X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FIntrinsicLowering.cpp;h=c38d4fbc5aa47452315a1a3744f39bceaa11f18e;hb=4ed2667c138442319f0c3dcc019def8735233a90;hp=256f0ffa56758ddebaa847fd2d38ee4e717e5be7;hpb=51b8d54922350b7e1c2cd5a5183ef2c5f5d1b1d5;p=oota-llvm.git

diff --git a/lib/CodeGen/IntrinsicLowering.cpp b/lib/CodeGen/IntrinsicLowering.cpp
index 256f0ffa567..c38d4fbc5aa 100644
--- a/lib/CodeGen/IntrinsicLowering.cpp
+++ b/lib/CodeGen/IntrinsicLowering.cpp
@@ -2,35 +2,59 @@
 //
 //                     The LLVM Compiler Infrastructure
 //
-// This file was developed by the LLVM research group and is distributed under
-// the University of Illinois Open Source License. See LICENSE.TXT for details.
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
-// This file implements the default intrinsic lowering implementation.
+// This file implements the IntrinsicLowering class.
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/CodeGen/IntrinsicLowering.h"
-#include "llvm/Constants.h"
-#include "llvm/DerivedTypes.h"
-#include "llvm/Module.h"
-#include "llvm/Instructions.h"
-#include "llvm/Type.h"
-#include <iostream>
-
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/IRBuilder.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/Type.h"
+#include "llvm/Support/CallSite.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/raw_ostream.h"
 using namespace llvm;
 
 template <class ArgIt>
-static Function *EnsureFunctionExists(Module &M, const char *Name,
-                                      ArgIt ArgBegin, ArgIt ArgEnd,
-                                      const Type *RetTy) {
-  if (Function *F = M.getNamedFunction(Name)) return F;
-  // It doesn't already exist in the program, insert a new definition now.
-  std::vector<const Type *> ParamTys;
+static void EnsureFunctionExists(Module &M, const char *Name,
+                                 ArgIt ArgBegin, ArgIt ArgEnd,
+                                 Type *RetTy) {
+  // Insert a correctly-typed definition now.
+  std::vector<Type *> ParamTys;
   for (ArgIt I = ArgBegin; I != ArgEnd; ++I)
     ParamTys.push_back(I->getType());
-  return M.getOrInsertFunction(Name, FunctionType::get(RetTy, ParamTys, false));
+  M.getOrInsertFunction(Name, FunctionType::get(RetTy, ParamTys, false));
+}
+
+static void EnsureFPIntrinsicsExist(Module &M, Function *Fn,
+                                    const char *FName,
+                                    const char *DName, const char *LDName) {
+  // Insert definitions for all the floating point types.
+  switch((int)Fn->arg_begin()->getType()->getTypeID()) {
+  case Type::FloatTyID:
+    EnsureFunctionExists(M, FName, Fn->arg_begin(), Fn->arg_end(),
+                         Type::getFloatTy(M.getContext()));
+    break;
+  case Type::DoubleTyID:
+    EnsureFunctionExists(M, DName, Fn->arg_begin(), Fn->arg_end(),
+                         Type::getDoubleTy(M.getContext()));
+    break;
+  case Type::X86_FP80TyID:
+  case Type::FP128TyID:
+  case Type::PPC_FP128TyID:
+    EnsureFunctionExists(M, LDName, Fn->arg_begin(), Fn->arg_end(),
+                         Fn->arg_begin()->getType());
+    break;
+  }
 }
 
