instcombine: Migrate strncpy optimizations

This patch migrates the strncpy optimizations from the simplify-libcalls
pass into the instcombine library call simplifier.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167102 91177308-0d34-0410-b5e6-96231b3b80d8

6 files changed, 175 insertions, 90 deletions

diff --git a/lib/Transforms/Scalar/SimplifyLibCalls.cpp b/lib/Transforms/Scalar/SimplifyLibCalls.cpp
index f7593e7ea8..4251e60a91 100644
--- a/lib/Transforms/Scalar/SimplifyLibCalls.cpp
+++ b/lib/Transforms/Scalar/SimplifyLibCalls.cpp
@@ -135,56 +135,6 @@ static bool IsOnlyUsedInEqualityComparison(Value *V, Value *With) {
 
 namespace {
 //===---------------------------------------===//
-// 'strncpy' Optimizations
-
-struct StrNCpyOpt : public LibCallOptimization {
-  virtual Value *CallOptimizer(Function *Callee, CallInst *CI, IRBuilder<> &B) {
-    FunctionType *FT = Callee->getFunctionType();
-    if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-        FT->getParamType(0) != FT->getParamType(1) ||
-        FT->getParamType(0) != B.getInt8PtrTy() ||
-        !FT->getParamType(2)->isIntegerTy())
-      return 0;
-
-    Value *Dst = CI->getArgOperand(0);
-    Value *Src = CI->getArgOperand(1);
-    Value *LenOp = CI->getArgOperand(2);
-
-    // See if we can get the length of the input string.
-    uint64_t SrcLen = GetStringLength(Src);
-    if (SrcLen == 0) return 0;
-    --SrcLen;
-
-    if (SrcLen == 0) {
-      // strncpy(x, "", y) -> memset(x, '\0', y, 1)
-      B.CreateMemSet(Dst, B.getInt8('\0'), LenOp, 1);
-      return Dst;
-    }
-
-    uint64_t Len;
-    if (ConstantInt *LengthArg = dyn_cast<ConstantInt>(LenOp))
-      Len = LengthArg->getZExtValue();
-    else
-      return 0;
-
-    if (Len == 0) return Dst; // strncpy(x, y, 0) -> x
-
-    // These optimizations require DataLayout.
-    if (!TD) return 0;
-
-    // Let strncpy handle the zero padding
-    if (Len > SrcLen+1) return 0;
-
-    Type *PT = FT->getParamType(0);
-    // strncpy(x, s, c) -> memcpy(x, s, c, 1) [s and c are constant]
-    B.CreateMemCpy(Dst, Src,
-                   ConstantInt::get(TD->getIntPtrType(PT), Len), 1);
-
-    return Dst;
-  }
-};
-
-//===---------------------------------------===//
 // 'strlen' Optimizations
 
 struct StrLenOpt : public LibCallOptimization {
@@ -1196,7 +1146,6 @@ namespace {
 
     StringMap<LibCallOptimization*> Optimizations;
     // String and Memory LibCall Optimizations
-    StrNCpyOpt StrNCpy;
     StrLenOpt StrLen; StrPBrkOpt StrPBrk;
     StrToOpt StrTo; StrSpnOpt StrSpn; StrCSpnOpt StrCSpn; StrStrOpt StrStr;
     MemCmpOpt MemCmp; MemCpyOpt MemCpy; MemMoveOpt MemMove; MemSetOpt MemSet;
@@ -1266,7 +1215,6 @@ void SimplifyLibCalls::AddOpt(LibFunc::Func F1, LibFunc::Func F2,
 /// we know.
 void SimplifyLibCalls::InitOptimizations() {
   // String and Memory LibCall Optimizations
-  Optimizations["strncpy"] = &StrNCpy;
   Optimizations["strlen"] = &StrLen;
   Optimizations["strpbrk"] = &StrPBrk;
   Optimizations["strtol"] = &StrTo;
diff --git a/lib/Transforms/Utils/SimplifyLibCalls.cpp b/lib/Transforms/Utils/SimplifyLibCalls.cpp
index bc360010de..cc03573326 100644
--- a/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -628,6 +628,53 @@ struct StpCpyOpt: public LibCallOptimization {
   }
 };
 
