
-! use your version of this file under the terms of the MPL, indicate your
-! decision by deleting the provisions above and replace them with the notice
-! and other provisions required by the GPL or the LGPL. If you do not delete
-! the provisions above, a recipient may use your version of this file under
-! the terms of any one of the MPL, the GPL or the LGPL.
-!
-! ***** END LICENSE BLOCK *****
-! $Id: montmulfv8.il,v 1.3 2004-04-27 23:04:36 gerv%gerv.net Exp $
-
-!
-! double upper32(double /*frs1*/);
-!
-        .inline upper32,8
-        std     %o0,[%sp+0x48]
-        ldd     [%sp+0x48],%f10
-
-	fdtox	%f10,%f10
-	fitod	%f10,%f0
-        .end
-
-!
-! double lower32(double /*frs1*/, double /* Zero */);
-!
-        .inline lower32,8
-        std     %o0,[%sp+0x48]
-        ldd     [%sp+0x48],%f10
-        std     %o2,[%sp+0x48]
-        ldd     [%sp+0x48],%f12
-
-	fdtox	%f10,%f10
-	fmovs	%f12,%f10
-	fxtod	%f10,%f0
-        .end
-
-!
-! double mod(double /*x*/, double /*1/m*/, double /*m*/);
-!
-        .inline mod,12
-        std     %o0,[%sp+0x48]
-        ldd     [%sp+0x48],%f2
-        std     %o2,[%sp+0x48]
-        ldd     [%sp+0x48],%f4
-        std     %o4,[%sp+0x48]
-        ldd     [%sp+0x48],%f6
-
-	fmuld	%f2,%f4,%f4
-	fdtox	%f4,%f4
-	fxtod	%f4,%f4
-	fmuld	%f4,%f6,%f4
-	fsubd	%f2,%f4,%f0
-        .end
-
-
-!
-! void i16_to_d16_and_d32x4(double * /*1/(2^16)*/, double * /* 2^16*/,
-!			    double * /* 0 */,
-!			    double * /*result16*/, double * /* result32 */
-!			    float *  /*source - should be unsigned int*
-!		            	       converted to float* */);
-!
-        .inline i16_to_d16_and_d32x4,24
-        ldd     [%o0],%f2  ! 1/(2^16)
-        ldd     [%o1],%f4  ! 2^16
-	ldd	[%o2],%f22
-
-	fmovd	%f22,%f6
-	ld	[%o5],%f7
-	fmovd	%f22,%f10
-	ld	[%o5+4],%f11
-	fmovd	%f22,%f14
-	ld	[%o5+8],%f15
-	fmovd	%f22,%f18
-	ld	[%o5+12],%f19
-	fxtod	%f6,%f6
-	std	%f6,[%o4]
-	fxtod	%f10,%f10
-	std	%f10,[%o4+8]
-	fxtod	%f14,%f14
-	std	%f14,[%o4+16]
-	fxtod	%f18,%f18
-	std	%f18,[%o4+24]
-	fmuld	%f2,%f6,%f8
-	fmuld	%f2,%f10,%f12
-	fmuld	%f2,%f14,%f16
-	fmuld	%f2,%f18,%f20
-	fdtox	%f8,%f8
-	fdtox	%f12,%f12
-	fdtox	%f16,%f16
-	fdtox	%f20,%f20
-	fxtod	%f8,%f8
-	std	%f8,[%o3+8]
-	fxtod	%f12,%f12
-	std	%f12,[%o3+24]
-	fxtod	%f16,%f16
-	std	%f16,[%o3+40]
-	fxtod	%f20,%f20
-	std	%f20,[%o3+56]
-	fmuld	%f8,%f4,%f8
-	fmuld	%f12,%f4,%f12
-	fmuld	%f16,%f4,%f16
-	fmuld	%f20,%f4,%f20
-	fsubd	%f6,%f8,%f8
-	std	%f8,[%o3]
-	fsubd	%f10,%f12,%f12
-	std	%f12,[%o3+16]
-	fsubd	%f14,%f16,%f16
-	std	%f16,[%o3+32]
-	fsubd	%f18,%f20,%f20
-	std	%f20,[%o3+48]
-        .end
-
-

mozilla/security/nss/lib/freebl/mpi/montmulfv9.il

@@ -1,126 +0,0 @@
-!  
-! ***** BEGIN LICENSE BLOCK *****
-! Version: MPL 1.1/GPL 2.0/LGPL 2.1
-!
-! The contents of this file are subject to the Mozilla Public License Version
-! 1.1 (the "License"); you may not use this file except in compliance with
-! the License. You may obtain a copy of the License at
-! http://www.mozilla.org/MPL/
-!
-! Software distributed under the License is distributed on an "AS IS" basis,
-! WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-! for the specific language governing rights and limitations under the
-! License.
-!
-! The Original Code is inline macros for SPARC Montgomery multiply functions.
-!
-! The Initial Developer of the Original Code is
-! Sun Microsystems Inc.
-! Portions created by the Initial Developer are Copyright (C) 1999-2000
-! the Initial Developer. All Rights Reserved.
-!
-! Contributor(s):
-!
-! Alternatively, the contents of this file may be used under the terms of
-! either the GNU General Public License Version 2 or later (the "GPL"), or
-! the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-! in which case the provisions of the GPL or the LGPL are applicable instead
-! of those above. If you wish to allow use of your version of this file only
-! under the terms of either the GPL or the LGPL, and not to allow others to
-! use your version of this file under the terms of the MPL, indicate your
-! decision by deleting the provisions above and replace them with the notice
-! and other provisions required by the GPL or the LGPL. If you do not delete
-! the provisions above, a recipient may use your version of this file under
-! the terms of any one of the MPL, the GPL or the LGPL.
-!
-! ***** END LICENSE BLOCK *****
-! $Id: montmulfv9.il,v 1.3 2004-04-27 23:04:36 gerv%gerv.net Exp $
-
-!
-! double upper32(double /*frs1*/);
-!
-        .inline upper32,8
-	fdtox	%f0,%f10
-	fitod	%f10,%f0
-        .end
-
-!
-! double lower32(double /*frs1*/, double /* Zero */);
-!
-        .inline lower32,8
-	fdtox	%f0,%f10
-	fmovs	%f2,%f10
-	fxtod	%f10,%f0
-        .end
-
-!
-! double mod(double /*x*/, double /*1/m*/, double /*m*/);
-!
-        .inline mod,12
-	fmuld	%f0,%f2,%f2
-	fdtox	%f2,%f2
-	fxtod	%f2,%f2
-	fmuld	%f2,%f4,%f2
-	fsubd	%f0,%f2,%f0
-        .end
-
-
-!
-! void i16_to_d16_and_d32x4(double * /*1/(2^16)*/, double * /* 2^16*/,
-!			    double * /* 0 */,
-!			    double * /*result16*/, double * /* result32 */
-!			    float *  /*source - should be unsigned int*
-!		            	       converted to float* */);
-!
-        .inline i16_to_d16_and_d32x4,24
-        ldd     [%o0],%f2  ! 1/(2^16)
-        ldd     [%o1],%f4  ! 2^16
-	ldd	[%o2],%f22
-
-	fmovd	%f22,%f6
-	ld	[%o5],%f7
-	fmovd	%f22,%f10
-	ld	[%o5+4],%f11
-	fmovd	%f22,%f14
-	ld	[%o5+8],%f15
-	fmovd	%f22,%f18
-	ld	[%o5+12],%f19
-	fxtod	%f6,%f6
-	std	%f6,[%o4]
-	fxtod	%f10,%f10
-	std	%f10,[%o4+8]
-	fxtod	%f14,%f14
-	std	%f14,[%o4+16]
-	fxtod	%f18,%f18
-	std	%f18,[%o4+24]
-	fmuld	%f2,%f6,%f8
-	fmuld	%f2,%f10,%f12
-	fmuld	%f2,%f14,%f16
-	fmuld	%f2,%f18,%f20
-	fdtox	%f8,%f8
-	fdtox	%f12,%f12
-	fdtox	%f16,%f16
-	fdtox	%f20,%f20
-	fxtod	%f8,%f8
-	std	%f8,[%o3+8]
-	fxtod	%f12,%f12
-	std	%f12,[%o3+24]
-	fxtod	%f16,%f16
-	std	%f16,[%o3+40]
-	fxtod	%f20,%f20
-	std	%f20,[%o3+56]
-	fmuld	%f8,%f4,%f8
-	fmuld	%f12,%f4,%f12
-	fmuld	%f16,%f4,%f16
-	fmuld	%f20,%f4,%f20
-	fsubd	%f6,%f8,%f8
-	std	%f8,[%o3]
-	fsubd	%f10,%f12,%f12
-	std	%f12,[%o3+16]
-	fsubd	%f14,%f16,%f16
-	std	%f16,[%o3+32]
-	fsubd	%f18,%f20,%f20
-	std	%f20,[%o3+48]
-        .end
-
-

mozilla/security/nss/lib/freebl/mpi/mp_gf2m-priv.h

@@ -1,102 +0,0 @@
-/*
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Multi-precision Binary Polynomial Arithmetic Library.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems, Inc.
- * Portions created by the Initial Developer are Copyright (C) 2003
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Sheueling Chang Shantz <sheueling.chang@sun.com> and
- *   Douglas Stebila <douglas@stebila.ca> of Sun Laboratories.
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#ifndef _MP_GF2M_PRIV_H_
-#define _MP_GF2M_PRIV_H_
-
-#include "mpi-priv.h"
-
-extern const mp_digit mp_gf2m_sqr_tb[16];
-
-#if defined(MP_USE_UINT_DIGIT)
-#define MP_DIGIT_BITS 32
-#else
-#define MP_DIGIT_BITS 64
-#endif
-
-/* Platform-specific macros for fast binary polynomial squaring. */
-#if MP_DIGIT_BITS == 32
-#define gf2m_SQR1(w) \
-    mp_gf2m_sqr_tb[(w) >> 28 & 0xF] << 24 | mp_gf2m_sqr_tb[(w) >> 24 & 0xF] << 16 | \
-    mp_gf2m_sqr_tb[(w) >> 20 & 0xF] <<  8 | mp_gf2m_sqr_tb[(w) >> 16 & 0xF]
-#define gf2m_SQR0(w) \
-    mp_gf2m_sqr_tb[(w) >> 12 & 0xF] << 24 | mp_gf2m_sqr_tb[(w) >>  8 & 0xF] << 16 | \
-    mp_gf2m_sqr_tb[(w) >>  4 & 0xF] <<  8 | mp_gf2m_sqr_tb[(w)       & 0xF]
-#else
-#define gf2m_SQR1(w) \
-    mp_gf2m_sqr_tb[(w) >> 60 & 0xF] << 56 | mp_gf2m_sqr_tb[(w) >> 56 & 0xF] << 48 | \
-    mp_gf2m_sqr_tb[(w) >> 52 & 0xF] << 40 | mp_gf2m_sqr_tb[(w) >> 48 & 0xF] << 32 | \
-    mp_gf2m_sqr_tb[(w) >> 44 & 0xF] << 24 | mp_gf2m_sqr_tb[(w) >> 40 & 0xF] << 16 | \
-    mp_gf2m_sqr_tb[(w) >> 36 & 0xF] <<  8 | mp_gf2m_sqr_tb[(w) >> 32 & 0xF]
-#define gf2m_SQR0(w) \
-    mp_gf2m_sqr_tb[(w) >> 28 & 0xF] << 56 | mp_gf2m_sqr_tb[(w) >> 24 & 0xF] << 48 | \
-    mp_gf2m_sqr_tb[(w) >> 20 & 0xF] << 40 | mp_gf2m_sqr_tb[(w) >> 16 & 0xF] << 32 | \
-    mp_gf2m_sqr_tb[(w) >> 12 & 0xF] << 24 | mp_gf2m_sqr_tb[(w) >>  8 & 0xF] << 16 | \
-    mp_gf2m_sqr_tb[(w) >>  4 & 0xF] <<  8 | mp_gf2m_sqr_tb[(w)       & 0xF]
-#endif
-
-/* Multiply two binary polynomials mp_digits a, b.
- * Result is a polynomial with degree < 2 * MP_DIGIT_BITS - 1.
- * Output in two mp_digits rh, rl.
- */
-void s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b);
-
-/* Compute xor-multiply of two binary polynomials  (a1, a0) x (b1, b0)  
- * result is a binary polynomial in 4 mp_digits r[4].
- * The caller MUST ensure that r has the right amount of space allocated.
- */
-void s_bmul_2x2(mp_digit *r, const mp_digit a1, const mp_digit a0, const mp_digit b1,
-	const mp_digit b0);
-
-/* Compute xor-multiply of two binary polynomials  (a2, a1, a0) x (b2, b1, b0)  
- * result is a binary polynomial in 6 mp_digits r[6].
- * The caller MUST ensure that r has the right amount of space allocated.
- */
-void s_bmul_3x3(mp_digit *r, const mp_digit a2, const mp_digit a1, const mp_digit a0, 
-	const mp_digit b2, const mp_digit b1, const mp_digit b0);
-
-/* Compute xor-multiply of two binary polynomials  (a3, a2, a1, a0) x (b3, b2, b1, b0)  
- * result is a binary polynomial in 8 mp_digits r[8].
- * The caller MUST ensure that r has the right amount of space allocated.
- */
-void s_bmul_4x4(mp_digit *r, const mp_digit a3, const mp_digit a2, const mp_digit a1, 
-	const mp_digit a0, const mp_digit b3, const mp_digit b2, const mp_digit b1, 
-	const mp_digit b0);
-
-#endif /* _MP_GF2M_PRIV_H_ */

mozilla/security/nss/lib/freebl/mpi/mp_gf2m.c

@@ -1,600 +0,0 @@
-/*
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Multi-precision Binary Polynomial Arithmetic Library.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems, Inc.
- * Portions created by the Initial Developer are Copyright (C) 2003
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Sheueling Chang Shantz <sheueling.chang@sun.com> and
- *   Douglas Stebila <douglas@stebila.ca> of Sun Laboratories.
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#include "mp_gf2m.h"
-#include "mp_gf2m-priv.h"
-#include "mplogic.h"
-#include "mpi-priv.h"
-
-const mp_digit mp_gf2m_sqr_tb[16] =
-{
-      0,     1,     4,     5,    16,    17,    20,    21,
-     64,    65,    68,    69,    80,    81,    84,    85
-};
-
-/* Multiply two binary polynomials mp_digits a, b.
- * Result is a polynomial with degree < 2 * MP_DIGIT_BITS - 1.
- * Output in two mp_digits rh, rl.
- */
-#if MP_DIGIT_BITS == 32
-void 
-s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b)
-{
-    register mp_digit h, l, s;
-    mp_digit tab[8], top2b = a >> 30; 
-    register mp_digit a1, a2, a4;
-
-    a1 = a & (0x3FFFFFFF); a2 = a1 << 1; a4 = a2 << 1;
-
-    tab[0] =  0; tab[1] = a1;    tab[2] = a2;    tab[3] = a1^a2;
-    tab[4] = a4; tab[5] = a1^a4; tab[6] = a2^a4; tab[7] = a1^a2^a4;
-
-    s = tab[b       & 0x7]; l  = s;
-    s = tab[b >>  3 & 0x7]; l ^= s <<  3; h  = s >> 29;
-    s = tab[b >>  6 & 0x7]; l ^= s <<  6; h ^= s >> 26;
-    s = tab[b >>  9 & 0x7]; l ^= s <<  9; h ^= s >> 23;
-    s = tab[b >> 12 & 0x7]; l ^= s << 12; h ^= s >> 20;
-    s = tab[b >> 15 & 0x7]; l ^= s << 15; h ^= s >> 17;
-    s = tab[b >> 18 & 0x7]; l ^= s << 18; h ^= s >> 14;
-    s = tab[b >> 21 & 0x7]; l ^= s << 21; h ^= s >> 11;
-    s = tab[b >> 24 & 0x7]; l ^= s << 24; h ^= s >>  8;
-    s = tab[b >> 27 & 0x7]; l ^= s << 27; h ^= s >>  5;
-    s = tab[b >> 30      ]; l ^= s << 30; h ^= s >>  2;
-
-    /* compensate for the top two bits of a */
-
-    if (top2b & 01) { l ^= b << 30; h ^= b >> 2; } 
-    if (top2b & 02) { l ^= b << 31; h ^= b >> 1; } 
-
-    *rh = h; *rl = l;
-} 
-#else
-void 
-s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b)
-{
-    register mp_digit h, l, s;
-    mp_digit tab[16], top3b = a >> 61;
-    register mp_digit a1, a2, a4, a8;
-
-    a1 = a & (0x1FFFFFFFFFFFFFFF); a2 = a1 << 1; 
-    a4 = a2 << 1; a8 = a4 << 1;
-    tab[ 0] = 0;     tab[ 1] = a1;       tab[ 2] = a2;       tab[ 3] = a1^a2;
-    tab[ 4] = a4;    tab[ 5] = a1^a4;    tab[ 6] = a2^a4;    tab[ 7] = a1^a2^a4;
-    tab[ 8] = a8;    tab[ 9] = a1^a8;    tab[10] = a2^a8;    tab[11] = a1^a2^a8;
-    tab[12] = a4^a8; tab[13] = a1^a4^a8; tab[14] = a2^a4^a8; tab[15] = a1^a2^a4^a8;
-
-    s = tab[b       & 0xF]; l  = s;
-    s = tab[b >>  4 & 0xF]; l ^= s <<  4; h  = s >> 60;
-    s = tab[b >>  8 & 0xF]; l ^= s <<  8; h ^= s >> 56;
-    s = tab[b >> 12 & 0xF]; l ^= s << 12; h ^= s >> 52;
-    s = tab[b >> 16 & 0xF]; l ^= s << 16; h ^= s >> 48;
-    s = tab[b >> 20 & 0xF]; l ^= s << 20; h ^= s >> 44;
-    s = tab[b >> 24 & 0xF]; l ^= s << 24; h ^= s >> 40;
-    s = tab[b >> 28 & 0xF]; l ^= s << 28; h ^= s >> 36;
-    s = tab[b >> 32 & 0xF]; l ^= s << 32; h ^= s >> 32;
-    s = tab[b >> 36 & 0xF]; l ^= s << 36; h ^= s >> 28;
-    s = tab[b >> 40 & 0xF]; l ^= s << 40; h ^= s >> 24;
-    s = tab[b >> 44 & 0xF]; l ^= s << 44; h ^= s >> 20;
-    s = tab[b >> 48 & 0xF]; l ^= s << 48; h ^= s >> 16;
-    s = tab[b >> 52 & 0xF]; l ^= s << 52; h ^= s >> 12;
-    s = tab[b >> 56 & 0xF]; l ^= s << 56; h ^= s >>  8;
-    s = tab[b >> 60      ]; l ^= s << 60; h ^= s >>  4;
-
-    /* compensate for the top three bits of a */
-
-    if (top3b & 01) { l ^= b << 61; h ^= b >> 3; } 
-    if (top3b & 02) { l ^= b << 62; h ^= b >> 2; } 
-    if (top3b & 04) { l ^= b << 63; h ^= b >> 1; } 
-
-    *rh = h; *rl = l;
-} 
-#endif
-
-/* Compute xor-multiply of two binary polynomials  (a1, a0) x (b1, b0)  
- * result is a binary polynomial in 4 mp_digits r[4].
- * The caller MUST ensure that r has the right amount of space allocated.
- */
-void 
-s_bmul_2x2(mp_digit *r, const mp_digit a1, const mp_digit a0, const mp_digit b1,
-           const mp_digit b0)
-{
-    mp_digit m1, m0;
-    /* r[3] = h1, r[2] = h0; r[1] = l1; r[0] = l0 */
-    s_bmul_1x1(r+3, r+2, a1, b1);
-    s_bmul_1x1(r+1, r, a0, b0);
-    s_bmul_1x1(&m1, &m0, a0 ^ a1, b0 ^ b1);
-    /* Correction on m1 ^= l1 ^ h1; m0 ^= l0 ^ h0; */
-    r[2] ^= m1 ^ r[1] ^ r[3];  /* h0 ^= m1 ^ l1 ^ h1; */
-    r[1]  = r[3] ^ r[2] ^ r[0] ^ m1 ^ m0;  /* l1 ^= l0 ^ h0 ^ m0; */
-}
-
-/* Compute xor-multiply of two binary polynomials  (a2, a1, a0) x (b2, b1, b0)  
- * result is a binary polynomial in 6 mp_digits r[6].
- * The caller MUST ensure that r has the right amount of space allocated.
- */
-void 
-s_bmul_3x3(mp_digit *r, const mp_digit a2, const mp_digit a1, const mp_digit a0, 
-	const mp_digit b2, const mp_digit b1, const mp_digit b0)
-{
-	mp_digit zm[4];
-
-	s_bmul_1x1(r+5, r+4, a2, b2);         /* fill top 2 words */
-	s_bmul_2x2(zm, a1, a2^a0, b1, b2^b0); /* fill middle 4 words */
-	s_bmul_2x2(r, a1, a0, b1, b0);        /* fill bottom 4 words */
-
-	zm[3] ^= r[3];
-	zm[2] ^= r[2]; 
-	zm[1] ^= r[1] ^ r[5];
-	zm[0] ^= r[0] ^ r[4];
-
-	r[5]  ^= zm[3];
-	r[4]  ^= zm[2];
-	r[3]  ^= zm[1];
-	r[2]  ^= zm[0];
-}
-
-/* Compute xor-multiply of two binary polynomials  (a3, a2, a1, a0) x (b3, b2, b1, b0)  
- * result is a binary polynomial in 8 mp_digits r[8].
- * The caller MUST ensure that r has the right amount of space allocated.
- */
-void s_bmul_4x4(mp_digit *r, const mp_digit a3, const mp_digit a2, const mp_digit a1, 
-	const mp_digit a0, const mp_digit b3, const mp_digit b2, const mp_digit b1, 
-	const mp_digit b0)
-{
-	mp_digit zm[4];
-
-	s_bmul_2x2(r+4, a3, a2, b3, b2);            /* fill top 4 words */
-	s_bmul_2x2(zm, a3^a1, a2^a0, b3^b1, b2^b0); /* fill middle 4 words */
-	s_bmul_2x2(r, a1, a0, b1, b0);              /* fill bottom 4 words */
-
-	zm[3] ^= r[3] ^ r[7]; 
-	zm[2] ^= r[2] ^ r[6]; 
-	zm[1] ^= r[1] ^ r[5]; 
-	zm[0] ^= r[0] ^ r[4]; 
-
-	r[5]  ^= zm[3];    
-	r[4]  ^= zm[2];
-	r[3]  ^= zm[1];    
-	r[2]  ^= zm[0];
-}
-
-/* Compute addition of two binary polynomials a and b,
- * store result in c; c could be a or b, a and b could be equal; 
- * c is the bitwise XOR of a and b.
- */
-mp_err
-mp_badd(const mp_int *a, const mp_int *b, mp_int *c)
-{
-    mp_digit *pa, *pb, *pc;
-    mp_size ix;
-    mp_size used_pa, used_pb;
-    mp_err res = MP_OKAY;
-
-    /* Add all digits up to the precision of b.  If b had more
-     * precision than a initially, swap a, b first
-     */
-    if (MP_USED(a) >= MP_USED(b)) {
-        pa = MP_DIGITS(a);
-        pb = MP_DIGITS(b);
-        used_pa = MP_USED(a);
-        used_pb = MP_USED(b);
-    } else {
-        pa = MP_DIGITS(b);
-        pb = MP_DIGITS(a);
-        used_pa = MP_USED(b);
-        used_pb = MP_USED(a);
-    }
-
-    /* Make sure c has enough precision for the output value */
-    MP_CHECKOK( s_mp_pad(c, used_pa) );
-
-    /* Do word-by-word xor */
-    pc = MP_DIGITS(c);
-    for (ix = 0; ix < used_pb; ix++) {
-        (*pc++) = (*pa++) ^ (*pb++);
-    }
-
-    /* Finish the rest of digits until we're actually done */
-    for (; ix < used_pa; ++ix) {
-        *pc++ = *pa++;
-    }
-
-    MP_USED(c) = used_pa;
-    MP_SIGN(c) = ZPOS;
-    s_mp_clamp(c);
-
-CLEANUP:
-    return res;
-} 
-
-#define s_mp_div2(a) MP_CHECKOK( mpl_rsh((a), (a), 1) );
-
-/* Compute binary polynomial multiply d = a * b */
-static void 
-s_bmul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *d)
-{
-    mp_digit a_i, a0b0, a1b1, carry = 0;
-    while (a_len--) {
-        a_i = *a++;
-        s_bmul_1x1(&a1b1, &a0b0, a_i, b);
-        *d++ = a0b0 ^ carry;
-        carry = a1b1;
-    }
-    *d = carry;
-}
-
-/* Compute binary polynomial xor multiply accumulate d ^= a * b */
-static void 
-s_bmul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *d)
-{
-    mp_digit a_i, a0b0, a1b1, carry = 0;
-    while (a_len--) {
-        a_i = *a++;
-        s_bmul_1x1(&a1b1, &a0b0, a_i, b);
-        *d++ ^= a0b0 ^ carry;
-        carry = a1b1;
-    }
-    *d ^= carry;
-}
-
-/* Compute binary polynomial xor multiply c = a * b.  
- * All parameters may be identical.
- */
-mp_err 
-mp_bmul(const mp_int *a, const mp_int *b, mp_int *c)
-{
-    mp_digit *pb, b_i;
-    mp_int tmp;
-    mp_size ib, a_used, b_used;
-    mp_err res = MP_OKAY;
-
-    MP_DIGITS(&tmp) = 0;
-
-    ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);
-
-    if (a == c) {
-        MP_CHECKOK( mp_init_copy(&tmp, a) );
-        if (a == b)
-            b = &tmp;
-        a = &tmp;
-    } else if (b == c) {
-        MP_CHECKOK( mp_init_copy(&tmp, b) );
-        b = &tmp;
-    }
-
-    if (MP_USED(a) < MP_USED(b)) {
-        const mp_int *xch = b;      /* switch a and b if b longer */
-        b = a;
-        a = xch;
-    }
-
-    MP_USED(c) = 1; MP_DIGIT(c, 0) = 0;
-    MP_CHECKOK( s_mp_pad(c, USED(a) + USED(b)) );
-
-    pb = MP_DIGITS(b);
-    s_bmul_d(MP_DIGITS(a), MP_USED(a), *pb++, MP_DIGITS(c));
-
-    /* Outer loop:  Digits of b */
-    a_used = MP_USED(a);
-    b_used = MP_USED(b);
-	MP_USED(c) = a_used + b_used;
-    for (ib = 1; ib < b_used; ib++) {
-        b_i = *pb++;
-
-        /* Inner product:  Digits of a */
-        if (b_i)
-            s_bmul_d_add(MP_DIGITS(a), a_used, b_i, MP_DIGITS(c) + ib);
-        else
-            MP_DIGIT(c, ib + a_used) = b_i;
-    }
-
-    s_mp_clamp(c);
-
-    SIGN(c) = ZPOS;
-
-CLEANUP:
-    mp_clear(&tmp);
-    return res;
-}
-
-
-/* Compute modular reduction of a and store result in r.  
- * r could be a. 
- * For modular arithmetic, the irreducible polynomial f(t) is represented 
- * as an array of int[], where f(t) is of the form: 
- *     f(t) = t^p[0] + t^p[1] + ... + t^p[k]
- * where m = p[0] > p[1] > ... > p[k] = 0.
- */
-mp_err
-mp_bmod(const mp_int *a, const unsigned int p[], mp_int *r)
-{
-    int j, k;
-    int n, dN, d0, d1;
-    mp_digit zz, *z, tmp;
-    mp_size used;
-    mp_err res = MP_OKAY;
-
-    /* The algorithm does the reduction in place in r, 
-     * if a != r, copy a into r first so reduction can be done in r
-     */
-    if (a != r) {
-        MP_CHECKOK( mp_copy(a, r) );
-    }
-    z = MP_DIGITS(r);
-
-    /* start reduction */
-    dN = p[0] / MP_DIGIT_BITS;
-    used = MP_USED(r);
-
-    for (j = used - 1; j > dN;) {
-
-        zz = z[j];
-        if (zz == 0) {
-            j--; continue;
-        }
-        z[j] = 0;
-
-        for (k = 1; p[k] > 0; k++) {
-            /* reducing component t^p[k] */
-            n = p[0] - p[k];
-            d0 = n % MP_DIGIT_BITS;  
-            d1 = MP_DIGIT_BITS - d0;
-            n /= MP_DIGIT_BITS;
-            z[j-n] ^= (zz>>d0);
-            if (d0) 
-                z[j-n-1] ^= (zz<<d1);
-        }
-
-        /* reducing component t^0 */
-        n = dN;  
-        d0 = p[0] % MP_DIGIT_BITS;
-        d1 = MP_DIGIT_BITS - d0;
-        z[j-n] ^= (zz >> d0);
-        if (d0) 
-            z[j-n-1] ^= (zz << d1);
-
-    }
-
-    /* final round of reduction */
-    while (j == dN) {
-
-        d0 = p[0] % MP_DIGIT_BITS;
-        zz = z[dN] >> d0;  
-        if (zz == 0) break;
-        d1 = MP_DIGIT_BITS - d0;
-
-        /* clear up the top d1 bits */
-        if (d0) z[dN] = (z[dN] << d1) >> d1; 
-        *z ^= zz; /* reduction t^0 component */
-
-        for (k = 1; p[k] > 0; k++) {
-            /* reducing component t^p[k]*/
-            n = p[k] / MP_DIGIT_BITS;
-            d0 = p[k] % MP_DIGIT_BITS;
-            d1 = MP_DIGIT_BITS - d0;
-            z[n] ^= (zz << d0);
-            tmp = zz >> d1;
-            if (d0 && tmp)
-                z[n+1] ^= tmp;
-        }
-    }
-
-    s_mp_clamp(r);
-CLEANUP:
-    return res;
-}
-
-/* Compute the product of two polynomials a and b, reduce modulo p, 
- * Store the result in r.  r could be a or b; a could be b.
- */
-mp_err 
-mp_bmulmod(const mp_int *a, const mp_int *b, const unsigned int p[], mp_int *r)
-{
-    mp_err res;
-    
-    if (a == b) return mp_bsqrmod(a, p, r);
-    if ((res = mp_bmul(a, b, r) ) != MP_OKAY)
-	return res;
-    return mp_bmod(r, p, r);
-}
-
-/* Compute binary polynomial squaring c = a*a mod p .  
- * Parameter r and a can be identical.
- */
-
-mp_err 
-mp_bsqrmod(const mp_int *a, const unsigned int p[], mp_int *r)
-{
-    mp_digit *pa, *pr, a_i;
-    mp_int tmp;
-    mp_size ia, a_used;
-    mp_err res;
-
-    ARGCHK(a != NULL && r != NULL, MP_BADARG);
-    MP_DIGITS(&tmp) = 0;
-
-    if (a == r) {
-        MP_CHECKOK( mp_init_copy(&tmp, a) );
-        a = &tmp;
-    }
-
-    MP_USED(r) = 1; MP_DIGIT(r, 0) = 0;
-    MP_CHECKOK( s_mp_pad(r, 2*USED(a)) );
-
-    pa = MP_DIGITS(a);
-    pr = MP_DIGITS(r);
-    a_used = MP_USED(a);
-	MP_USED(r) = 2 * a_used;
-
-    for (ia = 0; ia < a_used; ia++) {
-        a_i = *pa++;
-        *pr++ = gf2m_SQR0(a_i);
-        *pr++ = gf2m_SQR1(a_i);
-    }
-
-    MP_CHECKOK( mp_bmod(r, p, r) );
-    s_mp_clamp(r);
-    SIGN(r) = ZPOS;
-
-CLEANUP:
-    mp_clear(&tmp);
-    return res;
-}
-
-/* Compute binary polynomial y/x mod p, y divided by x, reduce modulo p.
- * Store the result in r. r could be x or y, and x could equal y.
- * Uses algorithm Modular_Division_GF(2^m) from 
- *     Chang-Shantz, S.  "From Euclid's GCD to Montgomery Multiplication to 
- *     the Great Divide".
- */
-int 
-mp_bdivmod(const mp_int *y, const mp_int *x, const mp_int *pp, 
-    const unsigned int p[], mp_int *r)
-{
-    mp_int aa, bb, uu;
-    mp_int *a, *b, *u, *v;
-    mp_err res = MP_OKAY;
-
-    MP_DIGITS(&aa) = 0;
-    MP_DIGITS(&bb) = 0;
-    MP_DIGITS(&uu) = 0;
-
-    MP_CHECKOK( mp_init_copy(&aa, x) );
-    MP_CHECKOK( mp_init_copy(&uu, y) );
-    MP_CHECKOK( mp_init_copy(&bb, pp) );
-    MP_CHECKOK( s_mp_pad(r, USED(pp)) );
-    MP_USED(r) = 1; MP_DIGIT(r, 0) = 0;
-
-    a = &aa; b= &bb; u=&uu; v=r;
-    /* reduce x and y mod p */
-    MP_CHECKOK( mp_bmod(a, p, a) );
-    MP_CHECKOK( mp_bmod(u, p, u) );
-
-    while (!mp_isodd(a)) {
-        s_mp_div2(a);
-        if (mp_isodd(u)) {
-            MP_CHECKOK( mp_badd(u, pp, u) );
-        }
-        s_mp_div2(u);
-    }
-
-    do {
-        if (mp_cmp_mag(b, a) > 0) {
-            MP_CHECKOK( mp_badd(b, a, b) );
-            MP_CHECKOK( mp_badd(v, u, v) );
-            do {
-                s_mp_div2(b);
-                if (mp_isodd(v)) {
-                    MP_CHECKOK( mp_badd(v, pp, v) );
-                }
-                s_mp_div2(v);
-            } while (!mp_isodd(b));
-        }
-        else if ((MP_DIGIT(a,0) == 1) && (MP_USED(a) == 1))
-            break;
-        else {
-            MP_CHECKOK( mp_badd(a, b, a) );
-            MP_CHECKOK( mp_badd(u, v, u) );
-            do {
-                s_mp_div2(a);
-                if (mp_isodd(u)) {
-                    MP_CHECKOK( mp_badd(u, pp, u) );
-                }
-                s_mp_div2(u);
-            } while (!mp_isodd(a));
-        }
-    } while (1);
-
-    MP_CHECKOK( mp_copy(u, r) );
-
-CLEANUP:
-    return res;
-
-}
-
-/* Convert the bit-string representation of a polynomial a into an array
- * of integers corresponding to the bits with non-zero coefficient.
- * Up to max elements of the array will be filled.  Return value is total
- * number of coefficients that would be extracted if array was large enough.
- */
-int
-mp_bpoly2arr(const mp_int *a, unsigned int p[], int max)
-{
-    int i, j, k;
-    mp_digit top_bit, mask;
-
-    top_bit = 1;
-    top_bit <<= MP_DIGIT_BIT - 1;
-
-    for (k = 0; k < max; k++) p[k] = 0;
-    k = 0;
-
-    for (i = MP_USED(a) - 1; i >= 0; i--) {
-        mask = top_bit;
-        for (j = MP_DIGIT_BIT - 1; j >= 0; j--) {
-            if (MP_DIGITS(a)[i] & mask) {
-                if (k < max) p[k] = MP_DIGIT_BIT * i + j;
-                k++;
-            }
-            mask >>= 1;
-        }
-    }
-
-    return k;
-}
-
-/* Convert the coefficient array representation of a polynomial to a 
- * bit-string.  The array must be terminated by 0.
- */
-mp_err
-mp_barr2poly(const unsigned int p[], mp_int *a)
-{
-
-    mp_err res = MP_OKAY;
-    int i;
-
-    mp_zero(a);
-    for (i = 0; p[i] > 0; i++) {
-	MP_CHECKOK( mpl_set_bit(a, p[i], 1) );
-    }
-    MP_CHECKOK( mpl_set_bit(a, 0, 1) );
-	
-CLEANUP:
-    return res;
-}

