Skip to content

Commit f3d2609

Browse files
alexey-bataevalexey-bataev
committed
[SLP]Improve/fix subvectors in gather/buildvector nodes handling
SLP vectorizer has an estimation for gather/buildvector nodes, which contain some scalar loads. SLP vectorizer performs pretty similar (but large in SLOCs) estimation, which not always correct. Instead, this patch implements clustering analysis and actual node allocation with the full analysis for the vectorized clustered scalars (not only loads, but also some other instructions) with the correct cost estimation and vector insert instructions. Improves overall vectorization quality and simplifies analysis/estimations. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: #104144
1 parent f142f8a commit f3d2609

27 files changed

+718
-785
lines changed

‎llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

Lines changed: 164 additions & 183 deletions
Large diffs are not rendered by default.

‎llvm/test/Transforms/PhaseOrdering/AArch64/slpordering.ll

Lines changed: 37 additions & 37 deletions
Original file line numberDiff line numberDiff line change
@@ -18,62 +18,62 @@ define i32 @slpordering(ptr noundef %p1, i32 noundef %ip1, ptr noundef %p2, i32
1818
; CHECK-NEXT: [[IDX_EXT63:%.*]] = sext i32 [[IP2]] to i64
1919
; CHECK-NEXT: [[RRRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 4
2020
; CHECK-NEXT: [[RRRAYIDX5:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 4
21-
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i8>, ptr [[P1]], align 1, !tbaa [[TBAA0:![0-9]+]]
22-
; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i8>, ptr [[P2]], align 1, !tbaa [[TBAA0]]
23-
; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3]], align 1, !tbaa [[TBAA0]]
24-
; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5]], align 1, !tbaa [[TBAA0]]
2521
; CHECK-NEXT: [[RDD_PTR:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[IDX_EXT]]
2622
; CHECK-NEXT: [[RDD_PTR64:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[IDX_EXT63]]
2723
; CHECK-NEXT: [[RRRAYIDX3_1:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR]], i64 4
2824
; CHECK-NEXT: [[RRRAYIDX5_1:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR64]], i64 4
29-
; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i8>, ptr [[RDD_PTR]], align 1, !tbaa [[TBAA0]]
30-
; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i8>, ptr [[RDD_PTR64]], align 1, !tbaa [[TBAA0]]
31-
; CHECK-NEXT: [[TMP6:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_1]], align 1, !tbaa [[TBAA0]]
32-
; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_1]], align 1, !tbaa [[TBAA0]]
3325
; CHECK-NEXT: [[RDD_PTR_1:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR]], i64 [[IDX_EXT]]
3426
; CHECK-NEXT: [[RDD_PTR64_1:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR64]], i64 [[IDX_EXT63]]
3527
; CHECK-NEXT: [[RRRAYIDX3_2:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR_1]], i64 4
3628
; CHECK-NEXT: [[RRRAYIDX5_2:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR64_1]], i64 4
37-
; CHECK-NEXT: [[TMP8:%.*]] = load <4 x i8>, ptr [[RDD_PTR_1]], align 1, !tbaa [[TBAA0]]
38-
; CHECK-NEXT: [[TMP9:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_1]], align 1, !tbaa [[TBAA0]]
39-
; CHECK-NEXT: [[TMP10:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_2]], align 1, !tbaa [[TBAA0]]
40-
; CHECK-NEXT: [[TMP11:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_2]], align 1, !tbaa [[TBAA0]]
4129
; CHECK-NEXT: [[RDD_PTR_2:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR_1]], i64 [[IDX_EXT]]
4230
; CHECK-NEXT: [[RDD_PTR64_2:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR64_1]], i64 [[IDX_EXT63]]
4331
; CHECK-NEXT: [[RRRAYIDX3_3:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR_2]], i64 4
4432
; CHECK-NEXT: [[RRRAYIDX5_3:%.*]] = getelementptr inbounds i8, ptr [[RDD_PTR64_2]], i64 4
33+
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i8>, ptr [[P1]], align 1, !tbaa [[TBAA0:![0-9]+]]
34+
; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i8>, ptr [[P2]], align 1, !tbaa [[TBAA0]]
35+
; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3]], align 1, !tbaa [[TBAA0]]
36+
; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5]], align 1, !tbaa [[TBAA0]]
37+
; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i8>, ptr [[RDD_PTR]], align 1, !tbaa [[TBAA0]]
38+
; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i8>, ptr [[RDD_PTR64]], align 1, !tbaa [[TBAA0]]
39+
; CHECK-NEXT: [[TMP6:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_1]], align 1, !tbaa [[TBAA0]]
40+
; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_1]], align 1, !tbaa [[TBAA0]]
41+
; CHECK-NEXT: [[TMP8:%.*]] = load <4 x i8>, ptr [[RDD_PTR_1]], align 1, !tbaa [[TBAA0]]
42+
; CHECK-NEXT: [[TMP9:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_1]], align 1, !tbaa [[TBAA0]]
43+
; CHECK-NEXT: [[TMP10:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_2]], align 1, !tbaa [[TBAA0]]
44+
; CHECK-NEXT: [[TMP11:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_2]], align 1, !tbaa [[TBAA0]]
