[InstCombine] mempcpy(d,s,n) to memcpy(d,s,n) + n

Summary:
Back-end currently expands mempcpy, but middle-end should work with memcpy instead of mempcpy to enable more memcpy-optimization.

GCC backend emits mempcpy, so LLVM backend could form it too, if we know mempcpy libcall is better than memcpy + n.
https://godbolt.org/z/dOCG96

Reviewers: efriedma, spatel, craig.topper, RKSimon, jdoerfert

Reviewed By: efriedma

Subscribers: hjl.tools, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D65737

llvm-svn: 370593

1 files changed, 11 insertions, 0 deletions

diff --git a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
index 7ff1f4aef23..1f397c0dedb 100644
--- a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -998,6 +998,15 @@ Value *LibCallSimplifier::optimizeMemCpy(CallInst *CI, IRBuilder<> &B,
   return CI->getArgOperand(0);
 }
 
+Value *LibCallSimplifier::optimizeMemPCpy(CallInst *CI, IRBuilder<> &B) {
+  Value *Dst = CI->getArgOperand(0);
+  Value *N = CI->getArgOperand(2);
+  // mempcpy(x, y, n) -> llvm.memcpy(align 1 x, align 1 y, n), x + n
+  CallInst *NewCI = B.CreateMemCpy(Dst, 1, CI->getArgOperand(1), 1, N);
+  NewCI->setAttributes(CI->getAttributes());
+  return B.CreateInBoundsGEP(B.getInt8Ty(), Dst, N);
+}
+
 Value *LibCallSimplifier::optimizeMemMove(CallInst *CI, IRBuilder<> &B, bool isIntrinsic) {
   Value *Size = CI->getArgOperand(2);
   if (ConstantInt *LenC = dyn_cast<ConstantInt>(Size))
@@ -2624,6 +2633,8 @@ Value *LibCallSimplifier::optimizeStringMemoryLibCall(CallInst *CI,
       return optimizeMemCmp(CI, Builder);
     case LibFunc_memcpy:
       return optimizeMemCpy(CI, Builder);
+    case LibFunc_mempcpy:
+      return optimizeMemPCpy(CI, Builder);
     case LibFunc_memmove:
       return optimizeMemMove(CI, Builder);
     case LibFunc_memset: