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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
|
//TEST:SIMPLE(filecheck=CHECK_GLSL): -allow-glsl -stage compute -entry computeMain -target glsl -DTARGET_GLSL
//TEST:SIMPLE(filecheck=CHECK_SPV): -allow-glsl -stage compute -entry computeMain -target spirv -emit-spirv-directly -DTARGET_SPIRV
//TEST:SIMPLE(filecheck=CHECK_HLSL): -allow-glsl -stage compute -entry computeMain -target hlsl -DTARGET_HLSL
//TEST:SIMPLE(filecheck=CHECK_CUDA): -allow-glsl -stage compute -entry computeMain -target cuda -DTARGET_CUDA
// not testing cpp due to missing impl
//DISABLE_TEST:SIMPLE(filecheck=CHECK_CPP): -allow-glsl -stage compute -entry computeMain -target cpp -DTARGET_CPP
//TEST(compute, vulkan):COMPARE_COMPUTE(filecheck-buffer=BUF):-vk -compute -entry computeMain -allow-glsl
//TEST(compute, vulkan):COMPARE_COMPUTE(filecheck-buffer=BUF):-vk -compute -entry computeMain -allow-glsl -emit-spirv-directly
//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=BUF):-wgpu -compute -entry computeMain -allow-glsl -xslang -DWGPU
// Not testing because CI runners may not support Metal's intrinsics.
//DISABLE_TEST(compute):COMPARE_COMPUTE(filecheck-buffer=BUF):-metal -compute -entry computeMain -allow-glsl -xslang -DMETAL
#version 430
#if 1 \
&& !defined(TARGET_HLSL) \
&& !defined(TARGET_CUDA) \
&& !defined(WGPU) \
&& !defined(METAL)
// hlsl does not treat boolean types with subgroup.* as a logical operator
// cuda, wgsl, and metal is missing an implementation
#define TEST_when_logical_operators_are_implemented
#endif
//TEST_INPUT:ubuffer(data=[0 0], stride=4):out,name=outputBuffer
buffer MyBlockName2
{
uint data[];
} outputBuffer;
#define local_size_x_v 4
layout(local_size_x = local_size_x_v) in;
__generic<T : __BuiltinLogicalType>
bool test1Logical() {
return true
#if defined(TEST_when_logical_operators_are_implemented)
& subgroupInclusiveAnd(T(1)) == T(1)
& subgroupInclusiveOr(T(1)) == T(1)
& subgroupInclusiveXor(T(1)) == T(0)
#endif // #if defined(TEST_when_logical_operators_are_implemented)
;
}
__generic<T : __BuiltinLogicalType, let N : int>
bool testVLogical() {
typealias gvec = vector<T, N>;
return true
#if defined(TEST_when_logical_operators_are_implemented)
& subgroupInclusiveAnd(gvec(T(1))) == gvec(T(1))
& subgroupInclusiveOr(gvec(T(1))) == gvec(T(1))
& subgroupInclusiveXor(gvec(T(1))) == gvec(T(0))
#endif // #if defined(TEST_when_logical_operators_are_implemented)
;
}
bool testLogical() {
return true
& test1Logical<int>()
& testVLogical<int, 2>()
& testVLogical<int, 3>()
& testVLogical<int, 4>()
& test1Logical<int8_t>()
& testVLogical<int8_t, 2>()
& testVLogical<int8_t, 3>()
& testVLogical<int8_t, 4>()
& test1Logical<int16_t>()
& testVLogical<int16_t, 2>()
& testVLogical<int16_t, 3>()
& testVLogical<int16_t, 4>()
& test1Logical<int64_t>()
& testVLogical<int64_t, 2>()
& testVLogical<int64_t, 3>()
& testVLogical<int64_t, 4>()
& test1Logical<uint>()
& testVLogical<uint, 2>()
& testVLogical<uint, 3>()
& testVLogical<uint, 4>()
& test1Logical<uint8_t>()
& testVLogical<uint8_t, 2>()
& testVLogical<uint8_t, 3>()
& testVLogical<uint8_t, 4>()
& test1Logical<uint16_t>()
& testVLogical<uint16_t, 2>()
& testVLogical<uint16_t, 3>()
& testVLogical<uint16_t, 4>()
& test1Logical<uint64_t>()
& testVLogical<uint64_t, 2>()
& testVLogical<uint64_t, 3>()
& testVLogical<uint64_t, 4>()
& test1Logical<bool>()
& testVLogical<bool, 2>()
& testVLogical<bool, 3>()
& testVLogical<bool, 4>()
;
}
__generic<T : __BuiltinArithmeticType>
bool test1Arithmetic() {
return true
& subgroupInclusiveAdd(T(1)) == T(4)
& subgroupInclusiveMul(T(1)) == T(1)
// WGSL and Metal does not support inclusive min/max
#if !defined(WGPU) && !defined(METAL)
& subgroupInclusiveMin(T(1)) == T(1)
& subgroupInclusiveMax(T(1)) == T(1)
#endif
;
}
__generic<T : __BuiltinArithmeticType, let N : int>
bool testVArithmetic() {
typealias gvec = vector<T, N>;
return true
& subgroupInclusiveAdd(gvec(T(1))) == gvec(T(4))
// & subgroupInclusiveMul(gvec(T(1))) == gvec(T(1))
// WGSL and Metal does not support inclusive min/max
#if !defined(WGPU) && !defined(METAL)
& subgroupInclusiveMin(gvec(T(1))) == gvec(T(1))
& subgroupInclusiveMax(gvec(T(1))) == gvec(T(1))
#endif
;
}
bool testArithmetic() {
return true
& test1Arithmetic<float>()
& testVArithmetic<float, 2>()
// & testVArithmetic<float, 3>()
// & testVArithmetic<float, 4>()
// & test1Arithmetic<half>()
// & testVArithmetic<half, 2>()
// & testVArithmetic<half, 3>()
// & testVArithmetic<half, 4>()
// & test1Arithmetic<int>()
// & testVArithmetic<int, 2>()
// & testVArithmetic<int, 3>()
// & testVArithmetic<int, 4>()
// & test1Arithmetic<uint>()
// & testVArithmetic<uint, 2>()
// & testVArithmetic<uint, 3>()
// & testVArithmetic<uint, 4>()
// Disabled on WGPU and Metal as these built-in types are not supported as of time of writing.
#if !defined(WGPU) && !defined(METAL)
& test1Arithmetic<double>() // WARNING: intel GPU's lack FP64 support
& testVArithmetic<double, 2>()
& testVArithmetic<double, 3>()
& testVArithmetic<double, 4>()
& test1Arithmetic<uint8_t>()
& testVArithmetic<uint8_t, 2>()
& testVArithmetic<uint8_t, 3>()
& testVArithmetic<uint8_t, 4>()
& test1Arithmetic<uint64_t>()
& testVArithmetic<uint64_t, 2>()
& testVArithmetic<uint64_t, 3>()
& testVArithmetic<uint64_t, 4>()
& test1Arithmetic<int8_t>()
& testVArithmetic<int8_t, 2>()
& testVArithmetic<int8_t, 3>()
& testVArithmetic<int8_t, 4>()
& test1Arithmetic<int64_t>()
& testVArithmetic<int64_t, 2>()
& testVArithmetic<int64_t, 3>()
& testVArithmetic<int64_t, 4>()
#endif
// Disabled on WGPU as these built-in types are not supported as of time of writing.
#if !defined (WGPU)
& test1Arithmetic<uint16_t>()
& testVArithmetic<uint16_t, 2>()
& testVArithmetic<uint16_t, 3>()
& testVArithmetic<uint16_t, 4>()
& test1Arithmetic<int16_t>()
& testVArithmetic<int16_t, 2>()
& testVArithmetic<int16_t, 3>()
& testVArithmetic<int16_t, 4>()
#endif
;
}
[shader("compute")]
void computeMain()
{
bool res0 = true
// WGSL and Metal does not support bitwise inclusive intrinsics.
#if !defined(WGPU) && !defined(METAL)
& testLogical()
#endif
;
bool res1 = true
& testArithmetic()
;
if (gl_LocalInvocationID.x == 3) {
// seperate so if there is an erroneous error the "major"
// tests are issolated into 2 branches without polluting the
// file with a bunch of individual test values
outputBuffer.data[0] = res0;
outputBuffer.data[1] = res1;
}
// CHECK_GLSL: void main(
// CHECK_SPV: OpEntryPoint
// CHECK_HLSL: void computeMain(
// CHECK_CUDA: void computeMain(
// CHECK_CPP: void _computeMain(
// BUF: 1
// BUF-NEXT: 1
}
|
