[llvm] [SLP]Reduce number of alternate instruction, where possible (PR #123360)
David Green via llvm-commits
llvm-commits at lists.llvm.org
Fri Jan 31 06:51:12 PST 2025
================
@@ -31,82 +31,108 @@ define i32 @slpordering(ptr noundef %p1, i32 noundef %ip1, ptr noundef %p2, i32
; CHECK-NEXT: [[RRRAYIDX3_3:%.*]] = getelementptr inbounds nuw i8, ptr [[RDD_PTR_2]], i64 4
; CHECK-NEXT: [[RRRAYIDX5_3:%.*]] = getelementptr inbounds nuw i8, ptr [[RDD_PTR64_2]], i64 4
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i8>, ptr [[P1]], align 1, !tbaa [[TBAA0:![0-9]+]]
-; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i8>, ptr [[P2]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i8>, ptr [[RDD_PTR]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i8>, ptr [[RDD_PTR64]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP6:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_1]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_1]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP8:%.*]] = load <4 x i8>, ptr [[RDD_PTR_1]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP9:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_1]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP10:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_2]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP11:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_2]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP12:%.*]] = load <4 x i8>, ptr [[RDD_PTR_2]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <4 x i8> [[TMP0]], <4 x i8> [[TMP4]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP14:%.*]] = shufflevector <4 x i8> [[TMP8]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP15:%.*]] = shufflevector <16 x i8> [[TMP13]], <16 x i8> [[TMP14]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP16:%.*]] = shufflevector <4 x i8> [[TMP12]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP17:%.*]] = shufflevector <16 x i8> [[TMP15]], <16 x i8> [[TMP16]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT: [[TMP18:%.*]] = zext <16 x i8> [[TMP17]] to <16 x i32>
-; CHECK-NEXT: [[TMP19:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_2]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <4 x i8> [[TMP1]], <4 x i8> [[TMP5]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <4 x i8> [[TMP9]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP22:%.*]] = shufflevector <16 x i8> [[TMP20]], <16 x i8> [[TMP21]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP23:%.*]] = shufflevector <4 x i8> [[TMP19]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP24:%.*]] = shufflevector <16 x i8> [[TMP22]], <16 x i8> [[TMP23]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT: [[TMP25:%.*]] = zext <16 x i8> [[TMP24]] to <16 x i32>
-; CHECK-NEXT: [[TMP26:%.*]] = sub nsw <16 x i32> [[TMP18]], [[TMP25]]
-; CHECK-NEXT: [[TMP27:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_3]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP28:%.*]] = shufflevector <4 x i8> [[TMP2]], <4 x i8> [[TMP6]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP29:%.*]] = shufflevector <4 x i8> [[TMP10]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP30:%.*]] = shufflevector <16 x i8> [[TMP28]], <16 x i8> [[TMP29]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP31:%.*]] = shufflevector <4 x i8> [[TMP27]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP32:%.*]] = shufflevector <16 x i8> [[TMP30]], <16 x i8> [[TMP31]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT: [[TMP33:%.*]] = zext <16 x i8> [[TMP32]] to <16 x i32>
