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 <stdlib.h>
20 #include <string.h>
21 #include <sys/types.h>
22 #include <sys/stat.h>
23
24
25 #include "procs.h"
26
27 const char *loop_kernel_code =
28 "__kernel void test_loop(__global int *src, __global int *loopindx, __global int *loopcnt, __global int *dst)\n"
29 "{\n"
30 " int tid = get_global_id(0);\n"
31 " int n = get_global_size(0);\n"
32 " int i, j;\n"
33 "\n"
34 " dst[tid] = 0;\n"
35 " for (i=0,j=loopindx[tid]; i<loopcnt[tid]; i++,j++)\n"
36 " {\n"
37 " if (j >= n)\n"
38 " j = 0;\n"
39 " dst[tid] += src[j];\n"
40 " }\n"
41 "\n"
42 "}\n";
43
44
45 int
verify_loop(int * inptr,int * loopindx,int * loopcnt,int * outptr,int n)46 verify_loop(int *inptr, int *loopindx, int *loopcnt, int *outptr, int n)
47 {
48 int r, i, j, k;
49
50 for (i=0; i<n; i++)
51 {
52 r = 0;
53 for (j=0,k=loopindx[i]; j<loopcnt[i]; j++,k++)
54 {
55 if (k >= n)
56 k = 0;
57 r += inptr[k];
58 }
59
60 if (r != outptr[i])
61 {
62 log_error("LOOP test failed: %d found, expected %d\n", outptr[i], r);
63 return -1;
64 }
65 }
66
67 log_info("LOOP test passed\n");
68 return 0;
69 }
70
test_loop(cl_device_id device,cl_context context,cl_command_queue queue,int num_elements)71 int test_loop(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
72 {
73 cl_mem streams[4];
74 cl_int *input_ptr, *loop_indx, *loop_cnt, *output_ptr;
75 cl_program program;
76 cl_kernel kernel;
77 size_t threads[1];
78 int err, i;
79
80 size_t length = sizeof(cl_int) * num_elements;
81 input_ptr = (cl_int*)malloc(length);
82 loop_indx = (cl_int*)malloc(length);
83 loop_cnt = (cl_int*)malloc(length);
84 output_ptr = (cl_int*)malloc(length);
85
86 streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
87 if (!streams[0])
88 {
89 log_error("clCreateBuffer failed\n");
90 return -1;
91 }
92 streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
93 if (!streams[1])
94 {
95 log_error("clCreateBuffer failed\n");
96 return -1;
97 }
98 streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
99 if (!streams[2])
100 {
101 log_error("clCreateBuffer failed\n");
102 return -1;
103 }
104 streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
105 if (!streams[3])
106 {
107 log_error("clCreateBuffer failed\n");
108 return -1;
109 }
110
111 MTdata d = init_genrand( gRandomSeed );
112 for (i=0; i<num_elements; i++)
113 {
114 input_ptr[i] = (int)genrand_int32(d);
115 loop_indx[i] = (int)get_random_float(0, num_elements-1, d);
116 loop_cnt[i] = (int)get_random_float(0, num_elements/32, d);
117 }
118 free_mtdata(d); d = NULL;
119
120 err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr, 0, NULL, NULL);
121 if (err != CL_SUCCESS)
122 {
123 log_error("clEnqueueWriteBuffer failed\n");
124 return -1;
125 }
126 err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, loop_indx, 0, NULL, NULL);
127 if (err != CL_SUCCESS)
128 {
129 log_error("clEnqueueWriteBuffer failed\n");
130 return -1;
131 }
132 err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, loop_cnt, 0, NULL, NULL);
133 if (err != CL_SUCCESS)
134 {
135 log_error("clEnqueueWriteBuffer failed\n");
136 return -1;
137 }
138
139 err = create_single_kernel_helper(context, &program, &kernel, 1, &loop_kernel_code, "test_loop" );
140 if (err)
141 return -1;
142
143 err = clSetKernelArg(kernel, 0, sizeof streams[0], &streams[0]);
144 err |= clSetKernelArg(kernel, 1, sizeof streams[1], &streams[1]);
145 err |= clSetKernelArg(kernel, 2, sizeof streams[2], &streams[2]);
146 err |= clSetKernelArg(kernel, 3, sizeof streams[3], &streams[3]);
147 if (err != CL_SUCCESS)
148 {
149 log_error("clSetKernelArgs failed\n");
150 return -1;
151 }
152
153 threads[0] = (unsigned int)num_elements;
154 err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, threads, NULL, 0, NULL, NULL);
155 if (err != CL_SUCCESS)
156 {
157 log_error("clEnqueueNDRangeKernel failed\n");
158 return -1;
159 }
160
161 err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
162 if (err != CL_SUCCESS)
163 {
164 log_error("clReadArray failed\n");
165 return -1;
166 }
167
168 err = verify_loop(input_ptr, loop_indx, loop_cnt, output_ptr, num_elements);
169
170 // cleanup
171 clReleaseMemObject(streams[0]);
172 clReleaseMemObject(streams[1]);
173 clReleaseMemObject(streams[2]);
174 clReleaseMemObject(streams[3]);
175 clReleaseKernel(kernel);
176 clReleaseProgram(program);
177 free(input_ptr);
178 free(loop_indx);
179 free(loop_cnt);
180 free(output_ptr);
181
182 return err;
183 }
184
185
186