 /// ReplaceCallWith - This function is used when we want to lower an intrinsic
@@ -40,98 +64,203 @@ static Function *EnsureFunctionExists(Module &M, const char *Name,
 template <class ArgIt>
 static CallInst *ReplaceCallWith(const char *NewFn, CallInst *CI,
                                  ArgIt ArgBegin, ArgIt ArgEnd,
-                                 const Type *RetTy, Function *&FCache) {
-  if (!FCache) {
-    // If we haven't already looked up this function, check to see if the
-    // program already contains a function with this name.
-    Module *M = CI->getParent()->getParent()->getParent();
-    FCache = M->getNamedFunction(NewFn);
-    if (!FCache) {
-      // It doesn't already exist in the program, insert a new definition now.
-      std::vector<const Type *> ParamTys;
-      for (ArgIt I = ArgBegin; I != ArgEnd; ++I)
-        ParamTys.push_back((*I)->getType());
-      FCache = M->getOrInsertFunction(NewFn,
-                                     FunctionType::get(RetTy, ParamTys, false));
-    }
-   }
-
-  const FunctionType *FT = FCache->getFunctionType();
-  std::vector<Value*> Operands;
-  unsigned ArgNo = 0;
-  for (ArgIt I = ArgBegin; I != ArgEnd && ArgNo != FT->getNumParams();
-       ++I, ++ArgNo) {
-    Value *Arg = *I;
-    if (Arg->getType() != FT->getParamType(ArgNo))
-      Arg = new CastInst(Arg, FT->getParamType(ArgNo), Arg->getName(), CI);
-    Operands.push_back(Arg);
-  }
-  // Pass nulls into any additional arguments...
-  for (; ArgNo != FT->getNumParams(); ++ArgNo)
-    Operands.push_back(Constant::getNullValue(FT->getParamType(ArgNo)));
-
-  std::string Name = CI->getName(); CI->setName("");
-  if (FT->getReturnType() == Type::VoidTy) Name.clear();
-  CallInst *NewCI = new CallInst(FCache, Operands, Name, CI);
-  if (!CI->use_empty()) {
-    Value *V = NewCI;
-    if (CI->getType() != NewCI->getType())
-      V = new CastInst(NewCI, CI->getType(), Name, CI);
-    CI->replaceAllUsesWith(V);
-  }
+                                 Type *RetTy) {
+  // If we haven't already looked up this function, check to see if the
+  // program already contains a function with this name.
+  Module *M = CI->getParent()->getParent()->getParent();
+  // Get or insert the definition now.
+  std::vector<Type *> ParamTys;
+  for (ArgIt I = ArgBegin; I != ArgEnd; ++I)
+    ParamTys.push_back((*I)->getType());
+  Constant* FCache = M->getOrInsertFunction(NewFn,
+                                  FunctionType::get(RetTy, ParamTys, false));
+
+  IRBuilder<> Builder(CI->getParent(), CI);
+  SmallVector<Value *, 8> Args(ArgBegin, ArgEnd);
+  CallInst *NewCI = Builder.CreateCall(FCache, Args);
+  NewCI->setName(CI->getName());
+  if (!CI->use_empty())
+    CI->replaceAllUsesWith(NewCI);
   return NewCI;
 }
 
-void DefaultIntrinsicLowering::AddPrototypes(Module &M) {
+// VisualStudio defines setjmp as _setjmp
+#if defined(_MSC_VER) && defined(setjmp) && \
+                         !defined(setjmp_undefined_for_msvc)
+#  pragma push_macro("setjmp")
+#  undef setjmp
+#  define setjmp_undefined_for_msvc
+#endif
+
+void IntrinsicLowering::AddPrototypes(Module &M) {
+  LLVMContext &Context = M.getContext();
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
-    if (I->isExternal() && !I->use_empty())
+    if (I->isDeclaration() && !I->use_empty())
       switch (I->getIntrinsicID()) {
       default: break;
       case Intrinsic::setjmp:
         EnsureFunctionExists(M, "setjmp", I->arg_begin(), I->arg_end(),
-                             Type::IntTy);