-; CHECK-NEXT: [[TMP34:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_3]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT: [[TMP35:%.*]] = shufflevector <4 x i8> [[TMP3]], <4 x i8> [[TMP7]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP36:%.*]] = shufflevector <4 x i8> [[TMP11]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP37:%.*]] = shufflevector <16 x i8> [[TMP35]], <16 x i8> [[TMP36]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP38:%.*]] = shufflevector <4 x i8> [[TMP34]], <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP39:%.*]] = shufflevector <16 x i8> [[TMP37]], <16 x i8> [[TMP38]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT: [[TMP40:%.*]] = zext <16 x i8> [[TMP39]] to <16 x i32>
-; CHECK-NEXT: [[TMP41:%.*]] = sub nsw <16 x i32> [[TMP33]], [[TMP40]]
-; CHECK-NEXT: [[TMP42:%.*]] = shl nsw <16 x i32> [[TMP41]], splat (i32 16)
-; CHECK-NEXT: [[TMP43:%.*]] = add nsw <16 x i32> [[TMP42]], [[TMP26]]
-; CHECK-NEXT: [[TMP44:%.*]] = shufflevector <16 x i32> [[TMP43]], <16 x i32> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP45:%.*]] = shufflevector <16 x i32> [[TMP43]], <16 x i32> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP46:%.*]] = shufflevector <16 x i32> [[TMP43]], <16 x i32> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP47:%.*]] = shufflevector <16 x i32> [[TMP43]], <16 x i32> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP48:%.*]] = add nsw <16 x i32> [[TMP45]], [[TMP47]]
-; CHECK-NEXT: [[TMP49:%.*]] = sub nsw <16 x i32> [[TMP44]], [[TMP46]]
-; CHECK-NEXT: [[TMP50:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP51:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP52:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP53:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP54:%.*]] = add nsw <16 x i32> [[TMP51]], [[TMP53]]
-; CHECK-NEXT: [[TMP55:%.*]] = sub nsw <16 x i32> [[TMP50]], [[TMP52]]
-; CHECK-NEXT: [[TMP56:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP57:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP58:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP59:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP60:%.*]] = sub nsw <16 x i32> [[TMP57]], [[TMP59]]
-; CHECK-NEXT: [[TMP61:%.*]] = add nsw <16 x i32> [[TMP56]], [[TMP58]]
-; CHECK-NEXT: [[TMP62:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP63:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP64:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP65:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP66:%.*]] = add nsw <16 x i32> [[TMP63]], [[TMP65]]
-; CHECK-NEXT: [[TMP67:%.*]] = sub nsw <16 x i32> [[TMP62]], [[TMP64]]
-; CHECK-NEXT: [[TMP68:%.*]] = shufflevector <16 x i32> [[TMP66]], <16 x i32> [[TMP67]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; CHECK-NEXT: [[TMP69:%.*]] = lshr <16 x i32> [[TMP68]], splat (i32 15)
