File: builtins-amdgcn-wmma-w64.cl

package info (click to toggle)
llvm-toolchain-15 1%3A15.0.6-4
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,554,644 kB
  • sloc: cpp: 5,922,452; ansic: 1,012,136; asm: 674,362; python: 191,568; objc: 73,855; f90: 42,327; lisp: 31,913; pascal: 11,973; javascript: 10,144; sh: 9,421; perl: 7,447; ml: 5,527; awk: 3,523; makefile: 2,520; xml: 885; cs: 573; fortran: 567
file content (84 lines) | stat: -rw-r--r-- 3,131 bytes parent folder | download | duplicates (3)
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
// REQUIRES: amdgpu-registered-target
// RUN: %clang_cc1 -triple amdgcn-unknown-unknown -target-cpu gfx1100 -DWMMA_GFX1100_TESTS -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-GFX1100

typedef float  v4f   __attribute__((ext_vector_type(4)));
typedef float  v8f   __attribute__((ext_vector_type(8)));
typedef half   v8h   __attribute__((ext_vector_type(8)));
typedef half   v16h  __attribute__((ext_vector_type(16)));
typedef int    v2i   __attribute__((ext_vector_type(2)));
typedef int    v4i   __attribute__((ext_vector_type(4)));
typedef int    v8i   __attribute__((ext_vector_type(8)));
typedef short  v8s   __attribute__((ext_vector_type(8)));
typedef short  v16s  __attribute__((ext_vector_type(16)));

#ifdef WMMA_GFX1100_TESTS

// Wave64

//
// amdgcn_wmma_f32_16x16x16_f16
//

// CHECK-GFX1100-LABEL: @test_amdgcn_wmma_f32_16x16x16_f16_w64
// CHECK-GFX1100: call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32(<16 x half> %a, <16 x half> %b, <4 x float> %c)
void test_amdgcn_wmma_f32_16x16x16_f16_w64(global v4f* out, v16h a, v16h b, v4f c)
{
  *out = __builtin_amdgcn_wmma_f32_16x16x16_f16_w64(a, b, c);
}

//
// amdgcn_wmma_f32_16x16x16_bf16
//

// CHECK-GFX1100-LABEL: @test_amdgcn_wmma_f32_16x16x16_bf16_w64
// CHECK-GFX1100: call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32(<16 x i16> %a, <16 x i16> %b, <4 x float> %c)
void test_amdgcn_wmma_f32_16x16x16_bf16_w64(global v4f* out, v16s a, v16s b, v4f c)
{
  *out = __builtin_amdgcn_wmma_f32_16x16x16_bf16_w64(a, b, c);
}

//
// amdgcn_wmma_f16_16x16x16_f16
//

// CHECK-GFX1100-LABEL: @test_amdgcn_wmma_f16_16x16x16_f16_w64
// CHECK-GFX1100: call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16(<16 x half> %a, <16 x half> %b, <8 x half> %c, i1 true)
void test_amdgcn_wmma_f16_16x16x16_f16_w64(global v8h* out, v16h a, v16h b, v8h c)
{
  *out = __builtin_amdgcn_wmma_f16_16x16x16_f16_w64(a, b, c, true);
}

//
// amdgcn_wmma_bf16_16x16x16_bf16
//

// CHECK-GFX1100-LABEL: @test_amdgcn_wmma_bf16_16x16x16_bf16_w64
// CHECK-GFX1100: call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16(<16 x i16> %a, <16 x i16> %b, <8 x i16> %c, i1 true)
void test_amdgcn_wmma_bf16_16x16x16_bf16_w64(global v8s* out, v16s a, v16s b, v8s c)
{
  *out = __builtin_amdgcn_wmma_bf16_16x16x16_bf16_w64(a, b, c, true);
}

//
// amdgcn_wmma_i32_16x16x16_iu8
//

// CHECK-GFX1100-LABEL: @test_amdgcn_wmma_i32_16x16x16_iu8_w64
// CHECK-GFX1100: call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32(i1 true, <4 x i32> %a, i1 true, <4 x i32> %b, <4 x i32> %c, i1 false)
void test_amdgcn_wmma_i32_16x16x16_iu8_w64(global v4i* out, v4i a, v4i b, v4i c)
{
  *out = __builtin_amdgcn_wmma_i32_16x16x16_iu8_w64(true, a, true, b, c, false);
}

//
// amdgcn_wmma_i32_16x16x16_iu4
//

// CHECK-GFX1100-LABEL: @test_amdgcn_wmma_i32_16x16x16_iu4_w64
// CHECK-GFX1100: call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32(i1 true, <2 x i32> %a, i1 true, <2 x i32> %b, <4 x i32> %c, i1 false)
void test_amdgcn_wmma_i32_16x16x16_iu4_w64(global v4i* out, v2i a, v2i b, v4i c)
{
  *out = __builtin_amdgcn_wmma_i32_16x16x16_iu4_w64(true, a, true, b, c, false);
}

#endif