+                             Type::getInt32Ty(M.getContext()));
         break;
       case Intrinsic::longjmp:
         EnsureFunctionExists(M, "longjmp", I->arg_begin(), I->arg_end(),
-                             Type::VoidTy);
+                             Type::getVoidTy(M.getContext()));
         break;
       case Intrinsic::siglongjmp:
         EnsureFunctionExists(M, "abort", I->arg_end(), I->arg_end(),
-                             Type::VoidTy);
+                             Type::getVoidTy(M.getContext()));
         break;
       case Intrinsic::memcpy:
-        EnsureFunctionExists(M, "memcpy", I->arg_begin(), --I->arg_end(),
-                             I->arg_begin()->getType());
+        M.getOrInsertFunction("memcpy",
+          Type::getInt8PtrTy(Context),
+                              Type::getInt8PtrTy(Context), 
+                              Type::getInt8PtrTy(Context), 
+                              TD.getIntPtrType(Context), (Type *)0);
         break;
       case Intrinsic::memmove:
-        EnsureFunctionExists(M, "memmove", I->arg_begin(), --I->arg_end(),
-                             I->arg_begin()->getType());
+        M.getOrInsertFunction("memmove",
+          Type::getInt8PtrTy(Context),
+                              Type::getInt8PtrTy(Context), 
+                              Type::getInt8PtrTy(Context), 
+                              TD.getIntPtrType(Context), (Type *)0);
         break;
       case Intrinsic::memset:
-        M.getOrInsertFunction("memset", PointerType::get(Type::SByteTy),
-                              PointerType::get(Type::SByteTy),
-                              Type::IntTy, (--(--I->arg_end()))->getType(),
-                              (Type *)0);
-        break;
-      case Intrinsic::isunordered:
-        EnsureFunctionExists(M, "isunordered", I->arg_begin(), I->arg_end(),
-                             Type::BoolTy);
+        M.getOrInsertFunction("memset",
+          Type::getInt8PtrTy(Context),
+                              Type::getInt8PtrTy(Context), 
+                              Type::getInt32Ty(M.getContext()), 
+                              TD.getIntPtrType(Context), (Type *)0);
         break;
       case Intrinsic::sqrt:
-        if(I->arg_begin()->getType() == Type::FloatTy)
-          EnsureFunctionExists(M, "sqrtf", I->arg_begin(), I->arg_end(),
-                               Type::FloatTy);
-        else
-          EnsureFunctionExists(M, "sqrt", I->arg_begin(), I->arg_end(),
-                               Type::DoubleTy);
+        EnsureFPIntrinsicsExist(M, I, "sqrtf", "sqrt", "sqrtl");
+        break;
+      case Intrinsic::sin:
+        EnsureFPIntrinsicsExist(M, I, "sinf", "sin", "sinl");
+        break;
+      case Intrinsic::cos:
+        EnsureFPIntrinsicsExist(M, I, "cosf", "cos", "cosl");
+        break;
+      case Intrinsic::pow:
+        EnsureFPIntrinsicsExist(M, I, "powf", "pow", "powl");
+        break;
+      case Intrinsic::log:
+        EnsureFPIntrinsicsExist(M, I, "logf", "log", "logl");
+        break;
+      case Intrinsic::log2:
+        EnsureFPIntrinsicsExist(M, I, "log2f", "log2", "log2l");
+        break;
+      case Intrinsic::log10:
+        EnsureFPIntrinsicsExist(M, I, "log10f", "log10", "log10l");
+        break;
+      case Intrinsic::exp:
+        EnsureFPIntrinsicsExist(M, I, "expf", "exp", "expl");
+        break;
+      case Intrinsic::exp2:
+        EnsureFPIntrinsicsExist(M, I, "exp2f", "exp2", "exp2l");
         break;
       }
 }
 
