1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -lower-matrix-intrinsics -matrix-default-layout=row-major -S < %s | FileCheck --check-prefix=RM %s
3
4define <8 x double> @transpose(<8 x double> %a) {
5; RM-LABEL: @transpose(
6; RM-NEXT:  entry:
7; RM-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
8; RM-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
9; RM-NEXT:    [[TMP0:%.*]] = extractelement <4 x double> [[SPLIT]], i64 0
10; RM-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double [[TMP0]], i64 0
11; RM-NEXT:    [[TMP2:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 0
12; RM-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> [[TMP1]], double [[TMP2]], i64 1
13; RM-NEXT:    [[TMP4:%.*]] = extractelement <4 x double> [[SPLIT]], i64 1
14; RM-NEXT:    [[TMP5:%.*]] = insertelement <2 x double> undef, double [[TMP4]], i64 0
15; RM-NEXT:    [[TMP6:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 1
16; RM-NEXT:    [[TMP7:%.*]] = insertelement <2 x double> [[TMP5]], double [[TMP6]], i64 1
17; RM-NEXT:    [[TMP8:%.*]] = extractelement <4 x double> [[SPLIT]], i64 2
18; RM-NEXT:    [[TMP9:%.*]] = insertelement <2 x double> undef, double [[TMP8]], i64 0
19; RM-NEXT:    [[TMP10:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 2
20; RM-NEXT:    [[TMP11:%.*]] = insertelement <2 x double> [[TMP9]], double [[TMP10]], i64 1
21; RM-NEXT:    [[TMP12:%.*]] = extractelement <4 x double> [[SPLIT]], i64 3
22; RM-NEXT:    [[TMP13:%.*]] = insertelement <2 x double> undef, double [[TMP12]], i64 0
23; RM-NEXT:    [[TMP14:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 3
24; RM-NEXT:    [[TMP15:%.*]] = insertelement <2 x double> [[TMP13]], double [[TMP14]], i64 1
25; RM-NEXT:    [[TMP16:%.*]] = shufflevector <2 x double> [[TMP3]], <2 x double> [[TMP7]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
26; RM-NEXT:    [[TMP17:%.*]] = shufflevector <2 x double> [[TMP11]], <2 x double> [[TMP15]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
27; RM-NEXT:    [[TMP18:%.*]] = shufflevector <4 x double> [[TMP16]], <4 x double> [[TMP17]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
28; RM-NEXT:    ret <8 x double> [[TMP18]]
29;
30entry:
31  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %a, i32 2, i32 4)
32  ret <8 x double> %c
33}
34
35declare <8 x double> @llvm.matrix.transpose(<8 x double>, i32, i32)
36
37define <8 x double> @transpose_single_column(<8 x double> %a) {
38; RM-LABEL: @transpose_single_column(
39; RM-NEXT:  entry:
40; RM-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> undef, <1 x i32> zeroinitializer
41; RM-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 1>
42; RM-NEXT:    [[SPLIT2:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 2>
43; RM-NEXT:    [[SPLIT3:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 3>
44; RM-NEXT:    [[SPLIT4:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 4>
45; RM-NEXT:    [[SPLIT5:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 5>
46; RM-NEXT:    [[SPLIT6:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 6>
47; RM-NEXT:    [[SPLIT7:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <1 x i32> <i32 7>
48; RM-NEXT:    [[TMP0:%.*]] = extractelement <1 x double> [[SPLIT]], i64 0
49; RM-NEXT:    [[TMP1:%.*]] = insertelement <8 x double> undef, double [[TMP0]], i64 0
50; RM-NEXT:    [[TMP2:%.*]] = extractelement <1 x double> [[SPLIT1]], i64 0
51; RM-NEXT:    [[TMP3:%.*]] = insertelement <8 x double> [[TMP1]], double [[TMP2]], i64 1
52; RM-NEXT:    [[TMP4:%.*]] = extractelement <1 x double> [[SPLIT2]], i64 0
53; RM-NEXT:    [[TMP5:%.*]] = insertelement <8 x double> [[TMP3]], double [[TMP4]], i64 2
54; RM-NEXT:    [[TMP6:%.*]] = extractelement <1 x double> [[SPLIT3]], i64 0
55; RM-NEXT:    [[TMP7:%.*]] = insertelement <8 x double> [[TMP5]], double [[TMP6]], i64 3
56; RM-NEXT:    [[TMP8:%.*]] = extractelement <1 x double> [[SPLIT4]], i64 0
