X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FIntrinsicLowering.cpp;h=d894f664dcb3dd19c2abd33e0c3dd8a54f2704fb;hb=08368387a450dc2b5681000e2728ec702a8f1197;hp=6ec3a629fbd5d130336fccf563db13e740a4b0fa;hpb=c23197a26f34f559ea9797de51e187087c039c42;p=oota-llvm.git diff --git a/lib/CodeGen/IntrinsicLowering.cpp b/lib/CodeGen/IntrinsicLowering.cpp index 6ec3a629fbd..d894f664dcb 100644 --- a/lib/CodeGen/IntrinsicLowering.cpp +++ b/lib/CodeGen/IntrinsicLowering.cpp @@ -11,23 +11,25 @@ // //===----------------------------------------------------------------------===// -#include "llvm/Constants.h" -#include "llvm/DerivedTypes.h" -#include "llvm/Module.h" -#include "llvm/Type.h" #include "llvm/CodeGen/IntrinsicLowering.h" -#include "llvm/Support/IRBuilder.h" -#include "llvm/Support/ErrorHandling.h" -#include "llvm/Target/TargetData.h" #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 static void EnsureFunctionExists(Module &M, const char *Name, ArgIt ArgBegin, ArgIt ArgEnd, - const Type *RetTy) { + Type *RetTy) { // Insert a correctly-typed definition now. - std::vector ParamTys; + std::vector ParamTys; for (ArgIt I = ArgBegin; I != ArgEnd; ++I) ParamTys.push_back(I->getType()); M.getOrInsertFunction(Name, FunctionType::get(RetTy, ParamTys, false)); @@ -40,11 +42,11 @@ static void EnsureFPIntrinsicsExist(Module &M, Function *Fn, switch((int)Fn->arg_begin()->getType()->getTypeID()) { case Type::FloatTyID: EnsureFunctionExists(M, FName, Fn->arg_begin(), Fn->arg_end(), - Type::FloatTy); + Type::getFloatTy(M.getContext())); break; case Type::DoubleTyID: EnsureFunctionExists(M, DName, Fn->arg_begin(), Fn->arg_end(), - Type::DoubleTy); + Type::getDoubleTy(M.getContext())); break; case Type::X86_FP80TyID: case Type::FP128TyID: @@ -62,12 +64,12 @@ static void EnsureFPIntrinsicsExist(Module &M, Function *Fn, template static CallInst *ReplaceCallWith(const char *NewFn, CallInst *CI, ArgIt ArgBegin, ArgIt ArgEnd, - const Type *RetTy) { + 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 ParamTys; + std::vector ParamTys; for (ArgIt I = ArgBegin; I != ArgEnd; ++I) ParamTys.push_back((*I)->getType()); Constant* FCache = M->getOrInsertFunction(NewFn, @@ -75,47 +77,59 @@ static CallInst *ReplaceCallWith(const char *NewFn, CallInst *CI, IRBuilder<> Builder(CI->getParent(), CI); SmallVector Args(ArgBegin, ArgEnd); - CallInst *NewCI = Builder.CreateCall(FCache, Args.begin(), Args.end()); + CallInst *NewCI = Builder.CreateCall(FCache, Args); NewCI->setName(CI->getName()); if (!CI->use_empty()) CI->replaceAllUsesWith(NewCI); return NewCI; } +// 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->isDeclaration() && !I->use_empty()) switch (I->getIntrinsicID()) { default: break; case Intrinsic::setjmp: EnsureFunctionExists(M, "setjmp", I->arg_begin(), I->arg_end(), - Type::Int32Ty); + 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: - M.getOrInsertFunction("memcpy", PointerType::getUnqual(Type::Int8Ty), - PointerType::getUnqual(Type::Int8Ty), - PointerType::getUnqual(Type::Int8Ty), - TD.getIntPtrType(), (Type *)0); + M.getOrInsertFunction("memcpy", + Type::getInt8PtrTy(Context), + Type::getInt8PtrTy(Context), + Type::getInt8PtrTy(Context), + TD.getIntPtrType(Context), (Type *)0); break; case Intrinsic::memmove: - M.getOrInsertFunction("memmove", PointerType::getUnqual(Type::Int8Ty), - PointerType::getUnqual(Type::Int8Ty), - PointerType::getUnqual(Type::Int8Ty), - TD.getIntPtrType(), (Type *)0); + 