+/// LowerBSWAP - Emit the code to lower bswap of V before the specified
+/// instruction IP.
+static Value *LowerBSWAP(LLVMContext &Context, Value *V, Instruction *IP) {
+  assert(V->getType()->isIntegerTy() && "Can't bswap a non-integer type!");
+
+  unsigned BitSize = V->getType()->getPrimitiveSizeInBits();
+  
+  IRBuilder<> Builder(IP->getParent(), IP);
+
+  switch(BitSize) {
+  default: llvm_unreachable("Unhandled type size of value to byteswap!");
+  case 16: {
+    Value *Tmp1 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 8),
+                                    "bswap.2");
+    Value *Tmp2 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 8),
+                                     "bswap.1");
+    V = Builder.CreateOr(Tmp1, Tmp2, "bswap.i16");
+    break;
+  }
+  case 32: {
+    Value *Tmp4 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 24),
+                                    "bswap.4");
+    Value *Tmp3 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 8),
+                                    "bswap.3");
+    Value *Tmp2 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 8),
+                                     "bswap.2");
+    Value *Tmp1 = Builder.CreateLShr(V,ConstantInt::get(V->getType(), 24),
+                                     "bswap.1");
+    Tmp3 = Builder.CreateAnd(Tmp3,
+                         ConstantInt::get(Type::getInt32Ty(Context), 0xFF0000),
+                             "bswap.and3");
+    Tmp2 = Builder.CreateAnd(Tmp2,
+                           ConstantInt::get(Type::getInt32Ty(Context), 0xFF00),
+                             "bswap.and2");
+    Tmp4 = Builder.CreateOr(Tmp4, Tmp3, "bswap.or1");
+    Tmp2 = Builder.CreateOr(Tmp2, Tmp1, "bswap.or2");
+    V = Builder.CreateOr(Tmp4, Tmp2, "bswap.i32");
+    break;
+  }
+  case 64: {
+    Value *Tmp8 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 56),
+                                    "bswap.8");
+    Value *Tmp7 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 40),
+                                    "bswap.7");
+    Value *Tmp6 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 24),
+                                    "bswap.6");
+    Value *Tmp5 = Builder.CreateShl(V, ConstantInt::get(V->getType(), 8),
+                                    "bswap.5");
+    Value* Tmp4 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 8),
+                                     "bswap.4");
+    Value* Tmp3 = Builder.CreateLShr(V, 
+                                     ConstantInt::get(V->getType(), 24),
+                                     "bswap.3");
+    Value* Tmp2 = Builder.CreateLShr(V, 
+                                     ConstantInt::get(V->getType(), 40),
+                                     "bswap.2");
+    Value* Tmp1 = Builder.CreateLShr(V, 
+                                     ConstantInt::get(V->getType(), 56),
+                                     "bswap.1");
+    Tmp7 = Builder.CreateAnd(Tmp7,
+                             ConstantInt::get(Type::getInt64Ty(Context),
+                                              0xFF000000000000ULL),
+                             "bswap.and7");
+    Tmp6 = Builder.CreateAnd(Tmp6,
+                             ConstantInt::get(Type::getInt64Ty(Context),
+                                              0xFF0000000000ULL),
+                             "bswap.and6");
+    Tmp5 = Builder.CreateAnd(Tmp5,
+                        ConstantInt::get(Type::getInt64Ty(Context),
+                             0xFF00000000ULL),
+                             "bswap.and5");
+    Tmp4 = Builder.CreateAnd(Tmp4,
+                        ConstantInt::get(Type::getInt64Ty(Context),
+                             0xFF000000ULL),
+                             "bswap.and4");
+    Tmp3 = Builder.CreateAnd(Tmp3,
+                             ConstantInt::get(Type::getInt64Ty(Context),
+                             0xFF0000ULL),
+                             "bswap.and3");
+    Tmp2 = Builder.CreateAnd(Tmp2,
+                             ConstantInt::get(Type::getInt64Ty(Context),
+                             0xFF00ULL),
+                             "bswap.and2");
+    Tmp8 = Builder.CreateOr(Tmp8, Tmp7, "bswap.or1");
+    Tmp6 = Builder.CreateOr(Tmp6, Tmp5, "bswap.or2");
+    Tmp4 = Builder.CreateOr(Tmp4, Tmp3, "bswap.or3");
+    Tmp2 = Builder.CreateOr(Tmp2, Tmp1, "bswap.or4");
+    Tmp8 = Builder.CreateOr(Tmp8, Tmp6, "bswap.or5");
+    Tmp4 = Builder.CreateOr(Tmp4, Tmp2, "bswap.or6");
+    V = Builder.CreateOr(Tmp8, Tmp4, "bswap.i64");
+    break;
+  }
+  }
+  return V;
+}
+
 /// LowerCTPOP - Emit the code to lower ctpop of V before the specified
-/// instruction.
-static Value *LowerCTPOP(Value *V, Instruction *IP) {
-  assert(V->getType()->isInteger() && "Can't ctpop a non-integer type!");
+/// instruction IP.
+static Value *LowerCTPOP(LLVMContext &Context, Value *V, Instruction *IP) {
+  assert(V->getType()->isIntegerTy() && "Can't ctpop a non-integer type!");
 
