[PATCH] D103419: [VectorCombine] Fix alignment in single element store

[Qiu Chaofan via Phabricator via llvm-commits]

qiucf updated this revision to Diff 350760.
qiucf marked an inline comment as done.
qiucf added a comment.

Add explicit align to affected test.
Add comment for implicit alignment.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D103419/new/

https://reviews.llvm.org/D103419

Files:
  llvm/lib/Transforms/Vectorize/VectorCombine.cpp
  llvm/test/Transforms/VectorCombine/load-insert-store.ll


Index: llvm/test/Transforms/VectorCombine/load-insert-store.ll
===================================================================
--- llvm/test/Transforms/VectorCombine/load-insert-store.ll
+++ llvm/test/Transforms/VectorCombine/load-insert-store.ll
@@ -2,6 +2,8 @@
 ; RUN: opt -S -vector-combine -data-layout=e < %s | FileCheck %s
 ; RUN: opt -S -vector-combine -data-layout=E < %s | FileCheck %s
 
+; These vector load/store without align will have implicit alignment of vector
+; size, which may affect align of scalarized store.
 define void @insert_store(<16 x i8>* %q, i8 zeroext %s) {
 ; CHECK-LABEL: @insert_store(
 ; CHECK-NEXT:  entry:
@@ -20,16 +22,30 @@
 ; CHECK-LABEL: @insert_store_i16_align1(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[Q:%.*]], i32 0, i32 3
-; CHECK-NEXT:    store i16 [[S:%.*]], i16* [[TMP0]], align 1
+; CHECK-NEXT:    store i16 [[S:%.*]], i16* [[TMP0]], align 2
 ; CHECK-NEXT:    ret void
 ;
 entry:
-  %0 = load <8 x i16>, <8 x i16>* %q
+  %0 = load <8 x i16>, <8 x i16>* %q, align 16
   %vecins = insertelement <8 x i16> %0, i16 %s, i32 3
   store <8 x i16> %vecins, <8 x i16>* %q, align 1
   ret void
 }
 
+define void @insert_store_zero_offset(<8 x i16>* %q, i16 zeroext %s) {
+; CHECK-LABEL: @insert_store_zero_offset(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[Q:%.*]], i32 0, i32 0
+; CHECK-NEXT:    store i16 [[S:%.*]], i16* [[TMP0]], align 128
+; CHECK-NEXT:    ret void
+;
+entry:
+  %0 = load <8 x i16>, <8 x i16>* %q, align 128
+  %vecins = insertelement <8 x i16> %0, i16 %s, i32 0
+  store <8 x i16> %vecins, <8 x i16>* %q, align 128
+  ret void
+}
+
 ; To verify case when index is out of bounds
 define void @insert_store_outofbounds(<8 x i16>* %q, i16 zeroext %s) {
 ; CHECK-LABEL: @insert_store_outofbounds(
@@ -125,6 +141,24 @@
   ret void
 }
 
+define void @insert_store_nonconst_large_alignment(<4 x i32>* %q, i32 zeroext %s, i32 %idx) {
+; CHECK-LABEL: @insert_store_nonconst_large_alignment(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i32 [[IDX:%.*]], 4
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP]])
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds <4 x i32>, <4 x i32>* [[Q:%.*]], i32 0, i32 [[IDX]]
+; CHECK-NEXT:    store i32 [[S:%.*]], i32* [[TMP0]], align 4
+; CHECK-NEXT:    ret void
+;
+entry:
+  %cmp = icmp ult i32 %idx, 4
+  call void @llvm.assume(i1 %cmp)
+  %i = load <4 x i32>, <4 x i32>* %q, align 128
+  %vecins = insertelement <4 x i32> %i, i32 %s, i32 %idx
+  store <4 x i32> %vecins, <4 x i32>* %q, align 128
+  ret void
+}
+
 define void @insert_store_nonconst_index_known_valid_by_assume(<16 x i8>* %q, i8 zeroext %s, i32 %idx) {
 ; CHECK-LABEL: @insert_store_nonconst_index_known_valid_by_assume(
 ; CHECK-NEXT:  entry:
Index: llvm/lib/Transforms/Vectorize/VectorCombine.cpp
===================================================================
--- llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -831,8 +831,15 @@
     Builder.Insert(GEP);
     StoreInst *NSI = Builder.CreateStore(NewElement, GEP);
     NSI->copyMetadata(*SI);
-    if (SI->getAlign() < NSI->getAlign())
-      NSI->setAlignment(SI->getAlign());
+    Align NewAlignment = std::max(SI->getAlign(), Load->getAlign());
+    if (auto *C = dyn_cast<ConstantInt>(Idx))
+      NewAlignment = commonAlignment(
+          NewAlignment,
+          C->getZExtValue() * DL.getTypeStoreSize(NewElement->getType()));
+    else
+      NewAlignment = commonAlignment(
+          NewAlignment, DL.getTypeStoreSize(NewElement->getType()));
+    NSI->setAlignment(NewAlignment);
     replaceValue(I, *NSI);
     // Need erasing the store manually.
     I.eraseFromParent();


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D103419.350760.patch
Type: text/x-patch
Size: 3812 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210609/ebeeac8b/attachment.bin>