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

Reviewed By: fhahn

Differential Revision: https://reviews.llvm.org/D87971

diff --git a/llvm/lib/Analysis/MemoryLocation.cpp b/llvm/lib/Analysis/MemoryLocation.cpp
index 976063c..17771f9 100644 (file)
--- a/llvm/lib/Analysis/MemoryLocation.cpp
+++ b/llvm/lib/Analysis/MemoryLocation.cpp
@@ -158,6 +158,7 @@ MemoryLocation MemoryLocation::getForArgument(const CallBase *Call,
       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");

diff --git a/llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll b/llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll
index 088752c..3b92bc2 100644 (file)
--- a/llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll
+++ b/llvm/test/Transforms/DeadStoreElimination/MSSA/memintrinsics.ll
@@ -2,48 +2,57 @@
 ; 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() {
+define void @test1(i8* noalias %A, i8* noalias %B) {
 ; CHECK-LABEL: @test1(
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i8(i8* [[A:%.*]], i8* [[B:%.*]], i8 12, i1 false)
 ; 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)
+  call void @llvm.memcpy.p0i8.p0i8.i8(i8* %A, i8* %B, i8 12, i1 false)
 
   ret void
 }
 
-define void @test2() {
+; FIXME: We should eliminate this dead store.
+define void @test2(i8* noalias %A, i8* noalias %B) {
 ; CHECK-LABEL: @test2(
+; CHECK-NEXT:    store i8 0, i8* [[A:%.*]], align 1
+; CHECK-NEXT:    call void @llvm.memmove.p0i8.p0i8.i8(i8* [[A]], i8* [[B:%.*]], i8 12, i1 false)
 ; 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 12, i1 false)
 
+  ret void
+}
+
+define void @test2a(i8* %A, i8* %B) {
+; CHECK-LABEL: @test2a(
+; CHECK-NEXT:    store i8 0, i8* [[A:%.*]], align 1
+; CHECK-NEXT:    call void @llvm.memmove.p0i8.p0i8.i8(i8* [[A]], i8* [[B:%.*]], i8 12, i1 false)
+; CHECK-NEXT:    ret void
+;
   store i8 0, i8* %A  ;; Written to by memmove
 
-  call void @llvm.memmove.p0i8.p0i8.i8(i8* %A, i8* %B, i8 -1, i1 false)
+  call void @llvm.memmove.p0i8.p0i8.i8(i8* %A, i8* %B, i8 12, i1 false)
 
   ret void
 }
 
-define void @test3() {
+define void @test3(i8* noalias %A) {
 ; CHECK-LABEL: @test3(
-; CHECK-NEXT:    [[B:%.*]] = alloca i8, align 1
+; CHECK-NEXT:    call void @llvm.memset.p0i8.i8(i8* [[A:%.*]], i8 0, i8 12, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %A = alloca i8
-  %B = alloca i8
-
   store i8 0, i8* %A  ;; Written to by memset
 
-  call void @llvm.memset.p0i8.i8(i8* %A, i8 0, i8 -1, i1 false)
+  call void @llvm.memset.p0i8.i8(i8* %A, i8 0, i8 12, i1 false)
 
   ret void
 }
@@ -97,3 +106,15 @@ define void @test6() {
 
   ret void
 }
+
+define void @test7(i8* noalias %A, i8* noalias %B) {
+; CHECK-LABEL: @test7(
+; CHECK-NEXT:    call void @llvm.memcpy.inline.p0i8.p0i8.i8(i8* [[A:%.*]], i8* [[B:%.*]], i8 12, i1 false)
+; CHECK-NEXT:    ret void
+;
+  store i8 0, i8* %A  ;; Written to by memcpy
+
+  call void @llvm.memcpy.inline.p0i8.p0i8.i8(i8* %A, i8* %B, i8 12, i1 false)
+
+  ret void
+}

diff --git a/llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll b/llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll
index ad88815..bb4a17f 100644 (file)
--- a/llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll
+++ b/llvm/test/Transforms/DeadStoreElimination/MSSA/memset-and-memcpy.ll
@@ -94,19 +94,29 @@ define void @test18_atomic(i8* %P, i8* %Q, i8* %R) nounwind ssp {
 }
 
 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
+  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
+  ret void
 }
 
 declare void @llvm.memcpy.inline.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64 immarg, i1 immarg)