   static const uint64_t MaskValues[6] = {
     0x5555555555555555ULL, 0x3333333333333333ULL,
@@ -139,120 +268,167 @@ static Value *LowerCTPOP(Value *V, Instruction *IP) {
     0x0000FFFF0000FFFFULL, 0x00000000FFFFFFFFULL
   };
 
-  const Type *DestTy = V->getType();
-
-  // Force to unsigned so that the shift rights are logical.
-  if (DestTy->isSigned())
-    V = new CastInst(V, DestTy->getUnsignedVersion(), V->getName(), IP);
+  IRBuilder<> Builder(IP->getParent(), IP);
 
   unsigned BitSize = V->getType()->getPrimitiveSizeInBits();
-  for (unsigned i = 1, ct = 0; i != BitSize; i <<= 1, ++ct) {
-    Value *MaskCst =
-      ConstantExpr::getCast(ConstantUInt::get(Type::ULongTy,
-                                              MaskValues[ct]), V->getType());
-    Value *LHS = BinaryOperator::createAnd(V, MaskCst, "cppop.and1", IP);
-    Value *VShift = new ShiftInst(Instruction::Shr, V,
-                      ConstantInt::get(Type::UByteTy, i), "ctpop.sh", IP);
-    Value *RHS = BinaryOperator::createAnd(VShift, MaskCst, "cppop.and2", IP);
-    V = BinaryOperator::createAdd(LHS, RHS, "ctpop.step", IP);
+  unsigned WordSize = (BitSize + 63) / 64;
+  Value *Count = ConstantInt::get(V->getType(), 0);
+
+  for (unsigned n = 0; n < WordSize; ++n) {
+    Value *PartValue = V;
+    for (unsigned i = 1, ct = 0; i < (BitSize>64 ? 64 : BitSize); 
+         i <<= 1, ++ct) {
+      Value *MaskCst = ConstantInt::get(V->getType(), MaskValues[ct]);
+      Value *LHS = Builder.CreateAnd(PartValue, MaskCst, "cppop.and1");
+      Value *VShift = Builder.CreateLShr(PartValue,
+                                        ConstantInt::get(V->getType(), i),
+                                         "ctpop.sh");
+      Value *RHS = Builder.CreateAnd(VShift, MaskCst, "cppop.and2");
+      PartValue = Builder.CreateAdd(LHS, RHS, "ctpop.step");
+    }
+    Count = Builder.CreateAdd(PartValue, Count, "ctpop.part");
+    if (BitSize > 64) {
+      V = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 64),
+                             "ctpop.part.sh");
+      BitSize -= 64;
+    }
   }
 
-  if (V->getType() != DestTy)
-    V = new CastInst(V, DestTy, V->getName(), IP);
-  return V;
+  return Count;
 }
 
 /// LowerCTLZ - Emit the code to lower ctlz of V before the specified
-/// instruction.
-static Value *LowerCTLZ(Value *V, Instruction *IP) {
-  const Type *DestTy = V->getType();
+/// instruction IP.
+static Value *LowerCTLZ(LLVMContext &Context, Value *V, Instruction *IP) {
 
-  // Force to unsigned so that the shift rights are logical.
-  if (DestTy->isSigned())
-    V = new CastInst(V, DestTy->getUnsignedVersion(), V->getName(), IP);
+  IRBuilder<> Builder(IP->getParent(), IP);
 
   unsigned BitSize = V->getType()->getPrimitiveSizeInBits();
-  for (unsigned i = 1; i != BitSize; i <<= 1) {
-    Value *ShVal = ConstantInt::get(Type::UByteTy, i);
-    ShVal = new ShiftInst(Instruction::Shr, V, ShVal, "ctlz.sh", IP);
-    V = BinaryOperator::createOr(V, ShVal, "ctlz.step", IP);
+  for (unsigned i = 1; i < BitSize; i <<= 1) {
+    Value *ShVal = ConstantInt::get(V->getType(), i);
+    ShVal = Builder.CreateLShr(V, ShVal, "ctlz.sh");
+    V = Builder.CreateOr(V, ShVal, "ctlz.step");
   }
 
-  if (V->getType() != DestTy)
-    V = new CastInst(V, DestTy, V->getName(), IP);
+  V = Builder.CreateNot(V);
+  return LowerCTPOP(Context, V, IP);
+}
 
-  V = BinaryOperator::createNot(V, "", IP);
-  return LowerCTPOP(V, IP);
+static void ReplaceFPIntrinsicWithCall(CallInst *CI, const char *Fname,
+                                       const char *Dname,
+                                       const char *LDname) {
+  CallSite CS(CI);
+  switch (CI->getArgOperand(0)->getType()->getTypeID()) {
+  default: llvm_unreachable("Invalid type in intrinsic");
+  case Type::FloatTyID:
+    ReplaceCallWith(Fname, CI, CS.arg_begin(), CS.arg_end(),
+                  Type::getFloatTy(CI->getContext()));
+    break;
+  case Type::DoubleTyID:
+    ReplaceCallWith(Dname, CI, CS.arg_begin(), CS.arg_end(),
+                  Type::getDoubleTy(CI->getContext()));
+    break;
+  case Type::X86_FP80TyID:
+  case Type::FP128TyID:
+  case Type::PPC_FP128TyID:
+    ReplaceCallWith(LDname, CI, CS.arg_begin(), CS.arg_end(),
+                  CI->getArgOperand(0)->getType());
+    break;
+  }
 }
 
-void DefaultIntrinsicLowering::LowerIntrinsicCall(CallInst *CI) {
-  Function *Callee = CI->getCalledFunction();
+void IntrinsicLowering::LowerIntrinsicCall(CallInst *CI) {
+  IRBuilder<> Builder(CI->getParent(), CI);
+  LLVMContext &Context = CI->getContext();
+
+  const Function *Callee = CI->getCalledFunction();
   assert(Callee && "Cannot lower an indirect call!");
 