+struct StrNCpyOpt : public LibCallOptimization {
+  virtual Value *callOptimizer(Function *Callee, CallInst *CI, IRBuilder<> &B) {
+    FunctionType *FT = Callee->getFunctionType();
+    if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
+        FT->getParamType(0) != FT->getParamType(1) ||
+        FT->getParamType(0) != B.getInt8PtrTy() ||
+        !FT->getParamType(2)->isIntegerTy())
+      return 0;
+
+    Value *Dst = CI->getArgOperand(0);
+    Value *Src = CI->getArgOperand(1);
+    Value *LenOp = CI->getArgOperand(2);
+
+    // See if we can get the length of the input string.
+    uint64_t SrcLen = GetStringLength(Src);
+    if (SrcLen == 0) return 0;
+    --SrcLen;
+
+    if (SrcLen == 0) {
+      // strncpy(x, "", y) -> memset(x, '\0', y, 1)
+      B.CreateMemSet(Dst, B.getInt8('\0'), LenOp, 1);
+      return Dst;
+    }
+
+    uint64_t Len;
+    if (ConstantInt *LengthArg = dyn_cast<ConstantInt>(LenOp))
+      Len = LengthArg->getZExtValue();
+    else
+      return 0;
+
+    if (Len == 0) return Dst; // strncpy(x, y, 0) -> x
+
+    // These optimizations require DataLayout.
+    if (!TD) return 0;
+
+    // Let strncpy handle the zero padding
+    if (Len > SrcLen+1) return 0;
+
+    Type *PT = FT->getParamType(0);
+    // strncpy(x, s, c) -> memcpy(x, s, c, 1) [s and c are constant]
+    B.CreateMemCpy(Dst, Src,
+                   ConstantInt::get(TD->getIntPtrType(PT), Len), 1);
+
+    return Dst;
+  }
+};
+
 } // End anonymous namespace.
 
 namespace llvm {
@@ -654,6 +701,7 @@ class LibCallSimplifierImpl {
   StrNCmpOpt StrNCmp;
   StrCpyOpt StrCpy;
   StpCpyOpt StpCpy;
+  StrNCpyOpt StrNCpy;
 
   void initOptimizations();
 public:
@@ -684,6 +732,7 @@ void LibCallSimplifierImpl::initOptimizations() {
   Optimizations["strncmp"] = &StrNCmp;
   Optimizations["strcpy"] = &StrCpy;
   Optimizations["stpcpy"] = &StpCpy;
+  Optimizations["strncpy"] = &StrNCpy;
 }
 
