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 "gl_headers.h"
17
18 #include "testBase.h"
19 #include "setup.h"
20 #include "harness/genericThread.h"
21
22 #ifndef GLsync
23 // For OpenGL before 3.2, we look for the ARB_sync extension and try to use that
24 #if !defined(_WIN32)
25 #include <inttypes.h>
26 #endif // !_WIN32
27 typedef int64_t GLint64;
28 typedef uint64_t GLuint64;
29 typedef struct __GLsync *GLsync;
30
31 typedef GLsync (*glFenceSyncPtr)(GLenum condition,GLbitfield flags);
32 glFenceSyncPtr glFenceSyncFunc;
33
34 typedef bool (*glIsSyncPtr)(GLsync sync);
35 glIsSyncPtr glIsSyncFunc;
36
37 typedef void (*glDeleteSyncPtr)(GLsync sync);
38 glDeleteSyncPtr glDeleteSyncFunc;
39
40 typedef GLenum (*glClientWaitSyncPtr)(GLsync sync,GLbitfield flags,GLuint64 timeout);
41 glClientWaitSyncPtr glClientWaitSyncFunc;
42
43 typedef void (*glWaitSyncPtr)(GLsync sync,GLbitfield flags,GLuint64 timeout);
44 glWaitSyncPtr glWaitSyncFunc;
45
46 typedef void (*glGetInteger64vPtr)(GLenum pname, GLint64 *params);
47 glGetInteger64vPtr glGetInteger64vFunc;
48
49 typedef void (*glGetSyncivPtr)(GLsync sync,GLenum pname,GLsizei bufSize,GLsizei *length,
50 GLint *values);
51 glGetSyncivPtr glGetSyncivFunc;
52
53 #define CHK_GL_ERR() printf("%s\n", gluErrorString(glGetError()))
54
InitSyncFns(void)55 static void InitSyncFns( void )
56 {
57 glFenceSyncFunc = (glFenceSyncPtr)glutGetProcAddress( "glFenceSync" );
58 glIsSyncFunc = (glIsSyncPtr)glutGetProcAddress( "glIsSync" );
59 glDeleteSyncFunc = (glDeleteSyncPtr)glutGetProcAddress( "glDeleteSync" );
60 glClientWaitSyncFunc = (glClientWaitSyncPtr)glutGetProcAddress( "glClientWaitSync" );
61 glWaitSyncFunc = (glWaitSyncPtr)glutGetProcAddress( "glWaitSync" );
62 glGetInteger64vFunc = (glGetInteger64vPtr)glutGetProcAddress( "glGetInteger64v" );
63 glGetSyncivFunc = (glGetSyncivPtr)glutGetProcAddress( "glGetSynciv" );
64 }
65
66 #define GL_MAX_SERVER_WAIT_TIMEOUT 0x9111
67
68 #define GL_OBJECT_TYPE 0x9112
69 #define GL_SYNC_CONDITION 0x9113
70 #define GL_SYNC_STATUS 0x9114
71 #define GL_SYNC_FLAGS 0x9115
72
73 #define GL_SYNC_FENCE 0x9116
74
75 #define GL_SYNC_GPU_COMMANDS_COMPLETE 0x9117
76
77 #define GL_UNSIGNALED 0x9118
78 #define GL_SIGNALED 0x9119
79
80 #define GL_SYNC_FLUSH_COMMANDS_BIT 0x00000001
81
82 #define GL_TIMEOUT_IGNORED 0xFFFFFFFFFFFFFFFFull
83
84 #define GL_ALREADY_SIGNALED 0x911A
85 #define GL_TIMEOUT_EXPIRED 0x911B
86 #define GL_CONDITION_SATISFIED 0x911C
87 #define GL_WAIT_FAILED 0x911D
88
89 #define USING_ARB_sync 1
90 #endif
91
92 typedef cl_event (CL_API_CALL *clCreateEventFromGLsyncKHR_fn)( cl_context context, GLsync sync, cl_int *errCode_ret) ;
93
94 clCreateEventFromGLsyncKHR_fn clCreateEventFromGLsyncKHR_ptr;
95
96
97 static const char *updateBuffersKernel[] = {
98 "__kernel void update( __global float4 * vertices, __global float4 *colors, int horizWrap, int rowIdx )\n"
99 "{\n"
100 " size_t tid = get_global_id(0);\n"
101 "\n"
102 " size_t xVal = ( tid & ( horizWrap - 1 ) );\n"
103 " vertices[ tid * 2 + 0 ] = (float4)( xVal, rowIdx*16.f, 0.0f, 1.f );\n"