+  CallSite CS(CI);
   switch (Callee->getIntrinsicID()) {
   case Intrinsic::not_intrinsic:
-    std::cerr << "Cannot lower a call to a non-intrinsic function '"
-              << Callee->getName() << "'!\n";
-    abort();
+    report_fatal_error("Cannot lower a call to a non-intrinsic function '"+
+                      Callee->getName() + "'!");
   default:
-    std::cerr << "Error: Code generator does not support intrinsic function '"
-              << Callee->getName() << "'!\n";
-    abort();
+    report_fatal_error("Code generator does not support intrinsic function '"+
+                      Callee->getName()+"'!");
+
+  case Intrinsic::expect: {
+    // Just replace __builtin_expect(exp, c) with EXP.
+    Value *V = CI->getArgOperand(0);
+    CI->replaceAllUsesWith(V);
+    break;
+  }
 
     // The setjmp/longjmp intrinsics should only exist in the code if it was
     // never optimized (ie, right out of the CFE), or if it has been hacked on
     // by the lowerinvoke pass.  In both cases, the right thing to do is to
     // convert the call to an explicit setjmp or longjmp call.
   case Intrinsic::setjmp: {
-    static Function *SetjmpFCache = 0;
-    Value *V = ReplaceCallWith("setjmp", CI, CI->op_begin()+1, CI->op_end(),
-                               Type::IntTy, SetjmpFCache);
-    if (CI->getType() != Type::VoidTy)
+    Value *V = ReplaceCallWith("setjmp", CI, CS.arg_begin(), CS.arg_end(),
+                               Type::getInt32Ty(Context));
+    if (!CI->getType()->isVoidTy())
       CI->replaceAllUsesWith(V);
     break;
   }
   case Intrinsic::sigsetjmp:
-     if (CI->getType() != Type::VoidTy)
+     if (!CI->getType()->isVoidTy())
        CI->replaceAllUsesWith(Constant::getNullValue(CI->getType()));
      break;
 
   case Intrinsic::longjmp: {
-    static Function *LongjmpFCache = 0;
-    ReplaceCallWith("longjmp", CI, CI->op_begin()+1, CI->op_end(),
-                    Type::VoidTy, LongjmpFCache);
+    ReplaceCallWith("longjmp", CI, CS.arg_begin(), CS.arg_end(),
+                    Type::getVoidTy(Context));
     break;
   }
 
   case Intrinsic::siglongjmp: {
     // Insert the call to abort
-    static Function *AbortFCache = 0;
-    ReplaceCallWith("abort", CI, CI->op_end(), CI->op_end(), Type::VoidTy,
-                    AbortFCache);
+    ReplaceCallWith("abort", CI, CS.arg_end(), CS.arg_end(), 
+                    Type::getVoidTy(Context));
     break;
   }
   case Intrinsic::ctpop:
-    CI->replaceAllUsesWith(LowerCTPOP(CI->getOperand(1), CI));
+    CI->replaceAllUsesWith(LowerCTPOP(Context, CI->getArgOperand(0), CI));
     break;
 
+  case Intrinsic::bswap:
+    CI->replaceAllUsesWith(LowerBSWAP(Context, CI->getArgOperand(0), CI));
+    break;
+    
   case Intrinsic::ctlz:
-    CI->replaceAllUsesWith(LowerCTLZ(CI->getOperand(1), CI));
+    CI->replaceAllUsesWith(LowerCTLZ(Context, CI->getArgOperand(0), CI));
     break;
+
   case Intrinsic::cttz: {
     // cttz(x) -> ctpop(~X & (X-1))
-    Value *Src = CI->getOperand(1);
-    Value *NotSrc = BinaryOperator::createNot(Src, Src->getName()+".not", CI);
-    Value *SrcM1  = ConstantInt::get(Src->getType(), 1);
-    SrcM1 = BinaryOperator::createSub(Src, SrcM1, "", CI);
-    Src = LowerCTPOP(BinaryOperator::createAnd(NotSrc, SrcM1, "", CI), CI);
+    Value *Src = CI->getArgOperand(0);
+    Value *NotSrc = Builder.CreateNot(Src);
+    NotSrc->setName(Src->getName() + ".not");
+    Value *SrcM1 = ConstantInt::get(Src->getType(), 1);
+    SrcM1 = Builder.CreateSub(Src, SrcM1);
+    Src = LowerCTPOP(Context, Builder.CreateAnd(NotSrc, SrcM1), CI);
     CI->replaceAllUsesWith(Src);
     break;
   }
 