mozilla/security/nss/lib/freebl/mpi/mp_gf2m.h

@@ -1,63 +0,0 @@
-/*
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Multi-precision Binary Polynomial Arithmetic Library.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems, Inc.
- * Portions created by the Initial Developer are Copyright (C) 2003
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Sheueling Chang Shantz <sheueling.chang@sun.com> and
- *   Douglas Stebila <douglas@stebila.ca> of Sun Laboratories.
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#ifndef _MP_GF2M_H_
-#define _MP_GF2M_H_
-
-#include "mpi.h"
-
-mp_err mp_badd(const mp_int *a, const mp_int *b, mp_int *c);
-mp_err mp_bmul(const mp_int *a, const mp_int *b, mp_int *c);
-
-/* For modular arithmetic, the irreducible polynomial f(t) is represented 
- * as an array of int[], where f(t) is of the form: 
- *     f(t) = t^p[0] + t^p[1] + ... + t^p[k]
- * where m = p[0] > p[1] > ... > p[k] = 0.
- */
-mp_err mp_bmod(const mp_int *a, const unsigned int p[], mp_int *r);
-mp_err mp_bmulmod(const mp_int *a, const mp_int *b, const unsigned int p[], 
-    mp_int *r);
-mp_err mp_bsqrmod(const mp_int *a, const unsigned int p[], mp_int *r);
-mp_err mp_bdivmod(const mp_int *y, const mp_int *x, const mp_int *pp, 
-    const unsigned int p[], mp_int *r);
-
-int mp_bpoly2arr(const mp_int *a, unsigned int p[], int max);
-mp_err mp_barr2poly(const unsigned int p[], mp_int *a);
-
-#endif /* _MP_GF2M_H_ */

mozilla/security/nss/lib/freebl/mpi/mpcpucache.c

@@ -1,760 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Netscape security libraries.
- *
- * The Initial Developer of the Original Code is
- * Red Hat, Inc
- * Portions created by the Initial Developer are Copyright (C) 2005
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Robert Relyea <rrelyea@redhat.com>
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#include "mpi.h"
-
-/*
- * This file implements a single function: mpi_getProcessorLineSize();
- * mpi_getProcessorLineSize() returns the size in bytes of the cache line
- * if a cache exists, or zero if there is no cache. If more than one
- * cache line exists, it should return the smallest line size (which is 
- * usually the L1 cache).
- *
- * mp_modexp uses this information to make sure that private key information
- * isn't being leaked through the cache.
- *
- * Currently the file returns good data for most modern x86 processors, and
- * reasonable data on 64-bit ppc processors. All other processors are assumed
- * to have a cache line size of 32 bytes unless modified by target.mk.
- * 
- */
-
-#if defined(i386) || defined(__i386) || defined(__X86__) || defined (_M_IX86)
-/* X86 processors have special instructions that tell us about the cache */
-#include "string.h"
-
-/* Generic CPUID function */
-#ifndef _WIN32
-static void cpuid(unsigned long op, unsigned long *eax, 
-	                 unsigned long *ebx, unsigned long *ecx, 
-                         unsigned long *edx)
-{
-/* sigh GCC isn't smart enough to save the ebx PIC register on it's own
- * in this case, so do it by hand. */
-	__asm__("pushl %%ebx\n\t"
-	        "cpuid\n\t"
-                "mov %%ebx,%1\n\t"
-                "popl %%ebx\n\t"
-		: "=a" (*eax),
-		  "=r" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (op));
-}
-
-/*
- * try flipping a processor flag to determine CPU type
- */
-static unsigned long changeFlag(unsigned long flag)
-{
-	unsigned long changedFlags, originalFlags;
-	__asm__("pushfl\n\t"            /* get the flags */
-	        "popl %0\n\t"
-	        "movl %0,%1\n\t"	/* save the original flags */
-	        "xor %0,%2\n\t"	/* flip the but */
-		"pushl %0\n\t"	/* set the flags */
-	        "popfl\n\t"
-		"pushfl\n\t"		/* get the flags again (for return) */
-		"popl %0\n\t"	
-		"pushl %1\n\t"		/* restore the original flags */
-		"popfl\n\t"
-		: "=r" (changedFlags),
-		  "=r" (originalFlags)
-		: "r" (flag));
-	return changedFlags ^ originalFlags;
-}
-
-#else
-
-/*
- * windows versions of the above assembler
- */
-#define wcpuid __asm __emit 0fh __asm __emit 0a2h
-static void cpuid(unsigned long op, unsigned long *Reax, unsigned long *Rebx, 
-                                    unsigned long *Recx, unsigned long *Redx)
-{
-        unsigned long  Leax, Lebx, Lecx, Ledx;
-        __asm {
-        pushad
-        mov     eax,op
-        wcpuid
-        mov     Leax,eax
-        mov     Lebx,ebx
-        mov     Lecx,ecx
-        mov     Ledx,edx
-        popad
-        }
-        *Reax = Leax;
-        *Rebx = Lebx;
-        *Recx = Lecx;
-        *Redx = Ledx;
-}
-
-static unsigned long changeFlag(unsigned long flag)
-{
-	unsigned long changedFlags, originalFlags;
-	__asm {
-		pushad
-		pushfd 	                /* get the flags */
-	        pop  eax
-	        mov  ecx,eax            /* save the original flags */
-	        mov  originalFlags,ecx  /* save the original flags */
-		mov  ebx,flag
-	        xor  eax,ebx            /* flip the but */
-		push eax                /* set the flags */
-	        popfd
-		pushfd                  /* get the flags again (for return) */
-		pop  eax	
-		push ecx                /* restore the original flags */
-		popfd
-		mov changedFlags,eax
-		popad
-	}
-	return changedFlags ^ originalFlags;
-}
-#endif
-
-#define AC_FLAG 0x40000
-#define ID_FLAG 0x200000
-
-/* 386 processors can't flip the AC_FLAG, intel AP Note AP-485 */
-static int is386()
-{
-    return changeFlag(AC_FLAG) == 0;
-}
-
-/* 486 processors can't flip the ID_FLAG, intel AP Note AP-485 */
-static int is486()
-{
-    return changeFlag(ID_FLAG) == 0;
-}
-
-
-/*
- * table for Intel Cache.
- * See Intel Application Note AP-485 for more information 
- */
-typedef enum {
-    Cache_NONE = 0,
-    Cache_UNKNOWN = 1,
-    Cache_TLB = 2,
-    Cache_Trace = 3,
-    Cache_L1 = 4,
-    Cache_L2 = 5 ,
-    Cache_L3 = 6
-} CacheType;
-
-#define DATA_INSTR 1
-#define DATA_DATA  2
-#define DATA_BOTH  3
-#define DATA_TRACE 4
-#define DATA_NONE  0
-
-#define TLB_4k	0x01
-#define TLB_2M  0x08
-#define TLB_4M  0x10
-#define TLB_4Mk 0x11
-#define TLB_ALL 0x19
-
-#define k * 1024
-#define M * (1024*1024)
-#define G * (1024*1024*1024)
-
-struct _cache {
-    CacheType type;
-    unsigned long data;
-#define pageSize size
-#define trcuops size
-    unsigned long size;
-    unsigned long association;
-#define tlbEntries lineSize
-    unsigned long lineSize;
-} CacheMap[] = {
-/* 00 */ {Cache_NONE,    DATA_NONE,    0,       0,    0   },
-/* 01 */ {Cache_TLB,     DATA_INSTR,   TLB_4k,  4,    32  },
-/* 02 */ {Cache_TLB,     DATA_INSTR,   TLB_4M,  0,    2   },
-/* 03 */ {Cache_TLB,     DATA_DATA,    TLB_4k,  4,    64  },
-/* 04 */ {Cache_TLB,     DATA_DATA,    TLB_4M,  4,    8   },
-/* 05 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 06 */ {Cache_L1,      DATA_INSTR,   8 k,     4,    32  },
-/* 07 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 08 */ {Cache_L1,      DATA_INSTR,   16 k,    4,    32  },
-/* 09 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 0a */ {Cache_L1,      DATA_DATA,    8 k,     4,    32  },
-/* 0b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 0c */ {Cache_L1,      DATA_DATA,    16 k,    4,    32  },
-/* 0d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 0e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 0f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 10 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 11 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 12 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 13 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 14 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 15 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 16 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 17 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 18 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 19 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 1a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 1b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 1c */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 1d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 1e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 1f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 20 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 21 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 22 */ {Cache_L3,      DATA_BOTH,    512 k,   8,    64  },
-/* 23 */ {Cache_L3,      DATA_BOTH,    1 M,     8,    64  },
-/* 24 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 25 */ {Cache_L3,      DATA_BOTH,    2 M,     8,    64  },
-/* 26 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 27 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 28 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 29 */ {Cache_L3,      DATA_BOTH,    4 M,     8,    64  },
-/* 2a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 2b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 2c */ {Cache_L1,      DATA_DATA,    32 k,    8,    64  },
-/* 2d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 2e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 2f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 30 */ {Cache_L1,      DATA_INSTR,   32 k,    8,    64  },
-/* 31 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 32 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 33 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 34 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 35 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 36 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 37 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 38 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 39 */ {Cache_L2,      DATA_BOTH,    128 k,   4,    64  },
-/* 3a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 3b */ {Cache_L2,      DATA_BOTH,    128 k,   2,    64  },
-/* 3c */ {Cache_L2,      DATA_BOTH,    256 k,   4,    64  },
-/* 3d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 3e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 3f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 40 */ {Cache_L2,      DATA_NONE,    0,       0,    0   },
-/* 41 */ {Cache_L2,      DATA_BOTH,    128 k,   4,    32  },
-/* 42 */ {Cache_L2,      DATA_BOTH,    256 k,   4,    32  },
-/* 43 */ {Cache_L2,      DATA_BOTH,    512 k,   4,    32  },
-/* 44 */ {Cache_L2,      DATA_BOTH,    1 M,     4,    32  },
-/* 45 */ {Cache_L2,      DATA_BOTH,    2 M,     4,    32  },
-/* 46 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 47 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 48 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 49 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 4a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 4b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 4c */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 4d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 4e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 4f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 50 */ {Cache_TLB,     DATA_INSTR,   TLB_ALL, 0,    64  },
-/* 51 */ {Cache_TLB,     DATA_INSTR,   TLB_ALL, 0,    128 },
-/* 52 */ {Cache_TLB,     DATA_INSTR,   TLB_ALL, 0,    256 },
-/* 53 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 54 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 55 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 56 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 57 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 58 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 59 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 5a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 5b */ {Cache_TLB,     DATA_DATA,    TLB_4Mk, 0,    64  },
-/* 5c */ {Cache_TLB,     DATA_DATA,    TLB_4Mk, 0,    128 },
-/* 5d */ {Cache_TLB,     DATA_DATA,    TLB_4Mk, 0,    256 },
-/* 5e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 5f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 60 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 61 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 62 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 63 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 64 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 65 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 66 */ {Cache_L1,      DATA_DATA,    8 k,     4,    64  },
-/* 67 */ {Cache_L1,      DATA_DATA,    16 k,    4,    64  },
-/* 68 */ {Cache_L1,      DATA_DATA,    32 k,    4,    64  },
-/* 69 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 6a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 6b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 6c */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 6d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 6e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 6f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 70 */ {Cache_Trace,   DATA_TRACE,   12 k,    8,    1   },
-/* 71 */ {Cache_Trace,   DATA_TRACE,   16 k,    8,    1   },
-/* 72 */ {Cache_Trace,   DATA_TRACE,   32 k,    8,    1   },
-/* 73 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 74 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 75 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 76 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 77 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 78 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 79 */ {Cache_L2,      DATA_BOTH,    128 k,   8,    64  },
-/* 7a */ {Cache_L2,      DATA_BOTH,    256 k,   8,    64  },
-/* 7b */ {Cache_L2,      DATA_BOTH,    512 k,   8,    64  },
-/* 7c */ {Cache_L2,      DATA_BOTH,    1 M,     8,    64  },
-/* 7d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 7e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 7f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 80 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 81 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 82 */ {Cache_L2,      DATA_BOTH,    256 k,   8,    32  },
-/* 83 */ {Cache_L2,      DATA_BOTH,    512 k,   8,    32  },
-/* 84 */ {Cache_L2,      DATA_BOTH,    1 M,     8,    32  },
-/* 85 */ {Cache_L2,      DATA_BOTH,    2 M,     8,    32  },
-/* 86 */ {Cache_L2,      DATA_BOTH,    512 k,   4,    64  },
-/* 87 */ {Cache_L2,      DATA_BOTH,    1 M,     8,    64  },
-/* 88 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 89 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 8a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 8b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 8c */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 8d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 8e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 8f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 90 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 91 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 92 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 93 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 94 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 95 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 96 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 97 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 98 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 99 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 9a */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 9b */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 9c */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 9d */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 9e */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* 9f */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a0 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a1 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a2 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a3 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a4 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a5 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a6 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a7 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a8 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* a9 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* aa */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ab */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ac */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ad */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ae */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* af */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b0 */ {Cache_TLB,     DATA_INSTR,   TLB_4k,  4,    128 },
-/* b1 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b2 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b3 */ {Cache_TLB,     DATA_DATA,    TLB_4k,  4,    128 },
-/* b4 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b5 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b6 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b7 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b8 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* b9 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ba */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* bb */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* bc */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* bd */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* be */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* bf */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c0 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c1 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c2 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c3 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c4 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c5 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c6 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c7 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c8 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* c9 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ca */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* cb */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* cc */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* cd */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ce */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* cf */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d0 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d1 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d2 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d3 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d4 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d5 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d6 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d7 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d8 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* d9 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* da */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* db */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* dc */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* dd */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* de */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* df */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e0 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e1 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e2 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e3 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e4 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e5 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e6 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e7 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e8 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* e9 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ea */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* eb */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ec */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ed */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ee */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ef */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f0 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f1 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f2 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f3 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f4 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f5 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f6 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f7 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f8 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* f9 */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* fa */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* fb */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* fc */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* fd */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* fe */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   },
-/* ff */ {Cache_UNKNOWN, DATA_NONE,    0,       0,    0   }
-};
-
-
-/*
- * use the above table to determine the CacheEntryLineSize.
- */
-static void
-getIntelCacheEntryLineSize(unsigned long val, int *level, 
-						unsigned long *lineSize)
-{
-    CacheType type;
-
-    type = CacheMap[val].type;
-    /* only interested in data caches */
-    /* NOTE val = 0x40 is a special value that means no L2 or L3 cache.
-     * this data check has the side effect of rejecting that entry. If
-     * that wasn't the case, we could have to reject it explicitly */
-    if ((CacheMap[val].data & DATA_DATA) != DATA_DATA) {
-	return;
-    }
-    /* look at the caches, skip types we aren't interested in.
-     * if we already have a value for a lower level cache, skip the
-     * current entry */
-    if (type == Cache_L1) {
-	*level = 1;
-	*lineSize = CacheMap[val].lineSize;
-    } else if ((*level >= 2) && type == Cache_L2) {
-	*level = 2;
-	*lineSize = CacheMap[val].lineSize;
-    } else if ((*level >= 3) && type == Cache_L3) {
-	*level = 3;
-	*lineSize = CacheMap[val].lineSize;
-    }
-    return;
-}
-
-
-static void
-getIntelRegisterCacheLineSize(unsigned long val, 
-			int *level, unsigned long *lineSize)
-{
-    getIntelCacheEntryLineSize(val >> 24 & 0xff, level, lineSize);
-    getIntelCacheEntryLineSize(val >> 16 & 0xff, level, lineSize);
-    getIntelCacheEntryLineSize(val >> 8 & 0xff, level, lineSize);
-    getIntelCacheEntryLineSize(val & 0xff, level, lineSize);
-}
-
-/*
- * returns '0' if no recognized cache is found, or if the cache
- * information is supported by this processor 
- */
-static unsigned long
-getIntelCacheLineSize(int cpuidLevel)
-{
-    int level = 4;
-    unsigned long lineSize = 0;
-    unsigned long eax, ebx, ecx, edx;
-    int repeat, count;
-
-    if (cpuidLevel < 2) {
-	return 0;
-    }
-
-    /* command '2' of the cpuid is intel's cache info call. Each byte of the
-     * 4 registers contain a potential descriptor for the cache. The CacheMap	
-     * table maps the cache entry with the processor cache. Register 'al'
-     * contains a count value that cpuid '2' needs to be called in order to 
-     * find all the cache descriptors. Only registers with the high bit set
-     * to 'zero' have valid descriptors. This code loops through all the
-     * required calls to cpuid '2' and passes any valid descriptors it finds
-     * to the getIntelRegisterCacheLineSize code, which breaks the registers
-     * down into their component descriptors. In the end the lineSize of the
-     * lowest level cache data cache is returned. */
-    cpuid(2, &eax, &ebx, &ecx, &edx);
-    repeat = eax & 0xf;
-    for (count = 0; count < repeat; count++) {
-	if ((eax & 0x80000000) == 0) {
-	    getIntelRegisterCacheLineSize(eax & 0xffffff00, &level, &lineSize);
-	}
-	if ((ebx & 0x80000000) == 0) {
-	    getIntelRegisterCacheLineSize(ebx, &level, &lineSize);
-	}
-	if ((ecx & 0x80000000) == 0) {
-	    getIntelRegisterCacheLineSize(ecx, &level, &lineSize);
-	}
-	if ((edx & 0x80000000) == 0) {
-	    getIntelRegisterCacheLineSize(edx, &level, &lineSize);
-	}
-	if (count+1 != repeat) {
-	    cpuid(2, &eax, &ebx, &ecx, &edx);
-	}
-    }
-    return lineSize;
-}
-
-/*
- * returns '0' if the cache info is not supported by this processor.
- * This is based on the AMD extended cache commands for cpuid. 
- * (see "AMD Processor Recognition Application Note" Publication 20734).
- * Some other processors use the identical scheme.
- * (see "Processor Recognition, Transmeta Corporation").
- */
-static unsigned long
-getOtherCacheLineSize(unsigned long cpuidLevel)
-{
-    unsigned long lineSize = 0;
-    unsigned long eax, ebx, ecx, edx;
-
-    /* get the Extended CPUID level */
-    cpuid(0x80000000, &eax, &ebx, &ecx, &edx);
-    cpuidLevel = eax;
-
-    if (cpuidLevel >= 0x80000005) {
-	cpuid(0x80000005, &eax, &ebx, &ecx, &edx);
-	lineSize = ecx & 0xff; /* line Size, L1 Data Cache */
-    }
-    return lineSize;
-}
-
-char *manMap[] = {
-#define INTEL     0
-    "GenuineIntel",
-#define AMD       1
-    "AuthenticAMD",
-#define CYRIX     2
-    "CyrixInstead",
-#define CENTAUR   2
-    "CentaurHauls",
-#define NEXGEN    3
-    "NexGenDriven",
-#define TRANSMETA 4
-    "GenuineTMx86",
-#define RISE      5
-    "RiseRiseRise",
-#define UMC       6
-    "UMC UMC UMC ",
-#define SIS       7
-    "Sis Sis Sis ",
-#define NATIONAL  8
-    "Geode by NSC",
-};
-
-int n_manufacturers = sizeof(manMap)/sizeof(manMap[0]);
-
-#define MAN_UNKNOWN 9
-
-
-unsigned long
-mpi_getProcessorLineSize()
-{
-    unsigned long eax, ebx, ecx, edx;
-    unsigned long cpuidLevel;
-    unsigned long cacheLineSize = 0;
-    int manufacturer = MAN_UNKNOWN;
-    int i;
-    char string[65];
-
-    if (is386()) {
-	return 0; /* 386 had no cache */
-    } if (is486()) {
-	return 32; /* really? need more info */
-    }
-
-    /* Pentium, cpuid command is available */
-    cpuid(0, &eax, &ebx, &ecx, &edx);
-    cpuidLevel = eax;
-    *(int *)string = ebx;
-    *(int *)&string[4] = edx;
-    *(int *)&string[8] = ecx;
-    string[12] = 0;
-
-    manufacturer = MAN_UNKNOWN;
-    for (i=0; i < n_manufacturers; i++) {
-	if ( strcmp(manMap[i],string) == 0) {
-	    manufacturer = i;
-	}
-    }
-
-    if (manufacturer == INTEL) {
-	cacheLineSize = getIntelCacheLineSize(cpuidLevel);
-    } else {
-	cacheLineSize = getOtherCacheLineSize(cpuidLevel);
-    }
-    /* doesn't support cache info based on cpuid. This means
-     * an old pentium class processor, which have cache lines of
-     * 32. If we learn differently, we can use a switch based on
-     * the Manufacturer id  */
-    if (cacheLineSize == 0) {
-	cacheLineSize = 32;
-    }
-    return cacheLineSize;
-}
-#define MPI_GET_PROCESSER_LINE_SIZE_DEFINED 1
-#endif
-
-#if defined(__ppc64__) 
-/*
- *  Sigh, The PPC has some really nice features to help us determine cache
- *  size, since it had lots of direct control functions to do so. The POWER
- *  processor even has an instruction to do this, but it was dropped in
- *  PowerPC. Unfortunately most of them are not available in user mode.
- *
- *  The dcbz function would be a great way to determine cache line size except
- *  1) it only works on write-back memory (it throws an exception otherwise), 
- *  and 2) because so many mac programs 'knew' the processor cache size was
- *  32 bytes, they used this instruction as a fast 'zero 32 bytes'. Now the new
- *  G5 processor has 128 byte cache, but dcbz only clears 32 bytes to keep
- *  these programs happy. dcbzl work if 64 bit instructions are supported.
- *  If you know 64 bit instructions are supported, and that stack is 
- *  write-back, you can use this code.
- */
-#include "memory.h"
-
-/* clear the cache line that contains 'array' */
-static inline void dcbzl(char *array)
-{
-	register char *a asm("r2") = array;
-	__asm__ __volatile__( "dcbzl %0,r0" : "=r" (a): "0"(a) );
-}
-
-
-#define PPC_DO_ALIGN(x,y) ((char *)\
-			((((long long) (x))+((y)-1))&~((y)-1)))
-
-#define PPC_MAX_LINE_SIZE 256
-unsigned long
-mpi_getProcessorLineSize()
-{
-    char testArray[2*PPC_MAX_LINE_SIZE+1];
-    char *test;
-    int i;
-
-    /* align the array on a maximum line size boundary, so we
-     * know we are starting to clear from the first address */
-    test = PPC_DO_ALIGN(testArray, PPC_MAX_LINE_SIZE); 
-    /* set all the values to 1's */
-    memset(test, 0xff, PPC_MAX_LINE_SIZE);
-    /* clear one cache block starting at 'test' */
-    dcbzl(test);
-
-    /* find the size of the cleared area, that's our block size */
-    for (i=PPC_MAX_LINE_SIZE; i != 0; i = i/2) {
-	if (test[i-1] == 0) {
-	    return i;
-	}
-    }
-    return 0;
-}
-
-#define MPI_GET_PROCESSER_LINE_SIZE_DEFINED 1
-#endif
-
-
-/*
- * put other processor and platform specific cache code here
- * return the smallest cache line size in bytes on the processor 
- * (usually the L1 cache). If the OS has a call, this would be
- * a greate place to put it.
- *
- * If there is no cache, return 0;
- * 
- * define MPI_GET_PROCESSOR_LINE_SIZE_DEFINED so the generic functions
- * below aren't compiled.
- *
- */
-
-
-/* target.mk can define LINE_SIZE if it's common for the family or OS */
-#if defined(LINE_SIZE) && !defined(MPI_GET_PROCESSOR_LINE_SIZE_DEFINED)
-
-unsigned long
-mpi_getProcessorLineSize()
-{
-   return LINE_SIZE;
-}
-#define MPI_GET_PROCESSER_LINE_SIZE_DEFINED 1
-#endif
-
-
-/* If no way to get the processor cache line size has been defined, assume
- * it's 32 bytes (most common value, does not significantly impact performance
- */ 
-#ifndef MPI_GET_PROCESSER_LINE_SIZE_DEFINED
-unsigned long
-mpi_getProcessorLineSize()
-{
-   return 32;
-}
-#endif
-
-#ifdef TEST_IT
-#include <stdio.h>
-
-main()
-{
-    printf("line size = %d\n", mpi_getProcessorLineSize());
-} 
-#endif

