1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -lower-matrix-intrinsics -S < %s | FileCheck %s
3; RUN: opt -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s
4
5define <9 x double> @strided_load_3x3(double* %in, i64 %stride) {
6; CHECK-LABEL: @strided_load_3x3(
7; CHECK-NEXT:  entry:
8; CHECK-NEXT:    [[VEC_START:%.*]] = mul i64 0, [[STRIDE:%.*]]
9; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, double* [[IN:%.*]], i64 [[VEC_START]]
10; CHECK-NEXT:    [[VEC_CAST:%.*]] = bitcast double* [[VEC_GEP]] to <3 x double>*
11; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, <3 x double>* [[VEC_CAST]], align 8
12; CHECK-NEXT:    [[VEC_START1:%.*]] = mul i64 1, [[STRIDE]]
13; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, double* [[IN]], i64 [[VEC_START1]]
14; CHECK-NEXT:    [[VEC_CAST3:%.*]] = bitcast double* [[VEC_GEP2]] to <3 x double>*
15; CHECK-NEXT:    [[COL_LOAD4:%.*]] = load <3 x double>, <3 x double>* [[VEC_CAST3]], align 8
16; CHECK-NEXT:    [[VEC_START5:%.*]] = mul i64 2, [[STRIDE]]
17; CHECK-NEXT:    [[VEC_GEP6:%.*]] = getelementptr double, double* [[IN]], i64 [[VEC_START5]]
18; CHECK-NEXT:    [[VEC_CAST7:%.*]] = bitcast double* [[VEC_GEP6]] to <3 x double>*
19; CHECK-NEXT:    [[COL_LOAD8:%.*]] = load <3 x double>, <3 x double>* [[VEC_CAST7]], align 8
20; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD4]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
21; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD8]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
22; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
23; CHECK-NEXT:    ret <9 x double> [[TMP2]]
24;
25entry:
26  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i64(double* %in, i64 %stride, i1 false, i32 3, i32 3)
27  ret <9 x double> %load
28}
29
30declare <9 x double> @llvm.matrix.column.major.load.v9f64.i64(double*, i64, i1, i32, i32)
31
32define <9 x double> @strided_load_9x1(double* %in, i64 %stride) {
33; CHECK-LABEL: @strided_load_9x1(
34; CHECK-NEXT:  entry:
35; CHECK-NEXT:    [[VEC_START:%.*]] = mul i64 0, [[STRIDE:%.*]]
36; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, double* [[IN:%.*]], i64 [[VEC_START]]
37; CHECK-NEXT:    [[VEC_CAST:%.*]] = bitcast double* [[VEC_GEP]] to <9 x double>*
38; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <9 x double>, <9 x double>* [[VEC_CAST]], align 8
39; CHECK-NEXT:    ret <9 x double> [[COL_LOAD]]
40;
41entry:
42  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i64(double* %in, i64 %stride, i1 false, i32 9, i32 1)
43  ret <9 x double> %load
44}
45
46declare <8 x double> @llvm.matrix.column.major.load.v8f64.i64(double*, i64, i1, i32, i32)
47
48define <8 x double> @strided_load_4x2(double* %in, i64 %stride) {
49; CHECK-LABEL: @strided_load_4x2(
50; CHECK-NEXT:  entry:
51; CHECK-NEXT:    [[VEC_START:%.*]] = mul i64 0, [[STRIDE:%.*]]
52; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, double* [[IN:%.*]], i64 [[VEC_START]]
53; CHECK-NEXT:    [[VEC_CAST:%.*]] = bitcast double* [[VEC_GEP]] to <4 x double>*
54; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <4 x double>, <4 x double>* [[VEC_CAST]], align 8
55; CHECK-NEXT:    [[VEC_START1:%.*]] = mul i64 1, [[STRIDE]]
56; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, double* [[IN]], i64 [[VEC_START1]]
57; CHECK-NEXT:    [[VEC_CAST3:%.*]] = bitcast double* [[VEC_GEP2]] to <4 x double>*
58; CHECK-NEXT:    [[COL_LOAD4:%.*]] = load <4 x double>, <4 x double>* [[VEC_CAST3]], align 8
59; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x double> [[COL_LOAD]], <4 x double> [[COL_LOAD4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
60; CHECK-NEXT:    ret <8 x double> [[TMP0]]
61;
62entry:
63  %load = call <8 x double> @llvm.matrix.column.major.load.v8f64.i64(double* %in, i64 %stride, i1 false, i32 4, i32 2)
64  ret <8 x double> %load
65}
66
67declare <8 x double> @llvm.matrix.column.major.load.v8f64.i32(double*, i32, i1, i32, i32)
68
69define <8 x double> @strided_load_4x2_stride_i32(double* %in, i32 %stride) {
70; CHECK-LABEL: @strided_load_4x2_stride_i32(
71; CHECK-NEXT:  entry:
72; CHECK-NEXT:    [[VEC_START:%.*]] = mul i32 0, [[STRIDE:%.*]]
73; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, double* [[IN:%.*]], i32 [[VEC_START]]
74; CHECK-NEXT:    [[VEC_CAST:%.*]] = bitcast double* [[VEC_GEP]] to <4 x double>*
75; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <4 x double>, <4 x double>* [[VEC_CAST]], align 8
76; CHECK-NEXT:    [[VEC_START1:%.*]] = mul i32 1, [[STRIDE]]
77; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, double* [[IN]], i32 [[VEC_START1]]
78; CHECK-NEXT:    [[VEC_CAST3:%.*]] = bitcast double* [[VEC_GEP2]] to <4 x double>*
79; CHECK-NEXT:    [[COL_LOAD4:%.*]] = load <4 x double>, <4 x double>* [[VEC_CAST3]], align 8
80; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x double> [[COL_LOAD]], <4 x double> [[COL_LOAD4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
81; CHECK-NEXT:    ret <8 x double> [[TMP0]]
82;
83entry:
84  %load = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(double* %in, i32 %stride, i1 false, i32 4, i32 2)
85  ret <8 x double> %load
86}
87