-; CHECK-NEXT: [[TMP70:%.*]] = and <16 x i32> [[TMP69]], splat (i32 65537)
-; CHECK-NEXT: [[TMP71:%.*]] = mul nuw <16 x i32> [[TMP70]], splat (i32 65535)
-; CHECK-NEXT: [[TMP72:%.*]] = add <16 x i32> [[TMP71]], [[TMP68]]
-; CHECK-NEXT: [[TMP73:%.*]] = xor <16 x i32> [[TMP72]], [[TMP71]]
-; CHECK-NEXT: [[TMP74:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP73]])
-; CHECK-NEXT: [[CONV118:%.*]] = and i32 [[TMP74]], 65535
-; CHECK-NEXT: [[SHR:%.*]] = lshr i32 [[TMP74]], 16
+; CHECK-NEXT: [[TMP1:%.*]] = zext <4 x i8> [[TMP0]] to <4 x i32>
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i8>, ptr [[P2]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP3:%.*]] = zext <4 x i8> [[TMP2]] to <4 x i32>
+; CHECK-NEXT: [[TMP4:%.*]] = sub nsw <4 x i32> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP6:%.*]] = zext <4 x i8> [[TMP5]] to <4 x i32>
+; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP8:%.*]] = zext <4 x i8> [[TMP7]] to <4 x i32>
+; CHECK-NEXT: [[TMP9:%.*]] = sub nsw <4 x i32> [[TMP6]], [[TMP8]]
+; CHECK-NEXT: [[TMP10:%.*]] = shl nsw <4 x i32> [[TMP9]], splat (i32 16)
+; CHECK-NEXT: [[TMP11:%.*]] = add nsw <4 x i32> [[TMP10]], [[TMP4]]
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <4 x i32> [[TMP11]], <4 x i32> poison, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
+; CHECK-NEXT: [[TMP13:%.*]] = add nsw <4 x i32> [[TMP11]], [[TMP12]]
+; CHECK-NEXT: [[TMP14:%.*]] = sub nsw <4 x i32> [[TMP11]], [[TMP12]]
+; CHECK-NEXT: [[TMP15:%.*]] = shufflevector <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], <4 x i32> <i32 3, i32 6, i32 1, i32 4>
+; CHECK-NEXT: [[TMP16:%.*]] = shufflevector <4 x i32> [[TMP15]], <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT: [[TMP17:%.*]] = add nsw <4 x i32> [[TMP15]], [[TMP16]]
+; CHECK-NEXT: [[TMP18:%.*]] = sub nsw <4 x i32> [[TMP15]], [[TMP16]]
+; CHECK-NEXT: [[TMP19:%.*]] = shufflevector <4 x i32> [[TMP17]], <4 x i32> [[TMP18]], <4 x i32> <i32 0, i32 1, i32 6, i32 7>
+; CHECK-NEXT: [[TMP20:%.*]] = load <4 x i8>, ptr [[RDD_PTR]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP21:%.*]] = zext <4 x i8> [[TMP20]] to <4 x i32>
+; CHECK-NEXT: [[TMP22:%.*]] = load <4 x i8>, ptr [[RDD_PTR64]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP23:%.*]] = zext <4 x i8> [[TMP22]] to <4 x i32>
+; CHECK-NEXT: [[TMP24:%.*]] = sub nsw <4 x i32> [[TMP21]], [[TMP23]]
+; CHECK-NEXT: [[TMP25:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_1]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP26:%.*]] = zext <4 x i8> [[TMP25]] to <4 x i32>
+; CHECK-NEXT: [[TMP27:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_1]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP28:%.*]] = zext <4 x i8> [[TMP27]] to <4 x i32>
+; CHECK-NEXT: [[TMP29:%.*]] = sub nsw <4 x i32> [[TMP26]], [[TMP28]]
+; CHECK-NEXT: [[TMP30:%.*]] = shl nsw <4 x i32> [[TMP29]], splat (i32 16)
+; CHECK-NEXT: [[TMP31:%.*]] = add nsw <4 x i32> [[TMP30]], [[TMP24]]
+; CHECK-NEXT: [[TMP32:%.*]] = shufflevector <4 x i32> [[TMP31]], <4 x i32> poison, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
+; CHECK-NEXT: [[TMP33:%.*]] = add nsw <4 x i32> [[TMP31]], [[TMP32]]
+; CHECK-NEXT: [[TMP34:%.*]] = sub nsw <4 x i32> [[TMP31]], [[TMP32]]