 Value *LibCallSimplifierImpl::optimizeCall(CallInst *CI) {
diff --git a/test/Transforms/InstCombine/strncpy-1.ll b/test/Transforms/InstCombine/strncpy-1.ll
new file mode 100644
index 0000000000..3ce2b9b5ee
--- /dev/null
+++ b/test/Transforms/InstCombine/strncpy-1.ll
@@ -0,0 +1,95 @@
+; Test that the strncpy library call simplifier works correctly.
+;
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+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"
+
+@hello = constant [6 x i8] c"hello\00"
+@null = constant [1 x i8] zeroinitializer
+@null_hello = constant [7 x i8] c"\00hello\00"
+@a = common global [32 x i8] zeroinitializer, align 1
+@b = common global [32 x i8] zeroinitializer, align 1
+
+declare i8* @strncpy(i8*, i8*, i32)
+declare i32 @puts(i8*)
+
+; Check a bunch of strncpy invocations together.
+
+define i32 @test_simplify1() {
+; CHECK: @test_simplify1
+; CHECK-NOT: call i8* @strncpy
+; CHECK: call i32 @puts
+  %target = alloca [1024 x i8]
+  %arg1 = getelementptr [1024 x i8]* %target, i32 0, i32 0
+  store i8 0, i8* %arg1
+
+  %arg2 = getelementptr [6 x i8]* @hello, i32 0, i32 0
+  %rslt1 = call i8* @strncpy(i8* %arg1, i8* %arg2, i32 6)
+
+  %arg3 = getelementptr [1 x i8]* @null, i32 0, i32 0
+  %rslt2 = call i8* @strncpy(i8* %rslt1, i8* %arg3, i32 42)
+
+  %arg4 = getelementptr [7 x i8]* @null_hello, i32 0, i32 0
+  %rslt3 = call i8* @strncpy(i8* %rslt2, i8* %arg4, i32 42)
+
+  call i32 @puts( i8* %rslt3 )
+  ret i32 0
+}
+
+; Check strncpy(x, "", y) -> memset(x, '\0', y, 1).
+
+define void @test_simplify2() {
+; CHECK: @test_simplify2
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [1 x i8]* @null, i32 0, i32 0
+
+  call i8* @strncpy(i8* %dst, i8* %src, i32 32)
+; CHECK: call void @llvm.memset.p0i8.i32
+  ret void
+}
+
+; Check strncpy(x, y, 0) -> x.
+
+define i8* @test_simplify3() {
+; CHECK: @test_simplify3
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8]* @hello, i32 0, i32 0
+
+  %ret = call i8* @strncpy(i8* %dst, i8* %src, i32 0)
+  ret i8* %ret
+; CHECK: ret i8* getelementptr inbounds ([32 x i8]* @a, i32 0, i32 0)
+}
+
+; Check  strncpy(x, s, c) -> memcpy(x, s, c, 1) [s and c are constant].
+
+define void @test_simplify4() {
+; CHECK: @test_simplify4
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8]* @hello, i32 0, i32 0
+
+  call i8* @strncpy(i8* %dst, i8* %src, i32 6)
+; CHECK: call void @llvm.memcpy.p0i8.p0i8.i32
+  ret void
+}
+
+; Check cases that shouldn't be simplified.
+
+define void @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
+
+  call i8* @strncpy(i8* %dst, i8* %src, i32 32)
+; CHECK: call i8* @strncpy
+  ret void
+}
+
+define void @test_no_simplify2() {
+; CHECK: @test_no_simplify2
+  %dst = getelementptr [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8]* @hello, i32 0, i32 0
+
+  call i8* @strncpy(i8* %dst, i8* %src, i32 8)
+; CHECK: call i8* @strncpy
+  ret void
+}
diff --git a/test/Transforms/InstCombine/strncpy-2.ll b/test/Transforms/InstCombine/strncpy-2.ll
new file mode 100644
index 0000000000..ac28ea6550
--- /dev/null
+++ b/test/Transforms/InstCombine/strncpy-2.ll
@@ -0,0 +1,22 @@
+; Test that the strncpy library call simplifier works correctly.
+;
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+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"
+
+@hello = constant [6 x i8] c"hello\00"
+@a = common global [32 x i8] zeroinitializer, align 1
+
+declare i16* @strncpy(i8*, i8*, i32)
+
+; Check that 'strncpy' functions with the wrong prototype aren't simplified.
+
+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* @strncpy(i8* %dst, i8* %src, i32 6)
+; CHECK: call i16* @strncpy
+  ret void
+}
diff --git a/test/Transforms/InstCombine/strncpy_chk-1.ll b/test/Transforms/InstCombine/strncpy_chk-1.ll
index ae7e2fb5f1..aadff4268e 100644
--- a/test/Transforms/InstCombine/strncpy_chk-1.ll
+++ b/test/Transforms/InstCombine/strncpy_chk-1.ll
@@ -7,27 +7,27 @@ target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f3
 
 @a = common global [60 x i8] zeroinitializer, align 1
 @b = common global [60 x i8] zeroinitializer, align 1
-@.str = private constant [8 x i8] c"abcdefg\00"
+@.str = private constant [12 x i8] c"abcdefghijk\00"
 
 ; Check cases where dstlen >= len
 
 define void @test_simplify1() {
 ; CHECK: @test_simplify1
   %dst = getelementptr inbounds [60 x i8]* @a, i32 0, i32 0
-  %src = getelementptr inbounds [8 x i8]* @.str, i32 0, i32 0
+  %src = getelementptr inbounds [12 x i8]* @.str, i32 0, i32 0
 
