instcombine: Migrate stpcpy optimizations

This patch migrates the stpcpy optimizations from the simplify-libcalls
pass into the instcombine library call simplifier.  Note that the
__stpcpy_chk simplifications were migrated in a previous commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167083 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Transforms/Scalar/SimplifyLibCalls.cpp

@@ -134,52 +134,6 @@ static bool IsOnlyUsedInEqualityComparison(Value *V, Value *With) {
 //===----------------------------------------------------------------------===//
 
 namespace {
-//===---------------------------------------===//
-// 'stpcpy' Optimizations
-
-struct StpCpyOpt: public LibCallOptimization {
-  bool OptChkCall;  // True if it's optimizing a __stpcpy_chk libcall.
-
-  StpCpyOpt(bool c) : OptChkCall(c) {}
-
-  virtual Value *CallOptimizer(Function *Callee, CallInst *CI, IRBuilder<> &B) {
-    // Verify the "stpcpy" function prototype.
-    unsigned NumParams = OptChkCall ? 3 : 2;
-    FunctionType *FT = Callee->getFunctionType();
-    if (FT->getNumParams() != NumParams ||
-        FT->getReturnType() != FT->getParamType(0) ||
-        FT->getParamType(0) != FT->getParamType(1) ||
-        FT->getParamType(0) != B.getInt8PtrTy())
-      return 0;
-
-    // These optimizations require DataLayout.
-    if (!TD) return 0;
-
-    Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
-    if (Dst == Src) {  // stpcpy(x,x)  -> x+strlen(x)
-      Value *StrLen = EmitStrLen(Src, B, TD, TLI);
-      return StrLen ? B.CreateInBoundsGEP(Dst, StrLen) : 0;
-    }
-
-    // See if we can get the length of the input string.
-    uint64_t Len = GetStringLength(Src);
-    if (Len == 0) return 0;
-
-    Type *PT = FT->getParamType(0);
-    Value *LenV = ConstantInt::get(TD->getIntPtrType(PT), Len);
-    Value *DstEnd = B.CreateGEP(Dst,
-                                ConstantInt::get(TD->getIntPtrType(PT),
-                                                 Len - 1));
-
-    // We have enough information to now generate the memcpy call to do the
-    // copy for us.  Make a memcpy to copy the nul byte with align = 1.
-    if (!OptChkCall || !EmitMemCpyChk(Dst, Src, LenV, CI->getArgOperand(2), B,
-                                      TD, TLI))
-      B.CreateMemCpy(Dst, Src, LenV, 1);
-    return DstEnd;
-  }
-};
-
 //===---------------------------------------===//
 // 'strncpy' Optimizations
 
@@ -1242,7 +1196,6 @@ namespace {
 
     StringMap<LibCallOptimization*> Optimizations;
     // String and Memory LibCall Optimizations
-    StpCpyOpt StpCpy; StpCpyOpt StpCpyChk;
     StrNCpyOpt StrNCpy;
     StrLenOpt StrLen; StrPBrkOpt StrPBrk;
     StrToOpt StrTo; StrSpnOpt StrSpn; StrCSpnOpt StrCSpn; StrStrOpt StrStr;
@@ -1261,8 +1214,8 @@ namespace {
     bool Modified;  // This is only used by doInitialization.
   public:
     static char ID; // Pass identification
-    SimplifyLibCalls() : FunctionPass(ID), StpCpy(false), StpCpyChk(true),
+    SimplifyLibCalls() : FunctionPass(ID), UnaryDoubleFP(false),
-                         UnaryDoubleFP(false), UnsafeUnaryDoubleFP(true) {
+                         UnsafeUnaryDoubleFP(true) {
       initializeSimplifyLibCallsPass(*PassRegistry::getPassRegistry());
     }
     void AddOpt(LibFunc::Func F, LibCallOptimization* Opt);
@@ -1314,7 +1267,6 @@ void SimplifyLibCalls::AddOpt(LibFunc::Func F1, LibFunc::Func F2,
 void SimplifyLibCalls::InitOptimizations() {
   // String and Memory LibCall Optimizations
   Optimizations["strncpy"] = &StrNCpy;
-  Optimizations["stpcpy"] = &StpCpy;
   Optimizations["strlen"] = &StrLen;
   Optimizations["strpbrk"] = &StrPBrk;
   Optimizations["strtol"] = &StrTo;
@@ -1332,9 +1284,6 @@ void SimplifyLibCalls::InitOptimizations() {
   Optimizations["memmove"] = &MemMove;
   AddOpt(LibFunc::memset, &MemSet);
 
