Skip to content

Commit 5dfc2d4

Browse files
artagnonartagnon
authored
[LV] Regen some tests with UTC (#152128)
1 parent e3467d8 commit 5dfc2d4

File tree

7 files changed

+504
-222
lines changed

7 files changed

+504
-222
lines changed

‎llvm/test/Transforms/LoopVectorize/AArch64/extractvalue-no-scalarization-required.ll

Lines changed: 49 additions & 34 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,4 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph" --version 5
12
; REQUIRES: asserts
23

34
; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-ios %s -S -debug -disable-output 2>&1 | FileCheck --check-prefix=CM %s
@@ -22,23 +23,31 @@
2223

2324
; Check that the extractvalue operands are actually free in vector code.
2425

25-
; FORCED: [[E1:%.+]] = extractvalue { i64, i64 } %sv, 0
26-
; FORCED-NEXT: %broadcast.splatinsert = insertelement <2 x i64> poison, i64 [[E1]], i64 0
27-
; FORCED-NEXT: %broadcast.splat = shufflevector <2 x i64> %broadcast.splatinsert, <2 x i64> poison, <2 x i32> zeroinitializer
28-
; FORCED-NEXT: [[E2:%.+]] = extractvalue { i64, i64 } %sv, 1
29-
; FORCED-NEXT: %broadcast.splatinsert1 = insertelement <2 x i64> poison, i64 [[E2]], i64 0
30-
; FORCED-NEXT: %broadcast.splat2 = shufflevector <2 x i64> %broadcast.splatinsert1, <2 x i64> poison, <2 x i32> zeroinitializer
31-
; FORCED-NEXT: [[ADD:%.+]] = add <2 x i64> %broadcast.splat, %broadcast.splat2
32-
33-
; FORCED-LABEL: vector.body: ; preds = %vector.body, %vector.ph
34-
; FORCED-NEXT: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
35-
; FORCED-NEXT: [[GEP:%.+]] = getelementptr i64, ptr %dst, i32 %index
36-
; FORCED-NEXT: store <2 x i64> [[ADD]], ptr [[GEP]], align 4
37-
; FORCED-NEXT: %index.next = add nuw i32 %index, 2
38-
; FORCED-NEXT: [[C:%.+]] = icmp eq i32 %index.next, 1000
39-
; FORCED-NEXT: br i1 [[C]], label %middle.block, label %vector.body
40-
4126
define void @test1(ptr %dst, {i64, i64} %sv) {
27+
; FORCED-LABEL: define void @test1(
28+
; FORCED-SAME: ptr [[DST:%.*]], { i64, i64 } [[SV:%.*]]) {
29+
; FORCED-NEXT: [[ENTRY:.*:]]
30+
; FORCED-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
31+
; FORCED: [[VECTOR_PH]]:
32+
; FORCED-NEXT: [[TMP0:%.*]] = extractvalue { i64, i64 } [[SV]], 0
33+
; FORCED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP0]], i64 0
34+
; FORCED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
35+
; FORCED-NEXT: [[TMP4:%.*]] = extractvalue { i64, i64 } [[SV]], 1
36+
; FORCED-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x i64> poison, i64 [[TMP4]], i64 0
37+
; FORCED-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT1]], <2 x i64> poison, <2 x i32> zeroinitializer
38+
; FORCED-NEXT: [[TMP1:%.*]] = add <2 x i64> [[BROADCAST_SPLAT]], [[BROADCAST_SPLAT2]]
39+
; FORCED-NEXT: br label %[[VECTOR_BODY:.*]]
40+
; FORCED: [[VECTOR_BODY]]:
41+
; FORCED-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
42+
; FORCED-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[DST]], i32 [[INDEX]]
43+
; FORCED-NEXT: store <2 x i64> [[TMP1]], ptr [[TMP2]], align 4
44+
; FORCED-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
45+
; FORCED-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
46+
; FORCED-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
47+
; FORCED: [[MIDDLE_BLOCK]]:
48+
; FORCED-NEXT: br [[EXIT:label %.*]]
49+
; FORCED: [[SCALAR_PH]]:
50+
;
4251
entry:
4352
br label %loop.body
4453