mozilla/security/nss/lib/freebl/mpi/mpi-config.h

@@ -1,112 +0,0 @@
-/* Default configuration for MPI library 
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1997
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Netscape Communications Corporation
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mpi-config.h,v 1.5 2004-04-25 15:03:10 gerv%gerv.net Exp $ */
-
-#ifndef MPI_CONFIG_H_
-#define MPI_CONFIG_H_
-
-/*
-  For boolean options, 
-  0 = no
-  1 = yes
-
-  Other options are documented individually.
-
- */
-
-#ifndef MP_IOFUNC
-#define MP_IOFUNC     0  /* include mp_print() ?                */
-#endif
-
-#ifndef MP_MODARITH
-#define MP_MODARITH   1  /* include modular arithmetic ?        */
-#endif
-
-#ifndef MP_NUMTH
-#define MP_NUMTH      1  /* include number theoretic functions? */
-#endif
-
-#ifndef MP_LOGTAB
-#define MP_LOGTAB     1  /* use table of logs instead of log()? */
-#endif
-
-#ifndef MP_MEMSET
-#define MP_MEMSET     1  /* use memset() to zero buffers?       */
-#endif
-
-#ifndef MP_MEMCPY
-#define MP_MEMCPY     1  /* use memcpy() to copy buffers?       */
-#endif
-
-#ifndef MP_CRYPTO
-#define MP_CRYPTO     1  /* erase memory on free?               */
-#endif
-
-#ifndef MP_ARGCHK
-/*
-  0 = no parameter checks
-  1 = runtime checks, continue execution and return an error to caller
-  2 = assertions; dump core on parameter errors
- */
-#ifdef DEBUG
-#define MP_ARGCHK     2  /* how to check input arguments        */
-#else
-#define MP_ARGCHK     1  /* how to check input arguments        */
-#endif
-#endif
-
-#ifndef MP_DEBUG
-#define MP_DEBUG      0  /* print diagnostic output?            */
-#endif
-
-#ifndef MP_DEFPREC
-#define MP_DEFPREC    64 /* default precision, in digits        */
-#endif
-
-#ifndef MP_MACRO
-#define MP_MACRO      0  /* use macros for frequent calls?      */
-#endif
-
-#ifndef MP_SQUARE
-#define MP_SQUARE     1  /* use separate squaring code?         */
-#endif
-
-#endif /* ifndef MPI_CONFIG_H_ */
-
-

mozilla/security/nss/lib/freebl/mpi/mpi-priv.h

@@ -1,289 +0,0 @@
-/*
- *  mpi-priv.h	- Private header file for MPI 
- *  Arbitrary precision integer arithmetic library
- *
- *  NOTE WELL: the content of this header file is NOT part of the "public"
- *  API for the MPI library, and may change at any time.  
- *  Application programs that use libmpi should NOT include this header file.
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1998
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Netscape Communications Corporation
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mpi-priv.h,v 1.18 2005-02-25 04:30:11 julien.pierre.bugs%sun.com Exp $ */
-#ifndef _MPI_PRIV_H_
-#define _MPI_PRIV_H_ 1
-
-#include "mpi.h"
-#include <stdlib.h>
-#include <string.h>
-#include <ctype.h>
-
-#if MP_DEBUG
-#include <stdio.h>
-
-#define DIAG(T,V) {fprintf(stderr,T);mp_print(V,stderr);fputc('\n',stderr);}
-#else
-#define DIAG(T,V)
-#endif
-
-/* If we aren't using a wired-in logarithm table, we need to include
-   the math library to get the log() function
- */
-
-/* {{{ s_logv_2[] - log table for 2 in various bases */
-
-#if MP_LOGTAB
-/*
-  A table of the logs of 2 for various bases (the 0 and 1 entries of
-  this table are meaningless and should not be referenced).  
-
-  This table is used to compute output lengths for the mp_toradix()
-  function.  Since a number n in radix r takes up about log_r(n)
-  digits, we estimate the output size by taking the least integer
-  greater than log_r(n), where:
-
-  log_r(n) = log_2(n) * log_r(2)
-
-  This table, therefore, is a table of log_r(2) for 2 <= r <= 36,
-  which are the output bases supported.  
- */
-
-extern const float s_logv_2[];
-#define LOG_V_2(R)  s_logv_2[(R)]
-
-#else
-
-/* 
-   If MP_LOGTAB is not defined, use the math library to compute the
-   logarithms on the fly.  Otherwise, use the table.
-   Pick which works best for your system.
- */
-
-#include <math.h>
-#define LOG_V_2(R)  (log(2.0)/log(R))
-
-#endif /* if MP_LOGTAB */
-
-/* }}} */
-
-/* {{{ Digit arithmetic macros */
-
-/*
-  When adding and multiplying digits, the results can be larger than
-  can be contained in an mp_digit.  Thus, an mp_word is used.  These
-  macros mask off the upper and lower digits of the mp_word (the
-  mp_word may be more than 2 mp_digits wide, but we only concern
-  ourselves with the low-order 2 mp_digits)
- */
-
-#define  CARRYOUT(W)  (mp_digit)((W)>>DIGIT_BIT)
-#define  ACCUM(W)     (mp_digit)(W)
-
-#define MP_MIN(a,b)   (((a) < (b)) ? (a) : (b))
-#define MP_MAX(a,b)   (((a) > (b)) ? (a) : (b))
-#define MP_HOWMANY(a,b) (((a) + (b) - 1)/(b))
-#define MP_ROUNDUP(a,b) (MP_HOWMANY(a,b) * (b))
-
-/* }}} */
-
-/* {{{ Comparison constants */
-
-#define  MP_LT       -1
-#define  MP_EQ        0
-#define  MP_GT        1
-
-/* }}} */
-
-/* {{{ private function declarations */
-
-/* 
-   If MP_MACRO is false, these will be defined as actual functions;
-   otherwise, suitable macro definitions will be used.  This works
-   around the fact that ANSI C89 doesn't support an 'inline' keyword
-   (although I hear C9x will ... about bloody time).  At present, the
-   macro definitions are identical to the function bodies, but they'll
-   expand in place, instead of generating a function call.
-
-   I chose these particular functions to be made into macros because
-   some profiling showed they are called a lot on a typical workload,
-   and yet they are primarily housekeeping.
- */
-#if MP_MACRO == 0
- void     s_mp_setz(mp_digit *dp, mp_size count); /* zero digits           */
- void     s_mp_copy(const mp_digit *sp, mp_digit *dp, mp_size count); /* copy */
- void    *s_mp_alloc(size_t nb, size_t ni);       /* general allocator     */
- void     s_mp_free(void *ptr);                   /* general free function */
-extern unsigned long mp_allocs;
-extern unsigned long mp_frees;
-extern unsigned long mp_copies;
-#else
-
- /* Even if these are defined as macros, we need to respect the settings
-    of the MP_MEMSET and MP_MEMCPY configuration options...
-  */
- #if MP_MEMSET == 0
-  #define  s_mp_setz(dp, count) \
-       {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=0;}
- #else
-  #define  s_mp_setz(dp, count) memset(dp, 0, (count) * sizeof(mp_digit))
- #endif /* MP_MEMSET */
-
- #if MP_MEMCPY == 0
-  #define  s_mp_copy(sp, dp, count) \
-       {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=(sp)[ix];}
- #else
-  #define  s_mp_copy(sp, dp, count) memcpy(dp, sp, (count) * sizeof(mp_digit))
- #endif /* MP_MEMCPY */
-
- #define  s_mp_alloc(nb, ni)  calloc(nb, ni)
- #define  s_mp_free(ptr) {if(ptr) free(ptr);}
-#endif /* MP_MACRO */
-
-mp_err   s_mp_grow(mp_int *mp, mp_size min);   /* increase allocated size */
-mp_err   s_mp_pad(mp_int *mp, mp_size min);    /* left pad with zeroes    */
-
-#if MP_MACRO == 0
- void     s_mp_clamp(mp_int *mp);               /* clip leading zeroes     */
-#else
- #define  s_mp_clamp(mp)\
-  { mp_size used = MP_USED(mp); \
-    while (used > 1 && DIGIT(mp, used - 1) == 0) --used; \
-    MP_USED(mp) = used; \
-  } 
-#endif /* MP_MACRO */
-
-void     s_mp_exch(mp_int *a, mp_int *b);      /* swap a and b in place   */
-
-mp_err   s_mp_lshd(mp_int *mp, mp_size p);     /* left-shift by p digits  */
-void     s_mp_rshd(mp_int *mp, mp_size p);     /* right-shift by p digits */
-mp_err   s_mp_mul_2d(mp_int *mp, mp_digit d);  /* multiply by 2^d in place */
-void     s_mp_div_2d(mp_int *mp, mp_digit d);  /* divide by 2^d in place  */
-void     s_mp_mod_2d(mp_int *mp, mp_digit d);  /* modulo 2^d in place     */
-void     s_mp_div_2(mp_int *mp);               /* divide by 2 in place    */
-mp_err   s_mp_mul_2(mp_int *mp);               /* multiply by 2 in place  */
-mp_err   s_mp_norm(mp_int *a, mp_int *b, mp_digit *pd); 
-                                               /* normalize for division  */
-mp_err   s_mp_add_d(mp_int *mp, mp_digit d);   /* unsigned digit addition */
-mp_err   s_mp_sub_d(mp_int *mp, mp_digit d);   /* unsigned digit subtract */
-mp_err   s_mp_mul_d(mp_int *mp, mp_digit d);   /* unsigned digit multiply */
-mp_err   s_mp_div_d(mp_int *mp, mp_digit d, mp_digit *r);
-		                               /* unsigned digit divide   */
-mp_err   s_mp_reduce(mp_int *x, const mp_int *m, const mp_int *mu);
-                                               /* Barrett reduction       */
-mp_err   s_mp_add(mp_int *a, const mp_int *b); /* magnitude addition      */
-mp_err   s_mp_add_3arg(const mp_int *a, const mp_int *b, mp_int *c);
-mp_err   s_mp_sub(mp_int *a, const mp_int *b); /* magnitude subtract      */
-mp_err   s_mp_sub_3arg(const mp_int *a, const mp_int *b, mp_int *c);
-mp_err   s_mp_add_offset(mp_int *a, mp_int *b, mp_size offset);
-                                               /* a += b * RADIX^offset   */
-mp_err   s_mp_mul(mp_int *a, const mp_int *b); /* magnitude multiply      */
-#if MP_SQUARE
-mp_err   s_mp_sqr(mp_int *a);                  /* magnitude square        */
-#else
-#define  s_mp_sqr(a) s_mp_mul(a, a)
-#endif
-mp_err   s_mp_div(mp_int *rem, mp_int *div, mp_int *quot); /* magnitude div */
-mp_err   s_mp_exptmod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c);
-mp_err   s_mp_2expt(mp_int *a, mp_digit k);    /* a = 2^k                 */
-int      s_mp_cmp(const mp_int *a, const mp_int *b); /* magnitude comparison */
-int      s_mp_cmp_d(const mp_int *a, mp_digit d); /* magnitude digit compare */
-int      s_mp_ispow2(const mp_int *v);         /* is v a power of 2?      */
-int      s_mp_ispow2d(mp_digit d);             /* is d a power of 2?      */
-
-int      s_mp_tovalue(char ch, int r);          /* convert ch to value    */
-char     s_mp_todigit(mp_digit val, int r, int low); /* convert val to digit */
-int      s_mp_outlen(int bits, int r);          /* output length in bytes */
-mp_digit s_mp_invmod_radix(mp_digit P);   /* returns (P ** -1) mod RADIX */
-mp_err   s_mp_invmod_odd_m( const mp_int *a, const mp_int *m, mp_int *c);
-mp_err   s_mp_invmod_2d(    const mp_int *a, mp_size k,       mp_int *c);
-mp_err   s_mp_invmod_even_m(const mp_int *a, const mp_int *m, mp_int *c);
-
-/* ------ mpv functions, operate on arrays of digits, not on mp_int's ------ */
-#if defined (__OS2__) && defined (__IBMC__)
-#define MPI_ASM_DECL __cdecl
-#else
-#define MPI_ASM_DECL
-#endif
-
-#ifdef MPI_AMD64
-
-mp_digit MPI_ASM_DECL s_mpv_mul_set_vec64(mp_digit*, mp_digit *, mp_size, mp_digit);
-mp_digit MPI_ASM_DECL s_mpv_mul_add_vec64(mp_digit*, const mp_digit*, mp_size, mp_digit);
-
-/* c = a * b */
-#define s_mpv_mul_d(a, a_len, b, c) \
-	((unsigned long*)c)[a_len] = s_mpv_mul_set_vec64(c, a, a_len, b)
-
-/* c += a * b */
-#define s_mpv_mul_d_add(a, a_len, b, c) \
-	((unsigned long*)c)[a_len] = s_mpv_mul_add_vec64(c, a, a_len, b)
-
-#else
-
-void     MPI_ASM_DECL s_mpv_mul_d(const mp_digit *a, mp_size a_len,
-                                        mp_digit b, mp_digit *c);
-void     MPI_ASM_DECL s_mpv_mul_d_add(const mp_digit *a, mp_size a_len,
-                                            mp_digit b, mp_digit *c);
-
-#endif
-
-void     MPI_ASM_DECL s_mpv_mul_d_add_prop(const mp_digit *a,
-                                                mp_size a_len, mp_digit b, 
-			                        mp_digit *c);
-void     MPI_ASM_DECL s_mpv_sqr_add_prop(const mp_digit *a,
-                                                mp_size a_len,
-                                                mp_digit *sqrs);
-
-mp_err   MPI_ASM_DECL s_mpv_div_2dx1d(mp_digit Nhi, mp_digit Nlo,
-                            mp_digit divisor, mp_digit *quot, mp_digit *rem);
-
-/* c += a * b * (MP_RADIX ** offset);  */
-#define s_mp_mul_d_add_offset(a, b, c, off) \
-(s_mpv_mul_d_add_prop(MP_DIGITS(a), MP_USED(a), b, MP_DIGITS(c) + off), MP_OKAY)
-
-typedef struct {
-  mp_int       N;	/* modulus N */
-  mp_digit     n0prime; /* n0' = - (n0 ** -1) mod MP_RADIX */
-  mp_size      b;	/* R == 2 ** b,  also b = # significant bits in N */
-} mp_mont_modulus;
-
-mp_err s_mp_mul_mont(const mp_int *a, const mp_int *b, mp_int *c, 
-	               mp_mont_modulus *mmm);
-mp_err s_mp_redc(mp_int *T, mp_mont_modulus *mmm);
-
-/* }}} */
-#endif
-

