[llvm-branch-commits] [llvm] [CGP] Simple loop strength reduction for vector values (PR #226197)

[Graham Hunter via llvm-branch-commits]

================
@@ -13,17 +13,28 @@ define void @init_array_of_ptrs_to_structs(ptr noalias %arc_ptrs, ptr %arc_new,
 ; CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[STEP]], 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:    [[TMP7:%.*]] = mul <vscale x 2 x i64> [[TMP0]], splat (i64 72)
+; CHECK-NEXT:    [[TMP8:%.*]] = ptrtoint ptr [[ARC_NEW]] to i64
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP8]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP9:%.*]] = add <vscale x 2 x i64> [[TMP7]], [[DOTSPLAT]]
+; CHECK-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 1, [[TMP11]]
+; CHECK-NEXT:    [[DOTSPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP5]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP0]], %[[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP9]], %[[ENTRY]] ], [ [[TMP10:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr <vscale x 2 x i64> [[VEC_IND]] to <vscale x 2 x ptr>
----------------
huntergr-arm wrote:

LSR only deals with SCEVable types, so won't touch this. I think CGP runs after LSR anyway.

I've converted it to use a vector of ptrs anyway, as Sander suggested.

https://github.com/llvm/llvm-project/pull/226197