@@ -70,25 +79,31 @@ declare float @powf(float, float) readnone nounwind
7079

7180
; CM: LV: Scalar loop costs: 14.
7281

73-
; FORCED-LABEL: define void @test_getVectorCallCost
74-
75-
; FORCED: [[E1:%.+]] = extractvalue { float, float } %sv, 0
76-
; FORCED-NEXT: %broadcast.splatinsert = insertelement <2 x float> poison, float [[E1]], i64 0
77-
; FORCED-NEXT: %broadcast.splat = shufflevector <2 x float> %broadcast.splatinsert, <2 x float> poison, <2 x i32> zeroinitializer
78-
; FORCED-NEXT: [[E2:%.+]] = extractvalue { float, float } %sv, 1
79-
; FORCED-NEXT: %broadcast.splatinsert1 = insertelement <2 x float> poison, float [[E2]], i64 0
80-
; FORCED-NEXT: %broadcast.splat2 = shufflevector <2 x float> %broadcast.splatinsert1, <2 x float> poison, <2 x i32> zeroinitializer
81-
82-
; FORCED-LABEL: vector.body: ; preds = %vector.body, %vector.ph
83-
; FORCED-NEXT: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
84-
; FORCED-NEXT: [[GEP1:%.+]] = getelementptr float, ptr %dst, i32 %index
85-
; FORCED-NEXT: [[POW:%.+]] = call <2 x float> @llvm.pow.v2f32(<2 x float> %broadcast.splat, <2 x float> %broadcast.splat2)
86-
; FORCED-NEXT: store <2 x float> [[POW]], ptr [[GEP1]], align 4
87-
; FORCED-NEXT: %index.next = add nuw i32 %index, 2
88-
; FORCED-NEXT: [[C:%.+]] = icmp eq i32 %index.next, 1000
89-
; FORCED-NEXT: br i1 [[C]], label %middle.block, label %vector.body
90-
9182
define void @test_getVectorCallCost(ptr %dst, {float, float} %sv) {
83+
; FORCED-LABEL: define void @test_getVectorCallCost(
84+
; FORCED-SAME: ptr [[DST:%.*]], { float, float } [[SV:%.*]]) {
85+
; FORCED-NEXT: [[ENTRY:.*:]]
86+
; FORCED-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
87+
; FORCED: [[VECTOR_PH]]:
88+
; FORCED-NEXT: [[TMP0:%.*]] = extractvalue { float, float } [[SV]], 0
89+
; FORCED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i64 0
90+
; FORCED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
91+
; FORCED-NEXT: [[TMP4:%.*]] = extractvalue { float, float } [[SV]], 1
92+
; FORCED-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x float> poison, float [[TMP4]], i64 0
93+
; FORCED-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT1]], <2 x float> poison, <2 x i32> zeroinitializer
94+
; FORCED-NEXT: br label %[[VECTOR_BODY:.*]]
95+
; FORCED: [[VECTOR_BODY]]:
96+
; FORCED-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
97+
; FORCED-NEXT: [[TMP1:%.*]] = getelementptr float, ptr [[DST]], i32 [[INDEX]]
98+
; FORCED-NEXT: [[TMP2:%.*]] = call <2 x float> @llvm.pow.v2f32(<2 x float> [[BROADCAST_SPLAT]], <2 x float> [[BROADCAST_SPLAT2]])
99+
; FORCED-NEXT: store <2 x float> [[TMP2]], ptr [[TMP1]], align 4
100+
; FORCED-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
101+
; FORCED-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
102+
; FORCED-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
103+
; FORCED: [[MIDDLE_BLOCK]]:
104+
; FORCED-NEXT: br [[EXIT:label %.*]]
105+
; FORCED: [[SCALAR_PH]]:
106+
;
92107
entry:
93108
br label %loop.body
94109

