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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
|
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "common.h"
#include "function_list.h"
#include "test_functions.h"
#include "utility.h"
#include <cinttypes>
#include <cstring>
namespace {
cl_int BuildKernelFn(cl_uint job_id, cl_uint thread_id UNUSED, void *p)
{
BuildKernelInfo &info = *(BuildKernelInfo *)p;
auto generator = [](const std::string &kernel_name, const char *builtin,
cl_uint vector_size_index) {
return GetUnaryKernel(kernel_name, builtin, ParameterType::Float,
ParameterType::UInt, vector_size_index);
};
return BuildKernels(info, job_id, generator);
}
} // anonymous namespace
int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
{
int error;
Programs programs;
KernelMatrix kernels;
const unsigned thread_id = 0; // Test is currently not multithreaded.
float maxError = 0.0f;
int ftz = f->ftz || gForceFTZ || 0 == (CL_FP_DENORM & gFloatCapabilities);
float maxErrorVal = 0.0f;
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
int scale = (int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(double)) + 1);
logFunctionInfo(f->name, sizeof(cl_float), relaxedMode);
float float_ulps;
if (gIsEmbedded)
float_ulps = f->float_embedded_ulps;
else
float_ulps = f->float_ulps;
// Init the kernels
BuildKernelInfo build_info{ 1, kernels, programs, f->nameInCode,
relaxedMode };
if ((error = ThreadPool_Do(BuildKernelFn,
gMaxVectorSizeIndex - gMinVectorSizeIndex,
&build_info)))
return error;
for (uint64_t i = 0; i < (1ULL << 32); i += step)
{
// Init input array
uint32_t *p = (uint32_t *)gIn;
if (gWimpyMode)
{
for (size_t j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j * scale;
}
else
{
for (size_t j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j;
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
}
// Write garbage into output arrays
for (auto j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
if (gHostFill)
{
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer[j],
CL_FALSE, 0, BUFFER_SIZE,
gOut[j], 0, NULL, NULL)))
{
vlog_error(
"\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
return error;
}
}
else
{
if ((error = clEnqueueFillBuffer(gQueue, gOutBuffer[j],
&pattern, sizeof(pattern), 0,
BUFFER_SIZE, 0, NULL, NULL)))
{
vlog_error("Error: clEnqueueFillBuffer failed! err: %d\n",
error);
return error;
}
}
}
// Run the kernels
for (auto j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_float);
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j][thread_id], 0,
sizeof(gOutBuffer[j]), &gOutBuffer[j])))
{
LogBuildError(programs[j]);
return error;
}
if ((error = clSetKernelArg(kernels[j][thread_id], 1,
sizeof(gInBuffer), &gInBuffer)))
{
LogBuildError(programs[j]);
return error;
}
if ((error = clEnqueueNDRangeKernel(gQueue, kernels[j][thread_id],
1, NULL, &localCount, NULL, 0,
NULL, NULL)))
{
vlog_error("FAILED -- could not execute kernel\n");
return error;
}
}
// Get that moving
if ((error = clFlush(gQueue))) vlog("clFlush failed\n");
// Calculate the correctly rounded reference result
float *r = (float *)gOut_Ref;
cl_uint *s = (cl_uint *)gIn;
for (size_t j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] = (float)f->func.f_u(s[j]);
// Read the data back
for (auto j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
return error;
}
}
if (gSkipCorrectnessTesting) break;
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
for (size_t j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (auto k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
uint32_t *q = (uint32_t *)(gOut[k]);
// If we aren't getting the correctly rounded result
if (t[j] != q[j])
{
float test = ((float *)q)[j];
double correct = f->func.f_u(s[j]);
float err = Ulp_Error(test, correct);
int fail = !(fabsf(err) <= float_ulps);
if (fail)
{
if (ftz || relaxedMode)
{
// retry per section 6.5.3.2
if (IsFloatResultSubnormal(correct, float_ulps))
{
fail = fail && (test != 0.0f);
if (!fail) err = 0.0f;
}
}
}
if (fabsf(err) > maxError)
{
maxError = fabsf(err);
maxErrorVal = s[j];
}
if (fail)
{
vlog_error(
"\n%s%s: %f ulp error at 0x%8.8x: *%a vs. %a\n",
f->name, sizeNames[k], err, ((uint32_t *)gIn)[j],
((float *)gOut_Ref)[j], test);
return -1;
}
}
}
}
if (0 == (i & 0x0fffffff))
{
if (gVerboseBruteForce)
{
vlog("base:%14" PRIu64 " step:%10" PRIu64
" bufferSize:%10d \n",
i, step, BUFFER_SIZE);
}
else
{
vlog(".");
}
fflush(stdout);
}
}
if (!gSkipCorrectnessTesting)
{
if (gWimpyMode)
vlog("Wimp pass");
else
vlog("passed");
vlog("\t%8.2f @ %a", maxError, maxErrorVal);
}
vlog("\n");
return CL_SUCCESS;
}
|