-; CHECK-NEXT: call i8* @strncpy
-  call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 8, i32 60)
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32
+  call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 12, i32 60)
   ret void
 }
 
 define void @test_simplify2() {
 ; CHECK: @test_simplify2
   %dst = getelementptr inbounds [60 x i8]* @a, i32 0, i32 0
-  %src = getelementptr inbounds [8 x i8]* @.str, i32 0, i32 0
+  %src = getelementptr inbounds [12 x i8]* @.str, i32 0, i32 0
 
-; CHECK-NEXT: call i8* @strncpy
-  call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 8, i32 8)
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32
+  call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 12, i32 12)
   ret void
 }
 
@@ -37,7 +37,7 @@ define void @test_simplify3() {
   %src = getelementptr inbounds [60 x i8]* @b, i32 0, i32 0
 
 ; CHECK-NEXT: call i8* @strncpy
-  call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 8, i32 60)
+  call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 12, i32 60)
   ret void
 }
 
@@ -46,7 +46,7 @@ define void @test_simplify3() {
 define void @test_no_simplify1() {
 ; CHECK: @test_no_simplify1
   %dst = getelementptr inbounds [60 x i8]* @a, i32 0, i32 0
-  %src = getelementptr inbounds [8 x i8]* @.str, i32 0, i32 0
+  %src = getelementptr inbounds [12 x i8]* @.str, i32 0, i32 0
 
 ; CHECK-NEXT: call i8* @__strncpy_chk
   call i8* @__strncpy_chk(i8* %dst, i8* %src, i32 8, i32 4)
diff --git a/test/Transforms/SimplifyLibCalls/StrNCpy.ll b/test/Transforms/SimplifyLibCalls/StrNCpy.ll
deleted file mode 100644
index 4e47b31a6a..0000000000
--- a/test/Transforms/SimplifyLibCalls/StrNCpy.ll
+++ /dev/null
@@ -1,29 +0,0 @@
-; Test that the StrNCpyOptimizer works correctly
-; RUN: opt < %s -simplify-libcalls -S | \
-; RUN:   not grep "call.*strncpy"
-
-; This transformation requires the pointer size, as it assumes that size_t is
-; the size of a pointer.
-target datalayout = "-p:64:64:64"
-
-@hello = constant [6 x i8] c"hello\00"		; <[6 x i8]*> [#uses=1]
-@null = constant [1 x i8] zeroinitializer		; <[1 x i8]*> [#uses=1]
-@null_hello = constant [7 x i8] c"\00hello\00"		; <[7 x i8]*> [#uses=1]
-
-declare i8* @strncpy(i8*, i8*, i32)
-
-declare i32 @puts(i8*)
-
-define i32 @main() {
-	%target = alloca [1024 x i8]		; <[1024 x i8]*> [#uses=1]
-	%arg1 = getelementptr [1024 x i8]* %target, i32 0, i32 0		; <i8*> [#uses=2]
-	store i8 0, i8* %arg1
-	%arg2 = getelementptr [6 x i8]* @hello, i32 0, i32 0		; <i8*> [#uses=1]
-	%rslt1 = call i8* @strncpy( i8* %arg1, i8* %arg2, i32 6 )		; <i8*> [#uses=1]
-	%arg3 = getelementptr [1 x i8]* @null, i32 0, i32 0		; <i8*> [#uses=1]
-	%rslt2 = call i8* @strncpy( i8* %rslt1, i8* %arg3, i32 42 )		; <i8*> [#uses=1]
-	%arg4 = getelementptr [7 x i8]* @null_hello, i32 0, i32 0		; <i8*> [#uses=1]
-	%rslt3 = call i8* @strncpy( i8* %rslt2, i8* %arg4, i32 42 )		; <i8*> [#uses=1]
-	call i32 @puts( i8* %rslt3 )		; <i32>:1 [#uses=0]
-	ret i32 0
-}