‎llvm/test/Transforms/LoopVectorize/AArch64/outer_loop_test1_no_explicit_vect_width.ll

Lines changed: 67 additions & 58 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,4 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph" --version 5
12
; RUN: opt -S -passes=loop-vectorize -enable-vplan-native-path -mtriple aarch64-gnu-linux < %s | FileCheck %s
23

34
; extern int arr[8][8];
@@ -16,36 +17,6 @@
1617
; }
1718
;
1819

19-
; CHECK-LABEL: @foo_i32(
20-
; CHECK-LABEL: vector.ph:
21-
; CHECK: %[[SplatVal:.*]] = insertelement <4 x i32> poison, i32 %n, i64 0
22-
; CHECK: %[[Splat:.*]] = shufflevector <4 x i32> %[[SplatVal]], <4 x i32> poison, <4 x i32> zeroinitializer
23-
24-
; CHECK-LABEL: vector.body:
25-
; CHECK: %[[Ind:.*]] = phi i64 [ 0, %vector.ph ], [ %[[IndNext:.*]], %[[ForInc:.*]] ]
26-
; CHECK: %[[VecInd:.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %vector.ph ], [ %[[VecIndNext:.*]], %[[ForInc]] ]
27-
; CHECK: %[[AAddr:.*]] = getelementptr inbounds [8 x i32], ptr @arr2, i64 0, <4 x i64> %[[VecInd]]
28-
; CHECK: %[[VecIndTr:.*]] = trunc <4 x i64> %[[VecInd]] to <4 x i32>
29-
; CHECK: call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %[[VecIndTr]], <4 x ptr> %[[AAddr]], i32 4, <4 x i1> splat (i1 true))
30-
; CHECK: %[[VecIndTr2:.*]] = trunc <4 x i64> %[[VecInd]] to <4 x i32>
31-
; CHECK: %[[StoreVal:.*]] = add nsw <4 x i32> %[[VecIndTr2]], %[[Splat]]
32-
; CHECK: br label %[[InnerLoop:.+]]
33-
34-
; CHECK: [[InnerLoop]]:
35-
; CHECK: %[[InnerPhi:.*]] = phi <4 x i64> [ zeroinitializer, %vector.body ], [ %[[InnerPhiNext:.*]], %[[InnerLoop]] ]
36-
; CHECK: %[[AAddr2:.*]] = getelementptr inbounds [8 x [8 x i32]], ptr @arr, i64 0, <4 x i64> %[[InnerPhi]], <4 x i64> %[[VecInd]]
37-
; CHECK: call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %[[StoreVal]], <4 x ptr> %[[AAddr2]], i32 4, <4 x i1> splat (i1 true))
38-
; CHECK: %[[InnerPhiNext]] = add nuw nsw <4 x i64> %[[InnerPhi]], splat (i64 1)
39-
; CHECK: %[[VecCond:.*]] = icmp eq <4 x i64> %[[InnerPhiNext]], splat (i64 8)
40-
; CHECK: %[[InnerCond:.*]] = extractelement <4 x i1> %[[VecCond]], i32 0
41-
; CHECK: br i1 %[[InnerCond]], label %[[ForInc]], label %[[InnerLoop]]
42-
43-
; CHECK: [[ForInc]]:
44-
; CHECK: %[[IndNext]] = add nuw i64 %[[Ind]], 4
45-
; CHECK: %[[VecIndNext]] = add <4 x i64> %[[VecInd]], splat (i64 4)
46-
; CHECK: %[[Cmp:.*]] = icmp eq i64 %[[IndNext]], 8
47-
; CHECK: br i1 %[[Cmp]], label %middle.block, label %vector.body
48-
4920
@arr2 = external global [8 x i32], align 16
5021
@arr = external global [8 x [8 x i32]], align 16
5122

