[MemLoc] Support lllvm.memcpy.inline in MemoryLocation::getForArgument

[fhahn]

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

111

why pass -1 as length? That make the function be UB?

It would probably also be better to just pass in 2 noalias pointers, rather than allocas. That way the memcpy call won't be removed?

llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll

103

nit: tab? Same below.

[xbolva00]

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

111

-1 is also used at line 18.

[fhahn]

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

111

yeah, I don't think it really makes sense there either?

[xbolva00]

llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll

103

Original test case ("ret void" line) is wrongly formatted.

[xbolva00]

Addressed review comments.

[xbolva00]

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

24

hmm, no aliasing pointers - shouldn't we remove this store?

[fhahn]

LGTM, thanks for also improving the test case!

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

24

Yes that looks like a missed optimization that was hidden by the fact that the store got removed because it was to an alloca that was never read! Thank you very much for improving the test so that it exposes this. Not sure what is going on, but it might be good to add a TODO/FIXME comment.

[xbolva00]

Thanks!

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

24

Added FIXME.

llvm/lib/Analysis/MemoryLocation.cpp

MemoryLocation MemoryLocation::getForArgument(const CallBase *Call,
   if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(Call)) {
     const DataLayout &DL = II->getModule()->getDataLayout();
 
     switch (II->getIntrinsicID()) {
     default:
       break;
     case Intrinsic::memset:
     case Intrinsic::memcpy:
+    case Intrinsic::memcpy_inline:
     case Intrinsic::memmove:
       assert((ArgIdx == 0 || ArgIdx == 1) &&
              "Invalid argument index for memory intrinsic");
       if (ConstantInt *LenCI = dyn_cast<ConstantInt>(II->getArgOperand(2)))
         return MemoryLocation(Arg, LocationSize::precise(LenCI->getZExtValue()),
                               AATags);
       break;
 

llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -dse < %s | FileCheck %s
 
 declare void @llvm.memcpy.p0i8.p0i8.i8(i8* nocapture, i8* nocapture, i8, i1) nounwind
+declare void @llvm.memcpy.inline.p0i8.p0i8.i8(i8* nocapture, i8* nocapture, i8, i1) nounwind
 declare void @llvm.memmove.p0i8.p0i8.i8(i8* nocapture, i8* nocapture, i8, i1) nounwind
 declare void @llvm.memset.p0i8.i8(i8* nocapture, i8, i8, i1) nounwind
 
 define void @test1() {
 ; CHECK-LABEL: @test1(
 ; CHECK-NEXT:    ret void
 ;
   %A = alloca i8
   %B = alloca i8
 
   store i8 0, i8* %A  ;; Written to by memcpy
 
   call void @llvm.memcpy.p0i8.p0i8.i8(i8* %A, i8* %B, i8 -1, i1 false)
 
   ret void
 }
 
 define void @test2() {
 ; CHECK-LABEL: @test2(
 ; CHECK-NEXT:    ret void
 ;
   %A = alloca i8
   %B = alloca i8
 
   store i8 0, i8* %A  ;; Written to by memmove
 
   call void @llvm.memmove.p0i8.p0i8.i8(i8* %A, i8* %B, i8 -1, i1 false)
;
   %B = alloca i16, i16 1024, align 2
 
   store atomic i16 0, i16* %A unordered, align 2 ;; Written to by memset
 
   call void @llvm.memset.element.unordered.atomic.p0i16.i16(i16* align 2 %A, i8 0, i16 1024, i32 2)
 
   ret void
 }
+
+define void @test7() {
+; CHECK-LABEL: @test7(
+; CHECK-NEXT:    ret void
+;
+  %A = alloca i8
+  %B = alloca i8
+
+  store i8 0, i8* %A  ;; Written to by memcpy
+
+  call void @llvm.memcpy.inline.p0i8.p0i8.i8(i8* %A, i8* %B, i8 -1, i1 false)
+
+  ret void
+}

llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll

 ; CHECK-NEXT:    ret void
 ;
   tail call void @llvm.memcpy.element.unordered.atomic.p0i8.p0i8.i64(i8* align 1 %P, i8* align 1 %Q, i64 12, i32 1)
   tail call void @llvm.memcpy.element.unordered.atomic.p0i8.p0i8.i64(i8* align 1 %P, i8* align 1 %R, i64 12, i32 1)
   ret void
 }
 
 define void @test_memset_memcpy_inline(i8* noalias %P, i8* noalias %Q) {
+; CHECK-LABEL: @test_memset_memcpy_inline(
+; CHECK-NEXT:    tail call void @llvm.memcpy.inline.p0i8.p0i8.i64(i8* align 1 [[P:%.*]], i8* align 1 [[Q:%.*]], i64 12, i1 false)
+; CHECK-NEXT:    ret void
+;
   tail call void @llvm.memset.p0i8.i64(i8* %P, i8 42, i64 8, i1 false)
   tail call void @llvm.memcpy.inline.p0i8.p0i8.i64(i8* align 1 %P, i8* align 1 %Q, i64 12, i1 false)
   ret void
 }
 
 define void @test_store_memcpy_inline(i8* noalias %P, i8* noalias %Q) {
+; CHECK-LABEL: @test_store_memcpy_inline(
+; CHECK-NEXT:    [[P_4:%.*]] = getelementptr i8, i8* [[P:%.*]], i64 4
+; CHECK-NEXT:    store i8 4, i8* [[P_4]], align 1
+; CHECK-NEXT:    tail call void @llvm.memcpy.inline.p0i8.p0i8.i64(i8* align 1 [[P]], i8* align 1 [[Q:%.*]], i64 4, i1 false)
+; CHECK-NEXT:    ret void
+;
   store i8 0, i8* %P
   %P.1 = getelementptr i8, i8* %P, i64 1
   store i8 1, i8* %P.1
   %P.4 = getelementptr i8, i8* %P, i64 4
   store i8 4, i8* %P.4
   tail call void @llvm.memcpy.inline.p0i8.p0i8.i64(i8* align 1 %P, i8* align 1 %Q, i64 4, i1 false)
   ret void
 }
 
 declare void @llvm.memcpy.inline.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64 immarg, i1 immarg)