mozilla/security/nss/lib/freebl/mpi/mpi.h

@@ -1,349 +0,0 @@
-/*
- *  mpi.h
- *
- *  Arbitrary precision integer arithmetic library
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1998
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Netscape Communications Corporation
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mpi.h,v 1.22.8.1 2005-06-09 20:44:58 relyea%netscape.com Exp $ */
-
-#ifndef _H_MPI_
-#define _H_MPI_
-
-#include "mpi-config.h"
-
-#if MP_DEBUG
-#undef MP_IOFUNC
-#define MP_IOFUNC 1
-#endif
-
-#if MP_IOFUNC
-#include <stdio.h>
-#include <ctype.h>
-#endif
-
-#include <limits.h>
-
-#if defined(BSDI)
-#undef ULLONG_MAX
-#endif
-
-#if defined( macintosh )
-#include <Types.h>
-#elif defined( _WIN32_WCE)
-/* #include <sys/types.h> What do we need here ?? */
-#else
-#include <sys/types.h>
-#endif
-
-#define  MP_NEG    1
-#define  MP_ZPOS   0
-
-#define  MP_OKAY          0 /* no error, all is well */
-#define  MP_YES           0 /* yes (boolean result)  */
-#define  MP_NO           -1 /* no (boolean result)   */
-#define  MP_MEM          -2 /* out of memory         */
-#define  MP_RANGE        -3 /* argument out of range */
-#define  MP_BADARG       -4 /* invalid parameter     */
-#define  MP_UNDEF        -5 /* answer is undefined   */
-#define  MP_LAST_CODE    MP_UNDEF
-
-typedef unsigned int      mp_sign;
-typedef unsigned int      mp_size;
-typedef int               mp_err;
-
-#define MP_32BIT_MAX 4294967295U
-
-#if !defined(ULONG_MAX) 
-#error "ULONG_MAX not defined"
-#elif !defined(UINT_MAX)
-#error "UINT_MAX not defined"
-#elif !defined(USHRT_MAX)
-#error "USHRT_MAX not defined"
-#endif
-
-#if defined(ULONG_LONG_MAX)			/* GCC, HPUX */
-#define MP_ULONG_LONG_MAX ULONG_LONG_MAX
-#elif defined(ULLONG_MAX)			/* Solaris */
-#define MP_ULONG_LONG_MAX ULLONG_MAX
-/* MP_ULONG_LONG_MAX was defined to be ULLONG_MAX */
-#elif defined(ULONGLONG_MAX)			/* IRIX, AIX */
-#define MP_ULONG_LONG_MAX ULONGLONG_MAX
-#endif
-
-/* We only use unsigned long for mp_digit iff long is more than 32 bits. */
-#if !defined(MP_USE_UINT_DIGIT) && ULONG_MAX > MP_32BIT_MAX
-typedef unsigned long     mp_digit;
-#define MP_DIGIT_MAX      ULONG_MAX
-#define MP_DIGIT_FMT      "%016lX"   /* printf() format for 1 digit */
-#define MP_HALF_DIGIT_MAX UINT_MAX
-#undef  MP_NO_MP_WORD
-#define MP_NO_MP_WORD 1
-#undef  MP_USE_LONG_DIGIT
-#define MP_USE_LONG_DIGIT 1
-#undef  MP_USE_LONG_LONG_DIGIT
-
-#elif !defined(MP_USE_UINT_DIGIT) && defined(MP_ULONG_LONG_MAX) 
-typedef unsigned long long mp_digit;
-#define MP_DIGIT_MAX       MP_ULONG_LONG_MAX
-#define MP_DIGIT_FMT      "%016llX"  /* printf() format for 1 digit */
-#define MP_HALF_DIGIT_MAX  UINT_MAX
-#undef  MP_NO_MP_WORD
-#define MP_NO_MP_WORD 1
-#undef  MP_USE_LONG_LONG_DIGIT
-#define MP_USE_LONG_LONG_DIGIT 1
-#undef  MP_USE_LONG_DIGIT
-
-#else
-typedef unsigned int      mp_digit;
-#define MP_DIGIT_MAX      UINT_MAX
-#define MP_DIGIT_FMT      "%08X"     /* printf() format for 1 digit */
-#define MP_HALF_DIGIT_MAX USHRT_MAX
-#undef  MP_USE_UINT_DIGIT
-#define MP_USE_UINT_DIGIT 1
-#undef  MP_USE_LONG_LONG_DIGIT
-#undef  MP_USE_LONG_DIGIT
-#endif
-
-#if !defined(MP_NO_MP_WORD) 
-#if  defined(MP_USE_UINT_DIGIT) && \
-    (defined(MP_ULONG_LONG_MAX) || (ULONG_MAX > UINT_MAX))
-
-#if (ULONG_MAX > UINT_MAX)
-typedef unsigned long     mp_word;
-typedef          long     mp_sword;
-#define MP_WORD_MAX       ULONG_MAX
-
-#else
-typedef unsigned long long mp_word;
-typedef          long long mp_sword;
-#define MP_WORD_MAX       MP_ULONG_LONG_MAX
-#endif
-
-#else 
-#define MP_NO_MP_WORD 1
-#endif
-#endif /* !defined(MP_NO_MP_WORD) */
-
-#if !defined(MP_WORD_MAX) && defined(MP_DEFINE_SMALL_WORD)
-typedef unsigned int      mp_word;
-typedef          int      mp_sword;
-#define MP_WORD_MAX       UINT_MAX
-#endif
-
-#define MP_DIGIT_BIT      (CHAR_BIT*sizeof(mp_digit))
-#define MP_WORD_BIT       (CHAR_BIT*sizeof(mp_word))
-#define MP_RADIX          (1+(mp_word)MP_DIGIT_MAX)
-
-#define MP_HALF_DIGIT_BIT (MP_DIGIT_BIT/2)
-#define MP_HALF_RADIX     (1+(mp_digit)MP_HALF_DIGIT_MAX)
-/* MP_HALF_RADIX really ought to be called MP_SQRT_RADIX, but it's named 
-** MP_HALF_RADIX because it's the radix for MP_HALF_DIGITs, and it's 
-** consistent with the other _HALF_ names.
-*/
-
-
-/* Macros for accessing the mp_int internals           */
-#define  MP_SIGN(MP)     ((MP)->sign)
-#define  MP_USED(MP)     ((MP)->used)
-#define  MP_ALLOC(MP)    ((MP)->alloc)
-#define  MP_DIGITS(MP)   ((MP)->dp)
-#define  MP_DIGIT(MP,N)  (MP)->dp[(N)]
-
-/* This defines the maximum I/O base (minimum is 2)   */
-#define MP_MAX_RADIX         64
-
-typedef struct {
-  mp_sign       sign;    /* sign of this quantity      */
-  mp_size       alloc;   /* how many digits allocated  */
-  mp_size       used;    /* how many digits used       */
-  mp_digit     *dp;      /* the digits themselves      */
-} mp_int;
-
-/* Default precision       */
-mp_size mp_get_prec(void);
-void    mp_set_prec(mp_size prec);
-
-/* Memory management       */
-mp_err mp_init(mp_int *mp);
-mp_err mp_init_size(mp_int *mp, mp_size prec);
-mp_err mp_init_copy(mp_int *mp, const mp_int *from);
-mp_err mp_copy(const mp_int *from, mp_int *to);
-void   mp_exch(mp_int *mp1, mp_int *mp2);
-void   mp_clear(mp_int *mp);
-void   mp_zero(mp_int *mp);
-void   mp_set(mp_int *mp, mp_digit d);
-mp_err mp_set_int(mp_int *mp, long z);
-#define mp_set_long(mp,z) mp_set_int(mp,z)
-mp_err mp_set_ulong(mp_int *mp, unsigned long z);
-
-/* Single digit arithmetic */
-mp_err mp_add_d(const mp_int *a, mp_digit d, mp_int *b);
-mp_err mp_sub_d(const mp_int *a, mp_digit d, mp_int *b);
-mp_err mp_mul_d(const mp_int *a, mp_digit d, mp_int *b);
-mp_err mp_mul_2(const mp_int *a, mp_int *c);
-mp_err mp_div_d(const mp_int *a, mp_digit d, mp_int *q, mp_digit *r);
-mp_err mp_div_2(const mp_int *a, mp_int *c);
-mp_err mp_expt_d(const mp_int *a, mp_digit d, mp_int *c);
-
-/* Sign manipulations      */
-mp_err mp_abs(const mp_int *a, mp_int *b);
-mp_err mp_neg(const mp_int *a, mp_int *b);
-
-/* Full arithmetic         */
-mp_err mp_add(const mp_int *a, const mp_int *b, mp_int *c);
-mp_err mp_sub(const mp_int *a, const mp_int *b, mp_int *c);
-mp_err mp_mul(const mp_int *a, const mp_int *b, mp_int *c);
-#if MP_SQUARE
-mp_err mp_sqr(const mp_int *a, mp_int *b);
-#else
-#define mp_sqr(a, b) mp_mul(a, a, b)
-#endif
-mp_err mp_div(const mp_int *a, const mp_int *b, mp_int *q, mp_int *r);
-mp_err mp_div_2d(const mp_int *a, mp_digit d, mp_int *q, mp_int *r);
-mp_err mp_expt(mp_int *a, mp_int *b, mp_int *c);
-mp_err mp_2expt(mp_int *a, mp_digit k);
-mp_err mp_sqrt(const mp_int *a, mp_int *b);
-
-/* Modular arithmetic      */
-#if MP_MODARITH
-mp_err mp_mod(const mp_int *a, const mp_int *m, mp_int *c);
-mp_err mp_mod_d(const mp_int *a, mp_digit d, mp_digit *c);
-mp_err mp_addmod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c);
-mp_err mp_submod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c);
-mp_err mp_mulmod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c);
-#if MP_SQUARE
-mp_err mp_sqrmod(const mp_int *a, const mp_int *m, mp_int *c);
-#else
-#define mp_sqrmod(a, m, c) mp_mulmod(a, a, m, c)
-#endif
-mp_err mp_exptmod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c);
-mp_err mp_exptmod_d(const mp_int *a, mp_digit d, const mp_int *m, mp_int *c);
-#endif /* MP_MODARITH */
-
-/* Comparisons             */
-int    mp_cmp_z(const mp_int *a);
-int    mp_cmp_d(const mp_int *a, mp_digit d);
-int    mp_cmp(const mp_int *a, const mp_int *b);
-int    mp_cmp_mag(mp_int *a, mp_int *b);
-int    mp_cmp_int(const mp_int *a, long z);
-int    mp_isodd(const mp_int *a);
-int    mp_iseven(const mp_int *a);
-
-/* Number theoretic        */
-#if MP_NUMTH
-mp_err mp_gcd(mp_int *a, mp_int *b, mp_int *c);
-mp_err mp_lcm(mp_int *a, mp_int *b, mp_int *c);
-mp_err mp_xgcd(const mp_int *a, const mp_int *b, mp_int *g, mp_int *x, mp_int *y);
-mp_err mp_invmod(const mp_int *a, const mp_int *m, mp_int *c);
-mp_err mp_invmod_xgcd(const mp_int *a, const mp_int *m, mp_int *c);
-#endif /* end MP_NUMTH */
-
-/* Input and output        */
-#if MP_IOFUNC
-void   mp_print(mp_int *mp, FILE *ofp);
-#endif /* end MP_IOFUNC */
-
-/* Base conversion         */
-mp_err mp_read_raw(mp_int *mp, char *str, int len);
-int    mp_raw_size(mp_int *mp);
-mp_err mp_toraw(mp_int *mp, char *str);
-mp_err mp_read_radix(mp_int *mp, const char *str, int radix);
-mp_err mp_read_variable_radix(mp_int *a, const char * str, int default_radix);
-int    mp_radix_size(mp_int *mp, int radix);
-mp_err mp_toradix(mp_int *mp, char *str, int radix);
-int    mp_tovalue(char ch, int r);
-
-#define mp_tobinary(M, S)  mp_toradix((M), (S), 2)
-#define mp_tooctal(M, S)   mp_toradix((M), (S), 8)
-#define mp_todecimal(M, S) mp_toradix((M), (S), 10)
-#define mp_tohex(M, S)     mp_toradix((M), (S), 16)
-
-/* Error strings           */
-const  char  *mp_strerror(mp_err ec);
-
-/* Octet string conversion functions */
-mp_err mp_read_unsigned_octets(mp_int *mp, const unsigned char *str, mp_size len);
-int    mp_unsigned_octet_size(const mp_int *mp);
-mp_err mp_to_unsigned_octets(const mp_int *mp, unsigned char *str, mp_size maxlen);
-mp_err mp_to_signed_octets(const mp_int *mp, unsigned char *str, mp_size maxlen);
-mp_err mp_to_fixlen_octets(const mp_int *mp, unsigned char *str, mp_size len);
-
-/* Miscellaneous */
-mp_size mp_trailing_zeros(const mp_int *mp);
-
-#define MP_CHECKOK(x)  if (MP_OKAY > (res = (x))) goto CLEANUP
-#define MP_CHECKERR(x) if (MP_OKAY > (res = (x))) goto CLEANUP
-
-#if defined(MP_API_COMPATIBLE)
-#define NEG             MP_NEG
-#define ZPOS            MP_ZPOS
-#define DIGIT_MAX       MP_DIGIT_MAX
-#define DIGIT_BIT       MP_DIGIT_BIT
-#define DIGIT_FMT       MP_DIGIT_FMT
-#define RADIX           MP_RADIX
-#define MAX_RADIX       MP_MAX_RADIX
-#define SIGN(MP)        MP_SIGN(MP)
-#define USED(MP)        MP_USED(MP)
-#define ALLOC(MP)       MP_ALLOC(MP)
-#define DIGITS(MP)      MP_DIGITS(MP)
-#define DIGIT(MP,N)     MP_DIGIT(MP,N)
-
-#if MP_ARGCHK == 1
-#define  ARGCHK(X,Y)  {if(!(X)){return (Y);}}
-#elif MP_ARGCHK == 2
-#include <assert.h>
-#define  ARGCHK(X,Y)  assert(X)
-#else
-#define  ARGCHK(X,Y)  /*  */
-#endif
-#endif /* defined MP_API_COMPATIBLE */
-
-/*
- * mpi_getProcessorLineSize() returns the size in bytes of the cache line
- * if a cache exists, or zero if there is no cache. If more than one
- * cache line exists, it should return the smallest line size (which is 
- * usually the L1 cache).
- *
- * mp_modexp uses this information to make sure that private key information
- * isn't being leaked through the cache.
- *
- * see mpcpucache.c for the implementation.
- */
-unsigned long mpi_getProcessorLineSize();
-
-#endif /* end _H_MPI_ */

mozilla/security/nss/lib/freebl/mpi/mpi_amd64.c

@@ -1,65 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Solaris software cryptographic token.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems, Inc.
- * Portions created by the Initial Developer are Copyright (C) 2005
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Sun Microsystems, Inc.
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#ifndef MPI_AMD64
-#error This file only works on AMD64 platforms.
-#endif
-
-#include <mpi-priv.h>
-
-/*
- * MPI glue
- *
- */
-
-/* Presently, this is only used by the Montgomery arithmetic code. */
-/* c += a * b */
-void MPI_ASM_DECL s_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len,
-                                       mp_digit b, mp_digit *c)
-{
-  mp_digit w;
-  mp_digit d;
-
-  d = s_mpv_mul_add_vec64(c, a, a_len, b);
-  c += a_len;
-  while (d) {
-    w = c[0] + d;
-    d = (w < c[0] || w < d);
-    *c++ = w;
-  }
-}
-

mozilla/security/nss/lib/freebl/mpi/mpi_amd64_gas.s

@@ -1,418 +0,0 @@
-# ***** BEGIN LICENSE BLOCK *****
-# Version: MPL 1.1/GPL 2.0/LGPL 2.1
-# 
-# The contents of this file are subject to the Mozilla Public License Version
-# 1.1 (the "License"); you may not use this file except in compliance with
-# the License. You may obtain a copy of the License at
-# http://www.mozilla.org/MPL/
-# 
-# Software distributed under the License is distributed on an "AS IS" basis,
-# WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-# for the specific language governing rights and limitations under the
-# License.
-# 
-# The Original Code is the Solaris software cryptographic token.
-# 
-# The Initial Developer of the Original Code is
-# Sun Microsystems, Inc.
-# Portions created by the Initial Developer are Copyright (C) 2005
-# the Initial Developer. All Rights Reserved.
-# 
-# Contributor(s):
-#   Sun Microsystems, Inc.
-# 
-# Alternatively, the contents of this file may be used under the terms of
-# either the GNU General Public License Version 2 or later (the "GPL"), or
-# the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-# in which case the provisions of the GPL or the LGPL are applicable instead
-# of those above. If you wish to allow use of your version of this file only
-# under the terms of either the GPL or the LGPL, and not to allow others to
-# use your version of this file under the terms of the MPL, indicate your
-# decision by deleting the provisions above and replace them with the notice
-# and other provisions required by the GPL or the LGPL. If you do not delete
-# the provisions above, a recipient may use your version of this file under
-# the terms of any one of the MPL, the GPL or the LGPL.
-# 
-# ***** END LICENSE BLOCK ***** */
-
-
-# ------------------------------------------------------------------------
-#
-#  Implementation of s_mpv_mul_set_vec which exploits
-#  the 64X64->128 bit  unsigned multiply instruction.
-#
-# ------------------------------------------------------------------------
-
-# r = a * digit, r and a are vectors of length len
-# returns the carry digit
-# r and a are 64 bit aligned.
-#
-# uint64_t
-# s_mpv_mul_set_vec64(uint64_t *r, uint64_t *a, int len, uint64_t digit)
-#
-
-.text; .align 16; .globl s_mpv_mul_set_vec64; .type s_mpv_mul_set_vec64, @function; s_mpv_mul_set_vec64:
-
-	xorq	%rax, %rax		# if (len == 0) return (0)
-	testq	%rdx, %rdx
-	jz	.L17
-
-	movq	%rdx, %r8		# Use r8 for len; %rdx is used by mul
-	xorq	%r9, %r9		# cy = 0
-
-.L15:
-	cmpq	$8, %r8			# 8 - len
-	jb	.L16
-	movq	0(%rsi), %rax		# rax = a[0]
-	movq	8(%rsi), %r11		# prefetch a[1]
-	mulq	%rcx			# p = a[0] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 0(%rdi)		# r[0] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	16(%rsi), %r11		# prefetch a[2]
-	mulq	%rcx			# p = a[1] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 8(%rdi)		# r[1] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	24(%rsi), %r11		# prefetch a[3]
-	mulq	%rcx			# p = a[2] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 16(%rdi)		# r[2] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	32(%rsi), %r11		# prefetch a[4]
-	mulq	%rcx			# p = a[3] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 24(%rdi)		# r[3] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	40(%rsi), %r11		# prefetch a[5]
-	mulq	%rcx			# p = a[4] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 32(%rdi)		# r[4] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	48(%rsi), %r11		# prefetch a[6]
-	mulq	%rcx			# p = a[5] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 40(%rdi)		# r[5] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	56(%rsi), %r11		# prefetch a[7]
-	mulq	%rcx			# p = a[6] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 48(%rdi)		# r[6] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	mulq	%rcx			# p = a[7] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 56(%rdi)		# r[7] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	addq	$64, %rsi
-	addq	$64, %rdi
-	subq	$8, %r8
-
-	jz	.L17
-	jmp	.L15
-
-.L16:
-	movq	0(%rsi), %rax
-	mulq	%rcx			# p = a[0] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 0(%rdi)		# r[0] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	8(%rsi), %rax
-	mulq	%rcx			# p = a[1] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 8(%rdi)		# r[1] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	16(%rsi), %rax
-	mulq	%rcx			# p = a[2] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 16(%rdi)		# r[2] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	24(%rsi), %rax
-	mulq	%rcx			# p = a[3] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 24(%rdi)		# r[3] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	32(%rsi), %rax
-	mulq	%rcx			# p = a[4] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 32(%rdi)		# r[4] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	40(%rsi), %rax
-	mulq	%rcx			# p = a[5] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 40(%rdi)		# r[5] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	48(%rsi), %rax
-	mulq	%rcx			# p = a[6] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 48(%rdi)		# r[6] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-
-.L17:
-	movq	%r9, %rax
-	ret
-
-.size s_mpv_mul_set_vec64, [.-s_mpv_mul_set_vec64]
-
-# ------------------------------------------------------------------------
-#
-#  Implementation of s_mpv_mul_add_vec which exploits
-#  the 64X64->128 bit  unsigned multiply instruction.
-#
-# ------------------------------------------------------------------------
-
-# r += a * digit, r and a are vectors of length len
-# returns the carry digit
-# r and a are 64 bit aligned.
-#
-# uint64_t
-# s_mpv_mul_add_vec64(uint64_t *r, uint64_t *a, int len, uint64_t digit)
-#
-
-.text; .align 16; .globl s_mpv_mul_add_vec64; .type s_mpv_mul_add_vec64, @function; s_mpv_mul_add_vec64:
-
-	xorq	%rax, %rax		# if (len == 0) return (0)
-	testq	%rdx, %rdx
-	jz	.L27
-
-	movq	%rdx, %r8		# Use r8 for len; %rdx is used by mul
-	xorq	%r9, %r9		# cy = 0
-
-.L25:
-	cmpq	$8, %r8			# 8 - len
-	jb	.L26
-	movq	0(%rsi), %rax		# rax = a[0]
-	movq	0(%rdi), %r10		# r10 = r[0]
-	movq	8(%rsi), %r11		# prefetch a[1]
-	mulq	%rcx			# p = a[0] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[0]
-	movq	8(%rdi), %r10		# prefetch r[1]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 0(%rdi)		# r[0] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	16(%rsi), %r11		# prefetch a[2]
-	mulq	%rcx			# p = a[1] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[1]
-	movq	16(%rdi), %r10		# prefetch r[2]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 8(%rdi)		# r[1] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	24(%rsi), %r11		# prefetch a[3]
-	mulq	%rcx			# p = a[2] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[2]
-	movq	24(%rdi), %r10		# prefetch r[3]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 16(%rdi)		# r[2] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	32(%rsi), %r11		# prefetch a[4]
-	mulq	%rcx			# p = a[3] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[3]
-	movq	32(%rdi), %r10		# prefetch r[4]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 24(%rdi)		# r[3] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	40(%rsi), %r11		# prefetch a[5]
-	mulq	%rcx			# p = a[4] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[4]
-	movq	40(%rdi), %r10		# prefetch r[5]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 32(%rdi)		# r[4] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	48(%rsi), %r11		# prefetch a[6]
-	mulq	%rcx			# p = a[5] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[5]
-	movq	48(%rdi), %r10		# prefetch r[6]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 40(%rdi)		# r[5] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	movq	56(%rsi), %r11		# prefetch a[7]
-	mulq	%rcx			# p = a[6] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[6]
-	movq	56(%rdi), %r10		# prefetch r[7]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 48(%rdi)		# r[6] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	movq	%r11, %rax
-	mulq	%rcx			# p = a[7] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[7]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 56(%rdi)		# r[7] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-
-	addq	$64, %rsi
-	addq	$64, %rdi
-	subq	$8, %r8
-
-	jz	.L27
-	jmp	.L25
-
-.L26:
-	movq	0(%rsi), %rax
-	movq	0(%rdi), %r10
-	mulq	%rcx			# p = a[0] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[0]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 0(%rdi)		# r[0] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	8(%rsi), %rax
-	movq	8(%rdi), %r10
-	mulq	%rcx			# p = a[1] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[1]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 8(%rdi)		# r[1] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	16(%rsi), %rax
-	movq	16(%rdi), %r10
-	mulq	%rcx			# p = a[2] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[2]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 16(%rdi)		# r[2] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	24(%rsi), %rax
-	movq	24(%rdi), %r10
-	mulq	%rcx			# p = a[3] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[3]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 24(%rdi)		# r[3] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	32(%rsi), %rax
-	movq	32(%rdi), %r10
-	mulq	%rcx			# p = a[4] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[4]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 32(%rdi)		# r[4] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	40(%rsi), %rax
-	movq	40(%rdi), %r10
-	mulq	%rcx			# p = a[5] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[5]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 40(%rdi)		# r[5] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	48(%rsi), %rax
-	movq	48(%rdi), %r10
-	mulq	%rcx			# p = a[6] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		# p += r[6]
-	addq	%r9, %rax
-	adcq	$0, %rdx		# p += cy
-	movq	%rax, 48(%rdi)		# r[6] = lo(p)
-	movq	%rdx, %r9		# cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-
-.L27:
-	movq	%r9, %rax
-	ret
-        
-.size s_mpv_mul_add_vec64, [.-s_mpv_mul_add_vec64]

mozilla/security/nss/lib/freebl/mpi/mpi_amd64_sun.s

@@ -1,418 +0,0 @@
-/ ***** BEGIN LICENSE BLOCK *****
-/ Version: MPL 1.1/GPL 2.0/LGPL 2.1
-/ 
-/ The contents of this file are subject to the Mozilla Public License Version
-/ 1.1 (the "License"); you may not use this file except in compliance with
-/ the License. You may obtain a copy of the License at
-/ http://www.mozilla.org/MPL/
-/ 
-/ Software distributed under the License is distributed on an "AS IS" basis,
-/ WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-/ for the specific language governing rights and limitations under the
-/ License.
-/ 
-/ The Original Code is the Solaris software cryptographic token.
-/ 
-/ The Initial Developer of the Original Code is
-/ Sun Microsystems, Inc.
-/ Portions created by the Initial Developer are Copyright (C) 2005
-/ the Initial Developer. All Rights Reserved.
-/ 
-/ Contributor(s):
-/   Sun Microsystems, Inc.
-/ 
-/ Alternatively, the contents of this file may be used under the terms of
-/ either the GNU General Public License Version 2 or later (the "GPL"), or
-/ the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-/ in which case the provisions of the GPL or the LGPL are applicable instead
-/ of those above. If you wish to allow use of your version of this file only
-/ under the terms of either the GPL or the LGPL, and not to allow others to
-/ use your version of this file under the terms of the MPL, indicate your
-/ decision by deleting the provisions above and replace them with the notice
-/ and other provisions required by the GPL or the LGPL. If you do not delete
-/ the provisions above, a recipient may use your version of this file under
-/ the terms of any one of the MPL, the GPL or the LGPL.
-/ 
-/ ***** END LICENSE BLOCK ***** */
-
-
-/ ------------------------------------------------------------------------
-/
-/  Implementation of s_mpv_mul_set_vec which exploits
-/  the 64X64->128 bit  unsigned multiply instruction.
-/
-/ ------------------------------------------------------------------------
-
-/ r = a * digit, r and a are vectors of length len
-/ returns the carry digit
-/ r and a are 64 bit aligned.
-/
-/ uint64_t
-/ s_mpv_mul_set_vec64(uint64_t *r, uint64_t *a, int len, uint64_t digit)
-/
-
-.text; .align 16; .globl s_mpv_mul_set_vec64; .type s_mpv_mul_set_vec64, @function; s_mpv_mul_set_vec64:
-
-	xorq	%rax, %rax		/ if (len == 0) return (0)
-	testq	%rdx, %rdx
-	jz	.L17
-
-	movq	%rdx, %r8		/ Use r8 for len; %rdx is used by mul
-	xorq	%r9, %r9		/ cy = 0
-
-.L15:
-	cmpq	$8, %r8			/ 8 - len
-	jb	.L16
-	movq	0(%rsi), %rax		/ rax = a[0]
-	movq	8(%rsi), %r11		/ prefetch a[1]
-	mulq	%rcx			/ p = a[0] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 0(%rdi)		/ r[0] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	16(%rsi), %r11		/ prefetch a[2]
-	mulq	%rcx			/ p = a[1] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 8(%rdi)		/ r[1] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	24(%rsi), %r11		/ prefetch a[3]
-	mulq	%rcx			/ p = a[2] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 16(%rdi)		/ r[2] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	32(%rsi), %r11		/ prefetch a[4]
-	mulq	%rcx			/ p = a[3] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 24(%rdi)		/ r[3] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	40(%rsi), %r11		/ prefetch a[5]
-	mulq	%rcx			/ p = a[4] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 32(%rdi)		/ r[4] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	48(%rsi), %r11		/ prefetch a[6]
-	mulq	%rcx			/ p = a[5] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 40(%rdi)		/ r[5] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	56(%rsi), %r11		/ prefetch a[7]
-	mulq	%rcx			/ p = a[6] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 48(%rdi)		/ r[6] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	mulq	%rcx			/ p = a[7] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 56(%rdi)		/ r[7] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	addq	$64, %rsi
-	addq	$64, %rdi
-	subq	$8, %r8
-
-	jz	.L17
-	jmp	.L15
-
-.L16:
-	movq	0(%rsi), %rax
-	mulq	%rcx			/ p = a[0] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 0(%rdi)		/ r[0] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	8(%rsi), %rax
-	mulq	%rcx			/ p = a[1] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 8(%rdi)		/ r[1] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	16(%rsi), %rax
-	mulq	%rcx			/ p = a[2] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 16(%rdi)		/ r[2] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	24(%rsi), %rax
-	mulq	%rcx			/ p = a[3] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 24(%rdi)		/ r[3] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	32(%rsi), %rax
-	mulq	%rcx			/ p = a[4] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 32(%rdi)		/ r[4] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	40(%rsi), %rax
-	mulq	%rcx			/ p = a[5] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 40(%rdi)		/ r[5] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-	movq	48(%rsi), %rax
-	mulq	%rcx			/ p = a[6] * digit
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 48(%rdi)		/ r[6] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L17
-
-
-.L17:
-	movq	%r9, %rax
-	ret
-
-.size s_mpv_mul_set_vec64, [.-s_mpv_mul_set_vec64]
-
-/ ------------------------------------------------------------------------
-/
-/  Implementation of s_mpv_mul_add_vec which exploits
-/  the 64X64->128 bit  unsigned multiply instruction.
-/
-/ ------------------------------------------------------------------------
-
-/ r += a * digit, r and a are vectors of length len
-/ returns the carry digit
-/ r and a are 64 bit aligned.
-/
-/ uint64_t
-/ s_mpv_mul_add_vec64(uint64_t *r, uint64_t *a, int len, uint64_t digit)
-/
-
-.text; .align 16; .globl s_mpv_mul_add_vec64; .type s_mpv_mul_add_vec64, @function; s_mpv_mul_add_vec64:
-
-	xorq	%rax, %rax		/ if (len == 0) return (0)
-	testq	%rdx, %rdx
-	jz	.L27
-
-	movq	%rdx, %r8		/ Use r8 for len; %rdx is used by mul
-	xorq	%r9, %r9		/ cy = 0
-
-.L25:
-	cmpq	$8, %r8			/ 8 - len
-	jb	.L26
-	movq	0(%rsi), %rax		/ rax = a[0]
-	movq	0(%rdi), %r10		/ r10 = r[0]
-	movq	8(%rsi), %r11		/ prefetch a[1]
-	mulq	%rcx			/ p = a[0] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[0]
-	movq	8(%rdi), %r10		/ prefetch r[1]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 0(%rdi)		/ r[0] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	16(%rsi), %r11		/ prefetch a[2]
-	mulq	%rcx			/ p = a[1] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[1]
-	movq	16(%rdi), %r10		/ prefetch r[2]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 8(%rdi)		/ r[1] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	24(%rsi), %r11		/ prefetch a[3]
-	mulq	%rcx			/ p = a[2] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[2]
-	movq	24(%rdi), %r10		/ prefetch r[3]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 16(%rdi)		/ r[2] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	32(%rsi), %r11		/ prefetch a[4]
-	mulq	%rcx			/ p = a[3] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[3]
-	movq	32(%rdi), %r10		/ prefetch r[4]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 24(%rdi)		/ r[3] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	40(%rsi), %r11		/ prefetch a[5]
-	mulq	%rcx			/ p = a[4] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[4]
-	movq	40(%rdi), %r10		/ prefetch r[5]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 32(%rdi)		/ r[4] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	48(%rsi), %r11		/ prefetch a[6]
-	mulq	%rcx			/ p = a[5] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[5]
-	movq	48(%rdi), %r10		/ prefetch r[6]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 40(%rdi)		/ r[5] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	movq	56(%rsi), %r11		/ prefetch a[7]
-	mulq	%rcx			/ p = a[6] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[6]
-	movq	56(%rdi), %r10		/ prefetch r[7]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 48(%rdi)		/ r[6] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	movq	%r11, %rax
-	mulq	%rcx			/ p = a[7] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[7]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 56(%rdi)		/ r[7] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-
-	addq	$64, %rsi
-	addq	$64, %rdi
-	subq	$8, %r8
-
-	jz	.L27
-	jmp	.L25
-
-.L26:
-	movq	0(%rsi), %rax
-	movq	0(%rdi), %r10
-	mulq	%rcx			/ p = a[0] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[0]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 0(%rdi)		/ r[0] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	8(%rsi), %rax
-	movq	8(%rdi), %r10
-	mulq	%rcx			/ p = a[1] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[1]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 8(%rdi)		/ r[1] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	16(%rsi), %rax
-	movq	16(%rdi), %r10
-	mulq	%rcx			/ p = a[2] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[2]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 16(%rdi)		/ r[2] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	24(%rsi), %rax
-	movq	24(%rdi), %r10
-	mulq	%rcx			/ p = a[3] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[3]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 24(%rdi)		/ r[3] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	32(%rsi), %rax
-	movq	32(%rdi), %r10
-	mulq	%rcx			/ p = a[4] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[4]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 32(%rdi)		/ r[4] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	40(%rsi), %rax
-	movq	40(%rdi), %r10
-	mulq	%rcx			/ p = a[5] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[5]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 40(%rdi)		/ r[5] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-	movq	48(%rsi), %rax
-	movq	48(%rdi), %r10
-	mulq	%rcx			/ p = a[6] * digit
-	addq	%r10, %rax
-	adcq	$0, %rdx		/ p += r[6]
-	addq	%r9, %rax
-	adcq	$0, %rdx		/ p += cy
-	movq	%rax, 48(%rdi)		/ r[6] = lo(p)
-	movq	%rdx, %r9		/ cy = hi(p)
-	decq	%r8
-	jz	.L27
-
-
-.L27:
-	movq	%r9, %rax
-	ret
-        
-.size s_mpv_mul_add_vec64, [.-s_mpv_mul_add_vec64]

mozilla/security/nss/lib/freebl/mpi/mpi_hp.c

@@ -1,115 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Netscape security libraries.
- *
- * The Initial Developer of the Original Code is
- * Netscape Communications Corporation.
- * Portions created by the Initial Developer are Copyright (C) 2000
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mpi_hp.c,v 1.5 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-/* This file contains routines that perform vector multiplication.  */
-
-#include "mpi-priv.h"
-#include <unistd.h>
-
-#include <stddef.h>
-/* #include <sys/systeminfo.h> */
-#include <strings.h>
-
-extern void multacc512( 
-   int             length,        /* doublewords in multiplicand vector. */
-   const mp_digit *scalaraddr,    /* Address of scalar. */
-   const mp_digit *multiplicand,  /* The multiplicand vector. */
-   mp_digit *      result);       /* Where to accumulate the result. */
-
-extern void maxpy_little(
-   int             length,        /* doublewords in multiplicand vector. */
-   const mp_digit *scalaraddr,    /* Address of scalar. */
-   const mp_digit *multiplicand,  /* The multiplicand vector. */
-   mp_digit *      result);       /* Where to accumulate the result. */
-
-extern void add_diag_little(
-   int            length,       /* doublewords in input vector. */
-   const mp_digit *root,         /* The vector to square. */
-   mp_digit *      result);      /* Where to accumulate the result. */
-
-void 
-s_mpv_sqr_add_prop(const mp_digit *pa, mp_size a_len, mp_digit *ps)
-{
-    add_diag_little(a_len, pa, ps);
-}
-
-#define MAX_STACK_DIGITS 258
-#define MULTACC512_LEN   (512 / MP_DIGIT_BIT)
-#define HP_MPY_ADD_FN    (a_len == MULTACC512_LEN ? multacc512 : maxpy_little)
-
-/* c = a * b */
-void 
-s_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    mp_digit x[MAX_STACK_DIGITS];
-    mp_digit *px = x;
-    size_t   xSize = 0;
-
-    if (a == c) {
-	if (a_len > MAX_STACK_DIGITS) {
-	    xSize = sizeof(mp_digit) * (a_len + 2);
-	    px = malloc(xSize);
-	    if (!px)
-		return;
-	}
-	memcpy(px, a, a_len * sizeof(*a));
-	a = px;
-    }
-    s_mp_setz(c, a_len + 1);
-    HP_MPY_ADD_FN(a_len, &b, a, c);
-    if (px != x && px) {
-	memset(px, 0, xSize);
-	free(px);
-    }
-}
-
-/* c += a * b, where a is a_len words long. */
-void     
-s_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    c[a_len] = 0;	/* so carry propagation stops here. */
-    HP_MPY_ADD_FN(a_len, &b, a, c);
-}
-
-/* c += a * b, where a is y words long. */
-void     
-s_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, 
-			 mp_digit *c)
-{
-    HP_MPY_ADD_FN(a_len, &b, a, c);
-}
-

mozilla/security/nss/lib/freebl/mpi/mpi_i86pc.s

@@ -1,342 +0,0 @@
- /
- / The contents of this file are subject to the Mozilla Public
- / License Version 1.1 (the "License"); you may not use this file
- / except in compliance with the License. You may obtain a copy of
- / the License at http://www.mozilla.org/MPL/
- / 
- / Software distributed under the License is distributed on an "AS
- / IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
- / implied. See the License for the specific language governing
- / rights and limitations under the License.
- / 
- / The Original Code is the Netscape security libraries.
- / 
- / The Initial Developer of the Original Code is Netscape
- / Communications Corporation.	Portions created by Netscape are 
- / Copyright (C) 2001 Netscape Communications Corporation.  All
- / Rights Reserved.
- / 
- / Contributor(s):
- / 
- / Alternatively, the contents of this file may be used under the
- / terms of the GNU General Public License Version 2 or later (the
- / "GPL"), in which case the provisions of the GPL are applicable 
- / instead of those above. If you wish to allow use of your 
- / version of this file only under the terms of the GPL and not to
- / allow others to use your version of this file under the MPL,
- / indicate your decision by deleting the provisions above and
- / replace them with the notice and other provisions required by
- / the GPL.  If you do not delete the provisions above, a recipient
- / may use your version of this file under either the MPL or the
- / GPL.
- /  $Id: mpi_i86pc.s,v 1.1 2001-04-27 20:47:39 nelsonb%netscape.com Exp $
- /
-
-.text
-
- /  ebp - 36:	caller's esi
- /  ebp - 32:	caller's edi
- /  ebp - 28:	
- /  ebp - 24:	
- /  ebp - 20:	
- /  ebp - 16:	
- /  ebp - 12:	
- /  ebp - 8:	
- /  ebp - 4:	
- /  ebp + 0:	caller's ebp
- /  ebp + 4:	return address
- /  ebp + 8:	a	argument
- /  ebp + 12:	a_len	argument
- /  ebp + 16:	b	argument
- /  ebp + 20:	c	argument
- /  registers:
- / 	eax:
- /	ebx:	carry
- /	ecx:	a_len
- /	edx:
- /	esi:	a ptr
- /	edi:	c ptr
-.globl	s_mpv_mul_d
-.type	s_mpv_mul_d,@function
-s_mpv_mul_d:
-    push   %ebp
-    mov    %esp,%ebp
-    sub    $28,%esp
-    push   %edi
-    push   %esi
-    push   %ebx
-    movl   $0,%ebx		/ carry = 0
-    mov    12(%ebp),%ecx	/ ecx = a_len
-    mov    20(%ebp),%edi
-    cmp    $0,%ecx
-    je     L2			/ jmp if a_len == 0
-    mov    8(%ebp),%esi		/ esi = a
-    cld
-L1:
-    lodsl			/ eax = [ds:esi]; esi += 4
-    mov    16(%ebp),%edx	/ edx = b
-    mull   %edx			/ edx:eax = Phi:Plo = a_i * b
-
-    add    %ebx,%eax		/ add carry (%ebx) to edx:eax
-    adc    $0,%edx
-    mov    %edx,%ebx		/ high half of product becomes next carry
-
-    stosl			/ [es:edi] = ax; edi += 4;
-    dec    %ecx			/ --a_len
-    jnz    L1			/ jmp if a_len != 0
-L2:
-    mov    %ebx,0(%edi)		/ *c = carry
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- /  ebp - 36:	caller's esi
- /  ebp - 32:	caller's edi
- /  ebp - 28:	
- /  ebp - 24:	
- /  ebp - 20:	
- /  ebp - 16:	
- /  ebp - 12:	
- /  ebp - 8:	
- /  ebp - 4:	
- /  ebp + 0:	caller's ebp
- /  ebp + 4:	return address
- /  ebp + 8:	a	argument
- /  ebp + 12:	a_len	argument
- /  ebp + 16:	b	argument
- /  ebp + 20:	c	argument
- /  registers:
- / 	eax:
- /	ebx:	carry
- /	ecx:	a_len
- /	edx:
- /	esi:	a ptr
- /	edi:	c ptr
-.globl	s_mpv_mul_d_add
-.type	s_mpv_mul_d_add,@function
-s_mpv_mul_d_add:
-    push   %ebp
-    mov    %esp,%ebp
-    sub    $28,%esp
-    push   %edi
-    push   %esi
-    push   %ebx
-    movl   $0,%ebx		/ carry = 0
-    mov    12(%ebp),%ecx	/ ecx = a_len
-    mov    20(%ebp),%edi
-    cmp    $0,%ecx
-    je     L4			/ jmp if a_len == 0
-    mov    8(%ebp),%esi		/ esi = a
-    cld
-L3:
-    lodsl			/ eax = [ds:esi]; esi += 4
-    mov    16(%ebp),%edx	/ edx = b
-    mull   %edx			/ edx:eax = Phi:Plo = a_i * b
-
-    add    %ebx,%eax		/ add carry (%ebx) to edx:eax
-    adc    $0,%edx
-    mov    0(%edi),%ebx		/ add in current word from *c
-    add    %ebx,%eax		
-    adc    $0,%edx
-    mov    %edx,%ebx		/ high half of product becomes next carry
-
-    stosl			/ [es:edi] = ax; edi += 4;
-    dec    %ecx			/ --a_len
-    jnz    L3			/ jmp if a_len != 0
-L4:
-    mov    %ebx,0(%edi)		/ *c = carry
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- /  ebp - 36:	caller's esi
- /  ebp - 32:	caller's edi
- /  ebp - 28:	
- /  ebp - 24:	
- /  ebp - 20:	
- /  ebp - 16:	
- /  ebp - 12:	
- /  ebp - 8:	
- /  ebp - 4:	
- /  ebp + 0:	caller's ebp
- /  ebp + 4:	return address
- /  ebp + 8:	a	argument
- /  ebp + 12:	a_len	argument
- /  ebp + 16:	b	argument
- /  ebp + 20:	c	argument
- /  registers:
- / 	eax:
- /	ebx:	carry
- /	ecx:	a_len
- /	edx:
- /	esi:	a ptr
- /	edi:	c ptr
-.globl	s_mpv_mul_d_add_prop
-.type	s_mpv_mul_d_add_prop,@function
-s_mpv_mul_d_add_prop:
-    push   %ebp
-    mov    %esp,%ebp
-    sub    $28,%esp
-    push   %edi
-    push   %esi
-    push   %ebx
-    movl   $0,%ebx		/ carry = 0
-    mov    12(%ebp),%ecx	/ ecx = a_len
-    mov    20(%ebp),%edi
-    cmp    $0,%ecx
-    je     L6			/ jmp if a_len == 0
-    cld
-    mov    8(%ebp),%esi		/ esi = a
-L5:
-    lodsl			/ eax = [ds:esi]; esi += 4
-    mov    16(%ebp),%edx	/ edx = b
-    mull   %edx			/ edx:eax = Phi:Plo = a_i * b
-
-    add    %ebx,%eax		/ add carry (%ebx) to edx:eax
-    adc    $0,%edx
-    mov    0(%edi),%ebx		/ add in current word from *c
-    add    %ebx,%eax		
-    adc    $0,%edx
-    mov    %edx,%ebx		/ high half of product becomes next carry
-
-    stosl			/ [es:edi] = ax; edi += 4;
-    dec    %ecx			/ --a_len
-    jnz    L5			/ jmp if a_len != 0
-L6:
-    cmp    $0,%ebx		/ is carry zero?
-    jz     L8
-    mov    0(%edi),%eax		/ add in current word from *c
-    add	   %ebx,%eax
-    stosl			/ [es:edi] = ax; edi += 4;
-    jnc    L8
-L7:
-    mov    0(%edi),%eax		/ add in current word from *c
-    adc	   $0,%eax
-    stosl			/ [es:edi] = ax; edi += 4;
-    jc     L7
-L8:
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- /  ebp - 20:	caller's esi
- /  ebp - 16:	caller's edi
- /  ebp - 12:	
- /  ebp - 8:	carry
- /  ebp - 4:	a_len	local
- /  ebp + 0:	caller's ebp
- /  ebp + 4:	return address
- /  ebp + 8:	pa	argument
- /  ebp + 12:	a_len	argument
- /  ebp + 16:	ps	argument
- /  ebp + 20:	
- /  registers:
- / 	eax:
- /	ebx:	carry
- /	ecx:	a_len
- /	edx:
- /	esi:	a ptr
- /	edi:	c ptr
-
-.globl	s_mpv_sqr_add_prop
-.type	s_mpv_sqr_add_prop,@function
-s_mpv_sqr_add_prop:
-     push   %ebp
-     mov    %esp,%ebp
-     sub    $12,%esp
-     push   %edi
-     push   %esi
-     push   %ebx
-     movl   $0,%ebx		/ carry = 0
-     mov    12(%ebp),%ecx	/ a_len
-     mov    16(%ebp),%edi	/ edi = ps
-     cmp    $0,%ecx
-     je     L11			/ jump if a_len == 0
-     cld
-     mov    8(%ebp),%esi	/ esi = pa
-L10:
-     lodsl			/ %eax = [ds:si]; si += 4;
-     mull   %eax
-
-     add    %ebx,%eax		/ add "carry"
-     adc    $0,%edx
-     mov    0(%edi),%ebx
-     add    %ebx,%eax		/ add low word from result
-     mov    4(%edi),%ebx
-     stosl			/ [es:di] = %eax; di += 4;
-     adc    %ebx,%edx		/ add high word from result
-     movl   $0,%ebx
-     mov    %edx,%eax
-     adc    $0,%ebx
-     stosl			/ [es:di] = %eax; di += 4;
-     dec    %ecx		/ --a_len
-     jnz    L10			/ jmp if a_len != 0
-L11:
-    cmp    $0,%ebx		/ is carry zero?
-    jz     L14
-    mov    0(%edi),%eax		/ add in current word from *c
-    add	   %ebx,%eax
-    stosl			/ [es:edi] = ax; edi += 4;
-    jnc    L14
-L12:
-    mov    0(%edi),%eax		/ add in current word from *c
-    adc	   $0,%eax
-    stosl			/ [es:edi] = ax; edi += 4;
-    jc     L12
-L14:
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- /
- / Divide 64-bit (Nhi,Nlo) by 32-bit divisor, which must be normalized
- / so its high bit is 1.   This code is from NSPR.
- /
- / mp_err s_mpv_div_2dx1d(mp_digit Nhi, mp_digit Nlo, mp_digit divisor,
- / 		          mp_digit *qp, mp_digit *rp)
-
- /  esp +  0:   Caller's ebx
- /  esp +  4:	return address
- /  esp +  8:	Nhi	argument
- /  esp + 12:	Nlo	argument
- /  esp + 16:	divisor	argument
- /  esp + 20:	qp	argument
- /  esp + 24:   rp	argument
- /  registers:
- / 	eax:
- /	ebx:	carry
- /	ecx:	a_len
- /	edx:
- /	esi:	a ptr
- /	edi:	c ptr
- / 
-
-.globl	s_mpv_div_2dx1d
-.type	s_mpv_div_2dx1d,@function
-s_mpv_div_2dx1d:
-       push   %ebx
-       mov    8(%esp),%edx
-       mov    12(%esp),%eax
-       mov    16(%esp),%ebx
-       div    %ebx
-       mov    20(%esp),%ebx
-       mov    %eax,0(%ebx)
-       mov    24(%esp),%ebx
-       mov    %edx,0(%ebx)
-       xor    %eax,%eax		/ return zero
-       pop    %ebx
-       ret    
-       nop
-  

mozilla/security/nss/lib/freebl/mpi/mpi_mips.s

@@ -1,502 +0,0 @@
-/*
- * The contents of this file are subject to the Mozilla Public
- * License Version 1.1 (the "License"); you may not use this file
- * except in compliance with the License. You may obtain a copy of
- * the License at http://www.mozilla.org/MPL/
- * 
- * Software distributed under the License is distributed on an "AS
- * IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
- * implied. See the License for the specific language governing
- * rights and limitations under the License.
- * 
- * The Original Code is the Netscape security libraries.
- * 
- * The Initial Developer of the Original Code is Netscape
- * Communications Corporation.	Portions created by Netscape are 
- * Copyright (C) 2000 Netscape Communications Corporation.  All
- * Rights Reserved.
- * 
- * Contributor(s):
- * 
- * Alternatively, the contents of this file may be used under the
- * terms of the GNU General Public License Version 2 or later (the
- * "GPL"), in which case the provisions of the GPL are applicable 
- * instead of those above.	If you wish to allow use of your 
- * version of this file only under the terms of the GPL and not to
- * allow others to use your version of this file under the MPL,
- * indicate your decision by deleting the provisions above and
- * replace them with the notice and other provisions required by
- * the GPL.  If you do not delete the provisions above, a recipient
- * may use your version of this file under either the MPL or the
- * GPL.
- *  $Id: mpi_mips.s,v 1.2 2000-08-31 02:40:32 nelsonb%netscape.com Exp $
- */
-#include <regdef.h>
-        .set    noreorder
-        .set    noat
-
-        .section        .text, 1, 0x00000006, 4, 4
-.text:
-        .section        .text
-
-        .ent    s_mpv_mul_d_add
-        .globl  s_mpv_mul_d_add
-
-s_mpv_mul_d_add: 
- #/* c += a * b */
- #void s_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, 
- #			      mp_digit *c)
- #{
- #  mp_digit   a0, a1;	regs a4, a5
- #  mp_digit   c0, c1;  regs a6, a7
- #  mp_digit   cy = 0;  reg t2
- #  mp_word    w0, w1;  regs t0, t1
- #
- #  if (a_len) {
-	beq	a1,zero,.L.1
-	move	t2,zero		# cy = 0
-	dsll32	a2,a2,0		# "b" is sometimes negative (?!?!)
-	dsrl32	a2,a2,0		# This clears the upper 32 bits.
- #    a0 = a[0];
-	lwu	a4,0(a0)
- #    w0 = ((mp_word)b * a0);
-	dmultu	a2,a4
- #    if (--a_len) {
-	addiu	a1,a1,-1
-	beq	a1,zero,.L.2
- #      while (a_len >= 2) {
-	sltiu	t3,a1,2
-	bne	t3,zero,.L.3
- #	  a1     = a[1];
-	lwu	a5,4(a0)
-.L.4:
- #	  a_len -= 2;
-        addiu	a1,a1,-2
- #	  c0     = c[0];
-	lwu	a6,0(a3)
- #	  w0    += cy;
-	mflo	t0
-	daddu	t0,t0,t2
- #	  w0    += c0;
-	daddu	t0,t0,a6
- #	  w1     = (mp_word)b * a1; 
-	dmultu	a2,a5			#
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  a0     = a[2];
-	lwu	a4,8(a0)
- #	  a     += 2;
-	addiu	a0,a0,8
- #	  c1     = c[1];
-	lwu	a7,4(a3)
- #	  w1    += cy;
-	mflo	t1
-	daddu	t1,t1,t2
- #	  w1    += c1;
-	daddu	t1,t1,a7
- #	  w0     = (mp_word)b * a0;
-	dmultu	a2,a4			#
- #	  cy     = CARRYOUT(w1);
-	dsrl32	t2,t1,0
- #	  c[1]   = ACCUM(w1);
-	sw	t1,4(a3)
- #	  c     += 2;
-	addiu	a3,a3,8
-	sltiu	t3,a1,2
-	beq	t3,zero,.L.4
- #	  a1     = a[1];
-	lwu	a5,4(a0)
- #      }
-.L.3:
- #      c0       = c[0];
-	lwu	a6,0(a3)
- #      w0      += cy;
- #      if (a_len) {
-	mflo	t0
-	beq	a1,zero,.L.5
-	daddu	t0,t0,t2
- #	  w1     = (mp_word)b * a1; 
-	dmultu	a2,a5
- #	  w0    += c0;
-	daddu	t0,t0,a6		#
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  c1     = c[1];
-	lwu	a7,4(a3)
- #	  w1    += cy;
-	mflo	t1
-	daddu	t1,t1,t2
- #	  w1    += c1;
-	daddu	t1,t1,a7
- #	  c[1]   = ACCUM(w1);
-	sw	t1,4(a3)
- #	  cy     = CARRYOUT(w1);
-	dsrl32	t2,t1,0
- #	  c     += 1;
-	b	.L.6
-	addiu	a3,a3,4
- #      } else {
-.L.5:
- #	  w0    += c0;
-	daddu	t0,t0,a6
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  cy     = CARRYOUT(w0);
-	b	.L.6
-	dsrl32	t2,t0,0
- #      }
- #    } else {
-.L.2:
- #      c0     = c[0];
-	lwu	a6,0(a3)
- #      w0    += c0;
-	mflo	t0
-	daddu	t0,t0,a6
- #      c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #      cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #    }
-.L.6:
- #    c[1] = cy;
-	jr	ra
-	sw	t2,4(a3)
- #  }
-.L.1:
-	jr	ra
-	nop
- #}
- #
-        .end    s_mpv_mul_d_add
-
-        .ent    s_mpv_mul_d_add_prop
-        .globl  s_mpv_mul_d_add_prop
-
-s_mpv_mul_d_add_prop: 
- #/* c += a * b */
- #void s_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, 
- #			      mp_digit *c)
- #{
- #  mp_digit   a0, a1;	regs a4, a5
- #  mp_digit   c0, c1;  regs a6, a7
- #  mp_digit   cy = 0;  reg t2
- #  mp_word    w0, w1;  regs t0, t1
- #
- #  if (a_len) {
-	beq	a1,zero,.M.1
-	move	t2,zero		# cy = 0
-	dsll32	a2,a2,0		# "b" is sometimes negative (?!?!)
-	dsrl32	a2,a2,0		# This clears the upper 32 bits.
- #    a0 = a[0];
-	lwu	a4,0(a0)
- #    w0 = ((mp_word)b * a0);
-	dmultu	a2,a4
- #    if (--a_len) {
-	addiu	a1,a1,-1
-	beq	a1,zero,.M.2
- #      while (a_len >= 2) {
-	sltiu	t3,a1,2
-	bne	t3,zero,.M.3
- #	  a1     = a[1];
-	lwu	a5,4(a0)
-.M.4:
- #	  a_len -= 2;
-        addiu	a1,a1,-2
- #	  c0     = c[0];
-	lwu	a6,0(a3)
- #	  w0    += cy;
-	mflo	t0
-	daddu	t0,t0,t2
- #	  w0    += c0;
-	daddu	t0,t0,a6
- #	  w1     = (mp_word)b * a1; 
-	dmultu	a2,a5			#
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  a0     = a[2];
-	lwu	a4,8(a0)
- #	  a     += 2;
-	addiu	a0,a0,8
- #	  c1     = c[1];
-	lwu	a7,4(a3)
- #	  w1    += cy;
-	mflo	t1
-	daddu	t1,t1,t2
- #	  w1    += c1;
-	daddu	t1,t1,a7
- #	  w0     = (mp_word)b * a0;
-	dmultu	a2,a4			#
- #	  cy     = CARRYOUT(w1);
-	dsrl32	t2,t1,0
- #	  c[1]   = ACCUM(w1);
-	sw	t1,4(a3)
- #	  c     += 2;
-	addiu	a3,a3,8
-	sltiu	t3,a1,2
-	beq	t3,zero,.M.4
- #	  a1     = a[1];
-	lwu	a5,4(a0)
- #      }
-.M.3:
- #      c0       = c[0];
-	lwu	a6,0(a3)
- #      w0      += cy;
- #      if (a_len) {
-	mflo	t0
-	beq	a1,zero,.M.5
-	daddu	t0,t0,t2
- #	  w1     = (mp_word)b * a1; 
-	dmultu	a2,a5
- #	  w0    += c0;
-	daddu	t0,t0,a6		#
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  c1     = c[1];
-	lwu	a7,4(a3)
- #	  w1    += cy;
-	mflo	t1
-	daddu	t1,t1,t2
- #	  w1    += c1;
-	daddu	t1,t1,a7
- #	  c[1]   = ACCUM(w1);
-	sw	t1,4(a3)
- #	  cy     = CARRYOUT(w1);
-	dsrl32	t2,t1,0
- #	  c     += 1;
-	b	.M.6
-	addiu	a3,a3,8
- #      } else {
-.M.5:
- #	  w0    += c0;
-	daddu	t0,t0,a6
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
-	b	.M.6
-	addiu	a3,a3,4
- #      }
- #    } else {
-.M.2:
- #      c0     = c[0];
-	lwu	a6,0(a3)
- #      w0    += c0;
-	mflo	t0
-	daddu	t0,t0,a6
- #      c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #      cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
-	addiu	a3,a3,4
- #    }
-.M.6:
-
- #    while (cy) {
-	beq	t2,zero,.M.1
-	nop
-.M.7:
- #      mp_word w = (mp_word)*c + cy;
-	lwu	a6,0(a3)
-	daddu	t2,t2,a6
- #      *c++ = ACCUM(w);
-	sw	t2,0(a3)
- #      cy = CARRYOUT(w);
-	dsrl32	t2,t2,0
-	bne	t2,zero,.M.7
-	addiu	a3,a3,4
-
- #  }
-.M.1:
-	jr	ra
-	nop
- #}
- #
-        .end    s_mpv_mul_d_add_prop
-
-        .ent    s_mpv_mul_d
-        .globl  s_mpv_mul_d
-
-s_mpv_mul_d: 
- #/* c = a * b */
- #void s_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, 
- #			      mp_digit *c)
- #{
- #  mp_digit   a0, a1;	regs a4, a5
- #  mp_digit   cy = 0;  reg t2
- #  mp_word    w0, w1;  regs t0, t1
- #
- #  if (a_len) {
-	beq	a1,zero,.N.1
-	move	t2,zero		# cy = 0
-	dsll32	a2,a2,0		# "b" is sometimes negative (?!?!)
-	dsrl32	a2,a2,0		# This clears the upper 32 bits.
- #    a0 = a[0];
-	lwu	a4,0(a0)
- #    w0 = ((mp_word)b * a0);
-	dmultu	a2,a4
- #    if (--a_len) {
-	addiu	a1,a1,-1
-	beq	a1,zero,.N.2
- #      while (a_len >= 2) {
-	sltiu	t3,a1,2
-	bne	t3,zero,.N.3
- #	  a1     = a[1];
-	lwu	a5,4(a0)
-.N.4:
- #	  a_len -= 2;
-        addiu	a1,a1,-2
- #	  w0    += cy;
-	mflo	t0
-	daddu	t0,t0,t2
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #	  w1     = (mp_word)b * a1; 
-	dmultu	a2,a5	
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  a0     = a[2];
-	lwu	a4,8(a0)
- #	  a     += 2;
-	addiu	a0,a0,8
- #	  w1    += cy;
-	mflo	t1
-	daddu	t1,t1,t2
- #	  cy     = CARRYOUT(w1);
-	dsrl32	t2,t1,0
- #	  w0     = (mp_word)b * a0;
-	dmultu	a2,a4	
- #	  c[1]   = ACCUM(w1);
-	sw	t1,4(a3)
- #	  c     += 2;
-	addiu	a3,a3,8
-	sltiu	t3,a1,2
-	beq	t3,zero,.N.4
- #	  a1     = a[1];
-	lwu	a5,4(a0)
- #      }
-.N.3:
- #      w0      += cy;
- #      if (a_len) {
-	mflo	t0
-	beq	a1,zero,.N.5
-	daddu	t0,t0,t2
- #	  w1     = (mp_word)b * a1; 
-	dmultu	a2,a5			#
- #	  cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  w1    += cy;
-	mflo	t1
-	daddu	t1,t1,t2
- #	  c[1]   = ACCUM(w1);
-	sw	t1,4(a3)
- #	  cy     = CARRYOUT(w1);
-	dsrl32	t2,t1,0
- #	  c     += 1;
-	b	.N.6
-	addiu	a3,a3,4
- #      } else {
-.N.5:
- #	  c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #	  cy     = CARRYOUT(w0);
-	b	.N.6
-	dsrl32	t2,t0,0
- #      }
- #    } else {
-.N.2:
-	mflo	t0
- #      c[0]   = ACCUM(w0);
-	sw	t0,0(a3)
- #      cy     = CARRYOUT(w0);
-	dsrl32	t2,t0,0
- #    }
-.N.6:
- #    c[1] = cy;
-	jr	ra
-	sw	t2,4(a3)
- #  }
-.N.1:
-	jr	ra
-	nop
- #}
- #
-        .end    s_mpv_mul_d
-
-
-        .ent    s_mpv_sqr_add_prop
-        .globl  s_mpv_sqr_add_prop
- #void   s_mpv_sqr_add_prop(const mp_digit *a, mp_size a_len, mp_digit *sqrs);
- #	registers
- #	a0		*a
- #	a1		a_len
- #	a2		*sqr
- #	a3		digit from *a, a_i
- #	a4		square of digit from a
- #	a5,a6		next 2 digits in sqr
- #	a7,t0		carry 
-s_mpv_sqr_add_prop:
-	move	a7,zero
-	move	t0,zero
-	lwu	a3,0(a0)
-	addiu	a1,a1,-1	# --a_len
-	dmultu	a3,a3
-	beq	a1,zero,.P.3	# jump if we've already done the only sqr
-	addiu	a0,a0,4		# ++a
-.P.2:
-        lwu	a5,0(a2)
-        lwu	a6,4(a2)
-	addiu	a2,a2,8		# sqrs += 2;
-	dsll32	a6,a6,0
-	daddu	a5,a5,a6
-	lwu	a3,0(a0)
-	addiu	a0,a0,4		# ++a
-	mflo	a4
-	daddu	a6,a5,a4
-	sltu	a7,a6,a5	# a7 = a6 < a5	detect overflow
-	dmultu	a3,a3
-	daddu	a4,a6,t0
-	sltu	t0,a4,a6
-	add	t0,t0,a7
-	sw	a4,-8(a2)
-	addiu	a1,a1,-1	# --a_len
-	dsrl32	a4,a4,0
-	bne	a1,zero,.P.2	# loop if a_len > 0
-	sw	a4,-4(a2)
-.P.3:
-        lwu	a5,0(a2)
-        lwu	a6,4(a2)
-	addiu	a2,a2,8		# sqrs += 2;
-	dsll32	a6,a6,0
-	daddu	a5,a5,a6
-	mflo	a4
-	daddu	a6,a5,a4
-	sltu	a7,a6,a5	# a7 = a6 < a5	detect overflow
-	daddu	a4,a6,t0
-	sltu	t0,a4,a6
-	add	t0,t0,a7
-	sw	a4,-8(a2)
-	beq	t0,zero,.P.9	# jump if no carry
-	dsrl32	a4,a4,0
-.P.8:
-	sw	a4,-4(a2)
-	/* propagate final carry */
-	lwu	a5,0(a2)
-	daddu	a6,a5,t0
-	sltu	t0,a6,a5
-	bne	t0,zero,.P.8	# loop if carry persists
-	addiu	a2,a2,4		# sqrs++
-.P.9:
-	jr	ra
-	sw	a4,-4(a2)
-
-        .end    s_mpv_sqr_add_prop