-  // _chk variants of String and Memory LibCall Optimizations.
-  Optimizations["__stpcpy_chk"] = &StpCpyChk;
-
   // Math Library Optimizations
   Optimizations["cosf"] = &Cos;
   Optimizations["cos"] = &Cos;

lib/Transforms/Utils/SimplifyLibCalls.cpp

@@ -592,6 +592,42 @@ struct StrCpyOpt : public LibCallOptimization {
   }
 };
 
+struct StpCpyOpt: public LibCallOptimization {
+  virtual Value *callOptimizer(Function *Callee, CallInst *CI, IRBuilder<> &B) {
+    // Verify the "stpcpy" function prototype.
+    FunctionType *FT = Callee->getFunctionType();
+    if (FT->getNumParams() != 2 ||
+        FT->getReturnType() != FT->getParamType(0) ||
+        FT->getParamType(0) != FT->getParamType(1) ||
+        FT->getParamType(0) != B.getInt8PtrTy())
+      return 0;
+
+    // These optimizations require DataLayout.
+    if (!TD) return 0;
+
+    Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
+    if (Dst == Src) {  // stpcpy(x,x)  -> x+strlen(x)
+      Value *StrLen = EmitStrLen(Src, B, TD, TLI);
+      return StrLen ? B.CreateInBoundsGEP(Dst, StrLen) : 0;
+    }
+
+    // See if we can get the length of the input string.
+    uint64_t Len = GetStringLength(Src);
+    if (Len == 0) return 0;
+
+    Type *PT = FT->getParamType(0);
+    Value *LenV = ConstantInt::get(TD->getIntPtrType(PT), Len);
+    Value *DstEnd = B.CreateGEP(Dst,
+                                ConstantInt::get(TD->getIntPtrType(PT),
+                                                 Len - 1));
+
+    // We have enough information to now generate the memcpy call to do the
+    // copy for us.  Make a memcpy to copy the nul byte with align = 1.
+    B.CreateMemCpy(Dst, Src, LenV, 1);
+    return DstEnd;
+  }
+};
+
 } // End anonymous namespace.
 
 namespace llvm {
@@ -617,6 +653,7 @@ class LibCallSimplifierImpl {
   StrCmpOpt StrCmp;
   StrNCmpOpt StrNCmp;
   StrCpyOpt StrCpy;
+  StpCpyOpt StpCpy;
 
   void initOptimizations();
 public:
@@ -646,6 +683,7 @@ void LibCallSimplifierImpl::initOptimizations() {
   Optimizations["strcmp"] = &StrCmp;
   Optimizations["strncmp"] = &StrNCmp;
   Optimizations["strcpy"] = &StrCpy;
+  Optimizations["stpcpy"] = &StpCpy;
 }
 
 Value *LibCallSimplifierImpl::optimizeCall(CallInst *CI) {

test/Transforms/InstCombine/stpcpy-1.ll

@@ -0,0 +1,46 @@
+; Test that the stpcpy library call simplifier works correctly.
+; RUN: opt < %s -instcombine -S | FileCheck %s
+;
+; This transformation requires the pointer size, as it assumes that size_t is
+; the size of a pointer.
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"
+
+@hello = constant [6 x i8] c"hello\00"
+@a = common global [32 x i8] zeroinitializer, align 1
+@b = common global [32 x i8] zeroinitializer, align 1
+
+declare i8* @stpcpy(i8*, i8*)
+
+define i8* @test_simplify1() {
+; CHECK: @test_simplify1
+
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8]* @hello, i32 0, i32 0
+
+  %ret = call i8* @stpcpy(i8* %dst, i8* %src)
+; CHECK: @llvm.memcpy.p0i8.p0i8.i32
+; CHECK-NEXT: getelementptr inbounds ([32 x i8]* @a, i32 0, i32 5)
+  ret i8* %ret
+}
+
+define i8* @test_simplify2() {
+; CHECK: @test_simplify2
+
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+
+  %ret = call i8* @stpcpy(i8* %dst, i8* %dst)
+; CHECK: [[LEN:%[a-z]+]] = call i32 @strlen
+; CHECK-NEXT: getelementptr inbounds [32 x i8]* @a, i32 0, i32 [[LEN]]
+  ret i8* %ret
+}
+
+define i8* @test_no_simplify1() {
+; CHECK: @test_no_simplify1
+
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [32 x i8]* @b, i32 0, i32 0
+
+  %ret = call i8* @stpcpy(i8* %dst, i8* %src)
+; CHECK: call i8* @stpcpy
+  ret i8* %ret
+}

