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
|
/*
* render_opencl.c
*
* Created on: 26.01.2018
* Author: Superleo1810
*/
#include "render_opencl.h"
void init_opencl(config_t *cfg)
{
config_opencl = cfg;
output = (cl_uint *) malloc((config_opencl->width) * (config_opencl->height) * sizeof(cl_uchar4));
context = NULL;
width_cl = config_opencl->width; // Leave it in, cl needs different endianness
cl_platform_id platform_id;
cl_uint ret_num_devices;
cl_uint ret_num_platforms;
size_t device_list_size;
cl_program program;
size_t cl_src_sz;
clGetPlatformIDs(1, &platform_id, &ret_num_platforms);
clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_DEFAULT, 1, &device_id, &ret_num_devices);
context = clCreateContext(NULL, 1, &device_id, NULL, NULL, &ret);
ret = clGetContextInfo(context, CL_CONTEXT_DEVICES, 0, NULL, &device_list_size);
devices = (cl_device_id *)malloc(device_list_size);
num_devices = (cl_uint)(device_list_size/sizeof(cl_device_id));
clGetContextInfo(context, CL_CONTEXT_DEVICES, device_list_size, devices, NULL);
FILE *fp;
char *cl_src, *path, *flags = (char *)malloc(200 * sizeof(char));
flags[0] = '\0';
switch(config_opencl->config_opencl.fpu)
{
case OPENCL_FPU_32:
switch(config_opencl->config_opencl.set_func)
{
case SFUNC_JULIA:
path = "cl/julia32.cl";
break;
case SFUNC_MANDELBROT:
default:
path = "cl/mandelbrot32.cl";
break;
}
break;
case OPENCL_FPU_64:
switch(config_opencl->config_opencl.set_func)
{
case SFUNC_JULIA:
path = "cl/julia64.cl";
break;
case SFUNC_MANDELBROT:
default:
path = "cl/mandelbrot64.cl";
break;
}
int khrFP64 = 0;
int amdFP64 = 0;
for (cl_uint i = 0; i < num_devices; i++)
{
char deviceExtensions[8192];
ret = clGetDeviceInfo(devices[i], CL_DEVICE_EXTENSIONS,
sizeof(deviceExtensions), deviceExtensions, 0);
if (strstr(deviceExtensions, "cl_khr_fp64"))
{
khrFP64++;
}
else
{
if (strstr(deviceExtensions, "cl_amd_fp64"))
{
amdFP64++;
}
}
}
if (khrFP64 == num_devices)
{
flags = strcat(flags, "-D KHR_DP_EXTENSION ");
}
else if (amdFP64 == num_devices)
{
flags = strcat(flags, "");
}
break;
case OPENCL_FPU_128:
printf("128 bit precision not implemented yet\n");
break;
}
cl_src = (char *)malloc(MAX_SOURCE_SIZE * sizeof(char));
fp = fopen(path, "r");
cl_src_sz = fread(cl_src, 1, MAX_SOURCE_SIZE, fp);
fclose(fp);
for (cl_uint i = 0; i < num_devices; i++)
{
cl_command_queue_properties prop = 0;
// if (sampleArgs->timing)
// {
// prop |= CL_QUEUE_PROFILING_ENABLE;
// }
commandQueue[i] = clCreateCommandQueue(context, devices[i], prop, &ret);
outputBuffer[i] = clCreateBuffer(context,
CL_MEM_WRITE_ONLY | CL_MEM_ALLOC_HOST_PTR,
(sizeof(cl_uint) * (config_opencl->width)
* (config_opencl->height)) / num_devices, NULL, &ret);
}
if (config_opencl->config_opencl.fma)
{
flags = strcat(flags, "-D MUL_ADD=fma ");
}
else
{
flags = strcat(flags, "-D MUL_ADD=mad ");
}
printf("flags: %s\n", flags);
program = clCreateProgramWithSource(context, 1, (const char **)&cl_src, (const size_t *)&cl_src_sz, &ret);
ret = clBuildProgram(program, num_devices, devices, flags, NULL, NULL);
for (cl_uint i = 0; i < num_devices; i++)
{
kernel_vector[i] = clCreateKernel(program, "calculate", &ret);
}
}
void render_opencl(d64 x_min, d64 y_min, d64 x_max, d64 y_max)
{
cl_event events[MAX_DEVICES];