4545
; CHECK-NEXT: [[TMP12:%.*]] = load <4 x i8>, ptr [[RDD_PTR_2]], align 1, !tbaa [[TBAA0]]
46-
; CHECK-NEXT: [[TMP13:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_2]], align 1, !tbaa [[TBAA0]]
47-
; CHECK-NEXT: [[TMP14:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_3]], align 1, !tbaa [[TBAA0]]
48-
; CHECK-NEXT: [[TMP15:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_3]], align 1, !tbaa [[TBAA0]]
49-
; CHECK-NEXT: [[TMP16:%.*]] = 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>
50-
; CHECK-NEXT: [[TMP17:%.*]] = 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>
51-
; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <16 x i8> [[TMP16]], <16 x i8> [[TMP17]], <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>
52-
; CHECK-NEXT: [[TMP19:%.*]] = 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>
53-
; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <16 x i8> [[TMP18]], <16 x i8> [[TMP19]], <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>
54-
; CHECK-NEXT: [[TMP21:%.*]] = zext <16 x i8> [[TMP20]] to <16 x i32>
55-
; CHECK-NEXT: [[TMP22:%.*]] = 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>
56-
; CHECK-NEXT: [[TMP23:%.*]] = 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>
57-
; 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 16, i32 17, i32 18, i32 19, i32 poison, i32 poison, i32 poison, i32 poison>
58-
; CHECK-NEXT: [[TMP25:%.*]] = shufflevector <4 x i8> [[TMP13]], <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>
59-
; CHECK-NEXT: [[TMP26:%.*]] = shufflevector <16 x i8> [[TMP24]], <16 x i8> [[TMP25]], <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>
60-
; CHECK-NEXT: [[TMP27:%.*]] = zext <16 x i8> [[TMP26]] to <16 x i32>
61-
; CHECK-NEXT: [[TMP28:%.*]] = sub nsw <16 x i32> [[TMP21]], [[TMP27]]
62-
; CHECK-NEXT: [[TMP29:%.*]] = 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>
63-
; CHECK-NEXT: [[TMP30:%.*]] = 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>
64-
; CHECK-NEXT: [[TMP31:%.*]] = shufflevector <16 x i8> [[TMP29]], <16 x i8> [[TMP30]], <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>
65-
; CHECK-NEXT: [[TMP32:%.*]] = shufflevector <4 x i8> [[TMP14]], <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>
66-
; CHECK-NEXT: [[TMP33:%.*]] = shufflevector <16 x i8> [[TMP31]], <16 x i8> [[TMP32]], <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>
67-
; CHECK-NEXT: [[TMP34:%.*]] = zext <16 x i8> [[TMP33]] to <16 x i32>
46+
; 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>
47+
; 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>
48+
; 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>
49+
; 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>
50+
; 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>
51+
; CHECK-NEXT: [[TMP18:%.*]] = zext <16 x i8> [[TMP17]] to <16 x i32>
52+
; CHECK-NEXT: [[TMP19:%.*]] = load <4 x i8>, ptr [[RDD_PTR64_2]], align 1, !tbaa [[TBAA0]]
53+
; 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>
54+
; 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>
55+
; 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>
56+
; 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>
57+
; 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>
58+
; CHECK-NEXT: [[TMP25:%.*]] = zext <16 x i8> [[TMP24]] to <16 x i32>
59+
; CHECK-NEXT: [[TMP26:%.*]] = sub nsw <16 x i32> [[TMP18]], [[TMP25]]
60+
; CHECK-NEXT: [[TMP27:%.*]] = load <4 x i8>, ptr [[RRRAYIDX3_3]], align 1, !tbaa [[TBAA0]]
61+
; 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>
62+
; 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>
63+
; 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>
64+
; 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>
65+
; 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>
66+
; CHECK-NEXT: [[TMP33:%.*]] = zext <16 x i8> [[TMP32]] to <16 x i32>
67+
; CHECK-NEXT: [[TMP34:%.*]] = load <4 x i8>, ptr [[RRRAYIDX5_3]], align 1, !tbaa [[TBAA0]]
6868
; 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>
6969
; 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>
7070
; 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>
71-
; CHECK-NEXT: [[TMP38:%.*]] = shufflevector <4 x i8> [[TMP15]], <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>
71+
; 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>
7272
; 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>
7373
; CHECK-NEXT: [[TMP40:%.*]] = zext <16 x i8> [[TMP39]] to <16 x i32>
74-
; CHECK-NEXT: [[TMP41:%.*]] = sub nsw <16 x i32> [[TMP34]], [[TMP40]]
74+
; CHECK-NEXT: [[TMP41:%.*]] = sub nsw <16 x i32> [[TMP33]], [[TMP40]]
7575
; CHECK-NEXT: [[TMP42:%.*]] = shl nsw <16 x i32> [[TMP41]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
76-
; CHECK-NEXT: [[TMP43:%.*]] = add nsw <16 x i32> [[TMP42]], [[TMP28]]
76+
; CHECK-NEXT: [[TMP43:%.*]] = add nsw <16 x i32> [[TMP42]], [[TMP26]]
7777
; 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>
7878
; 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>
7979
; 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>