@@ -54,6 +25,40 @@
5425

5526
; Function Attrs: norecurse nounwind uwtable
5627
define void @foo_i32(i32 %n) {
28+
; CHECK-LABEL: define void @foo_i32(
29+
; CHECK-SAME: i32 [[N:%.*]]) {
30+
; CHECK-NEXT: [[ENTRY:.*:]]
31+
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
32+
; CHECK: [[VECTOR_PH]]:
33+
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[N]], i64 0
34+
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
35+
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
36+
; CHECK: [[VECTOR_BODY]]:
37+
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_LATCH:.*]] ]
38+
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_LATCH]] ]
39+
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [8 x i32], ptr @arr2, i64 0, <4 x i64> [[VEC_IND]]
40+
; CHECK-NEXT: [[TMP1:%.*]] = trunc <4 x i64> [[VEC_IND]] to <4 x i32>
41+
; CHECK-NEXT: call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> [[TMP1]], <4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true))
42+
; CHECK-NEXT: [[TMP8:%.*]] = trunc <4 x i64> [[VEC_IND]] to <4 x i32>
43+
; CHECK-NEXT: [[TMP2:%.*]] = add nsw <4 x i32> [[TMP8]], [[BROADCAST_SPLAT]]
44+
; CHECK-NEXT: br label %[[FOR_BODY31:.*]]
45+
; CHECK: [[FOR_BODY31]]:
46+
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, %[[VECTOR_BODY]] ], [ [[TMP4:%.*]], %[[FOR_BODY31]] ]
47+
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [8 x [8 x i32]], ptr @arr, i64 0, <4 x i64> [[VEC_PHI]], <4 x i64> [[VEC_IND]]
48+
; CHECK-NEXT: call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> [[TMP2]], <4 x ptr> [[TMP3]], i32 4, <4 x i1> splat (i1 true))
49+
; CHECK-NEXT: [[TMP4]] = add nuw nsw <4 x i64> [[VEC_PHI]], splat (i64 1)
50+
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq <4 x i64> [[TMP4]], splat (i64 8)
51+
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x i1> [[TMP5]], i32 0
52+
; CHECK-NEXT: br i1 [[TMP6]], label %[[VECTOR_LATCH]], label %[[FOR_BODY31]]
53+
; CHECK: [[VECTOR_LATCH]]:
54+
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
55+
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
56+
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 8
57+
; CHECK-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
58+
; CHECK: [[MIDDLE_BLOCK]]:
59+
; CHECK-NEXT: br i1 true, [[FOR_END10:label %.*]], label %[[SCALAR_PH]]
60+
; CHECK: [[SCALAR_PH]]:
61+
;
5762
entry:
5863
br label %for.body
5964