cl_int eventStatus = CL_QUEUED;
size_t globalThreads[1];
size_t localThreads[1];
size_t kernelWorkGroupSize;
cl_kernel kernel;
cl_double x_min_cl = (cl_double) x_min;
cl_double y_min_cl = (cl_double) y_min;
cl_double x_max_cl = (cl_double) x_max;
cl_double y_max_cl = (cl_double) y_max;
cl_double y_max_t;
cl_float y_max_t_f;
cl_double x_delta = ((x_max_cl - x_min_cl) / (cl_double) config_opencl->width);
cl_double y_delta = -((y_max_cl - y_min_cl) / (cl_double) config_opencl->height);
cl_float x_delta_f = (float) x_delta;
cl_float y_delta_f = (float) y_delta;
cl_float x_min_f = (float) x_min;
globalThreads[0] = ((config_opencl->width) * (config_opencl->height))
/ num_devices;
localThreads[0] = 256;
globalThreads[0] >>= 2;
for (cl_uint i = 0; i < num_devices; i++)
{
kernel = kernel_vector[i];
ret = clGetKernelWorkGroupInfo(kernel, devices[i],
CL_KERNEL_WORK_GROUP_SIZE, sizeof(size_t), &kernelWorkGroupSize,
0);
if ((cl_uint) (localThreads[0]) > kernelWorkGroupSize)
{
localThreads[0] = kernelWorkGroupSize;
}
y_max_t = (((y_min_cl + y_max_cl) / 2.0) + (y_max_cl - y_min_cl) / 2.0 - ((cl_double) i * (y_max_cl - y_min_cl)) / (cl_double) num_devices);
ret = clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *) &outputBuffer[i]);
y_max_t_f = (float) y_max_t;
//printf("x_delta: %f, y_delta: %f, x_delta_f: %f, y_delta_f: %f, x_min_f: %f, y_max_t: %f, y_max_t_f: %f\n", x_delta, y_delta, x_delta_f, y_delta_f, x_min_f, y_max_t, y_max_t_f);
switch (config_opencl->config_opencl.fpu)
{
case OPENCL_FPU_32: // lel
ret = clSetKernelArg(kernel, 1, sizeof(cl_float), (void *) &x_min_f);
ret = clSetKernelArg(kernel, 2, sizeof(cl_float), (void *) &y_max_t_f);
ret = clSetKernelArg(kernel, 3, sizeof(cl_float), (void *) &x_delta_f);
ret = clSetKernelArg(kernel, 4, sizeof(cl_float), (void *) &y_delta_f);
break;
case OPENCL_FPU_64:
ret = clSetKernelArg(kernel, 1, sizeof(cl_double), (void *) &x_min_cl);
ret = clSetKernelArg(kernel, 2, sizeof(cl_double), (void *) &y_max_t);
ret = clSetKernelArg(kernel, 3, sizeof(cl_double), (void *) &x_delta);
ret = clSetKernelArg(kernel, 4, sizeof(cl_double), (void *) &y_delta);
break;
case OPENCL_FPU_128:
break;
}
ret = clSetKernelArg(kernel, 5, sizeof(cl_uint),
(void *) &config_opencl->iterations);
ret = clSetKernelArg(kernel, 6, sizeof(cl_int),
(void *) &width_cl);
ret = clEnqueueNDRangeKernel(commandQueue[i], kernel, 1, NULL,
globalThreads, localThreads, 0, NULL, &events[i]);
}
for (cl_uint i = 0; i < num_devices; i++)
{
ret = clFlush(commandQueue[i]);
}
for (cl_uint i = 0; i < num_devices; i++)
{
ret = clWaitForEvents(1, &events[num_devices - i - 1]);
ret = clReleaseEvent(events[num_devices - i - 1]);
}
for (cl_uint i = 0; i < num_devices; i++)
{
ret = clEnqueueReadBuffer(commandQueue[i], outputBuffer[i],
CL_FALSE, 0,
(config_opencl->width * config_opencl->height * sizeof(cl_int))
/ num_devices,
config_opencl->arr
+ (config_opencl->width * config_opencl->height
/ num_devices) * i, 0,
NULL, &events[i]);
}
for (cl_uint i = 0; i < num_devices; i++)
{
ret = clFlush(commandQueue[i]);
}
for (cl_uint i = 0; i < num_devices; i++)
{
ret = clWaitForEvents(1, &events[num_devices - i - 1]);
ret = clReleaseEvent(events[num_devices - i - 1]);
}
}
void idle_opencl(void)
{
}
|