‎llvm/test/Transforms/SLPVectorizer/AArch64/getelementptr.ll

Lines changed: 6 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -169,11 +169,12 @@ define i32 @getelementptr_2x32(ptr nocapture readonly %g, i32 %n, i32 %x, i32 %y
169169
; CHECK-NEXT: [[TMP9:%.*]] = sext i32 [[T11]] to i64
170170
; CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds i32, ptr [[G]], i64 [[TMP9]]
171171
; CHECK-NEXT: [[T12:%.*]] = load i32, ptr [[ARRAYIDX15]], align 4
172-
; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i32> [[TMP6]], <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
173-
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[T10]], i64 2
174-
; CHECK-NEXT: [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[T12]], i64 3
175-
; CHECK-NEXT: [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP12]])
176-
; CHECK-NEXT: [[OP_RDX]] = add i32 [[TMP13]], [[SUM_032]]
172+
; CHECK-NEXT: [[TMP10:%.*]] = insertelement <4 x i32> poison, i32 [[T10]], i64 2
173+
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[T12]], i64 3
174+
; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x i32> [[TMP6]], <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
175+
; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <4 x i32> [[TMP12]], <4 x i32> [[TMP11]], <4 x i32> <i32 0, i32 1, i32 6, i32 7>
176+
; CHECK-NEXT: [[TMP14:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP13]])
177+
; CHECK-NEXT: [[OP_RDX]] = add i32 [[TMP14]], [[SUM_032]]
177178
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
178179
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[N]]
179180
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]]

0 commit comments

Comments
 (0)