mozilla/security/nss/lib/freebl/mpi/mpi_sparc.c

@@ -1,361 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the Netscape security libraries.
- *
- * The Initial Developer of the Original Code is
- * Netscape Communications Corporation.
- * Portions created by the Initial Developer are Copyright (C) 2000
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mpi_sparc.c,v 1.6 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-/* Multiplication performance enhancements for sparc v8+vis CPUs. */
-
-#include "mpi-priv.h"
-#include <stddef.h>
-#include <sys/systeminfo.h>
-#include <strings.h>
-
-/* In the functions below, */
-/* vector y must be 8-byte aligned, and n must be even */
-/* returns carry out of high order word of result */
-/* maximum n is 256 */
-
-/* vector x += vector y * scaler a; where y is of length n words. */
-extern mp_digit mul_add_inp(mp_digit *x, const mp_digit *y, int n, mp_digit a);
-
-/* vector z = vector x + vector y * scaler a; where y is of length n words. */
-extern mp_digit mul_add(mp_digit *z, const mp_digit *x, const mp_digit *y, 
-			int n, mp_digit a);
-
-/* v8 versions of these functions run on any Sparc v8 CPU. */
-
-/* This trick works on Sparc V8 CPUs with the Workshop compilers. */
-#define MP_MUL_DxD(a, b, Phi, Plo) \
-  { unsigned long long product = (unsigned long long)a * b; \
-    Plo = (mp_digit)product; \
-    Phi = (mp_digit)(product >> MP_DIGIT_BIT); }
-
-/* c = a * b */
-static void 
-v8_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-#if !defined(MP_NO_MP_WORD)
-  mp_digit   d = 0;
-
-  /* Inner product:  Digits of a */
-  while (a_len--) {
-    mp_word w = ((mp_word)b * *a++) + d;
-    *c++ = ACCUM(w);
-    d = CARRYOUT(w);
-  }
-  *c = d;
-#else
-  mp_digit carry = 0;
-  while (a_len--) {
-    mp_digit a_i = *a++;
-    mp_digit a0b0, a1b1;
-
-    MP_MUL_DxD(a_i, b, a1b1, a0b0);
-
-    a0b0 += carry;
-    if (a0b0 < carry)
-      ++a1b1;
-    *c++ = a0b0;
-    carry = a1b1;
-  }
-  *c = carry;
-#endif
-}
-
-/* c += a * b */
-static void 
-v8_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-#if !defined(MP_NO_MP_WORD)
-  mp_digit   d = 0;
-
-  /* Inner product:  Digits of a */
-  while (a_len--) {
-    mp_word w = ((mp_word)b * *a++) + *c + d;
-    *c++ = ACCUM(w);
-    d = CARRYOUT(w);
-  }
-  *c = d;
-#else
-  mp_digit carry = 0;
-  while (a_len--) {
-    mp_digit a_i = *a++;
-    mp_digit a0b0, a1b1;
-
-    MP_MUL_DxD(a_i, b, a1b1, a0b0);
-
-    a0b0 += carry;
-    if (a0b0 < carry)
-      ++a1b1;
-    a0b0 += a_i = *c;
-    if (a0b0 < a_i)
-      ++a1b1;
-    *c++ = a0b0;
-    carry = a1b1;
-  }
-  *c = carry;
-#endif
-}
-
-/* Presently, this is only used by the Montgomery arithmetic code. */
-/* c += a * b */
-static void 
-v8_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-#if !defined(MP_NO_MP_WORD)
-  mp_digit   d = 0;
-
-  /* Inner product:  Digits of a */
-  while (a_len--) {
-    mp_word w = ((mp_word)b * *a++) + *c + d;
-    *c++ = ACCUM(w);
-    d = CARRYOUT(w);
-  }
-
-  while (d) {
-    mp_word w = (mp_word)*c + d;
-    *c++ = ACCUM(w);
-    d = CARRYOUT(w);
-  }
-#else
-  mp_digit carry = 0;
-  while (a_len--) {
-    mp_digit a_i = *a++;
-    mp_digit a0b0, a1b1;
-
-    MP_MUL_DxD(a_i, b, a1b1, a0b0);
-
-    a0b0 += carry;
-    if (a0b0 < carry)
-      ++a1b1;
-
-    a0b0 += a_i = *c;
-    if (a0b0 < a_i)
-      ++a1b1;
-
-    *c++ = a0b0;
-    carry = a1b1;
-  }
-  while (carry) {
-    mp_digit c_i = *c;
-    carry += c_i;
-    *c++ = carry;
-    carry = carry < c_i;
-  }
-#endif
-}
-
-/* vis versions of these functions run only on v8+vis or v9+vis CPUs. */
-
-/* c = a * b */
-static void 
-vis_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    mp_digit d;
-    mp_digit x[258];
-    if (a_len <= 256) {
-	if (a == c || ((ptrdiff_t)a & 0x7) != 0 || (a_len & 1) != 0) {
-	    mp_digit * px;
-	    px = (((ptrdiff_t)x & 0x7) != 0) ? x + 1 : x;
-	    memcpy(px, a, a_len * sizeof(*a));
-	    a = px;
-	    if (a_len & 1) {
-		px[a_len] = 0;
-	    }
-	}
-	s_mp_setz(c, a_len + 1);
-	d = mul_add_inp(c, a, a_len, b);
-	c[a_len] = d;
-    } else {
-	v8_mpv_mul_d(a, a_len, b, c);
-    }
-}
-
-/* c += a * b, where a is a_len words long. */
-static void     
-vis_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    mp_digit d;
-    mp_digit x[258];
-    if (a_len <= 256) {
-	if (((ptrdiff_t)a & 0x7) != 0 || (a_len & 1) != 0) {
-	    mp_digit * px;
-	    px = (((ptrdiff_t)x & 0x7) != 0) ? x + 1 : x;
-	    memcpy(px, a, a_len * sizeof(*a));
-	    a = px;
-	    if (a_len & 1) {
-		px[a_len] = 0;
-	    }
-	}
-	d = mul_add_inp(c, a, a_len, b);
-	c[a_len] = d;
-    } else {
-	v8_mpv_mul_d_add(a, a_len, b, c);
-    }
-}
-
-/* c += a * b, where a is y words long. */
-static void     
-vis_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, 
-			 mp_digit *c)
-{
-    mp_digit d;
-    mp_digit x[258];
-    if (a_len <= 256) {
-	if (((ptrdiff_t)a & 0x7) != 0 || (a_len & 1) != 0) {
-	    mp_digit * px;
-	    px = (((ptrdiff_t)x & 0x7) != 0) ? x + 1 : x;
-	    memcpy(px, a, a_len * sizeof(*a));
-	    a = px;
-	    if (a_len & 1) {
-		px[a_len] = 0;
-	    }
-	}
-	d = mul_add_inp(c, a, a_len, b);
-	if (d) {
-	    c += a_len;
-	    do {
-		mp_digit sum = d + *c;
-		*c++ = sum;
-		d = sum < d;
-	    } while (d);
-	}
-    } else {
-	v8_mpv_mul_d_add_prop(a, a_len, b, c);
-    }
-}
-
-#if defined(SOLARIS2_5)
-static int
-isSparcV8PlusVis(void)
-{
-    long buflen;
-    int  rv             = 0;    /* false */
-    char buf[256];
-    buflen = sysinfo(SI_MACHINE, buf, sizeof buf);
-    if (buflen > 0) {
-        rv = (!strcmp(buf, "sun4u") || !strcmp(buf, "sun4u1"));
-    }
-    return rv;
-}
-#else   /* SunOS2.6or higher has SI_ISALIST */
-
-static int
-isSparcV8PlusVis(void)
-{
-    long buflen;
-    int  rv             = 0;    /* false */
-    char buf[256];
-    buflen = sysinfo(SI_ISALIST, buf, sizeof buf);
-    if (buflen > 0) {
-#if defined(MP_USE_LONG_DIGIT)
-        char * found = strstr(buf, "sparcv9+vis");
-#else
-        char * found = strstr(buf, "sparcv8plus+vis");
-#endif
-        rv = (found != 0);
-    }
-    return rv;
-}
-#endif
-
-typedef void MPVmpy(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c);
-
-/* forward static function declarations */
-static MPVmpy sp_mpv_mul_d;
-static MPVmpy sp_mpv_mul_d_add;
-static MPVmpy sp_mpv_mul_d_add_prop;
-
-static MPVmpy *p_mpv_mul_d		= &sp_mpv_mul_d;
-static MPVmpy *p_mpv_mul_d_add		= &sp_mpv_mul_d_add;
-static MPVmpy *p_mpv_mul_d_add_prop	= &sp_mpv_mul_d_add_prop;
-
-static void
-initPtrs(void)
-{
-    if (isSparcV8PlusVis()) {
-	p_mpv_mul_d = 		&vis_mpv_mul_d;
-	p_mpv_mul_d_add = 	&vis_mpv_mul_d_add;
-	p_mpv_mul_d_add_prop = 	&vis_mpv_mul_d_add_prop;
-    } else {
-	p_mpv_mul_d = 		&v8_mpv_mul_d;
-	p_mpv_mul_d_add = 	&v8_mpv_mul_d_add;
-	p_mpv_mul_d_add_prop = 	&v8_mpv_mul_d_add_prop;
-    }
-}
-
-static void 
-sp_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    initPtrs();
-    (* p_mpv_mul_d)(a, a_len, b, c);
-}
-
-static void 
-sp_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    initPtrs();
-    (* p_mpv_mul_d_add)(a, a_len, b, c);
-}
-
-static void 
-sp_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    initPtrs();
-    (* p_mpv_mul_d_add_prop)(a, a_len, b, c);
-}
-
-
-/* This is the external interface */
-
-void 
-s_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    (* p_mpv_mul_d)(a, a_len, b, c);
-}
-
-void 
-s_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    (* p_mpv_mul_d_add)(a, a_len, b, c);
-}
-
-void 
-s_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *c)
-{
-    (* p_mpv_mul_d_add_prop)(a, a_len, b, c);
-}
-
-

mozilla/security/nss/lib/freebl/mpi/mpi_x86.asm

@@ -1,356 +0,0 @@
-;
-;  mpi_x86.asm - assembly language implementation of s_mpv_ functions.
-; 
-; ***** BEGIN LICENSE BLOCK *****
-; Version: MPL 1.1/GPL 2.0/LGPL 2.1
-;
-; The contents of this file are subject to the Mozilla Public License Version
-; 1.1 (the "License"); you may not use this file except in compliance with
-; the License. You may obtain a copy of the License at
-; http://www.mozilla.org/MPL/
-;
-; Software distributed under the License is distributed on an "AS IS" basis,
-; WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
-; for the specific language governing rights and limitations under the
-; License.
-;
-; The Original Code is the Netscape security libraries.
-;
-; The Initial Developer of the Original Code is
-; Netscape Communications Corporation.
-; Portions created by the Initial Developer are Copyright (C) 2000
-; the Initial Developer. All Rights Reserved.
-;
-; Contributor(s):
-;
-; Alternatively, the contents of this file may be used under the terms of
-; either the GNU General Public License Version 2 or later (the "GPL"), or
-; the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
-; in which case the provisions of the GPL or the LGPL are applicable instead
-; of those above. If you wish to allow use of your version of this file only
-; under the terms of either the GPL or the LGPL, and not to allow others to
-; use your version of this file under the terms of the MPL, indicate your
-; decision by deleting the provisions above and replace them with the notice
-; and other provisions required by the GPL or the LGPL. If you do not delete
-; the provisions above, a recipient may use your version of this file under
-; the terms of any one of the MPL, the GPL or the LGPL.
-;
-; ***** END LICENSE BLOCK *****
-
-;   $Id: mpi_x86.asm,v 1.2 2005-02-02 22:28:22 gerv%gerv.net Exp $
-
-	.386p
-	.MODEL	FLAT
-	ASSUME  CS: FLAT, DS: FLAT, SS: FLAT
-_TEXT   SEGMENT
-
-;   ebp - 36:	caller's esi
-;   ebp - 32:	caller's edi
-;   ebp - 28:	
-;   ebp - 24:	
-;   ebp - 20:	
-;   ebp - 16:	
-;   ebp - 12:	
-;   ebp - 8:	
-;   ebp - 4:	
-;   ebp + 0:	caller's ebp
-;   ebp + 4:	return address
-;   ebp + 8:	a	argument
-;   ebp + 12:	a_len	argument
-;   ebp + 16:	b	argument
-;   ebp + 20:	c	argument
-;   registers:
-;  	eax:
-; 	ebx:	carry
-; 	ecx:	a_len
-; 	edx:
-; 	esi:	a ptr
-; 	edi:	c ptr
-
-public	_s_mpv_mul_d
-_s_mpv_mul_d PROC NEAR
-    push   ebp
-    mov    ebp,esp
-    sub    esp,28
-    push   edi
-    push   esi
-    push   ebx
-    mov    ebx,0		; carry = 0
-    mov    ecx,[ebp+12]		; ecx = a_len
-    mov    edi,[ebp+20]
-    cmp    ecx,0
-    je     L_2			; jmp if a_len == 0
-    mov    esi,[ebp+8]		; esi = a
-    cld
-L_1:
-    lodsd			; eax = [ds:esi]; esi += 4
-    mov    edx,[ebp+16]		; edx = b
-    mul    edx			; edx:eax = Phi:Plo = a_i * b
-
-    add    eax,ebx		; add carry (ebx) to edx:eax
-    adc    edx,0
-    mov    ebx,edx		; high half of product becomes next carry
-
-    stosd			; [es:edi] = ax; edi += 4;
-    dec    ecx			; --a_len
-    jnz    L_1			; jmp if a_len != 0
-L_2:
-    mov    [edi],ebx		; *c = carry
-    pop    ebx
-    pop    esi
-    pop    edi
-    leave  
-    ret    
-    nop
-_s_mpv_mul_d ENDP
-
-;   ebp - 36:	caller's esi
-;   ebp - 32:	caller's edi
-;   ebp - 28:	
-;   ebp - 24:	
-;   ebp - 20:	
-;   ebp - 16:	
-;   ebp - 12:	
-;   ebp - 8:	
-;   ebp - 4:	
-;   ebp + 0:	caller's ebp
-;   ebp + 4:	return address
-;   ebp + 8:	a	argument
-;   ebp + 12:	a_len	argument
-;   ebp + 16:	b	argument
-;   ebp + 20:	c	argument
-;   registers:
-;  	eax:
-; 	ebx:	carry
-; 	ecx:	a_len
-; 	edx:
-; 	esi:	a ptr
-; 	edi:	c ptr
-public	_s_mpv_mul_d_add
-_s_mpv_mul_d_add PROC NEAR
-    push   ebp
-    mov    ebp,esp
-    sub    esp,28
-    push   edi
-    push   esi
-    push   ebx
-    mov    ebx,0		; carry = 0
-    mov    ecx,[ebp+12]		; ecx = a_len
-    mov    edi,[ebp+20]
-    cmp    ecx,0
-    je     L_4			; jmp if a_len == 0
-    mov    esi,[ebp+8]		; esi = a
-    cld
-L_3:
-    lodsd			; eax = [ds:esi]; esi += 4
-    mov    edx,[ebp+16]		; edx = b
-    mul    edx			; edx:eax = Phi:Plo = a_i * b
-
-    add    eax,ebx		; add carry (ebx) to edx:eax
-    adc    edx,0
-    mov    ebx,[edi]		; add in current word from *c
-    add    eax,ebx		
-    adc    edx,0
-    mov    ebx,edx		; high half of product becomes next carry
-
-    stosd			; [es:edi] = ax; edi += 4;
-    dec    ecx			; --a_len
-    jnz    L_3			; jmp if a_len != 0
-L_4:
-    mov    [edi],ebx		; *c = carry
-    pop    ebx
-    pop    esi
-    pop    edi
-    leave  
-    ret    
-    nop
-_s_mpv_mul_d_add ENDP
-
-;   ebp - 36:	caller's esi
-;   ebp - 32:	caller's edi
-;   ebp - 28:	
-;   ebp - 24:	
-;   ebp - 20:	
-;   ebp - 16:	
-;   ebp - 12:	
-;   ebp - 8:	
-;   ebp - 4:	
-;   ebp + 0:	caller's ebp
-;   ebp + 4:	return address
-;   ebp + 8:	a	argument
-;   ebp + 12:	a_len	argument
-;   ebp + 16:	b	argument
-;   ebp + 20:	c	argument
-;   registers:
-;  	eax:
-; 	ebx:	carry
-; 	ecx:	a_len
-; 	edx:
-; 	esi:	a ptr
-; 	edi:	c ptr
-public	_s_mpv_mul_d_add_prop
-_s_mpv_mul_d_add_prop PROC NEAR
-    push   ebp
-    mov    ebp,esp
-    sub    esp,28
-    push   edi
-    push   esi
-    push   ebx
-    mov    ebx,0		; carry = 0
-    mov    ecx,[ebp+12]		; ecx = a_len
-    mov    edi,[ebp+20]
-    cmp    ecx,0
-    je     L_6			; jmp if a_len == 0
-    cld
-    mov    esi,[ebp+8]		; esi = a
-L_5:
-    lodsd			; eax = [ds:esi]; esi += 4
-    mov    edx,[ebp+16]		; edx = b
-    mul    edx			; edx:eax = Phi:Plo = a_i * b
-
-    add    eax,ebx		; add carry (ebx) to edx:eax
-    adc    edx,0
-    mov    ebx,[edi]		; add in current word from *c
-    add    eax,ebx		
-    adc    edx,0
-    mov    ebx,edx		; high half of product becomes next carry
-
-    stosd			; [es:edi] = ax; edi += 4;
-    dec    ecx			; --a_len
-    jnz    L_5			; jmp if a_len != 0
-L_6:
-    cmp    ebx,0		; is carry zero?
-    jz     L_8
-    mov    eax,[edi]		; add in current word from *c
-    add    eax,ebx
-    stosd			; [es:edi] = ax; edi += 4;
-    jnc    L_8
-L_7:
-    mov    eax,[edi]		; add in current word from *c
-    adc    eax,0
-    stosd			; [es:edi] = ax; edi += 4;
-    jc     L_7
-L_8:
-    pop    ebx
-    pop    esi
-    pop    edi
-    leave  
-    ret    
-    nop
-_s_mpv_mul_d_add_prop ENDP
-
-;   ebp - 20:	caller's esi
-;   ebp - 16:	caller's edi
-;   ebp - 12:	
-;   ebp - 8:	carry
-;   ebp - 4:	a_len	local
-;   ebp + 0:	caller's ebp
-;   ebp + 4:	return address
-;   ebp + 8:	pa	argument
-;   ebp + 12:	a_len	argument
-;   ebp + 16:	ps	argument
-;   ebp + 20:	
-;   registers:
-;  	eax:
-; 	ebx:	carry
-; 	ecx:	a_len
-; 	edx:
-; 	esi:	a ptr
-; 	edi:	c ptr
-
-public	_s_mpv_sqr_add_prop
-_s_mpv_sqr_add_prop PROC NEAR
-     push   ebp
-     mov    ebp,esp
-     sub    esp,12
-     push   edi
-     push   esi
-     push   ebx
-     mov    ebx,0		; carry = 0
-     mov    ecx,[ebp+12]	; a_len
-     mov    edi,[ebp+16]	; edi = ps
-     cmp    ecx,0
-     je     L_11		; jump if a_len == 0
-     cld
-     mov    esi,[ebp+8]		; esi = pa
-L_10:
-     lodsd			; eax = [ds:si]; si += 4;
-     mul    eax
-
-     add    eax,ebx		; add "carry"
-     adc    edx,0
-     mov    ebx,[edi]
-     add    eax,ebx		; add low word from result
-     mov    ebx,[edi+4]
-     stosd			; [es:di] = eax; di += 4;
-     adc    edx,ebx		; add high word from result
-     mov    ebx,0
-     mov    eax,edx
-     adc    ebx,0
-     stosd			; [es:di] = eax; di += 4;
-     dec    ecx			; --a_len
-     jnz    L_10		; jmp if a_len != 0
-L_11:
-    cmp    ebx,0		; is carry zero?
-    jz     L_14
-    mov    eax,[edi]		; add in current word from *c
-    add    eax,ebx
-    stosd			; [es:edi] = ax; edi += 4;
-    jnc    L_14
-L_12:
-    mov    eax,[edi]		; add in current word from *c
-    adc    eax,0
-    stosd			; [es:edi] = ax; edi += 4;
-    jc     L_12
-L_14:
-    pop    ebx
-    pop    esi
-    pop    edi
-    leave  
-    ret    
-    nop
-_s_mpv_sqr_add_prop ENDP
-
-; 
-;  Divide 64-bit (Nhi,Nlo) by 32-bit divisor, which must be normalized
-;  so its high bit is 1.   This code is from NSPR.
-; 
-;  mp_err s_mpv_div_2dx1d(mp_digit Nhi, mp_digit Nlo, mp_digit divisor,
-;  		          mp_digit *qp, mp_digit *rp)
-
-;  Dump of assembler code for function s_mpv_div_2dx1d:
-;  
-;   esp +  0:   Caller's ebx
-;   esp +  4:	return address
-;   esp +  8:	Nhi	argument
-;   esp + 12:	Nlo	argument
-;   esp + 16:	divisor	argument
-;   esp + 20:	qp	argument
-;   esp + 24:   rp	argument
-;   registers:
-;  	eax:
-; 	ebx:	carry
-; 	ecx:	a_len
-; 	edx:
-; 	esi:	a ptr
-; 	edi:	c ptr
-;  
-public	_s_mpv_div_2dx1d
-_s_mpv_div_2dx1d PROC NEAR
-       push   ebx
-       mov    edx,[esp+8]
-       mov    eax,[esp+12]
-       mov    ebx,[esp+16]
-       div    ebx
-       mov    ebx,[esp+20]
-       mov    [ebx],eax
-       mov    ebx,[esp+24]
-       mov    [ebx],edx
-       xor    eax,eax		; return zero
-       pop    ebx
-       ret    
-       nop
-_s_mpv_div_2dx1d ENDP
-
-_TEXT	ENDS
-END