+  case Intrinsic::stacksave:
+  case Intrinsic::stackrestore: {
+    if (!Warned)
+      errs() << "WARNING: this target does not support the llvm.stack"
+             << (Callee->getIntrinsicID() == Intrinsic::stacksave ?
+               "save" : "restore") << " intrinsic.\n";
+    Warned = true;
+    if (Callee->getIntrinsicID() == Intrinsic::stacksave)
+      CI->replaceAllUsesWith(Constant::getNullValue(CI->getType()));
+    break;
+  }
+    
   case Intrinsic::returnaddress:
   case Intrinsic::frameaddress:
-    std::cerr << "WARNING: this target does not support the llvm."
-              << (Callee->getIntrinsicID() == Intrinsic::returnaddress ?
-                  "return" : "frame") << "address intrinsic.\n";
+    errs() << "WARNING: this target does not support the llvm."
+           << (Callee->getIntrinsicID() == Intrinsic::returnaddress ?
+             "return" : "frame") << "address intrinsic.\n";
     CI->replaceAllUsesWith(ConstantPointerNull::get(
                                             cast<PointerType>(CI->getType())));
     break;
@@ -263,70 +439,134 @@ void DefaultIntrinsicLowering::LowerIntrinsicCall(CallInst *CI) {
   case Intrinsic::pcmarker:
     break;    // Simply strip out pcmarker on unsupported architectures
   case Intrinsic::readcyclecounter: {
-    std::cerr << "WARNING: this target does not support the llvm.readcyclecounter"
-              << " intrinsic.  It is being lowered to a constant 0\n";
-    CI->replaceAllUsesWith(ConstantUInt::get(Type::ULongTy, 0));
+    errs() << "WARNING: this target does not support the llvm.readcyclecoun"
+           << "ter intrinsic.  It is being lowered to a constant 0\n";
+    CI->replaceAllUsesWith(ConstantInt::get(Type::getInt64Ty(Context), 0));
     break;
   }
 
-  case Intrinsic::dbg_stoppoint:
-  case Intrinsic::dbg_region_start:
-  case Intrinsic::dbg_region_end:
   case Intrinsic::dbg_declare:
-  case Intrinsic::dbg_func_start:
-    if (CI->getType() != Type::VoidTy)
-      CI->replaceAllUsesWith(Constant::getNullValue(CI->getType()));
     break;    // Simply strip out debugging intrinsics
 
+  case Intrinsic::eh_typeid_for:
+    // Return something different to eh_selector.
+    CI->replaceAllUsesWith(ConstantInt::get(CI->getType(), 1));
+    break;
+
+  case Intrinsic::annotation:
+  case Intrinsic::ptr_annotation:
+    // Just drop the annotation, but forward the value
+    CI->replaceAllUsesWith(CI->getOperand(0));
+    break;
+
+  case Intrinsic::var_annotation:
+    break;   // Strip out annotate intrinsic
+    
   case Intrinsic::memcpy: {
-    // The memcpy intrinsic take an extra alignment argument that the memcpy
-    // libc function does not.
-    static Function *MemcpyFCache = 0;
-    ReplaceCallWith("memcpy", CI, CI->op_begin()+1, CI->op_end()-1,
-                    (*(CI->op_begin()+1))->getType(), MemcpyFCache);
+    Type *IntPtr = TD.getIntPtrType(Context);
+    Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr,
+                                        /* isSigned */ false);
+    Value *Ops[3];
+    Ops[0] = CI->getArgOperand(0);
+    Ops[1] = CI->getArgOperand(1);
+    Ops[2] = Size;
+    ReplaceCallWith("memcpy", CI, Ops, Ops+3, CI->getArgOperand(0)->getType());
     break;
   }
   case Intrinsic::memmove: {
-    // The memmove intrinsic take an extra alignment argument that the memmove
-    // libc function does not.
-    static Function *MemmoveFCache = 0;
-    ReplaceCallWith("memmove", CI, CI->op_begin()+1, CI->op_end()-1,
-                    (*(CI->op_begin()+1))->getType(), MemmoveFCache);
+    Type *IntPtr = TD.getIntPtrType(Context);
+    Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr,
+                                        /* isSigned */ false);
+    Value *Ops[3];
+    Ops[0] = CI->getArgOperand(0);
+    Ops[1] = CI->getArgOperand(1);
+    Ops[2] = Size;
+    ReplaceCallWith("memmove", CI, Ops, Ops+3, CI->getArgOperand(0)->getType());
     break;
   }
   case Intrinsic::memset: {
-    // The memset intrinsic take an extra alignment argument that the memset
-    // libc function does not.
-    static Function *MemsetFCache = 0;
-    ReplaceCallWith("memset", CI, CI->op_begin()+1, CI->op_end()-1,
-                    (*(CI->op_begin()+1))->getType(), MemsetFCache);
+    Value *Op0 = CI->getArgOperand(0);
+    Type *IntPtr = TD.getIntPtrType(Op0->getType());
+    Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr,
+                                        /* isSigned */ false);
+    Value *Ops[3];
+    Ops[0] = Op0;
+    // Extend the amount to i32.
+    Ops[1] = Builder.CreateIntCast(CI->getArgOperand(1),
+                                   Type::getInt32Ty(Context),
+                                   /* isSigned */ false);
+    Ops[2] = Size;
+    ReplaceCallWith("memset", CI, Ops, Ops+3, CI->getArgOperand(0)->getType());
     break;
   }
-  case Intrinsic::isunordered: {
-    Value *L = CI->getOperand(1);
-    Value *R = CI->getOperand(2);
-
-    Value *LIsNan = new SetCondInst(Instruction::SetNE, L, L, "LIsNan", CI);
-    Value *RIsNan = new SetCondInst(Instruction::SetNE, R, R, "RIsNan", CI);
-    CI->replaceAllUsesWith(
-      BinaryOperator::create(Instruction::Or, LIsNan, RIsNan,
-                             "isunordered", CI));
+  case Intrinsic::sqrt: {
+    ReplaceFPIntrinsicWithCall(CI, "sqrtf", "sqrt", "sqrtl");
     break;
   }
-  case Intrinsic::sqrt: {
-    static Function *sqrtFCache = 0;
-    static Function *sqrtfFCache = 0;
-    if(CI->getType() == Type::FloatTy)
-      ReplaceCallWith("sqrtf", CI, CI->op_begin()+1, CI->op_end(),
-                      Type::FloatTy, sqrtfFCache);
-    else
-      ReplaceCallWith("sqrt", CI, CI->op_begin()+1, CI->op_end(),
-                      Type::DoubleTy, sqrtFCache);
+  case Intrinsic::log: {
+    ReplaceFPIntrinsicWithCall(CI, "logf", "log", "logl");
+    break;
+  }
+  case Intrinsic::log2: {
+    ReplaceFPIntrinsicWithCall(CI, "log2f", "log2", "log2l");
     break;
   }
+  case Intrinsic::log10: {
+    ReplaceFPIntrinsicWithCall(CI, "log10f", "log10", "log10l");
+    break;
+  }
+  case Intrinsic::exp: {
+    ReplaceFPIntrinsicWithCall(CI, "expf", "exp", "expl");
+    break;
+  }
+  case Intrinsic::exp2: {
+    ReplaceFPIntrinsicWithCall(CI, "exp2f", "exp2", "exp2l");
+    break;
+  }
+  case Intrinsic::pow: {
+    ReplaceFPIntrinsicWithCall(CI, "powf", "pow", "powl");
+    break;
+  }
+  case Intrinsic::flt_rounds:
+     // Lower to "round to the nearest"
+     if (!CI->getType()->isVoidTy())
+       CI->replaceAllUsesWith(ConstantInt::get(CI->getType(), 1));
+     break;
+  case Intrinsic::invariant_start:
+  case Intrinsic::lifetime_start:
+    // Discard region information.
+    CI->replaceAllUsesWith(UndefValue::get(CI->getType()));
+    break;
+  case Intrinsic::invariant_end:
+  case Intrinsic::lifetime_end:
+    // Discard region information.
+    break;
   }
 
   assert(CI->use_empty() &&
          "Lowering should have eliminated any uses of the intrinsic call!");
   CI->eraseFromParent();
 }
+
+bool IntrinsicLowering::LowerToByteSwap(CallInst *CI) {
+  // Verify this is a simple bswap.
+  if (CI->getNumArgOperands() != 1 ||
+      CI->getType() != CI->getArgOperand(0)->getType() ||
+      !CI->getType()->isIntegerTy())
+    return false;
+
+  IntegerType *Ty = dyn_cast<IntegerType>(CI->getType());
+  if (!Ty)
+    return false;
+
+  // Okay, we can do this xform, do so now.
+  Module *M = CI->getParent()->getParent()->getParent();
+  Constant *Int = Intrinsic::getDeclaration(M, Intrinsic::bswap, Ty);
+
+  Value *Op = CI->getArgOperand(0);
+  Op = CallInst::Create(Int, Op, CI->getName(), CI);
+
+  CI->replaceAllUsesWith(Op);
+  CI->eraseFromParent();
+  return true;
+}