+; CHECK-NEXT: [[TMP35:%.*]] = shufflevector <4 x i32> [[TMP33]], <4 x i32> [[TMP34]], <4 x i32> <i32 3, i32 6, i32 1, i32 4>
+; CHECK-NEXT: [[TMP36:%.*]] = shufflevector <4 x i32> [[TMP35]], <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT: [[TMP37:%.*]] = add nsw <4 x i32> [[TMP35]], [[TMP36]]
+; CHECK-NEXT: [[TMP38:%.*]] = sub nsw <4 x i32> [[TMP35]], [[TMP36]]
+; CHECK-NEXT: [[TMP39:%.*]] = shufflevector <4 x i32> [[TMP37]], <4 x i32> [[TMP38]], <4 x i32> <i32 0, i32 1, i32 6, i32 7>
+; CHECK-NEXT: [[TMP40:%.*]] = load <4 x i8>, ptr [[RDD_PTR_1]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP41:%.*]] = zext <4 x i8> [[TMP40]] to <4 x i32>
+; CHECK-NEXT: [[TMP42:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_1]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP43:%.*]] = zext <4 x i8> [[TMP42]] to <4 x i32>
+; CHECK-NEXT: [[TMP44:%.*]] = sub nsw <4 x i32> [[TMP41]], [[TMP43]]
+; CHECK-NEXT: [[TMP45:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_2]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP46:%.*]] = zext <4 x i8> [[TMP45]] to <4 x i32>
+; CHECK-NEXT: [[TMP47:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_2]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP48:%.*]] = zext <4 x i8> [[TMP47]] to <4 x i32>
+; CHECK-NEXT: [[TMP49:%.*]] = sub nsw <4 x i32> [[TMP46]], [[TMP48]]
+; CHECK-NEXT: [[TMP50:%.*]] = shl nsw <4 x i32> [[TMP49]], splat (i32 16)
+; CHECK-NEXT: [[TMP51:%.*]] = add nsw <4 x i32> [[TMP50]], [[TMP44]]
+; CHECK-NEXT: [[TMP52:%.*]] = shufflevector <4 x i32> [[TMP51]], <4 x i32> poison, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
+; CHECK-NEXT: [[TMP53:%.*]] = add nsw <4 x i32> [[TMP51]], [[TMP52]]
+; CHECK-NEXT: [[TMP54:%.*]] = sub nsw <4 x i32> [[TMP51]], [[TMP52]]
+; CHECK-NEXT: [[TMP55:%.*]] = shufflevector <4 x i32> [[TMP53]], <4 x i32> [[TMP54]], <4 x i32> <i32 3, i32 6, i32 1, i32 4>
+; CHECK-NEXT: [[TMP56:%.*]] = shufflevector <4 x i32> [[TMP55]], <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT: [[TMP57:%.*]] = add nsw <4 x i32> [[TMP55]], [[TMP56]]
+; CHECK-NEXT: [[TMP58:%.*]] = sub nsw <4 x i32> [[TMP55]], [[TMP56]]
+; CHECK-NEXT: [[TMP59:%.*]] = shufflevector <4 x i32> [[TMP57]], <4 x i32> [[TMP58]], <4 x i32> <i32 0, i32 1, i32 6, i32 7>
+; CHECK-NEXT: [[TMP60:%.*]] = load <4 x i8>, ptr [[RDD_PTR_2]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP61:%.*]] = zext <4 x i8> [[TMP60]] to <4 x i32>
+; CHECK-NEXT: [[TMP62:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_2]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP63:%.*]] = zext <4 x i8> [[TMP62]] to <4 x i32>
+; CHECK-NEXT: [[TMP64:%.*]] = sub nsw <4 x i32> [[TMP61]], [[TMP63]]
+; CHECK-NEXT: [[TMP65:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_3]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP66:%.*]] = zext <4 x i8> [[TMP65]] to <4 x i32>
+; CHECK-NEXT: [[TMP67:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_3]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT: [[TMP68:%.*]] = zext <4 x i8> [[TMP67]] to <4 x i32>
+; CHECK-NEXT: [[TMP69:%.*]] = sub nsw <4 x i32> [[TMP66]], [[TMP68]]
+; CHECK-NEXT: [[TMP70:%.*]] = shl nsw <4 x i32> [[TMP69]], splat (i32 16)
+; CHECK-NEXT: [[TMP71:%.*]] = add nsw <4 x i32> [[TMP70]], [[TMP64]]