@@ -83,35 +88,39 @@ for.end10: ; preds = %for.inc8
8388
ret void
8489
}
8590

86-
; CHECK-LABEL: @foo_i64(
87-
; CHECK-LABEL: vector.ph:
88-
; CHECK: %[[SplatVal:.*]] = insertelement <2 x i64> poison, i64 %n, i64 0
89-
; CHECK: %[[Splat:.*]] = shufflevector <2 x i64> %[[SplatVal]], <2 x i64> poison, <2 x i32> zeroinitializer
90-
91-
; CHECK-LABEL: vector.body:
92-
; CHECK: %[[Ind:.*]] = phi i64 [ 0, %vector.ph ], [ %[[IndNext:.*]], %[[ForInc:.*]] ]
93-
; CHECK: %[[VecInd:.*]] = phi <2 x i64> [ <i64 0, i64 1>, %vector.ph ], [ %[[VecIndNext:.*]], %[[ForInc]] ]
94-
; CHECK: %[[AAddr:.*]] = getelementptr inbounds [8 x i64], ptr @arrX, i64 0, <2 x i64> %[[VecInd]]
95-
; CHECK: call void @llvm.masked.scatter.v2i64.v2p0(<2 x i64> %[[VecInd]], <2 x ptr> %[[AAddr]], i32 4, <2 x i1> splat (i1 true))
96-
; CHECK: %[[StoreVal:.*]] = add nsw <2 x i64> %[[VecInd]], %[[Splat]]
97-
; CHECK: br label %[[InnerLoop:.+]]
98-
99-
; CHECK: [[InnerLoop]]:
100-
; CHECK: %[[InnerPhi:.*]] = phi <2 x i64> [ zeroinitializer, %vector.body ], [ %[[InnerPhiNext:.*]], %[[InnerLoop]] ]
101-
; CHECK: %[[AAddr2:.*]] = getelementptr inbounds [8 x [8 x i64]], ptr @arrY, i64 0, <2 x i64> %[[InnerPhi]], <2 x i64> %[[VecInd]]
102-
; CHECK: call void @llvm.masked.scatter.v2i64.v2p0(<2 x i64> %[[StoreVal]], <2 x ptr> %[[AAddr2]], i32 4, <2 x i1> splat (i1 true))
103-
; CHECK: %[[InnerPhiNext]] = add nuw nsw <2 x i64> %[[InnerPhi]], splat (i64 1)
104-
; CHECK: %[[VecCond:.*]] = icmp eq <2 x i64> %[[InnerPhiNext]], splat (i64 8)
105-
; CHECK: %[[InnerCond:.*]] = extractelement <2 x i1> %[[VecCond]], i32 0
106-
; CHECK: br i1 %[[InnerCond]], label %[[ForInc]], label %[[InnerLoop]]
107-
108-
; CHECK: [[ForInc]]:
109-
; CHECK: %[[IndNext]] = add nuw i64 %[[Ind]], 2
110-
; CHECK: %[[VecIndNext]] = add <2 x i64> %[[VecInd]], splat (i64 2)
111-
; CHECK: %[[Cmp:.*]] = icmp eq i64 %[[IndNext]], 8
112-
; CHECK: br i1 %[[Cmp]], label %middle.block, label %vector.body
113-
; Function Attrs: norecurse nounwind uwtable
11491
define void @foo_i64(i64 %n) {
92+
; CHECK-LABEL: define void @foo_i64(
93+
; CHECK-SAME: i64 [[N:%.*]]) {
94+
; CHECK-NEXT: [[ENTRY:.*:]]
95+
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
96+
; CHECK: [[VECTOR_PH]]:
97+
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[N]], i64 0
98+
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
99+
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
100+
; CHECK: [[VECTOR_BODY]]:
101+
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_LATCH:.*]] ]
102+
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_LATCH]] ]
103+
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [8 x i64], ptr @arrX, i64 0, <2 x i64> [[VEC_IND]]
104+
; CHECK-NEXT: call void @llvm.masked.scatter.v2i64.v2p0(<2 x i64> [[VEC_IND]], <2 x ptr> [[TMP0]], i32 4, <2 x i1> splat (i1 true))
105+
; CHECK-NEXT: [[TMP1:%.*]] = add nsw <2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
106+
; CHECK-NEXT: br label %[[FOR_BODY31:.*]]
107+
; CHECK: [[FOR_BODY31]]:
108+
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, %[[VECTOR_BODY]] ], [ [[TMP3:%.*]], %[[FOR_BODY31]] ]
109+
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [8 x [8 x i64]], ptr @arrY, i64 0, <2 x i64> [[VEC_PHI]], <2 x i64> [[VEC_IND]]
110+
; CHECK-NEXT: call void @llvm.masked.scatter.v2i64.v2p0(<2 x i64> [[TMP1]], <2 x ptr> [[TMP2]], i32 4, <2 x i1> splat (i1 true))
111+
; CHECK-NEXT: [[TMP3]] = add nuw nsw <2 x i64> [[VEC_PHI]], splat (i64 1)
112+
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq <2 x i64> [[TMP3]], splat (i64 8)
113+
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
114+
; CHECK-NEXT: br i1 [[TMP5]], label %[[VECTOR_LATCH]], label %[[FOR_BODY31]]
115+
; CHECK: [[VECTOR_LATCH]]:
116+
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
117+
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
118+
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 8
119+
; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
120+
; CHECK: [[MIDDLE_BLOCK]]:
121+
; CHECK-NEXT: br i1 true, [[FOR_END10:label %.*]], label %[[SCALAR_PH]]
122+
; CHECK: [[SCALAR_PH]]:
123+
;
115124
entry:
116125
br label %for.body
117126