test/Transforms/InstCombine/stpcpy-2.ll

@@ -0,0 +1,22 @@
+; Test that the stpcpy library call simplifier works correctly.
+; RUN: opt < %s -instcombine -S | FileCheck %s
+;
+; This transformation requires the pointer size, as it assumes that size_t is
+; the size of a pointer.
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"
+
+@hello = constant [6 x i8] c"hello\00"
+@a = common global [32 x i8] zeroinitializer, align 1
+
+declare i16* @stpcpy(i8*, i8*)
+
+define void @test_no_simplify1() {
+; CHECK: @test_no_simplify1
+
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8]* @hello, i32 0, i32 0
+
+  call i16* @stpcpy(i8* %dst, i8* %src)
+; CHECK: call i16* @stpcpy
+  ret void
+}

test/Transforms/InstCombine/stpcpy_chk-1.ll

@@ -16,7 +16,7 @@ define void @test_simplify1() {
   %dst = getelementptr inbounds [60 x i8]* @a, i32 0, i32 0
   %src = getelementptr inbounds [12 x i8]* @.str, i32 0, i32 0
 
-; CHECK-NEXT: call i8* @stpcpy
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32
   call i8* @__stpcpy_chk(i8* %dst, i8* %src, i32 60)
   ret void
 }
@@ -26,7 +26,7 @@ define void @test_simplify2() {
   %dst = getelementptr inbounds [60 x i8]* @a, i32 0, i32 0
   %src = getelementptr inbounds [12 x i8]* @.str, i32 0, i32 0
 
-; CHECK-NEXT: call i8* @stpcpy
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32
   call i8* @__stpcpy_chk(i8* %dst, i8* %src, i32 12)
   ret void
 }
@@ -36,7 +36,7 @@ define void @test_simplify3() {
   %dst = getelementptr inbounds [60 x i8]* @a, i32 0, i32 0
   %src = getelementptr inbounds [12 x i8]* @.str, i32 0, i32 0
 
-; CHECK-NEXT: call i8* @stpcpy
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32
   call i8* @__stpcpy_chk(i8* %dst, i8* %src, i32 -1)
   ret void
 }

test/Transforms/SimplifyLibCalls/StpCpy.ll

@@ -1,43 +0,0 @@
-; Test that the StpCpyOptimizer works correctly
-; RUN: opt < %s -simplify-libcalls -S | FileCheck %s
-
-; This transformation requires the pointer size, as it assumes that size_t is
-; the size of a pointer.
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"
-
-@hello = constant [6 x i8] c"hello\00"
-
-declare i8* @stpcpy(i8*, i8*)
-
-declare i8* @__stpcpy_chk(i8*, i8*, i32) nounwind
-
-declare i32 @llvm.objectsize.i32(i8*, i1) nounwind readonly
-
-define i32 @t1() {
-; CHECK: @t1
-  %target = alloca [1024 x i8]
-  %arg1 = getelementptr [1024 x i8]* %target, i32 0, i32 0
-  %arg2 = getelementptr [6 x i8]* @hello, i32 0, i32 0
-  %rslt1 = call i8* @stpcpy( i8* %arg1, i8* %arg2 )
-; CHECK: @llvm.memcpy.p0i8.p0i8.i32
-  ret i32 0
-}
-
-define i32 @t2() {
-; CHECK: @t2
-  %target = alloca [1024 x i8]
-  %arg1 = getelementptr [1024 x i8]* %target, i32 0, i32 0
-  %arg2 = getelementptr [6 x i8]* @hello, i32 0, i32 0
-  %tmp1 = call i32 @llvm.objectsize.i32(i8* %arg1, i1 false)
-  %rslt1 = call i8* @__stpcpy_chk(i8* %arg1, i8* %arg2, i32 %tmp1)
-; CHECK: @__memcpy_chk
-  ret i32 0
-}
-
-define i8* @t3(i8* %arg) {
-; CHECK: @t3
-  %stpcpy = tail call i8* @stpcpy(i8* %arg, i8* %arg)
-; CHECK: [[LEN:%[a-z]+]] = call i32 @strlen(i8* %arg)
-; CHECK-NEXT: getelementptr inbounds i8* %arg, i32 [[LEN]]
-  ret i8* %stpcpy
-}