xref: /llvm-project/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backward.ll (revision dbe4143f23ff27a058c4e7c551a0fa70b93b438e) 
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -matrix-allow-contract=false -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s
3
4define <8 x double> @fadd_transpose(<8 x double> %a, <8 x double> %b) {
5; CHECK-LABEL: @fadd_transpose(
6; CHECK-NEXT:  entry:
7; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> poison, <2 x i32> <i32 0, i32 1>
8; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> poison, <2 x i32> <i32 2, i32 3>
9; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <8 x double> [[A]], <8 x double> poison, <2 x i32> <i32 4, i32 5>
10; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <8 x double> [[A]], <8 x double> poison, <2 x i32> <i32 6, i32 7>
11; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <8 x double> [[B:%.*]], <8 x double> poison, <2 x i32> <i32 0, i32 1>
12; CHECK-NEXT:    [[SPLIT5:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 2, i32 3>
13; CHECK-NEXT:    [[SPLIT6:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 4, i32 5>
14; CHECK-NEXT:    [[SPLIT7:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 6, i32 7>
15; CHECK-NEXT:    [[TMP0:%.*]] = fadd <2 x double> [[SPLIT]], [[SPLIT4]]
16; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x double> [[SPLIT1]], [[SPLIT5]]
17; CHECK-NEXT:    [[TMP2:%.*]] = fadd <2 x double> [[SPLIT2]], [[SPLIT6]]
18; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x double> [[SPLIT3]], [[SPLIT7]]
19; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 0
20; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> poison, double [[TMP4]], i64 0
21; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 0
22; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 1
23; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 0
24; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 2
25; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 0
26; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 3
27; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 1
28; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i64 0
29; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 1
30; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 1
31; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 1
32; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 2
33; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 1
34; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 3
35; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
36; CHECK-NEXT:    ret <8 x double> [[TMP20]]
37;
38entry:
39  %add = fadd <8 x double> %a, %b
40  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %add, i32 2, i32 4)
41  ret <8 x double> %c
42}
43
44define <8 x double> @load_fadd_transpose(ptr %A.Ptr, <8 x double> %b) {
45; CHECK-LABEL: @load_fadd_transpose(
46; CHECK-NEXT:  entry:
47; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <2 x double>, ptr [[A_PTR:%.*]], align 8
48; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[A_PTR]], i64 2
49; CHECK-NEXT:    [[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8
50; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[A_PTR]], i64 4
51; CHECK-NEXT:    [[COL_LOAD3:%.*]] = load <2 x double>, ptr [[VEC_GEP2]], align 8
52; CHECK-NEXT:    [[VEC_GEP4:%.*]] = getelementptr double, ptr [[A_PTR]], i64 6
53; CHECK-NEXT:    [[COL_LOAD5:%.*]] = load <2 x double>, ptr [[VEC_GEP4]], align 8
54; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[B:%.*]], <8 x double> poison, <2 x i32> <i32 0, i32 1>
55; CHECK-NEXT:    [[SPLIT6:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 2, i32 3>
56; CHECK-NEXT:    [[SPLIT7:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 4, i32 5>
57; CHECK-NEXT:    [[SPLIT8:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 6, i32 7>
58; CHECK-NEXT:    [[TMP0:%.*]] = fadd <2 x double> [[COL_LOAD]], [[SPLIT]]
59; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x double> [[COL_LOAD1]], [[SPLIT6]]
60; CHECK-NEXT:    [[TMP2:%.*]] = fadd <2 x double> [[COL_LOAD3]], [[SPLIT7]]
61; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x double> [[COL_LOAD5]], [[SPLIT8]]
62; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 0
63; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> poison, double [[TMP4]], i64 0
64; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 0
65; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 1
66; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 0
67; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 2
68; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 0
69; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 3
70; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 1
71; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i64 0
72; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 1
73; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 1
74; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 1
75; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 2
76; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 1
77; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 3
78; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
79; CHECK-NEXT:    ret <8 x double> [[TMP20]]
80;
81
82
83entry:
84  %a = load <8 x double>, ptr %A.Ptr, align 8
85  %add = fadd <8 x double> %a, %b
86  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %add, i32 2, i32 4)
87  ret <8 x double> %c
88}
89
90define <8 x double> @load_fneg_transpose(ptr %A.Ptr) {
91; CHECK-LABEL: @load_fneg_transpose(
92; CHECK-NEXT:  entry:
93; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <2 x double>, ptr [[A_PTR:%.*]], align 8
94; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[A_PTR]], i64 2
95; CHECK-NEXT:    [[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8
96; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[A_PTR]], i64 4
97; CHECK-NEXT:    [[COL_LOAD3:%.*]] = load <2 x double>, ptr [[VEC_GEP2]], align 8
98; CHECK-NEXT:    [[VEC_GEP4:%.*]] = getelementptr double, ptr [[A_PTR]], i64 6
99; CHECK-NEXT:    [[COL_LOAD5:%.*]] = load <2 x double>, ptr [[VEC_GEP4]], align 8
100; CHECK-NEXT:    [[TMP0:%.*]] = fneg <2 x double> [[COL_LOAD]]
101; CHECK-NEXT:    [[TMP1:%.*]] = fneg <2 x double> [[COL_LOAD1]]
102; CHECK-NEXT:    [[TMP2:%.*]] = fneg <2 x double> [[COL_LOAD3]]
103; CHECK-NEXT:    [[TMP3:%.*]] = fneg <2 x double> [[COL_LOAD5]]
104; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 0
105; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> poison, double [[TMP4]], i64 0
106; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 0
107; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 1
108; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 0
109; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 2
110; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 0
111; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 3
112; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 1
113; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i64 0
114; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 1
115; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 1
116; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 1
117; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 2
118; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 1
119; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 3
120; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
121; CHECK-NEXT:    ret <8 x double> [[TMP20]]
122;
123entry:
124  %a = load <8 x double>, ptr %A.Ptr, align 8
125  %neg = fneg <8 x double> %a
126  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %neg, i32 2, i32 4)
127  ret <8 x double> %c
128}
129declare <8 x double> @llvm.matrix.transpose(<8 x double>, i32, i32)
130