mozilla/security/nss/lib/freebl/mpi/mpi_x86.s

@@ -1,345 +0,0 @@
- #
- # The contents of this file are subject to the Mozilla Public
- # License Version 1.1 (the "License"); you may not use this file
- # except in compliance with the License. You may obtain a copy of
- # the License at http://www.mozilla.org/MPL/
- # 
- # Software distributed under the License is distributed on an "AS
- # IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
- # implied. See the License for the specific language governing
- # rights and limitations under the License.
- # 
- # The Original Code is the Netscape security libraries.
- # 
- # The Initial Developer of the Original Code is Netscape
- # Communications Corporation.	Portions created by Netscape are 
- # Copyright (C) 2000 Netscape Communications Corporation.  All
- # Rights Reserved.
- # 
- # Contributor(s):
- # 
- # Alternatively, the contents of this file may be used under the
- # terms of the GNU General Public License Version 2 or later (the
- # "GPL"), in which case the provisions of the GPL are applicable 
- # instead of those above.	If you wish to allow use of your 
- # version of this file only under the terms of the GPL and not to
- # allow others to use your version of this file under the MPL,
- # indicate your decision by deleting the provisions above and
- # replace them with the notice and other provisions required by
- # the GPL.  If you do not delete the provisions above, a recipient
- # may use your version of this file under either the MPL or the
- # GPL.
- #  $Id: mpi_x86.s,v 1.4 2003-10-24 04:47:23 wchang0222%aol.com Exp $
- #
-
-.text
-
- #  ebp - 36:	caller's esi
- #  ebp - 32:	caller's edi
- #  ebp - 28:	
- #  ebp - 24:	
- #  ebp - 20:	
- #  ebp - 16:	
- #  ebp - 12:	
- #  ebp - 8:	
- #  ebp - 4:	
- #  ebp + 0:	caller's ebp
- #  ebp + 4:	return address
- #  ebp + 8:	a	argument
- #  ebp + 12:	a_len	argument
- #  ebp + 16:	b	argument
- #  ebp + 20:	c	argument
- #  registers:
- # 	eax:
- #	ebx:	carry
- #	ecx:	a_len
- #	edx:
- #	esi:	a ptr
- #	edi:	c ptr
-.globl	s_mpv_mul_d
-.type	s_mpv_mul_d,@function
-s_mpv_mul_d:
-    push   %ebp
-    mov    %esp,%ebp
-    sub    $28,%esp
-    push   %edi
-    push   %esi
-    push   %ebx
-    movl   $0,%ebx		# carry = 0
-    mov    12(%ebp),%ecx	# ecx = a_len
-    mov    20(%ebp),%edi
-    cmp    $0,%ecx
-    je     2f			# jmp if a_len == 0
-    mov    8(%ebp),%esi		# esi = a
-    cld
-1:
-    lodsl			# eax = [ds:esi]; esi += 4
-    mov    16(%ebp),%edx	# edx = b
-    mull   %edx			# edx:eax = Phi:Plo = a_i * b
-
-    add    %ebx,%eax		# add carry (%ebx) to edx:eax
-    adc    $0,%edx
-    mov    %edx,%ebx		# high half of product becomes next carry
-
-    stosl			# [es:edi] = ax; edi += 4;
-    dec    %ecx			# --a_len
-    jnz    1b			# jmp if a_len != 0
-2:
-    mov    %ebx,0(%edi)		# *c = carry
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- #  ebp - 36:	caller's esi
- #  ebp - 32:	caller's edi
- #  ebp - 28:	
- #  ebp - 24:	
- #  ebp - 20:	
- #  ebp - 16:	
- #  ebp - 12:	
- #  ebp - 8:	
- #  ebp - 4:	
- #  ebp + 0:	caller's ebp
- #  ebp + 4:	return address
- #  ebp + 8:	a	argument
- #  ebp + 12:	a_len	argument
- #  ebp + 16:	b	argument
- #  ebp + 20:	c	argument
- #  registers:
- # 	eax:
- #	ebx:	carry
- #	ecx:	a_len
- #	edx:
- #	esi:	a ptr
- #	edi:	c ptr
-.globl	s_mpv_mul_d_add
-.type	s_mpv_mul_d_add,@function
-s_mpv_mul_d_add:
-    push   %ebp
-    mov    %esp,%ebp
-    sub    $28,%esp
-    push   %edi
-    push   %esi
-    push   %ebx
-    movl   $0,%ebx		# carry = 0
-    mov    12(%ebp),%ecx	# ecx = a_len
-    mov    20(%ebp),%edi
-    cmp    $0,%ecx
-    je     4f			# jmp if a_len == 0
-    mov    8(%ebp),%esi		# esi = a
-    cld
-3:
-    lodsl			# eax = [ds:esi]; esi += 4
-    mov    16(%ebp),%edx	# edx = b
-    mull   %edx			# edx:eax = Phi:Plo = a_i * b
-
-    add    %ebx,%eax		# add carry (%ebx) to edx:eax
-    adc    $0,%edx
-    mov    0(%edi),%ebx		# add in current word from *c
-    add    %ebx,%eax		
-    adc    $0,%edx
-    mov    %edx,%ebx		# high half of product becomes next carry
-
-    stosl			# [es:edi] = ax; edi += 4;
-    dec    %ecx			# --a_len
-    jnz    3b			# jmp if a_len != 0
-4:
-    mov    %ebx,0(%edi)		# *c = carry
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- #  ebp - 36:	caller's esi
- #  ebp - 32:	caller's edi
- #  ebp - 28:	
- #  ebp - 24:	
- #  ebp - 20:	
- #  ebp - 16:	
- #  ebp - 12:	
- #  ebp - 8:	
- #  ebp - 4:	
- #  ebp + 0:	caller's ebp
- #  ebp + 4:	return address
- #  ebp + 8:	a	argument
- #  ebp + 12:	a_len	argument
- #  ebp + 16:	b	argument
- #  ebp + 20:	c	argument
- #  registers:
- # 	eax:
- #	ebx:	carry
- #	ecx:	a_len
- #	edx:
- #	esi:	a ptr
- #	edi:	c ptr
-.globl	s_mpv_mul_d_add_prop
-.type	s_mpv_mul_d_add_prop,@function
-s_mpv_mul_d_add_prop:
-    push   %ebp
-    mov    %esp,%ebp
-    sub    $28,%esp
-    push   %edi
-    push   %esi
-    push   %ebx
-    movl   $0,%ebx		# carry = 0
-    mov    12(%ebp),%ecx	# ecx = a_len
-    mov    20(%ebp),%edi
-    cmp    $0,%ecx
-    je     6f			# jmp if a_len == 0
-    cld
-    mov    8(%ebp),%esi		# esi = a
-5:
-    lodsl			# eax = [ds:esi]; esi += 4
-    mov    16(%ebp),%edx	# edx = b
-    mull   %edx			# edx:eax = Phi:Plo = a_i * b
-
-    add    %ebx,%eax		# add carry (%ebx) to edx:eax
-    adc    $0,%edx
-    mov    0(%edi),%ebx		# add in current word from *c
-    add    %ebx,%eax		
-    adc    $0,%edx
-    mov    %edx,%ebx		# high half of product becomes next carry
-
-    stosl			# [es:edi] = ax; edi += 4;
-    dec    %ecx			# --a_len
-    jnz    5b			# jmp if a_len != 0
-6:
-    cmp    $0,%ebx		# is carry zero?
-    jz     8f
-    mov    0(%edi),%eax		# add in current word from *c
-    add	   %ebx,%eax
-    stosl			# [es:edi] = ax; edi += 4;
-    jnc    8f
-7:
-    mov    0(%edi),%eax		# add in current word from *c
-    adc	   $0,%eax
-    stosl			# [es:edi] = ax; edi += 4;
-    jc     7b
-8:
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- #  ebp - 20:	caller's esi
- #  ebp - 16:	caller's edi
- #  ebp - 12:	
- #  ebp - 8:	carry
- #  ebp - 4:	a_len	local
- #  ebp + 0:	caller's ebp
- #  ebp + 4:	return address
- #  ebp + 8:	pa	argument
- #  ebp + 12:	a_len	argument
- #  ebp + 16:	ps	argument
- #  ebp + 20:	
- #  registers:
- # 	eax:
- #	ebx:	carry
- #	ecx:	a_len
- #	edx:
- #	esi:	a ptr
- #	edi:	c ptr
-
-.globl	s_mpv_sqr_add_prop
-.type	s_mpv_sqr_add_prop,@function
-s_mpv_sqr_add_prop:
-     push   %ebp
-     mov    %esp,%ebp
-     sub    $12,%esp
-     push   %edi
-     push   %esi
-     push   %ebx
-     movl   $0,%ebx		# carry = 0
-     mov    12(%ebp),%ecx	# a_len
-     mov    16(%ebp),%edi	# edi = ps
-     cmp    $0,%ecx
-     je     11f			# jump if a_len == 0
-     cld
-     mov    8(%ebp),%esi	# esi = pa
-10:
-     lodsl			# %eax = [ds:si]; si += 4;
-     mull   %eax
-
-     add    %ebx,%eax		# add "carry"
-     adc    $0,%edx
-     mov    0(%edi),%ebx
-     add    %ebx,%eax		# add low word from result
-     mov    4(%edi),%ebx
-     stosl			# [es:di] = %eax; di += 4;
-     adc    %ebx,%edx		# add high word from result
-     movl   $0,%ebx
-     mov    %edx,%eax
-     adc    $0,%ebx
-     stosl			# [es:di] = %eax; di += 4;
-     dec    %ecx		# --a_len
-     jnz    10b			# jmp if a_len != 0
-11:
-    cmp    $0,%ebx		# is carry zero?
-    jz     14f
-    mov    0(%edi),%eax		# add in current word from *c
-    add	   %ebx,%eax
-    stosl			# [es:edi] = ax; edi += 4;
-    jnc    14f
-12:
-    mov    0(%edi),%eax		# add in current word from *c
-    adc	   $0,%eax
-    stosl			# [es:edi] = ax; edi += 4;
-    jc     12b
-14:
-    pop    %ebx
-    pop    %esi
-    pop    %edi
-    leave  
-    ret    
-    nop
-
- #
- # Divide 64-bit (Nhi,Nlo) by 32-bit divisor, which must be normalized
- # so its high bit is 1.   This code is from NSPR.
- #
- # mp_err s_mpv_div_2dx1d(mp_digit Nhi, mp_digit Nlo, mp_digit divisor,
- # 		          mp_digit *qp, mp_digit *rp)
-
- #  esp +  0:   Caller's ebx
- #  esp +  4:	return address
- #  esp +  8:	Nhi	argument
- #  esp + 12:	Nlo	argument
- #  esp + 16:	divisor	argument
- #  esp + 20:	qp	argument
- #  esp + 24:   rp	argument
- #  registers:
- # 	eax:
- #	ebx:	carry
- #	ecx:	a_len
- #	edx:
- #	esi:	a ptr
- #	edi:	c ptr
- # 
-
-.globl	s_mpv_div_2dx1d
-.type	s_mpv_div_2dx1d,@function
-s_mpv_div_2dx1d:
-       push   %ebx
-       mov    8(%esp),%edx
-       mov    12(%esp),%eax
-       mov    16(%esp),%ebx
-       div    %ebx
-       mov    20(%esp),%ebx
-       mov    %eax,0(%ebx)
-       mov    24(%esp),%ebx
-       mov    %edx,0(%ebx)
-       xor    %eax,%eax		# return zero
-       pop    %ebx
-       ret    
-       nop
-  
- # Magic indicating no need for an executable stack
-.section .note.GNU-stack, "", @progbits
-.previous

mozilla/security/nss/lib/freebl/mpi/mplogic.c

@@ -1,465 +0,0 @@
-/*
- *  mplogic.c
- *
- *  Bitwise logical operations on MPI values
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1998
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mplogic.c,v 1.15 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-#include "mpi-priv.h"
-#include "mplogic.h"
-
-/* {{{ Lookup table for population count */
-
-static unsigned char bitc[] = {
-   0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4, 
-   1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 
-   1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 
-   2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 
-   1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 
-   2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 
-   2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 
-   3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 
-   1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 
-   2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 
-   2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 
-   3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 
-   2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 
-   3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 
-   3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 
-   4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8
-};
-
-/* }}} */
-
-/*------------------------------------------------------------------------*/
-/*
-  mpl_not(a, b)    - compute b = ~a
-  mpl_and(a, b, c) - compute c = a & b
-  mpl_or(a, b, c)  - compute c = a | b
-  mpl_xor(a, b, c) - compute c = a ^ b
- */
-
-/* {{{ mpl_not(a, b) */
-
-mp_err mpl_not(mp_int *a, mp_int *b)
-{
-  mp_err   res;
-  unsigned int      ix;
-
-  ARGCHK(a != NULL && b != NULL, MP_BADARG);
-
-  if((res = mp_copy(a, b)) != MP_OKAY)
-    return res;
-
-  /* This relies on the fact that the digit type is unsigned */
-  for(ix = 0; ix < USED(b); ix++) 
-    DIGIT(b, ix) = ~DIGIT(b, ix);
-
-  s_mp_clamp(b);
-
-  return MP_OKAY;
-
-} /* end mpl_not() */
-
-/* }}} */
-
-/* {{{ mpl_and(a, b, c) */
-
-mp_err mpl_and(mp_int *a, mp_int *b, mp_int *c)
-{
-  mp_int  *which, *other;
-  mp_err   res;
-  unsigned int      ix;
-
-  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);
-
-  if(USED(a) <= USED(b)) {
-    which = a;
-    other = b;
-  } else {
-    which = b;
-    other = a;
-  }
-
-  if((res = mp_copy(which, c)) != MP_OKAY)
-    return res;
-
-  for(ix = 0; ix < USED(which); ix++)
-    DIGIT(c, ix) &= DIGIT(other, ix);
-
-  s_mp_clamp(c);
-
-  return MP_OKAY;
-
-} /* end mpl_and() */
-
-/* }}} */
-
-/* {{{ mpl_or(a, b, c) */
-
-mp_err mpl_or(mp_int *a, mp_int *b, mp_int *c)
-{
-  mp_int  *which, *other;
-  mp_err   res;
-  unsigned int      ix;
-
-  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);
-
-  if(USED(a) >= USED(b)) {
-    which = a;
-    other = b;
-  } else {
-    which = b;
-    other = a;
-  }
-
-  if((res = mp_copy(which, c)) != MP_OKAY)
-    return res;
-
-  for(ix = 0; ix < USED(which); ix++)
-    DIGIT(c, ix) |= DIGIT(other, ix);
-
-  return MP_OKAY;
-
-} /* end mpl_or() */
-
-/* }}} */
-
-/* {{{ mpl_xor(a, b, c) */
-
-mp_err mpl_xor(mp_int *a, mp_int *b, mp_int *c)
-{
-  mp_int  *which, *other;
-  mp_err   res;
-  unsigned int      ix;
-
-  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);
-
-  if(USED(a) >= USED(b)) {
-    which = a;
-    other = b;
-  } else {
-    which = b;
-    other = a;
-  }
-
-  if((res = mp_copy(which, c)) != MP_OKAY)
-    return res;
-
-  for(ix = 0; ix < USED(which); ix++)
-    DIGIT(c, ix) ^= DIGIT(other, ix);
-
-  s_mp_clamp(c);
-
-  return MP_OKAY;
-
-} /* end mpl_xor() */
-
-/* }}} */
-
-/*------------------------------------------------------------------------*/
-/*
-  mpl_rsh(a, b, d)     - b = a >> d
-  mpl_lsh(a, b, d)     - b = a << d
- */
-
-/* {{{ mpl_rsh(a, b, d) */
-
-mp_err mpl_rsh(const mp_int *a, mp_int *b, mp_digit d)
-{
-  mp_err   res;
-
-  ARGCHK(a != NULL && b != NULL, MP_BADARG);
-
-  if((res = mp_copy(a, b)) != MP_OKAY)
-    return res;
-
-  s_mp_div_2d(b, d);
-
-  return MP_OKAY;
-
-} /* end mpl_rsh() */
-
-/* }}} */
-
-/* {{{ mpl_lsh(a, b, d) */
-
-mp_err mpl_lsh(const mp_int *a, mp_int *b, mp_digit d)
-{
-  mp_err   res;
-
-  ARGCHK(a != NULL && b != NULL, MP_BADARG);
-
-  if((res = mp_copy(a, b)) != MP_OKAY)
-    return res;
-
-  return s_mp_mul_2d(b, d);
-
-} /* end mpl_lsh() */
-
-/* }}} */
-
-/*------------------------------------------------------------------------*/
-/*
-  mpl_num_set(a, num)
-
-  Count the number of set bits in the binary representation of a.
-  Returns MP_OKAY and sets 'num' to be the number of such bits, if
-  possible.  If num is NULL, the result is thrown away, but it is
-  not considered an error.
-
-  mpl_num_clear() does basically the same thing for clear bits.
- */
-
-/* {{{ mpl_num_set(a, num) */
-
-mp_err mpl_num_set(mp_int *a, int *num)
-{
-  unsigned int   ix;
-  int            db, nset = 0;
-  mp_digit       cur;
-  unsigned char  reg;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  for(ix = 0; ix < USED(a); ix++) {
-    cur = DIGIT(a, ix);
-    
-    for(db = 0; db < sizeof(mp_digit); db++) {
-      reg = (unsigned char)(cur >> (CHAR_BIT * db));
-
-      nset += bitc[reg];
-    }
-  }
-
-  if(num)
-    *num = nset;
-
-  return MP_OKAY;
-
-} /* end mpl_num_set() */
-
-/* }}} */
-
-/* {{{ mpl_num_clear(a, num) */
-
-mp_err mpl_num_clear(mp_int *a, int *num)
-{
-  unsigned int   ix;
-  int            db, nset = 0;
-  mp_digit       cur;
-  unsigned char  reg;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  for(ix = 0; ix < USED(a); ix++) {
-    cur = DIGIT(a, ix);
-    
-    for(db = 0; db < sizeof(mp_digit); db++) {
-      reg = (unsigned char)(cur >> (CHAR_BIT * db));
-
-      nset += bitc[UCHAR_MAX - reg];
-    }
-  }
-
-  if(num)
-    *num = nset;
-
-  return MP_OKAY;
-
-
-} /* end mpl_num_clear() */
-
-/* }}} */
-
-/*------------------------------------------------------------------------*/
-/*
-  mpl_parity(a)
-
-  Determines the bitwise parity of the value given.  Returns MP_EVEN
-  if an even number of digits are set, MP_ODD if an odd number are
-  set.
- */
-
-/* {{{ mpl_parity(a) */
-
-mp_err mpl_parity(mp_int *a)
-{
-  unsigned int ix;
-  int      par = 0;
-  mp_digit cur;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  for(ix = 0; ix < USED(a); ix++) {
-    int   shft = (sizeof(mp_digit) * CHAR_BIT) / 2;
-
-    cur = DIGIT(a, ix);
-
-    /* Compute parity for current digit */
-    while(shft != 0) {
-      cur ^= (cur >> shft);
-      shft >>= 1;
-    }
-    cur &= 1;
-
-    /* XOR with running parity so far   */
-    par ^= cur;
-  }
-
-  if(par)
-    return MP_ODD;
-  else
-    return MP_EVEN;
-
-} /* end mpl_parity() */
-
-/* }}} */
-
-/*
-  mpl_set_bit
-
-  Returns MP_OKAY or some error code.
-  Grows a if needed to set a bit to 1.
- */
-mp_err mpl_set_bit(mp_int *a, mp_size bitNum, mp_size value)
-{
-  mp_size      ix;
-  mp_err       rv;
-  mp_digit     mask;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  ix = bitNum / MP_DIGIT_BIT;
-  if (ix + 1 > MP_USED(a)) {
-    rv = s_mp_pad(a, ix + 1);
-    if (rv != MP_OKAY)
-      return rv;
-  }
-
-  bitNum = bitNum % MP_DIGIT_BIT;
-  mask = (mp_digit)1 << bitNum;
-  if (value)
-    MP_DIGIT(a,ix) |= mask;
-  else
-    MP_DIGIT(a,ix) &= ~mask;
-  s_mp_clamp(a);
-  return MP_OKAY;
-}
-
-/*
-  mpl_get_bit
-
-  returns 0 or 1 or some (negative) error code.
- */
-mp_err mpl_get_bit(const mp_int *a, mp_size bitNum)
-{
-  mp_size      bit, ix;
-  mp_err       rv;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  ix = bitNum / MP_DIGIT_BIT;
-  ARGCHK(ix <= MP_USED(a) - 1, MP_RANGE);
-
-  bit   = bitNum % MP_DIGIT_BIT;
-  rv = (mp_err)(MP_DIGIT(a, ix) >> bit) & 1;
-  return rv;
-}
-
-/*
-  mpl_get_bits
-  - Extracts numBits bits from a, where the least significant extracted bit
-  is bit lsbNum.  Returns a negative value if error occurs.
-  - Because sign bit is used to indicate error, maximum number of bits to 
-  be returned is the lesser of (a) the number of bits in an mp_digit, or
-  (b) one less than the number of bits in an mp_err.
-  - lsbNum + numbits can be greater than the number of significant bits in
-  integer a, as long as bit lsbNum is in the high order digit of a.
- */
-mp_err mpl_get_bits(const mp_int *a, mp_size lsbNum, mp_size numBits) 
-{
-  mp_size    rshift = (lsbNum % MP_DIGIT_BIT);
-  mp_size    lsWndx = (lsbNum / MP_DIGIT_BIT);
-  mp_digit * digit  = MP_DIGITS(a) + lsWndx;
-  mp_digit   mask   = ((1 << numBits) - 1);
-
-  ARGCHK(numBits < CHAR_BIT * sizeof mask, MP_BADARG);
-  ARGCHK(MP_HOWMANY(lsbNum, MP_DIGIT_BIT) <= MP_USED(a), MP_RANGE);
-
-  if ((numBits + lsbNum % MP_DIGIT_BIT <= MP_DIGIT_BIT) ||
-      (lsWndx + 1 >= MP_USED(a))) {
-    mask &= (digit[0] >> rshift);
-  } else {
-    mask &= ((digit[0] >> rshift) | (digit[1] << (MP_DIGIT_BIT - rshift)));
-  }
-  return (mp_err)mask;
-}
-
-/*
-  mpl_significant_bits
-  returns number of significnant bits in abs(a).
-  returns 1 if value is zero.
- */
-mp_err mpl_significant_bits(const mp_int *a)
-{
-  mp_err bits 	= 0;
-  int    ix;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  ix = MP_USED(a);
-  for (ix = MP_USED(a); ix > 0; ) {
-    mp_digit d;
-    d = MP_DIGIT(a, --ix);
-    if (d) {
-      while (d) {
-	++bits;
-	d >>= 1;
-      }
-      break;
-    }
-  }
-  bits += ix * MP_DIGIT_BIT;
-  if (!bits)
-    bits = 1;
-  return bits;
-}
-
-/*------------------------------------------------------------------------*/
-/* HERE THERE BE DRAGONS                                                  */

mozilla/security/nss/lib/freebl/mpi/mplogic.h

@@ -1,85 +0,0 @@
-/*
- *  mplogic.h
- *
- *  Bitwise logical operations on MPI values
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1998
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mplogic.h,v 1.7 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-#ifndef _H_MPLOGIC_
-#define _H_MPLOGIC_
-
-#include "mpi.h"
-
-/*
-  The logical operations treat an mp_int as if it were a bit vector,
-  without regard to its sign (an mp_int is represented in a signed
-  magnitude format).  Values are treated as if they had an infinite
-  string of zeros left of the most-significant bit.
- */
-
-/* Parity results                    */
-
-#define MP_EVEN       MP_YES
-#define MP_ODD        MP_NO
-
-/* Bitwise functions                 */
-
-mp_err mpl_not(mp_int *a, mp_int *b);            /* one's complement  */
-mp_err mpl_and(mp_int *a, mp_int *b, mp_int *c); /* bitwise AND       */
-mp_err mpl_or(mp_int *a, mp_int *b, mp_int *c);  /* bitwise OR        */
-mp_err mpl_xor(mp_int *a, mp_int *b, mp_int *c); /* bitwise XOR       */
-
-/* Shift functions                   */
-
-mp_err mpl_rsh(const mp_int *a, mp_int *b, mp_digit d);   /* right shift    */
-mp_err mpl_lsh(const mp_int *a, mp_int *b, mp_digit d);   /* left shift     */
-
-/* Bit count and parity              */
-
-mp_err mpl_num_set(mp_int *a, int *num);         /* count set bits    */
-mp_err mpl_num_clear(mp_int *a, int *num);       /* count clear bits  */
-mp_err mpl_parity(mp_int *a);                    /* determine parity  */
-
-/* Get & Set the value of a bit */
-
-mp_err mpl_set_bit(mp_int *a, mp_size bitNum, mp_size value);
-mp_err mpl_get_bit(const mp_int *a, mp_size bitNum);
-mp_err mpl_get_bits(const mp_int *a, mp_size lsbNum, mp_size numBits);
-mp_err mpl_significant_bits(const mp_int *a);
-
-#endif /* end _H_MPLOGIC_ */

