[llvm] [LAA] Detect cross-iteration WAW when writing to the same pointer (PR #187802)

[Igor Kirillov via llvm-commits]

================
@@ -162,35 +162,19 @@ exit:
 define void @store_to_addr_generated_from_invariant_addr(ptr noalias %p0, ptr noalias %p1, ptr noalias %p2, ptr %p3, i64 %N) {
 ; CHECK-LABEL: @store_to_addr_generated_from_invariant_addr(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], 1
-; CHECK-NEXT:    br label [[VECTOR_PH:%.*]]
-; CHECK:       vector.ph:
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 2 x ptr> poison, ptr [[P0:%.*]], i64 0
-; CHECK-NEXT:    [[BROADCAST_SPLAT1:%.*]] = shufflevector <vscale x 2 x ptr> [[BROADCAST_SPLATINSERT2]], <vscale x 2 x ptr> poison, <vscale x 2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP1:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
-; CHECK:       vector.body:
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP1]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[AVL:%.*]] = phi i64 [ [[TMP0]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
-; CHECK-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP4]], i64 0
-; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[P1:%.*]], <vscale x 2 x i64> [[VEC_IND]]
-; CHECK-NEXT:    call void @llvm.vp.scatter.nxv2p0.nxv2p0(<vscale x 2 x ptr> [[BROADCAST_SPLAT1]], <vscale x 2 x ptr> align 8 [[TMP5]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP3]])
+; CHECK:       loop:
----------------
igogo-x86 wrote:

Looks like a potential performance regression, not sure if it is possible to fix though.

https://github.com/llvm/llvm-project/pull/187802