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::getUnqual(Type::Int8Ty), - PointerType::getUnqual(Type::Int8Ty), - Type::Int32Ty, - TD.getIntPtrType(), (Type *)0); + M.getOrInsertFunction("memset", + Type::getInt8PtrTy(Context), + Type::getInt8PtrTy(Context), + Type::getInt32Ty(M.getContext()), + TD.getIntPtrType(Context), (Type *)0); break; case Intrinsic::sqrt: EnsureFPIntrinsicsExist(M, I, "sqrtf", "sqrt", "sqrtl"); @@ -149,8 +163,8 @@ void IntrinsicLowering::AddPrototypes(Module &M) { /// LowerBSWAP - Emit the code to lower bswap of V before the specified /// instruction IP. -static Value *LowerBSWAP(Value *V, Instruction *IP) { - assert(V->getType()->isInteger() && "Can't bswap a non-integer type!"); +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(); @@ -173,11 +187,13 @@ static Value *LowerBSWAP(Value *V, Instruction *IP) { "bswap.3"); Value *Tmp2 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 8), "bswap.2"); - Value *Tmp1 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 24), + Value *Tmp1 = Builder.CreateLShr(V,ConstantInt::get(V->getType(), 24), "bswap.1"); - Tmp3 = Builder.CreateAnd(Tmp3, ConstantInt::get(Type::Int32Ty, 0xFF0000), + Tmp3 = Builder.CreateAnd(Tmp3, + ConstantInt::get(Type::getInt32Ty(Context), 0xFF0000), "bswap.and3"); - Tmp2 = Builder.CreateAnd(Tmp2, ConstantInt::get(Type::Int32Ty, 0xFF00), + 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"); @@ -195,31 +211,38 @@ static Value *LowerBSWAP(Value *V, Instruction *IP) { "bswap.5"); Value* Tmp4 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 8), "bswap.4"); - Value* Tmp3 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 24), + Value* Tmp3 = Builder.CreateLShr(V, + ConstantInt::get(V->getType(), 24), "bswap.3"); - Value* Tmp2 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 40), + Value* Tmp2 = Builder.CreateLShr(V, + ConstantInt::get(V->getType(), 40), "bswap.2"); - Value* Tmp1 = Builder.CreateLShr(V, ConstantInt::get(V->getType(), 56), + Value* Tmp1 = Builder.CreateLShr(V, + ConstantInt::get(V->getType(), 56), "bswap.1"); Tmp7 = Builder.CreateAnd(Tmp7, - ConstantInt::get(Type::Int64Ty, + ConstantInt::get(Type::getInt64Ty(Context), 0xFF000000000000ULL), "bswap.and7"); Tmp6 = Builder.CreateAnd(Tmp6, - ConstantInt::get(Type::Int64Ty, + ConstantInt::get(Type::getInt64Ty(Context), 0xFF0000000000ULL), "bswap.and6"); Tmp5 = Builder.CreateAnd(Tmp5, - ConstantInt::get(Type::Int64Ty, 0xFF00000000ULL), + ConstantInt::get(Type::getInt64Ty(Context), + 0xFF00000000ULL), "bswap.and5"); Tmp4 = Builder.CreateAnd(Tmp4, - ConstantInt::get(Type::Int64Ty, 0xFF000000ULL), + ConstantInt::get(Type::getInt64Ty(Context), + 0xFF000000ULL), "bswap.and4"); Tmp3 = Builder.CreateAnd(Tmp3, - ConstantInt::get(Type::Int64Ty, 0xFF0000ULL), + ConstantInt::get(Type::getInt64Ty(Context), + 0xFF0000ULL), "bswap.and3"); Tmp2 = Builder.CreateAnd(Tmp2, - ConstantInt::get(Type::Int64Ty, 0xFF00ULL), + ConstantInt::get(Type::getInt64Ty(Context), + 0xFF00ULL), "bswap.and2"); Tmp8 = Builder.CreateOr(Tmp8, Tmp7, "bswap.or1"); Tmp6 = Builder.CreateOr(Tmp6, Tmp5, "bswap.or2"); @@ -236,8 +259,8 @@ static Value *LowerBSWAP(Value *V, Instruction *IP) { /// LowerCTPOP - Emit the code to lower ctpop of V before the specified /// instruction IP. -static Value *LowerCTPOP(Value *V, Instruction *IP) { - assert(V->getType()->isInteger() && "Can't ctpop a non-integer type!"); +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, @@ -258,7 +281,7 @@ static Value *LowerCTPOP(Value *V, Instruction *IP) { 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), + ConstantInt::get(V->getType(), i), "ctpop.sh"); Value *RHS = Builder.CreateAnd(VShift, MaskCst, "cppop.and2"); PartValue = Builder.CreateAdd(LHS, RHS, "ctpop.step"); @@ -276,7 +299,7 @@ static Value *LowerCTPOP(Value *V, Instruction *IP) { /// LowerCTLZ - Emit the code to lower ctlz of V before the specified /// instruction IP. -static Value *LowerCTLZ(Value *V, Instruction *IP) { +static Value *LowerCTLZ(LLVMContext &Context, Value *V, Instruction *IP) { IRBuilder<> Builder(IP->getParent(), IP); @@ -288,94 +311,103 @@ static Value *LowerCTLZ(Value *V, Instruction *IP) { } V = Builder.CreateNot(V); - return LowerCTPOP(V, IP); + return LowerCTPOP(Context, V, IP); } static void ReplaceFPIntrinsicWithCall(CallInst *CI, const char *Fname, const char *Dname, const char *LDname) { - switch (CI->getOperand(1)->getType()->getTypeID()) { + CallSite CS(CI); + switch (CI->getArgOperand(0)->getType()->getTypeID()) { default: llvm_unreachable("Invalid type in intrinsic"); case Type::FloatTyID: - ReplaceCallWith(Fname, CI, CI->op_begin() + 1, CI->op_end(), - Type::FloatTy); + ReplaceCallWith(Fname, CI, CS.arg_begin(), CS.arg_end(), + Type::getFloatTy(CI->getContext())); break; case Type::DoubleTyID: - ReplaceCallWith(Dname, CI, CI->op_begin() + 1, CI->op_end(), - Type::DoubleTy); + 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, CI->op_begin() + 1, CI->op_end(), - CI->getOperand(1)->getType()); + ReplaceCallWith(LDname, CI, CS.arg_begin(), CS.arg_end(), + CI->getArgOperand(0)->getType()); break; } } void IntrinsicLowering::LowerIntrinsicCall(CallInst *CI) { IRBuilder<> Builder(CI->getParent(), CI); - LLVMContext *Context = CI->getParent()->getContext(); + LLVMContext &Context = CI->getContext(); - Function *Callee = CI->getCalledFunction(); + const Function *Callee = CI->getCalledFunction(); assert(Callee && "Cannot lower an indirect call!"); + CallSite CS(CI); switch (Callee->getIntrinsicID()) { case Intrinsic::not_intrinsic: - llvm_report_error("Cannot lower a call to a non-intrinsic function '"+ + report_fatal_error("Cannot lower a call to a non-intrinsic function '"+ Callee->getName() + "'!"); default: - llvm_report_error("Code generator does not support intrinsic function '"+ + 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: { - Value *V = ReplaceCallWith("setjmp", CI, CI->op_begin() + 1, CI->op_end(), - Type::Int32Ty); - 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) - CI->replaceAllUsesWith(Context->getNullValue(CI->getType())); + if (!CI->getType()->isVoidTy()) + CI->replaceAllUsesWith(Constant::getNullValue(CI->getType())); break; case Intrinsic::longjmp: { - ReplaceCallWith("longjmp", CI, CI->op_begin() + 1, CI->op_end(), - Type::VoidTy); + ReplaceCallWith("longjmp", CI, CS.arg_begin(), CS.arg_end(), + Type::getVoidTy(Context)); break; } case Intrinsic::siglongjmp: { // Insert the call to abort - ReplaceCallWith("abort", CI, CI->op_end(), CI->op_end(), - Type::VoidTy); + 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(CI->getOperand(1), CI)); + 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 *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(Builder.CreateAnd(NotSrc, SrcM1), CI); + Src = LowerCTPOP(Context, Builder.CreateAnd(NotSrc, SrcM1), CI); CI->replaceAllUsesWith(Src); break; } @@ -383,19 +415,19 @@ void IntrinsicLowering::LowerIntrinsicCall(CallInst *CI) { case Intrinsic::stacksave: case Intrinsic::stackrestore: { if (!Warned) - cerr << "WARNING: this target does not support the llvm.stack" - << (Callee->getIntrinsicID() == Intrinsic::stacksave ? + 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(Context->getNullValue(CI->getType())); + CI->replaceAllUsesWith(Constant::getNullValue(CI->getType())); break; } case Intrinsic::returnaddress: case Intrinsic::frameaddress: - cerr << "WARNING: this target does not support the llvm." - << (Callee->getIntrinsicID() == Intrinsic::returnaddress ? + errs() << "WARNING: this target does not support the llvm." + << (Callee->getIntrinsicID() == Intrinsic::returnaddress ? "return" : "frame") << "address intrinsic.