‎llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-extractvalue.ll

Lines changed: 35 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -1,17 +1,43 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph" --version 5
12
; RUN: opt -S -passes=loop-vectorize < %s | FileCheck %s
23

34
target triple = "aarch64-unknown-linux-gnu"
45

56
define void @widen_extractvalue(ptr %dst, {i64, i64} %sv) #0 {
6-
; CHECK-LABEL: @widen_extractvalue(
7-
; CHECK: [[EXTRACT0:%.*]] = extractvalue { i64, i64 } [[SV:%.*]], 0
8-
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[EXTRACT0]], i64 0
9-
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
10-
; CHECK-NEXT: [[EXTRACT1:%.*]] = extractvalue { i64, i64 } [[SV]], 1
11-
; CHECK-NEXT: [[DOTSPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[EXTRACT1]], i64 0
12-
; CHECK-NEXT: [[DOTSPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
13-
; CHECK: [[ADD:%.*]] = add <vscale x 2 x i64> [[DOTSPLAT]], [[DOTSPLAT2]]
14-
; CHECK: vector.body:
7+
; CHECK-LABEL: define void @widen_extractvalue(
8+
; CHECK-SAME: ptr [[DST:%.*]], { i64, i64 } [[SV:%.*]]) #[[ATTR0:[0-9]+]] {
9+
; CHECK-NEXT: [[ENTRY:.*:]]
10+
; CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
11+
; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i32 [[TMP0]], 2
12+
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 1000, [[TMP1]]
13+
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
14+
; CHECK: [[VECTOR_PH]]:
15+
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
16+
; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i32 [[TMP2]], 2
17+
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 1000, [[TMP3]]
18+
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 1000, [[N_MOD_VF]]
19+
; CHECK-NEXT: [[TMP4:%.*]] = call i32 @llvm.vscale.i32()
20+
; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i32 [[TMP4]], 2
21+
; CHECK-NEXT: [[EXTRACT0:%.*]] = extractvalue { i64, i64 } [[SV]], 0
22+
; CHECK-NEXT: [[DOTSPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[EXTRACT0]], i64 0
23+
; CHECK-NEXT: [[DOTSPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
24+
; CHECK-NEXT: [[TMP10:%.*]] = extractvalue { i64, i64 } [[SV]], 1
25+
; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
26+
; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
27+
; CHECK-NEXT: [[TMP7:%.*]] = add <vscale x 2 x i64> [[DOTSPLAT2]], [[BROADCAST_SPLAT2]]
28+
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
29+
; CHECK: [[VECTOR_BODY]]:
30+
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
31+
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[DST]], i32 [[INDEX]]
32+
; CHECK-NEXT: store <vscale x 2 x i64> [[TMP7]], ptr [[TMP8]], align 8
33+
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP5]]
34+
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
35+
; CHECK-NEXT: br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
36+
; CHECK: [[MIDDLE_BLOCK]]:
37+
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 1000, [[N_VEC]]
38+
; CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
39+
; CHECK: [[SCALAR_PH]]:
40+
;
1541
entry:
1642
br label %loop.body
1743

@@ -38,4 +64,3 @@ attributes #0 = { "target-features"="+sve" }
3864
!3 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}
3965
!4 = !{!"llvm.loop.interleave.count", i32 1}
4066
!5 = !{!"llvm.loop.vectorize.enable", i1 true}
41-

0 commit comments

Comments
 (0)