104 " vertices[ tid * 2 + 1 ] = (float4)( xVal, rowIdx*16.f + 4.0f, 0.0f, 1.f );\n"
105 "\n"
106 " int rowV = rowIdx + 1;\n"
107 " colors[ tid * 2 + 0 ] = (float4)( ( rowV & 1 ) / 255.f, ( ( rowV & 2 ) >> 1 ) / 255.f, ( ( rowV & 4 ) >> 2 ) / 255.f, 1.f );\n"
108 " //colors[ tid * 2 + 0 ] = (float4)( (float)xVal/(float)horizWrap, 1.0f, 1.0f, 1.0f );\n"
109 " colors[ tid * 2 + 1 ] = colors[ tid * 2 + 0 ];\n"
110 "}\n" };
111
112 //Passthrough VertexShader
113 static const char vertexshader[] =
114 "uniform mat4 projMatrix;\n"
115 "attribute vec4 inPosition;\n"
116 "attribute vec4 inColor;\n"
117 "varying vec4 outColor;\n"
118 "void main (void) {\n"
119 " gl_Position = projMatrix*inPosition;\n"
120 " outColor = inColor;\n"
121 "}\n";
122
123 //Passthrough FragmentShader
124 static const char fragmentshader[] =
125 "varying vec4 outColor;\n"
126 "void main (void) {\n"
127 " gl_FragColor = outColor;\n"
128 "}\n";
129
createShaderProgram(GLint * posLoc,GLint * colLoc)130 GLuint createShaderProgram(GLint *posLoc, GLint *colLoc)
131 {
132 GLint logLength, status;
133 GLuint program = glCreateProgram();
134 GLuint vpShader;
135
136 char* vpstr = (char*)malloc(sizeof(vertexshader));
137 strcpy(vpstr, vertexshader);
138
139 vpShader = glCreateShader(GL_VERTEX_SHADER);
140 glShaderSource(vpShader, 1, (const GLchar **)&vpstr, NULL);
141 glCompileShader(vpShader);
142 glGetShaderiv(vpShader, GL_INFO_LOG_LENGTH, &logLength);
143 if (logLength > 0) {
144 GLchar *log = (GLchar*) malloc(logLength);
145 glGetShaderInfoLog(vpShader, logLength, &logLength, log);
146 log_info("Vtx Shader compile log:\n%s", log);
147 free(log);
148 }
149
150 free(vpstr);
151 vpstr = NULL;
152
153 glGetShaderiv(vpShader, GL_COMPILE_STATUS, &status);
154 if (status == 0)
155 {
156 log_error("Failed to compile vtx shader:\n");
157 return 0;
158 }
159
160 glAttachShader(program, vpShader);
161
162 GLuint fpShader;
163 char *fpstr = (char *)malloc(sizeof(fragmentshader));
164 strcpy(fpstr, fragmentshader);
165 fpShader = glCreateShader(GL_FRAGMENT_SHADER);
166 glShaderSource(fpShader, 1, (const GLchar **)&fpstr, NULL);
167 glCompileShader(fpShader);
168
169 free(fpstr);
170 fpstr = NULL;
171
172 glGetShaderiv(fpShader, GL_INFO_LOG_LENGTH, &logLength);
173 if (logLength > 0) {
174 GLchar *log = (GLchar*)malloc(logLength);
175 glGetShaderInfoLog(fpShader, logLength, &logLength, log);
176 log_info("Frag Shader compile log:\n%s", log);
177 free(log);
178 }
179
180 glAttachShader(program, fpShader);
181 glGetShaderiv(fpShader, GL_COMPILE_STATUS, &status);
182 if (status == 0)
183 {
184 log_error("Failed to compile frag shader:\n\n");
185 return 0;
186 }
187
188 glLinkProgram(program);
189 glGetProgramiv(program, GL_INFO_LOG_LENGTH, &logLength);
190 if (logLength > 0) {
191 GLchar *log = (GLchar*)malloc(logLength);
192 glGetProgramInfoLog(program, logLength, &logLength, log);
193 log_info("Program link log:\n%s", log);
194 free(log);
195 }
196
197 glGetProgramiv(program, GL_LINK_STATUS, &status);
