1 //
2 // Copyright (c) 2017 The Khronos Group Inc.
3 //
4 // Licensed under the Apache License, Version 2.0 (the "License");
5 // you may not use this file except in compliance with the License.
6 // You may obtain a copy of the License at
7 //
8 // http://www.apache.org/licenses/LICENSE-2.0
9 //
10 // Unless required by applicable law or agreed to in writing, software
11 // distributed under the License is distributed on an "AS IS" BASIS,
12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 // See the License for the specific language governing permissions and
14 // limitations under the License.
15 //
16 #include "harness/compat.h"
17
18 #include <stdio.h>
19 #include <string.h>
20 #include <sys/types.h>
21 #include <sys/stat.h>
22
23 #include "procs.h"
24
25 const char *mix_kernel_code =
26 "__kernel void test_mix(__global float *srcA, __global float *srcB, __global float *srcC, __global float *dst)\n"
27 "{\n"
28 " int tid = get_global_id(0);\n"
29 "\n"
30 " dst[tid] = mix(srcA[tid], srcB[tid], srcC[tid]);\n"
31 "}\n";
32
33 #define MAX_ERR 1e-3
34
35 float
verify_mix(float * inptrA,float * inptrB,float * inptrC,float * outptr,int n)36 verify_mix(float *inptrA, float *inptrB, float *inptrC, float *outptr, int n)
37 {
38 float r, delta, max_err = 0.0f;
39 int i;
40
41 for (i=0; i<n; i++)
42 {
43 r = inptrA[i] + ((inptrB[i] - inptrA[i]) * inptrC[i]);
44 delta = fabsf(r - outptr[i]) / r;
45 if(delta > max_err) max_err = delta;
46 }
47 return max_err;
48 }
49
50 int
test_mix(cl_device_id device,cl_context context,cl_command_queue queue,int num_elements)51 test_mix(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
52 {
53 cl_mem streams[4];
54 cl_float *input_ptr[3], *output_ptr, *p;
55 cl_program program;
56 cl_kernel kernel;
57 void *values[4];
58 size_t lengths[1];
59 size_t threads[1];
60 float max_err;
61 int err;
62 int i;
63 MTdata d;
64
65 input_ptr[0] = (cl_float*)malloc(sizeof(cl_float) * num_elements);
66 input_ptr[1] = (cl_float*)malloc(sizeof(cl_float) * num_elements);
67 input_ptr[2] = (cl_float*)malloc(sizeof(cl_float) * num_elements);
68 output_ptr = (cl_float*)malloc(sizeof(cl_float) * num_elements);
69 streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE,
70 sizeof(cl_float) * num_elements, NULL, NULL);
71 if (!streams[0])
72 {
73 log_error("clCreateBuffer failed\n");
74 return -1;
75 }
76 streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE,
77 sizeof(cl_float) * num_elements, NULL, NULL);
78 if (!streams[1])
79 {
80 log_error("clCreateBuffer failed\n");
81 return -1;
82 }
83 streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE,
84 sizeof(cl_float) * num_elements, NULL, NULL);
85 if (!streams[2])
86 {
87 log_error("clCreateBuffer failed\n");
88 return -1;
89 }
90
91 streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE,
92 sizeof(cl_float) * num_elements, NULL, NULL);
93 if (!streams[3])
94 {
95 log_error("clCreateBuffer failed\n");
96 return -1;
97 }
98
99 p = input_ptr[0];
100 d = init_genrand( gRandomSeed );
101 for (i=0; i<num_elements; i++)
102 {
103 p[i] = (float) genrand_real1(d);
104 }
105 p = input_ptr[1];
106 for (i=0; i<num_elements; i++)
107 {
108 p[i] = (float) genrand_real1(d);
109 }
110 p = input_ptr[2];
111 for (i=0; i<num_elements; i++)
112 {
113 p[i] = (float) genrand_real1(d);
114 }
115 free_mtdata(d); d = NULL;
116
117 err = clEnqueueWriteBuffer( queue, streams[0], true, 0, sizeof(cl_float)*num_elements, (void *)input_ptr[0], 0, NULL, NULL );
118 if (err != CL_SUCCESS)
119 {
120 log_error("clWriteArray failed\n");
121 return -1;
122 }
123 err = clEnqueueWriteBuffer( queue, streams[1], true, 0, sizeof(cl_float)*num_elements, (void *)input_ptr[1], 0, NULL, NULL );
124 if (err != CL_SUCCESS)
125 {
126 log_error("clWriteArray failed\n");
127 return -1;
128 }
129 err = clEnqueueWriteBuffer( queue, streams[2], true, 0, sizeof(cl_float)*num_elements, (void *)input_ptr[2], 0, NULL, NULL );
130 if (err != CL_SUCCESS)
131 {
132 log_error("clWriteArray failed\n");
133 return -1;
134 }
135
136 lengths[0] = strlen(mix_kernel_code);
137 err = create_single_kernel_helper( context, &program, &kernel, 1, &mix_kernel_code, "test_mix" );
138 test_error( err, "Unable to create test kernel" );
139
140
141 values[0] = streams[0];
142 values[1] = streams[1];
143 values[2] = streams[2];
144 values[3] = streams[3];
145 err = clSetKernelArg(kernel, 0, sizeof streams[0], &streams[0] );
146 err |= clSetKernelArg(kernel, 1, sizeof streams[1], &streams[1] );
147 err |= clSetKernelArg(kernel, 2, sizeof streams[2], &streams[2] );
148 err |= clSetKernelArg(kernel, 3, sizeof streams[3], &streams[3] );
149 if (err != CL_SUCCESS)
150 {
151 log_error("clSetKernelArgs failed\n");
152 return -1;
153 }
154
155 threads[0] = (size_t)num_elements;
156 err = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, NULL, 0, NULL, NULL );
157 if (err != CL_SUCCESS)
158 {
159 log_error("clEnqueueNDRangeKernel failed\n");
160 return -1;
161 }
162
163 err = clEnqueueReadBuffer( queue, streams[3], true, 0, sizeof(cl_float)*num_elements, (void *)output_ptr, 0, NULL, NULL );
164 if (err != CL_SUCCESS)
165 {
166 log_error("clEnqueueReadBuffer failed\n");
167 return -1;
168 }
169
170 max_err = verify_mix(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
171 if (max_err > MAX_ERR)
172 {
173 log_error("MIX test failed %g max err\n", max_err);
174 err = -1;
175 }
176 else
177 {
178 log_info("MIX test passed %g max err\n", max_err);
179 err = 0;
180 }
181
182 clReleaseMemObject(streams[0]);
183 clReleaseMemObject(streams[1]);
184 clReleaseMemObject(streams[2]);
185 clReleaseMemObject(streams[3]);
186 clReleaseKernel(kernel);
187 clReleaseProgram(program);
188 free(input_ptr[0]);
189 free(input_ptr[1]);
190 free(input_ptr[2]);
191 free(output_ptr);
192
193 return err;
194 }
195
196
197
198
199
200