\n"; CI->replaceAllUsesWith(ConstantPointerNull::get( cast(CI->getType()))); @@ -407,67 +439,63 @@ void IntrinsicLowering::LowerIntrinsicCall(CallInst *CI) { case Intrinsic::pcmarker: break; // Simply strip out pcmarker on unsupported architectures case Intrinsic::readcyclecounter: { - cerr << "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::Int64Ty, 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_func_start: case Intrinsic::dbg_declare: break; // Simply strip out debugging intrinsics - case Intrinsic::eh_exception: - case Intrinsic::eh_selector_i32: - case Intrinsic::eh_selector_i64: - CI->replaceAllUsesWith(Context->getNullValue(CI->getType())); - break; - - case Intrinsic::eh_typeid_for_i32: - case Intrinsic::eh_typeid_for_i64: + 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: { - const IntegerType *IntPtr = TD.getIntPtrType(); - Value *Size = Builder.CreateIntCast(CI->getOperand(3), IntPtr, + Type *IntPtr = TD.getIntPtrType(Context); + Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr, /* isSigned */ false); Value *Ops[3]; - Ops[0] = CI->getOperand(1); - Ops[1] = CI->getOperand(2); + Ops[0] = CI->getArgOperand(0); + Ops[1] = CI->getArgOperand(1); Ops[2] = Size; - ReplaceCallWith("memcpy", CI, Ops, Ops+3, CI->getOperand(1)->getType()); + ReplaceCallWith("memcpy", CI, Ops, Ops+3, CI->getArgOperand(0)->getType()); break; } case Intrinsic::memmove: { - const IntegerType *IntPtr = TD.getIntPtrType(); - Value *Size = Builder.CreateIntCast(CI->getOperand(3), IntPtr, + Type *IntPtr = TD.getIntPtrType(Context); + Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr, /* isSigned */ false); Value *Ops[3]; - Ops[0] = CI->getOperand(1); - Ops[1] = CI->getOperand(2); + Ops[0] = CI->getArgOperand(0); + Ops[1] = CI->getArgOperand(1); Ops[2] = Size; - ReplaceCallWith("memmove", CI, Ops, Ops+3, CI->getOperand(1)->getType()); + ReplaceCallWith("memmove", CI, Ops, Ops+3, CI->getArgOperand(0)->getType()); break; } case Intrinsic::memset: { - const IntegerType *IntPtr = TD.getIntPtrType(); - Value *Size = Builder.CreateIntCast(CI->getOperand(3), IntPtr, + Type *IntPtr = TD.getIntPtrType(Context); + Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr, /* isSigned */ false); Value *Ops[3]; - Ops[0] = CI->getOperand(1); + Ops[0] = CI->getArgOperand(0); // Extend the amount to i32. - Ops[1] = Builder.CreateIntCast(CI->getOperand(2), Type::Int32Ty, + Ops[1] = Builder.CreateIntCast(CI->getArgOperand(1), + Type::getInt32Ty(Context), /* isSigned */ false); Ops[2] = Size; - ReplaceCallWith("memset", CI, Ops, Ops+3, CI->getOperand(1)->getType()); + ReplaceCallWith("memset", CI, Ops, Ops+3, CI->getArgOperand(0)->getType()); break; } case Intrinsic::sqrt: { @@ -500,12 +528,44 @@ void IntrinsicLowering::LowerIntrinsicCall(CallInst *CI) { } case Intrinsic::flt_rounds: // Lower to "round to the nearest" - if (CI->getType() != Type::VoidTy) + 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(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; +}