1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
|
// RUN: %clang_cc1 %s -flax-vector-conversions=none -ffreestanding -triple=x86_64-unknown-unknown -target-feature +avx512f \
// RUN: -target-feature +amx-transpose -target-feature +amx-bf16 -target-feature +amx-fp16 -target-feature +amx-complex \
// RUN: -emit-llvm -o - -Werror -pedantic | FileCheck %s --check-prefixes=CHECK
#include <immintrin.h>
char buf[2048];
#define STRIDE 32
char buf2[2048];
void test_tile_2rpntlvwz0(__tile1024i dst0, __tile1024i dst1) {
//CHECK-LABEL: @test_tile_2rpntlvwz0
//CHECK: call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 0
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 1
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
__tile_2rpntlvwz0(&dst0, &dst1, buf, STRIDE);
}
void test_tile_2rpntlvwz0t1(__tile1024i dst0, __tile1024i dst1) {
//CHECK-LABEL: @test_tile_2rpntlvwz0t1
//CHECK: call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0t1.internal
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 0
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 1
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
__tile_2rpntlvwz0t1(&dst0, &dst1, buf, STRIDE);
}
void test_tile_2rpntlvwz1(__tile1024i dst0, __tile1024i dst1) {
//CHECK-LABEL: @test_tile_2rpntlvwz1
//CHECK: call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1.internal
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 0
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 1
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
__tile_2rpntlvwz1(&dst0, &dst1, buf, STRIDE);
}
void test_tile_2rpntlvwz1t1(__tile1024i dst0, __tile1024i dst1) {
//CHECK-LABEL: @test_tile_2rpntlvwz1t1
//CHECK: call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1t1.internal
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 0
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
//CHECK-NEXT: {{%.*}} = extractvalue { x86_amx, x86_amx } {{%.*}}, 1
//CHECK-NEXT: {{%.*}} = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
//CHECK-NEXT: store <256 x i32> {{%.*}}, ptr {{%.*}}
__tile_2rpntlvwz1t1(&dst0, &dst1, buf, STRIDE);
}
void test_tile_transposed(__tile1024i dst, __tile1024i src) {
//CHECK-LABEL: @test_tile_transposed
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.ttransposed.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_transposed(&dst, src);
}
void test_tile_tdpbf16ps(__tile1024i a, __tile1024i b, __tile1024i c) {
//CHECK-LABEL: @test_tile_tdpbf16ps
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.ttdpbf16ps.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_tdpbf16ps(&c, a, b);
}
void test_tile_tdpfp16ps(__tile1024i a, __tile1024i b, __tile1024i c) {
//CHECK-LABEL: @test_tile_tdpfp16ps
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.ttdpfp16ps.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_tdpfp16ps(&c, a, b);
}
void test_tile_tcmmimfp16ps(__tile1024i a, __tile1024i b, __tile1024i c) {
//CHECK-LABEL: @test_tile_tcmmimfp16ps
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.ttcmmimfp16ps.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_tcmmimfp16ps(&c, a, b);
}
void test_tile_tcmmrlfp16ps(__tile1024i a, __tile1024i b, __tile1024i c) {
//CHECK-LABEL: @test_tile_tcmmrlfp16ps
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.ttcmmrlfp16ps.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_tcmmrlfp16ps(&c, a, b);
}
void test_tile_conjtcmmimfp16ps(__tile1024i a, __tile1024i b, __tile1024i c) {
//CHECK-LABEL: @test_tile_conjtcmmimfp16ps
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.tconjtcmmimfp16ps.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_conjtcmmimfp16ps(&c, a, b);
}
void test_tile_conjtfp16(__tile1024i dst, __tile1024i src) {
//CHECK-LABEL: @test_tile_conjtfp16
//CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}})
//CHECK-DAG: call x86_amx @llvm.x86.tconjtfp16.internal
//CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}})
__tile_conjtfp16(&dst, src);
}
|
