[GlobalISel] Translate memset/memmove/memcpy from undef ptrs into nothing

[paquette]

Summary

If the source is undef, then just don't do anything.

This matches SelectionDAG's behaviour in SelectionDAG.cpp.

Also add a test showing that we do the right thing here. (irtranslator-memfunc-undef.ll)

[arsenm]

llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-memfunc-undef.ll

3 ↗

(On Diff #203891)

This negative test seems fragile to me. Better to check entry, check-next ret or something?

[paquette]

Improve test

[arsenm]

LGTM

llvm/trunk/lib/CodeGen/GlobalISel/IRTranslator.cpp

bool IRTranslator::translateGetElementPtr(const User &U,
 
   MIRBuilder.buildCopy(getOrCreateVReg(U), BaseReg);
   return true;
 }
 
 bool IRTranslator::translateMemfunc(const CallInst &CI,
                                     MachineIRBuilder &MIRBuilder,
                                     unsigned ID) {
+
+  // If the source is undef, then just emit a nop.
+  if (isa<UndefValue>(CI.getArgOperand(1))) {
+    switch (ID) {
+    case Intrinsic::memmove:
+    case Intrinsic::memcpy:
+    case Intrinsic::memset:
+      return true;
+    default:
+      break;
+    }
+  }
+
   LLT SizeTy = getLLTForType(*CI.getArgOperand(2)->getType(), *DL);
   Type *DstTy = CI.getArgOperand(0)->getType();
   if (cast<PointerType>(DstTy)->getAddressSpace() != 0 ||
       SizeTy.getSizeInBits() != DL->getPointerSizeInBits(0))
     return false;
 
   SmallVector<CallLowering::ArgInfo, 8> Args;
   for (int i = 0; i < 3; ++i) {

llvm/trunk/test/CodeGen/AArch64/GlobalISel/irtranslator-memfunc-undef.ll

+; RUN: llc -mtriple=aarch64-unknown-unknown -global-isel -verify-machineinstrs -stop-after=irtranslator %s -o - | FileCheck %s
+
+define void @memset() {
+  ; CHECK-LABEL: name: memset
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:_(p0) = G_FRAME_INDEX %stack.0.buf
+  ; CHECK-NEXT:   RET_ReallyLR
+entry:
+  %buf = alloca [512 x i8], align 1
+  %ptr = getelementptr inbounds [512 x i8], [512 x i8]* %buf, i32 0, i32 0
+  call void @llvm.memset.p0i8.i32(i8* %ptr, i8 undef, i32 512, i1 false)
+  ret void
+}
+
+define void @memcpy() {
+  ; CHECK-LABEL: name: memcpy
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:_(p0) = G_FRAME_INDEX %stack.0.buf
+  ; CHECK-NEXT:   RET_ReallyLR
+entry:
+  %buf = alloca [512 x i8], align 1
+  %ptr = getelementptr inbounds [512 x i8], [512 x i8]* %buf, i32 0, i32 0
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %ptr, i8* undef, i32 512, i1 false)
+  ret void
+}
+
+define void @memmove() {
+  ; CHECK-LABEL: name: memmove
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT: [[FRAME_INDEX:%[0-9]+]]:_(p0) = G_FRAME_INDEX %stack.0.buf
+  ; CHECK-NEXT: RET_ReallyLR
+entry:
+  %buf = alloca [512 x i8], align 1
+  %ptr = getelementptr inbounds [512 x i8], [512 x i8]* %buf, i32 0, i32 0
+  call void @llvm.memmove.p0i8.p0i8.i32(i8* %ptr, i8* undef, i32 512, i1 false)
+  ret void
+}
+
+declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i1) nounwind
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture writeonly, i8* nocapture readonly, i32, i1) nounwind
+declare void @llvm.memmove.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i1) nounwind