mozilla/security/nss/lib/freebl/mpi/mpprime.c

@@ -1,626 +0,0 @@
-/*
- *  mpprime.c
- *
- *  Utilities for finding and working with prime and pseudo-prime
- *  integers
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1997
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *   Netscape Communications Corporation
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#include "mpi-priv.h"
-#include "mpprime.h"
-#include "mplogic.h"
-#include <stdlib.h>
-#include <string.h>
-
-#define SMALL_TABLE 0 /* determines size of hard-wired prime table */
-
-#define RANDOM() rand()
-
-#include "primes.c"  /* pull in the prime digit table */
-
-/* 
-   Test if any of a given vector of digits divides a.  If not, MP_NO
-   is returned; otherwise, MP_YES is returned and 'which' is set to
-   the index of the integer in the vector which divided a.
- */
-mp_err    s_mpp_divp(mp_int *a, const mp_digit *vec, int size, int *which);
-
-/* {{{ mpp_divis(a, b) */
-
-/*
-  mpp_divis(a, b)
-
-  Returns MP_YES if a is divisible by b, or MP_NO if it is not.
- */
-
-mp_err  mpp_divis(mp_int *a, mp_int *b)
-{
-  mp_err  res;
-  mp_int  rem;
-
-  if((res = mp_init(&rem)) != MP_OKAY)
-    return res;
-
-  if((res = mp_mod(a, b, &rem)) != MP_OKAY)
-    goto CLEANUP;
-
-  if(mp_cmp_z(&rem) == 0)
-    res = MP_YES;
-  else
-    res = MP_NO;
-
-CLEANUP:
-  mp_clear(&rem);
-  return res;
-
-} /* end mpp_divis() */
-
-/* }}} */
-
-/* {{{ mpp_divis_d(a, d) */
-
-/*
-  mpp_divis_d(a, d)
-
-  Return MP_YES if a is divisible by d, or MP_NO if it is not.
- */
-
-mp_err  mpp_divis_d(mp_int *a, mp_digit d)
-{
-  mp_err     res;
-  mp_digit   rem;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  if(d == 0)
-    return MP_NO;
-
-  if((res = mp_mod_d(a, d, &rem)) != MP_OKAY)
-    return res;
-
-  if(rem == 0)
-    return MP_YES;
-  else
-    return MP_NO;
-
-} /* end mpp_divis_d() */
-
-/* }}} */
-
-/* {{{ mpp_random(a) */
-
-/*
-  mpp_random(a)
-
-  Assigns a random value to a.  This value is generated using the
-  standard C library's rand() function, so it should not be used for
-  cryptographic purposes, but it should be fine for primality testing,
-  since all we really care about there is good statistical properties.
-
-  As many digits as a currently has are filled with random digits.
- */
-
-mp_err  mpp_random(mp_int *a)
-
-{
-  mp_digit  next = 0;
-  unsigned int       ix, jx;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  for(ix = 0; ix < USED(a); ix++) {
-    for(jx = 0; jx < sizeof(mp_digit); jx++) {
-      next = (next << CHAR_BIT) | (RANDOM() & UCHAR_MAX);
-    }
-    DIGIT(a, ix) = next;
-  }
-
-  return MP_OKAY;
-
-} /* end mpp_random() */
-
-/* }}} */
-
-/* {{{ mpp_random_size(a, prec) */
-
-mp_err  mpp_random_size(mp_int *a, mp_size prec)
-{
-  mp_err   res;
-
-  ARGCHK(a != NULL && prec > 0, MP_BADARG);
-  
-  if((res = s_mp_pad(a, prec)) != MP_OKAY)
-    return res;
-
-  return mpp_random(a);
-
-} /* end mpp_random_size() */
-
-/* }}} */
-
-/* {{{ mpp_divis_vector(a, vec, size, which) */
-
-/*
-  mpp_divis_vector(a, vec, size, which)
-
-  Determines if a is divisible by any of the 'size' digits in vec.
-  Returns MP_YES and sets 'which' to the index of the offending digit,
-  if it is; returns MP_NO if it is not.
- */
-
-mp_err  mpp_divis_vector(mp_int *a, const mp_digit *vec, int size, int *which)
-{
-  ARGCHK(a != NULL && vec != NULL && size > 0, MP_BADARG);
-  
-  return s_mpp_divp(a, vec, size, which);
-
-} /* end mpp_divis_vector() */
-
-/* }}} */
-
-/* {{{ mpp_divis_primes(a, np) */
-
-/*
-  mpp_divis_primes(a, np)
-
-  Test whether a is divisible by any of the first 'np' primes.  If it
-  is, returns MP_YES and sets *np to the value of the digit that did
-  it.  If not, returns MP_NO.
- */
-mp_err  mpp_divis_primes(mp_int *a, mp_digit *np)
-{
-  int     size, which;
-  mp_err  res;
-
-  ARGCHK(a != NULL && np != NULL, MP_BADARG);
-
-  size = (int)*np;
-  if(size > prime_tab_size)
-    size = prime_tab_size;
-
-  res = mpp_divis_vector(a, prime_tab, size, &which);
-  if(res == MP_YES) 
-    *np = prime_tab[which];
-
-  return res;
-
-} /* end mpp_divis_primes() */
-
-/* }}} */
-
-/* {{{ mpp_fermat(a, w) */
-
-/*
-  Using w as a witness, try pseudo-primality testing based on Fermat's
-  little theorem.  If a is prime, and (w, a) = 1, then w^a == w (mod
-  a).  So, we compute z = w^a (mod a) and compare z to w; if they are
-  equal, the test passes and we return MP_YES.  Otherwise, we return
-  MP_NO.
- */
-mp_err  mpp_fermat(mp_int *a, mp_digit w)
-{
-  mp_int  base, test;
-  mp_err  res;
-  
-  if((res = mp_init(&base)) != MP_OKAY)
-    return res;
-
-  mp_set(&base, w);
-
-  if((res = mp_init(&test)) != MP_OKAY)
-    goto TEST;
-
-  /* Compute test = base^a (mod a) */
-  if((res = mp_exptmod(&base, a, a, &test)) != MP_OKAY)
-    goto CLEANUP;
-
-  
-  if(mp_cmp(&base, &test) == 0)
-    res = MP_YES;
-  else
-    res = MP_NO;
-
- CLEANUP:
-  mp_clear(&test);
- TEST:
-  mp_clear(&base);
-
-  return res;
-
-} /* end mpp_fermat() */
-
-/* }}} */
-
-/*
-  Perform the fermat test on each of the primes in a list until
-  a) one of them shows a is not prime, or 
-  b) the list is exhausted.
-  Returns:  MP_YES if it passes tests.
-	    MP_NO  if fermat test reveals it is composite
-	    Some MP error code if some other error occurs.
- */
-mp_err mpp_fermat_list(mp_int *a, const mp_digit *primes, mp_size nPrimes)
-{
-  mp_err rv = MP_YES;
-
-  while (nPrimes-- > 0 && rv == MP_YES) {
-    rv = mpp_fermat(a, *primes++);
-  }
-  return rv;
-}
-
-/* {{{ mpp_pprime(a, nt) */
-
-/*
-  mpp_pprime(a, nt)
-
-  Performs nt iteration of the Miller-Rabin probabilistic primality
-  test on a.  Returns MP_YES if the tests pass, MP_NO if one fails.
-  If MP_NO is returned, the number is definitely composite.  If MP_YES
-  is returned, it is probably prime (but that is not guaranteed).
- */
-
-mp_err  mpp_pprime(mp_int *a, int nt)
-{
-  mp_err   res;
-  mp_int   x, amo, m, z;	/* "amo" = "a minus one" */
-  int      iter;
-  unsigned int jx;
-  mp_size  b;
-
-  ARGCHK(a != NULL, MP_BADARG);
-
-  MP_DIGITS(&x) = 0;
-  MP_DIGITS(&amo) = 0;
-  MP_DIGITS(&m) = 0;
-  MP_DIGITS(&z) = 0;
-
-  /* Initialize temporaries... */
-  MP_CHECKOK( mp_init(&amo));
-  /* Compute amo = a - 1 for what follows...    */
-  MP_CHECKOK( mp_sub_d(a, 1, &amo) );
-
-  b = mp_trailing_zeros(&amo);
-  if (!b) { /* a was even ? */
-    res = MP_NO;
-    goto CLEANUP;
-  }
-
-  MP_CHECKOK( mp_init_size(&x, MP_USED(a)) );
-  MP_CHECKOK( mp_init(&z) );
-  MP_CHECKOK( mp_init(&m) );
-  MP_CHECKOK( mp_div_2d(&amo, b, &m, 0) );
-
-  /* Do the test nt times... */
-  for(iter = 0; iter < nt; iter++) {
-
-    /* Choose a random value for x < a          */
-    s_mp_pad(&x, USED(a));
-    mpp_random(&x);
-    MP_CHECKOK( mp_mod(&x, a, &x) );
-
-    /* Compute z = (x ** m) mod a               */
-    MP_CHECKOK( mp_exptmod(&x, &m, a, &z) );
-    
-    if(mp_cmp_d(&z, 1) == 0 || mp_cmp(&z, &amo) == 0) {
-      res = MP_YES;
-      continue;
-    }
-    
-    res = MP_NO;  /* just in case the following for loop never executes. */
-    for (jx = 1; jx < b; jx++) {
-      /* z = z^2 (mod a) */
-      MP_CHECKOK( mp_sqrmod(&z, a, &z) );
-      res = MP_NO;	/* previous line set res to MP_YES */
-
-      if(mp_cmp_d(&z, 1) == 0) {
-	break;
-      }
-      if(mp_cmp(&z, &amo) == 0) {
-	res = MP_YES;
-	break;
-      } 
-    } /* end testing loop */
-
-    /* If the test passes, we will continue iterating, but a failed
-       test means the candidate is definitely NOT prime, so we will
-       immediately break out of this loop
-     */
-    if(res == MP_NO)
-      break;
-
-  } /* end iterations loop */
-  
-CLEANUP:
-  mp_clear(&m);
-  mp_clear(&z);
-  mp_clear(&x);
-  mp_clear(&amo);
-  return res;
-
-} /* end mpp_pprime() */
-
-/* }}} */
-
-/* Produce table of composites from list of primes and trial value.  
-** trial must be odd. List of primes must not include 2.
-** sieve should have dimension >= MAXPRIME/2, where MAXPRIME is largest 
-** prime in list of primes.  After this function is finished,
-** if sieve[i] is non-zero, then (trial + 2*i) is composite.
-** Each prime used in the sieve costs one division of trial, and eliminates
-** one or more values from the search space. (3 eliminates 1/3 of the values
-** alone!)  Each value left in the search space costs 1 or more modular 
-** exponentations.  So, these divisions are a bargain!
-*/
-mp_err mpp_sieve(mp_int *trial, const mp_digit *primes, mp_size nPrimes, 
-		 unsigned char *sieve, mp_size nSieve)
-{
-  mp_err       res;
-  mp_digit     rem;
-  mp_size      ix;
-  unsigned long offset;
-
-  memset(sieve, 0, nSieve);
-
-  for(ix = 0; ix < nPrimes; ix++) {
-    mp_digit prime = primes[ix];
-    mp_size  i;
-    if((res = mp_mod_d(trial, prime, &rem)) != MP_OKAY) 
-      return res;
-
-    if (rem == 0) {
-      offset = 0;
-    } else {
-      offset = prime - (rem / 2);
-    }
-    for (i = offset; i < nSieve ; i += prime) {
-      sieve[i] = 1;
-    }
-  }
-
-  return MP_OKAY;
-}
-
-#define SIEVE_SIZE 32*1024
-
-mp_err mpp_make_prime(mp_int *start, mp_size nBits, mp_size strong,
-		      unsigned long * nTries)
-{
-  mp_digit      np;
-  mp_err        res;
-  int           i	= 0;
-  mp_int        trial;
-  mp_int        q;
-  mp_size       num_tests;
-  /*
-   * Always make sieve the last variabale allocated so that 
-   * Mac builds don't break by adding an extra variable
-   * on the stack. -javi
-   */
-#if defined(macintosh) || defined (XP_OS2) \
-    || (defined(HPUX) && defined(__ia64))
-  unsigned char *sieve;
-  
-  sieve = malloc(SIEVE_SIZE);
-  ARGCHK(sieve != NULL, MP_MEM);
-#else
-  unsigned char sieve[SIEVE_SIZE];
-#endif  
-
-  ARGCHK(start != 0, MP_BADARG);
-  ARGCHK(nBits > 16, MP_RANGE);
-
-  MP_DIGITS(&trial) = 0;
-  MP_DIGITS(&q) = 0;
-  MP_CHECKOK( mp_init(&trial) );
-  MP_CHECKOK( mp_init(&q)     );
-  /* values taken from table 4.4, HandBook of Applied Cryptography */
-  if (nBits >= 1300) {
-    num_tests = 2;
-  } else if (nBits >= 850) {
-    num_tests = 3;
-  } else if (nBits >= 650) {
-    num_tests = 4;
-  } else if (nBits >= 550) {
-    num_tests = 5;
-  } else if (nBits >= 450) {
-    num_tests = 6;
-  } else if (nBits >= 400) {
-    num_tests = 7;
-  } else if (nBits >= 350) {
-    num_tests = 8;
-  } else if (nBits >= 300) {
-    num_tests = 9;
-  } else if (nBits >= 250) {
-    num_tests = 12;
-  } else if (nBits >= 200) {
-    num_tests = 15;
-  } else if (nBits >= 150) {
-    num_tests = 18;
-  } else if (nBits >= 100) {
-    num_tests = 27;
-  } else
-    num_tests = 50;
-
-  if (strong) 
-    --nBits;
-  MP_CHECKOK( mpl_set_bit(start, nBits - 1, 1) );
-  MP_CHECKOK( mpl_set_bit(start,         0, 1) );
-  for (i = mpl_significant_bits(start) - 1; i >= nBits; --i) {
-    MP_CHECKOK( mpl_set_bit(start, i, 0) );
-  }
-  /* start sieveing with prime value of 3. */
-  MP_CHECKOK(mpp_sieve(start, prime_tab + 1, prime_tab_size - 1, 
-		       sieve, SIEVE_SIZE) );
-
-#ifdef DEBUG_SIEVE
-  res = 0;
-  for (i = 0; i < SIEVE_SIZE; ++i) {
-    if (!sieve[i])
-      ++res;
-  }
-  fprintf(stderr,"sieve found %d potential primes.\n", res);
-#define FPUTC(x,y) fputc(x,y)
-#else
-#define FPUTC(x,y) 
-#endif
-
-  res = MP_NO;
-  for(i = 0; i < SIEVE_SIZE; ++i) {
-    if (sieve[i])	/* this number is composite */
-      continue;
-    MP_CHECKOK( mp_add_d(start, 2 * i, &trial) );
-    FPUTC('.', stderr);
-    /* run a Fermat test */
-    res = mpp_fermat(&trial, 2);
-    if (res != MP_OKAY) {
-      if (res == MP_NO)
-	continue;	/* was composite */
-      goto CLEANUP;
-    }
-      
-    FPUTC('+', stderr);
-    /* If that passed, run some Miller-Rabin tests	*/
-    res = mpp_pprime(&trial, num_tests);
-    if (res != MP_OKAY) {
-      if (res == MP_NO)
-	continue;	/* was composite */
-      goto CLEANUP;
-    }
-    FPUTC('!', stderr);
-
-    if (!strong) 
-      break;	/* success !! */
-
-    /* At this point, we have strong evidence that our candidate
-       is itself prime.  If we want a strong prime, we need now
-       to test q = 2p + 1 for primality...
-     */
-    MP_CHECKOK( mp_mul_2(&trial, &q) );
-    MP_CHECKOK( mp_add_d(&q, 1, &q)  );
-
-    /* Test q for small prime divisors ... */
-    np = prime_tab_size;
-    res = mpp_divis_primes(&q, &np);
-    if (res == MP_YES) { /* is composite */
-      mp_clear(&q);
-      continue;
-    }
-    if (res != MP_NO) 
-      goto CLEANUP;
-
-    /* And test with Fermat, as with its parent ... */
-    res = mpp_fermat(&q, 2);
-    if (res != MP_YES) {
-      mp_clear(&q);
-      if (res == MP_NO)
-	continue;	/* was composite */
-      goto CLEANUP;
-    }
-
-    /* And test with Miller-Rabin, as with its parent ... */
-    res = mpp_pprime(&q, num_tests);
-    if (res != MP_YES) {
-      mp_clear(&q);
-      if (res == MP_NO)
-	continue;	/* was composite */
-      goto CLEANUP;
-    }
-
-    /* If it passed, we've got a winner */
-    mp_exch(&q, &trial);
-    mp_clear(&q);
-    break;
-
-  } /* end of loop through sieved values */
-  if (res == MP_YES) 
-    mp_exch(&trial, start);
-CLEANUP:
-  mp_clear(&trial);
-  mp_clear(&q);
-  if (nTries)
-    *nTries += i;
-#if defined(macintosh) || defined(XP_OS2) \
-    || (defined(HPUX) && defined(__ia64))
-  if (sieve != NULL) {
-  	memset(sieve, 0, SIEVE_SIZE);
-  	free (sieve);
-  }
-#endif    
-  return res;
-}
-
-/*========================================================================*/
-/*------------------------------------------------------------------------*/
-/* Static functions visible only to the library internally                */
-
-/* {{{ s_mpp_divp(a, vec, size, which) */
-
-/* 
-   Test for divisibility by members of a vector of digits.  Returns
-   MP_NO if a is not divisible by any of them; returns MP_YES and sets
-   'which' to the index of the offender, if it is.  Will stop on the
-   first digit against which a is divisible.
- */
-
-mp_err    s_mpp_divp(mp_int *a, const mp_digit *vec, int size, int *which)
-{
-  mp_err    res;
-  mp_digit  rem;
-
-  int     ix;
-
-  for(ix = 0; ix < size; ix++) {
-    if((res = mp_mod_d(a, vec[ix], &rem)) != MP_OKAY) 
-      return res;
-
-    if(rem == 0) {
-      if(which)
-	*which = ix;
-      return MP_YES;
-    }
-  }
-
-  return MP_NO;
-
-} /* end s_mpp_divp() */
-
-/* }}} */
-
-/*------------------------------------------------------------------------*/
-/* HERE THERE BE DRAGONS                                                  */

mozilla/security/nss/lib/freebl/mpi/mpprime.h

@@ -1,70 +0,0 @@
-/*
- *  mpprime.h
- *
- *  Utilities for finding and working with prime and pseudo-prime
- *  integers
- *
- * ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
- *
- * The Initial Developer of the Original Code is
- * Michael J. Fromberger.
- * Portions created by the Initial Developer are Copyright (C) 1997
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-
-#ifndef _H_MP_PRIME_
-#define _H_MP_PRIME_
-
-#include "mpi.h"
-
-extern const int prime_tab_size;   /* number of primes available */
-extern const mp_digit prime_tab[];
-
-/* Tests for divisibility    */
-mp_err  mpp_divis(mp_int *a, mp_int *b);
-mp_err  mpp_divis_d(mp_int *a, mp_digit d);
-
-/* Random selection          */
-mp_err  mpp_random(mp_int *a);
-mp_err  mpp_random_size(mp_int *a, mp_size prec);
-
-/* Pseudo-primality testing  */
-mp_err  mpp_divis_vector(mp_int *a, const mp_digit *vec, int size, int *which);
-mp_err  mpp_divis_primes(mp_int *a, mp_digit *np);
-mp_err  mpp_fermat(mp_int *a, mp_digit w);
-mp_err mpp_fermat_list(mp_int *a, const mp_digit *primes, mp_size nPrimes);
-mp_err  mpp_pprime(mp_int *a, int nt);
-mp_err mpp_sieve(mp_int *trial, const mp_digit *primes, mp_size nPrimes, 
-		 unsigned char *sieve, mp_size nSieve);
-mp_err mpp_make_prime(mp_int *start, mp_size nBits, mp_size strong,
-		      unsigned long * nTries);
-
-#endif /* end _H_MP_PRIME_ */

mozilla/security/nss/lib/freebl/mpi/mpv_sparc.c

@@ -1,253 +0,0 @@
-/* ***** BEGIN LICENSE BLOCK *****
- * Version: MPL 1.1/GPL 2.0/LGPL 2.1
- *
- * The contents of this file are subject to the Mozilla Public License Version
- * 1.1 (the "License"); you may not use this file except in compliance with
- * the License. You may obtain a copy of the License at
- * http://www.mozilla.org/MPL/
- *
- * Software distributed under the License is distributed on an "AS IS" basis,
- * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
- * for the specific language governing rights and limitations under the
- * License.
- *
- * The Original Code is a SPARC/VIS optimized multiply and add function.
- *
- * The Initial Developer of the Original Code is
- * Sun Microsystems Inc.
- * Portions created by the Initial Developer are Copyright (C) 1999-2000
- * the Initial Developer. All Rights Reserved.
- *
- * Contributor(s):
- *
- * Alternatively, the contents of this file may be used under the terms of
- * either the GNU General Public License Version 2 or later (the "GPL"), or
- * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
- * in which case the provisions of the GPL or the LGPL are applicable instead
- * of those above. If you wish to allow use of your version of this file only
- * under the terms of either the GPL or the LGPL, and not to allow others to
- * use your version of this file under the terms of the MPL, indicate your
- * decision by deleting the provisions above and replace them with the notice
- * and other provisions required by the GPL or the LGPL. If you do not delete
- * the provisions above, a recipient may use your version of this file under
- * the terms of any one of the MPL, the GPL or the LGPL.
- *
- * ***** END LICENSE BLOCK ***** */
-/* $Id: mpv_sparc.c,v 1.4 2004-04-27 23:04:36 gerv%gerv.net Exp $ */
-
-#include "vis_proto.h"
-
-/***************************************************************/
-
-typedef  int                t_s32;
-typedef  unsigned int       t_u32;
-#if defined(__sparcv9)
-typedef  long               t_s64;
-typedef  unsigned long      t_u64;
-#else
-typedef  long long          t_s64;
-typedef  unsigned long long t_u64;
-#endif
-typedef  double             t_d64;
-
-/***************************************************************/
-
-typedef union {
-  t_d64 d64;
-  struct {
-    t_s32 i0;
-    t_s32 i1;
-  } i32s;
-} d64_2_i32;
-
-/***************************************************************/
-
-#define BUFF_SIZE  256
-
-#define A_BITS  19
-#define A_MASK  ((1 << A_BITS) - 1)
-
-/***************************************************************/
-
-static t_u64 mask_cnst[] = {
-  0x8000000080000000ull
-};
-
-/***************************************************************/
-
-#define DEF_VARS(N)                     \
-  t_d64 *py = (t_d64*)y;                \
-  t_d64 mask = *((t_d64*)mask_cnst);    \
-  t_d64 ca = (1u << 31) - 1;            \
-  t_d64 da = (t_d64)a;                  \
-  t_s64 buff[N], s;                     \
-  d64_2_i32 dy
-
-/***************************************************************/
-
-#define MUL_U32_S64_2(i)                                \
-  dy.d64 = vis_fxnor(mask, py[i]);                      \
-  buff[2*(i)  ] = (ca - (t_d64)dy.i32s.i0) * da;        \
-  buff[2*(i)+1] = (ca - (t_d64)dy.i32s.i1) * da
-
-#define MUL_U32_S64_2_D(i)              \
-  dy.d64 = vis_fxnor(mask, py[i]);      \
-  d0 = ca - (t_d64)dy.i32s.i0;          \
-  d1 = ca - (t_d64)dy.i32s.i1;          \
-  buff[4*(i)  ] = (t_s64)(d0 * da);     \
-  buff[4*(i)+1] = (t_s64)(d0 * db);     \
-  buff[4*(i)+2] = (t_s64)(d1 * da);     \
-  buff[4*(i)+3] = (t_s64)(d1 * db)
-
-/***************************************************************/
-
-#define ADD_S64_U32(i)          \
-  s = buff[i] + x[i] + c;       \
-  z[i] = s;                     \
-  c = (s >> 32)
-
-#define ADD_S64_U32_D(i)                        \
-  s = buff[2*(i)] +(((t_s64)(buff[2*(i)+1]))<<A_BITS) + x[i] + uc;   \
-  z[i] = s;                                     \
-  uc = ((t_u64)s >> 32)
-
-/***************************************************************/
-
-#define MUL_U32_S64_8(i)        \
-  MUL_U32_S64_2(i);             \
-  MUL_U32_S64_2(i+1);           \
-  MUL_U32_S64_2(i+2);           \
-  MUL_U32_S64_2(i+3)
-
-#define MUL_U32_S64_D_8(i)      \
-  MUL_U32_S64_2_D(i);           \
-  MUL_U32_S64_2_D(i+1);         \
-  MUL_U32_S64_2_D(i+2);         \
-  MUL_U32_S64_2_D(i+3)
-
-/***************************************************************/
-
-#define ADD_S64_U32_8(i)        \
-  ADD_S64_U32(i);               \
-  ADD_S64_U32(i+1);             \
-  ADD_S64_U32(i+2);             \
-  ADD_S64_U32(i+3);             \
-  ADD_S64_U32(i+4);             \
-  ADD_S64_U32(i+5);             \
-  ADD_S64_U32(i+6);             \
-  ADD_S64_U32(i+7)
-
-#define ADD_S64_U32_D_8(i)      \
-  ADD_S64_U32_D(i);             \
-  ADD_S64_U32_D(i+1);           \
-  ADD_S64_U32_D(i+2);           \
-  ADD_S64_U32_D(i+3);           \
-  ADD_S64_U32_D(i+4);           \
-  ADD_S64_U32_D(i+5);           \
-  ADD_S64_U32_D(i+6);           \
-  ADD_S64_U32_D(i+7)
-
-/***************************************************************/
-
-t_u32 mul_add(t_u32 *z, t_u32 *x, t_u32 *y, int n, t_u32 a)
-{
-  if (a < (1 << A_BITS)) {
-
-    if (n == 8) {
-      DEF_VARS(8);
-      t_s32 c = 0;
-
-      MUL_U32_S64_8(0);
-      ADD_S64_U32_8(0);
-
-      return c;
-
-    } else if (n == 16) {
-      DEF_VARS(16);
-      t_s32 c = 0;
-
-      MUL_U32_S64_8(0);
-      MUL_U32_S64_8(4);
-      ADD_S64_U32_8(0);
-      ADD_S64_U32_8(8);
-
-      return c;
-
-    } else {
-      DEF_VARS(BUFF_SIZE);
-      t_s32 i, c = 0;
-
-#pragma pipeloop(0)
-      for (i = 0; i < (n+1)/2; i ++) {
-        MUL_U32_S64_2(i);
-      }
-
-#pragma pipeloop(0)
-      for (i = 0; i < n; i ++) {
-        ADD_S64_U32(i);
-      }
-
-      return c;
-
-    }
-  } else {
-
-    if (n == 8) {
-      DEF_VARS(2*8);
-      t_d64 d0, d1, db;
-      t_u32 uc = 0;
-
-      da = (t_d64)(a &  A_MASK);
-      db = (t_d64)(a >> A_BITS);
-
-      MUL_U32_S64_D_8(0);
-      ADD_S64_U32_D_8(0);
-
-      return uc;
-
-    } else if (n == 16) {
-      DEF_VARS(2*16);
-      t_d64 d0, d1, db;
-      t_u32 uc = 0;
-
-      da = (t_d64)(a &  A_MASK);
-      db = (t_d64)(a >> A_BITS);
-
-      MUL_U32_S64_D_8(0);
-      MUL_U32_S64_D_8(4);
-      ADD_S64_U32_D_8(0);
-      ADD_S64_U32_D_8(8);
-
-      return uc;
-
-    } else {
-      DEF_VARS(2*BUFF_SIZE);
-      t_d64 d0, d1, db;
-      t_u32 i, uc = 0;
-
-      da = (t_d64)(a &  A_MASK);
-      db = (t_d64)(a >> A_BITS);
-
-#pragma pipeloop(0)
-      for (i = 0; i < (n+1)/2; i ++) {
-        MUL_U32_S64_2_D(i);
-      }
-
-#pragma pipeloop(0)
-      for (i = 0; i < n; i ++) {
-        ADD_S64_U32_D(i);
-      }
-
-      return uc;
-    }
-  }
-}
-
-/***************************************************************/
-
-t_u32 mul_add_inp(t_u32 *x, t_u32 *y, int n, t_u32 a)
-{
-  return mul_add(x, x, y, n, a);
-}
-
-/***************************************************************/