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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
|
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
// REQUIRES: amdgpu-registered-target, x86-registered-target
// fp-contract, -no-enable-noundef-analysis and visibility are to just get the
// same output for openmp and hip.
// RUN: %clang_cc1 -x c++ -fopenmp -fvisibility=default -ffp-contract=off -no-enable-noundef-analysis \
// RUN: -internal-isystem %S/../../lib/Headers/openmp_wrappers \
// RUN: -include __clang_openmp_device_functions.h \
// RUN: -internal-isystem %S/../../lib/Headers/openmp_wrappers \
// RUN: -internal-isystem %S/Inputs/include \
// RUN: -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown -verify \
// RUN: -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm -fopenmp-is-target-device \
// RUN: -o - %s | FileCheck --check-prefixes=CHECK,OPENMP,OPENMP-CPP %s
// RUN: %clang_cc1 -x c -fopenmp -fvisibility=default -ffp-contract=off -no-enable-noundef-analysis \
// RUN: -internal-isystem %S/../../lib/Headers/openmp_wrappers \
// RUN: -include __clang_openmp_device_functions.h \
// RUN: -internal-isystem %S/../../lib/Headers/openmp_wrappers \
// RUN: -internal-isystem %S/Inputs/include \
// RUN: -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown -verify \
// RUN: -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm -fopenmp-is-target-device \
// RUN: -o - %s | FileCheck --check-prefixes=CHECK,OPENMP,OPENMP-C %s
// RUN: %clang_cc1 -x hip -fvisibility=default -ffp-contract=off -no-enable-noundef-analysis \
// RUN: -include __clang_hip_runtime_wrapper.h \
// RUN: -internal-isystem %S/../../lib/Headers/cuda_wrappers \
// RUN: -internal-isystem %S/Inputs/include \
// RUN: -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown -verify \
// RUN: -emit-llvm -fcuda-is-device -o - \
// RUN: -D__HIPCC_RTC__ %s | FileCheck -check-prefixes=CHECK,HIP %s
// expected-no-diagnostics
#ifdef __cplusplus
#include <cmath>
#else
#include <math.h>
typedef _Bool bool;
#endif
#ifdef _OPENMP
#define __device__
#endif
// static and overloadable to get the same function annotations between C, C++ and HIP.
#define TEST_FUNC_ATTRS static __device__ __attribute__((used,overloadable))
#ifdef _OPENMP
#pragma omp begin declare target
#endif
// This function is a hack to get the same IR out of HIP and OpenMP. The HIP
// headers declare __cxa_* functions with these attributes, such that the
// attribute groups are different. update_cc_test_checks isn't smart enough to
// strip attributes from the checks, or semantically compare them.
__attribute__((__visibility__("default")))
__attribute__((weak))
__attribute__((noreturn))
__attribute__((overloadable))
// OPENMP-LABEL: define weak hidden void @_Z20attribute_check_hackv
// OPENMP-SAME: () #[[ATTR0:[0-9]+]] {
// OPENMP-NEXT: entry:
// OPENMP-NEXT: call void @llvm.trap()
// OPENMP-NEXT: unreachable
//
void attribute_check_hack(void) {
__builtin_trap();
}
// CHECK-LABEL: define internal float @_ZL18test_ockl_acos_f32f
// CHECK-SAME: (float [[SRC:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[SRC_ADDR:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
// CHECK-NEXT: [[SRC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SRC_ADDR]] to ptr
// CHECK-NEXT: store float [[SRC]], ptr [[SRC_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[SRC_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[CALL:%.*]] = call float @__ocml_acos_f32(float [[TMP0]]) #[[ATTR4:[0-9]+]]
// CHECK-NEXT: ret float [[CALL]]
//
TEST_FUNC_ATTRS float test_ockl_acos_f32(float src) {
return __ocml_acos_f32(src);
}
// CHECK-LABEL: define internal float @_ZL15test_ockl_fdot2Dv2_DF16_S_fbi
// CHECK-SAME: (<2 x half> [[A:%.*]], <2 x half> [[B:%.*]], float [[C:%.*]], i1 zeroext [[S:%.*]], i32 [[S_INT:%.*]]) #[[ATTR2]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca <2 x half>, align 4, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca <2 x half>, align 4, addrspace(5)
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[S_ADDR:%.*]] = alloca i8, align 1, addrspace(5)
// CHECK-NEXT: [[S_INT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[X:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[Y:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
// CHECK-NEXT: [[S_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[S_ADDR]] to ptr
// CHECK-NEXT: [[S_INT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[S_INT_ADDR]] to ptr
// CHECK-NEXT: [[X_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[X]] to ptr
// CHECK-NEXT: [[Y_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[Y]] to ptr
// CHECK-NEXT: store <2 x half> [[A]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: store <2 x half> [[B]], ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: store float [[C]], ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[FROMBOOL:%.*]] = zext i1 [[S]] to i8
// CHECK-NEXT: store i8 [[FROMBOOL]], ptr [[S_ADDR_ASCAST]], align 1
// CHECK-NEXT: store i32 [[S_INT]], ptr [[S_INT_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = load <2 x half>, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = load i8, ptr [[S_ADDR_ASCAST]], align 1
// CHECK-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
// CHECK-NEXT: [[CALL:%.*]] = call float @__ockl_fdot2(<2 x half> [[TMP0]], <2 x half> [[TMP1]], float [[TMP2]], i1 zeroext [[TOBOOL]]) #[[ATTR4]]
// CHECK-NEXT: store float [[CALL]], ptr [[X_ASCAST]], align 4
// CHECK-NEXT: [[TMP4:%.*]] = load <2 x half>, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP5:%.*]] = load <2 x half>, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[S_INT_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TOBOOL1:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK-NEXT: [[CALL2:%.*]] = call float @__ockl_fdot2(<2 x half> [[TMP4]], <2 x half> [[TMP5]], float [[TMP6]], i1 zeroext [[TOBOOL1]]) #[[ATTR4]]
// CHECK-NEXT: store float [[CALL2]], ptr [[Y_ASCAST]], align 4
// CHECK-NEXT: [[TMP8:%.*]] = load float, ptr [[X_ASCAST]], align 4
// CHECK-NEXT: [[TMP9:%.*]] = load float, ptr [[Y_ASCAST]], align 4
// CHECK-NEXT: [[ADD:%.*]] = fadd float [[TMP8]], [[TMP9]]
// CHECK-NEXT: ret float [[ADD]]
//
TEST_FUNC_ATTRS float test_ockl_fdot2(__2f16 a, __2f16 b, float c, bool s, int s_int) {
float x = __ockl_fdot2(a, b, c, s);
float y = __ockl_fdot2(a, b, c, s_int);
return x + y;
}
#ifdef _OPENMP
#pragma omp end declare target
#endif
//// NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
// HIP: {{.*}}
// OPENMP-C: {{.*}}
// OPENMP-CPP: {{.*}}
|