57; RM-NEXT:    [[TMP9:%.*]] = insertelement <8 x double> [[TMP7]], double [[TMP8]], i64 4
58; RM-NEXT:    [[TMP10:%.*]] = extractelement <1 x double> [[SPLIT5]], i64 0
59; RM-NEXT:    [[TMP11:%.*]] = insertelement <8 x double> [[TMP9]], double [[TMP10]], i64 5
60; RM-NEXT:    [[TMP12:%.*]] = extractelement <1 x double> [[SPLIT6]], i64 0
61; RM-NEXT:    [[TMP13:%.*]] = insertelement <8 x double> [[TMP11]], double [[TMP12]], i64 6
62; RM-NEXT:    [[TMP14:%.*]] = extractelement <1 x double> [[SPLIT7]], i64 0
63; RM-NEXT:    [[TMP15:%.*]] = insertelement <8 x double> [[TMP13]], double [[TMP14]], i64 7
64; RM-NEXT:    ret <8 x double> [[TMP15]]
65;
66entry:
67  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %a, i32 8, i32 1)
68  ret <8 x double> %c
69}
70
71declare <12 x double> @llvm.matrix.transpose.v12f64(<12 x double>, i32, i32)
72
73define <12 x double> @transpose_double_3x4(<12 x double> %a) {
74; RM-LABEL: @transpose_double_3x4(
75; RM-NEXT:  entry:
76; RM-NEXT:    [[SPLIT:%.*]] = shufflevector <12 x double> [[A:%.*]], <12 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
77; RM-NEXT:    [[SPLIT1:%.*]] = shufflevector <12 x double> [[A]], <12 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
78; RM-NEXT:    [[SPLIT2:%.*]] = shufflevector <12 x double> [[A]], <12 x double> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
79; RM-NEXT:    [[TMP0:%.*]] = extractelement <4 x double> [[SPLIT]], i64 0
80; RM-NEXT:    [[TMP1:%.*]] = insertelement <3 x double> undef, double [[TMP0]], i64 0
81; RM-NEXT:    [[TMP2:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 0
82; RM-NEXT:    [[TMP3:%.*]] = insertelement <3 x double> [[TMP1]], double [[TMP2]], i64 1
83; RM-NEXT:    [[TMP4:%.*]] = extractelement <4 x double> [[SPLIT2]], i64 0
84; RM-NEXT:    [[TMP5:%.*]] = insertelement <3 x double> [[TMP3]], double [[TMP4]], i64 2
85; RM-NEXT:    [[TMP6:%.*]] = extractelement <4 x double> [[SPLIT]], i64 1
86; RM-NEXT:    [[TMP7:%.*]] = insertelement <3 x double> undef, double [[TMP6]], i64 0
87; RM-NEXT:    [[TMP8:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 1
88; RM-NEXT:    [[TMP9:%.*]] = insertelement <3 x double> [[TMP7]], double [[TMP8]], i64 1
89; RM-NEXT:    [[TMP10:%.*]] = extractelement <4 x double> [[SPLIT2]], i64 1
90; RM-NEXT:    [[TMP11:%.*]] = insertelement <3 x double> [[TMP9]], double [[TMP10]], i64 2
91; RM-NEXT:    [[TMP12:%.*]] = extractelement <4 x double> [[SPLIT]], i64 2
92; RM-NEXT:    [[TMP13:%.*]] = insertelement <3 x double> undef, double [[TMP12]], i64 0
93; RM-NEXT:    [[TMP14:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 2
94; RM-NEXT:    [[TMP15:%.*]] = insertelement <3 x double> [[TMP13]], double [[TMP14]], i64 1
95; RM-NEXT:    [[TMP16:%.*]] = extractelement <4 x double> [[SPLIT2]], i64 2
96; RM-NEXT:    [[TMP17:%.*]] = insertelement <3 x double> [[TMP15]], double [[TMP16]], i64 2
97; RM-NEXT:    [[TMP18:%.*]] = extractelement <4 x double> [[SPLIT]], i64 3
98; RM-NEXT:    [[TMP19:%.*]] = insertelement <3 x double> undef, double [[TMP18]], i64 0
99; RM-NEXT:    [[TMP20:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 3
100; RM-NEXT:    [[TMP21:%.*]] = insertelement <3 x double> [[TMP19]], double [[TMP20]], i64 1
101; RM-NEXT:    [[TMP22:%.*]] = extractelement <4 x double> [[SPLIT2]], i64 3
102; RM-NEXT:    [[TMP23:%.*]] = insertelement <3 x double> [[TMP21]], double [[TMP22]], i64 2
103; RM-NEXT:    [[TMP24:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> [[TMP11]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
104; RM-NEXT:    [[TMP25:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> [[TMP23]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
105; RM-NEXT:    [[TMP26:%.*]] = shufflevector <6 x double> [[TMP24]], <6 x double> [[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>
106; RM-NEXT:    ret <12 x double> [[TMP26]]
107;
108entry:
109  %c  = call <12 x double> @llvm.matrix.transpose.v12f64(<12 x double> %a, i32 3, i32 4)
110  ret <12 x double> %c
111}
112