+; CHECK-NEXT: [[TMP72:%.*]] = shufflevector <4 x i32> [[TMP71]], <4 x i32> poison, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
+; CHECK-NEXT: [[TMP73:%.*]] = add nsw <4 x i32> [[TMP71]], [[TMP72]]
+; CHECK-NEXT: [[TMP74:%.*]] = sub nsw <4 x i32> [[TMP71]], [[TMP72]]
+; CHECK-NEXT: [[TMP75:%.*]] = shufflevector <4 x i32> [[TMP73]], <4 x i32> [[TMP74]], <4 x i32> <i32 3, i32 6, i32 1, i32 4>
+; CHECK-NEXT: [[TMP76:%.*]] = shufflevector <4 x i32> [[TMP75]], <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT: [[TMP77:%.*]] = add nsw <4 x i32> [[TMP75]], [[TMP76]]
+; CHECK-NEXT: [[TMP78:%.*]] = sub nsw <4 x i32> [[TMP75]], [[TMP76]]
+; CHECK-NEXT: [[TMP79:%.*]] = shufflevector <4 x i32> [[TMP77]], <4 x i32> [[TMP78]], <4 x i32> <i32 0, i32 1, i32 6, i32 7>
+; CHECK-NEXT: [[TMP80:%.*]] = add nsw <4 x i32> [[TMP39]], [[TMP19]]
+; CHECK-NEXT: [[TMP81:%.*]] = sub nsw <4 x i32> [[TMP19]], [[TMP39]]
+; CHECK-NEXT: [[TMP82:%.*]] = add nsw <4 x i32> [[TMP79]], [[TMP59]]
+; CHECK-NEXT: [[TMP83:%.*]] = sub nsw <4 x i32> [[TMP59]], [[TMP79]]
+; CHECK-NEXT: [[TMP84:%.*]] = add nsw <4 x i32> [[TMP83]], [[TMP81]]
+; CHECK-NEXT: [[TMP85:%.*]] = add nsw <4 x i32> [[TMP82]], [[TMP80]]
+; CHECK-NEXT: [[TMP86:%.*]] = shufflevector <4 x i32> [[TMP84]], <4 x i32> [[TMP85]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT: [[TMP87:%.*]] = sub nsw <4 x i32> [[TMP80]], [[TMP82]]
+; CHECK-NEXT: [[TMP88:%.*]] = sub nsw <4 x i32> [[TMP81]], [[TMP83]]
+; CHECK-NEXT: [[TMP89:%.*]] = shufflevector <4 x i32> [[TMP87]], <4 x i32> [[TMP88]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT: [[TMP90:%.*]] = shufflevector <4 x i32> [[TMP84]], <4 x i32> [[TMP85]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP91:%.*]] = shufflevector <4 x i32> [[TMP87]], <4 x i32> [[TMP88]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP92:%.*]] = shufflevector <16 x i32> [[TMP90]], <16 x i32> [[TMP91]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; CHECK-NEXT: [[TMP93:%.*]] = lshr <16 x i32> [[TMP92]], splat (i32 15)
+; CHECK-NEXT: [[TMP94:%.*]] = and <16 x i32> [[TMP93]], splat (i32 65537)
+; CHECK-NEXT: [[TMP95:%.*]] = mul nuw <16 x i32> [[TMP94]], splat (i32 65535)
+; CHECK-NEXT: [[TMP96:%.*]] = shufflevector <16 x i32> [[TMP95]], <16 x i32> poison, <16 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7, i32 8, i32 12, i32 9, i32 13, i32 10, i32 14, i32 11, i32 15>
+; CHECK-NEXT: [[TMP97:%.*]] = shufflevector <8 x i32> [[TMP86]], <8 x i32> [[TMP89]], <16 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7, i32 8, i32 12, i32 9, i32 13, i32 10, i32 14, i32 11, i32 15>
+; CHECK-NEXT: [[TMP98:%.*]] = add <16 x i32> [[TMP96]], [[TMP97]]
+; CHECK-NEXT: [[TMP99:%.*]] = xor <16 x i32> [[TMP98]], [[TMP96]]
+; CHECK-NEXT: [[TMP100:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP99]])
+; CHECK-NEXT: [[CONV118:%.*]] = and i32 [[TMP100]], 65535
+; CHECK-NEXT: [[SHR:%.*]] = lshr i32 [[TMP100]], 16
----------------
davemgreen wrote:
This looks quite a lot worse btw. I'm not sure what is going on, but I know the alt instructions, while not useful directly for aarch64, can lead to better vectorization if we can shuffle nicely in the right places.
https://github.com/llvm/llvm-project/pull/123360
More information about the llvm-commits
mailing list