xref: /llvm-project/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll (revision 8b56da5e9f3ba737a5ff4bf5dee654416849042f) 
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s
3
4
5define <8 x float> @transpose(<8 x float> %a) {
6; CHECK-LABEL: @transpose(
7; CHECK-NEXT:  entry:
8; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <2 x i32> <i32 0, i32 1>
9; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <2 x i32> <i32 2, i32 3>
10; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <2 x i32> <i32 4, i32 5>
11; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <2 x i32> <i32 6, i32 7>
12; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <2 x float> [[SPLIT]], i64 0
13; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i64 0
14; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x float> [[SPLIT1]], i64 0
15; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP1]], float [[TMP2]], i64 1
16; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[SPLIT2]], i64 0
17; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> [[TMP3]], float [[TMP4]], i64 2
18; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x float> [[SPLIT3]], i64 0
19; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x float> [[TMP5]], float [[TMP6]], i64 3
20; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x float> [[SPLIT]], i64 1
21; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x float> poison, float [[TMP8]], i64 0
22; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x float> [[SPLIT1]], i64 1
23; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x float> [[TMP9]], float [[TMP10]], i64 1
24; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x float> [[SPLIT2]], i64 1
25; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x float> [[TMP11]], float [[TMP12]], i64 2
26; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x float> [[SPLIT3]], i64 1
27; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x float> [[TMP13]], float [[TMP14]], i64 3
28; CHECK-NEXT:    [[TMP16:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> [[TMP15]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
29; CHECK-NEXT:    ret <8 x float> [[TMP16]]
30;
31entry:
32  %c  = call <8 x float> @llvm.matrix.transpose(<8 x float> %a, i32 2, i32 4)
33  ret <8 x float> %c
34}
35
36declare <8 x float> @llvm.matrix.transpose(<8 x float>, i32, i32)
37
38define <8 x float> @transpose_single_column(<8 x float> %a) {
39; CHECK-LABEL: @transpose_single_column(
40; CHECK-NEXT:  entry:
41; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
42; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <8 x float> [[SPLIT]], i64 0
43; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <1 x float> poison, float [[TMP0]], i64 0
44; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x float> [[SPLIT]], i64 1
45; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <1 x float> poison, float [[TMP2]], i64 0
46; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <8 x float> [[SPLIT]], i64 2
47; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <1 x float> poison, float [[TMP4]], i64 0
48; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <8 x float> [[SPLIT]], i64 3
49; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <1 x float> poison, float [[TMP6]], i64 0
50; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <8 x float> [[SPLIT]], i64 4
51; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <1 x float> poison, float [[TMP8]], i64 0
52; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <8 x float> [[SPLIT]], i64 5
53; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <1 x float> poison, float [[TMP10]], i64 0
54; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <8 x float> [[SPLIT]], i64 6
55; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <1 x float> poison, float [[TMP12]], i64 0
56; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <8 x float> [[SPLIT]], i64 7
57; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <1 x float> poison, float [[TMP14]], i64 0
58; CHECK-NEXT:    [[TMP16:%.*]] = shufflevector <1 x float> [[TMP1]], <1 x float> [[TMP3]], <2 x i32> <i32 0, i32 1>
59; CHECK-NEXT:    [[TMP17:%.*]] = shufflevector <1 x float> [[TMP5]], <1 x float> [[TMP7]], <2 x i32> <i32 0, i32 1>
60; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <1 x float> [[TMP9]], <1 x float> [[TMP11]], <2 x i32> <i32 0, i32 1>
61; CHECK-NEXT:    [[TMP19:%.*]] = shufflevector <1 x float> [[TMP13]], <1 x float> [[TMP15]], <2 x i32> <i32 0, i32 1>
62; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <2 x float> [[TMP16]], <2 x float> [[TMP17]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
63; CHECK-NEXT:    [[TMP21:%.*]] = shufflevector <2 x float> [[TMP18]], <2 x float> [[TMP19]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
64; CHECK-NEXT:    [[TMP22:%.*]] = shufflevector <4 x float> [[TMP20]], <4 x float> [[TMP21]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
65; CHECK-NEXT:    ret <8 x float> [[TMP22]]
66;
67entry:
68  %c  = call <8 x float> @llvm.matrix.transpose(<8 x float> %a, i32 8, i32 1)
69  ret <8 x float> %c
70}
71
72declare <12 x float> @llvm.matrix.transpose.v12f32(<12 x float>, i32, i32)
73
74define <12 x float> @transpose_float_3x4(<12 x float> %a) {
75; CHECK-LABEL: @transpose_float_3x4(
76; CHECK-NEXT:  entry:
77; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <12 x float> [[A:%.*]], <12 x float> poison, <3 x i32> <i32 0, i32 1, i32 2>
78; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <3 x i32> <i32 3, i32 4, i32 5>
79; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <3 x i32> <i32 6, i32 7, i32 8>
80; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <3 x i32> <i32 9, i32 10, i32 11>
81; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <3 x float> [[SPLIT]], i64 0
82; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i64 0
83; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <3 x float> [[SPLIT1]], i64 0
84; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP1]], float [[TMP2]], i64 1
85; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <3 x float> [[SPLIT2]], i64 0
86; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> [[TMP3]], float [[TMP4]], i64 2
87; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <3 x float> [[SPLIT3]], i64 0
88; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x float> [[TMP5]], float [[TMP6]], i64 3
89; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <3 x float> [[SPLIT]], i64 1
90; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x float> poison, float [[TMP8]], i64 0
91; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <3 x float> [[SPLIT1]], i64 1
92; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x float> [[TMP9]], float [[TMP10]], i64 1
93; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <3 x float> [[SPLIT2]], i64 1
94; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x float> [[TMP11]], float [[TMP12]], i64 2
95; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <3 x float> [[SPLIT3]], i64 1
96; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x float> [[TMP13]], float [[TMP14]], i64 3
97; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <3 x float> [[SPLIT]], i64 2
98; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x float> poison, float [[TMP16]], i64 0
99; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <3 x float> [[SPLIT1]], i64 2
100; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x float> [[TMP17]], float [[TMP18]], i64 1
101; CHECK-NEXT:    [[TMP20:%.*]] = extractelement <3 x float> [[SPLIT2]], i64 2
102; CHECK-NEXT:    [[TMP21:%.*]] = insertelement <4 x float> [[TMP19]], float [[TMP20]], i64 2
103; CHECK-NEXT:    [[TMP22:%.*]] = extractelement <3 x float> [[SPLIT3]], i64 2
104; CHECK-NEXT:    [[TMP23:%.*]] = insertelement <4 x float> [[TMP21]], float [[TMP22]], i64 3
105; CHECK-NEXT:    [[TMP24:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> [[TMP15]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
106; CHECK-NEXT:    [[TMP25:%.*]] = shufflevector <4 x float> [[TMP23]], <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
107; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <8 x float> [[TMP24]], <8 x float> [[TMP25]], <12 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>
108; CHECK-NEXT:    ret <12 x float> [[TMP26]]
109;
110entry:
111  %c  = call <12 x float> @llvm.matrix.transpose.v12f32(<12 x float> %a, i32 3, i32 4)
112  ret <12 x float> %c
113}
114