198 if (status == 0)
199 {
200 log_error("Failed to link program\n");
201 return 0;
202 }
203
204 glValidateProgram(program);
205 glGetProgramiv(program, GL_INFO_LOG_LENGTH, &logLength);
206 if (logLength > 0) {
207 GLchar *log = (GLchar*)malloc(logLength);
208 glGetProgramInfoLog(program, logLength, &logLength, log);
209 log_info("Program validate log:\n%s", log);
210 free(log);
211 }
212
213 glGetProgramiv(program, GL_VALIDATE_STATUS, &status);
214 if (status == 0)
215 {
216 log_error("Failed to validate program\n");
217 return 0;
218 }
219
220 *posLoc = glGetAttribLocation(program, "inPosition");
221 *colLoc = glGetAttribLocation(program, "inColor");
222
223 return program;
224 }
225
destroyShaderProgram(GLuint program)226 void destroyShaderProgram(GLuint program)
227 {
228 GLuint shaders[2];
229 GLsizei count;
230 glUseProgram(0);
231 glGetAttachedShaders(program, 2, &count, shaders);
232 int i;
233 for(i = 0; i < count; i++)
234 {
235 glDetachShader(program, shaders[i]);
236 glDeleteShader(shaders[i]);
237 }
238 glDeleteProgram(program);
239 }
240
241 // This function queues up and runs the above CL kernel that writes the vertex data
run_cl_kernel(cl_kernel kernel,cl_command_queue queue,cl_mem stream0,cl_mem stream1,cl_int rowIdx,cl_event fenceEvent,size_t numThreads)242 cl_int run_cl_kernel( cl_kernel kernel, cl_command_queue queue, cl_mem stream0, cl_mem stream1,
243 cl_int rowIdx, cl_event fenceEvent, size_t numThreads )
244 {
245 cl_int error = clSetKernelArg( kernel, 3, sizeof( rowIdx ), &rowIdx );
246 test_error( error, "Unable to set kernel arguments" );
247
248 clEventWrapper acqEvent1, acqEvent2, kernEvent, relEvent1, relEvent2;
249 int numEvents = ( fenceEvent != NULL ) ? 1 : 0;
250 cl_event *fence_evt = ( fenceEvent != NULL ) ? &fenceEvent : NULL;
251
252 error = (*clEnqueueAcquireGLObjects_ptr)( queue, 1, &stream0, numEvents, fence_evt, &acqEvent1 );
253 test_error( error, "Unable to acquire GL obejcts");
254 error = (*clEnqueueAcquireGLObjects_ptr)( queue, 1, &stream1, numEvents, fence_evt, &acqEvent2 );
255 test_error( error, "Unable to acquire GL obejcts");
256
257 cl_event evts[ 2 ] = { acqEvent1, acqEvent2 };
258
259 error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, &numThreads, NULL, 2, evts, &kernEvent );
260 test_error( error, "Unable to execute test kernel" );
261
262 error = (*clEnqueueReleaseGLObjects_ptr)( queue, 1, &stream0, 1, &kernEvent, &relEvent1 );
263 test_error(error, "clEnqueueReleaseGLObjects failed");
264 error = (*clEnqueueReleaseGLObjects_ptr)( queue, 1, &stream1, 1, &kernEvent, &relEvent2 );
265 test_error(error, "clEnqueueReleaseGLObjects failed");
266
267 evts[ 0 ] = relEvent1;
268 evts[ 1 ] = relEvent2;
269 error = clWaitForEvents( 2, evts );
270 test_error( error, "Unable to wait for release events" );
271
272 return 0;
273 }
274
275 class RunThread : public genericThread
276 {
277 public:
278
279 cl_kernel mKernel;
280 cl_command_queue mQueue;
281 cl_mem mStream0, mStream1;
282 cl_int mRowIdx;
283 cl_event mFenceEvent;
284 size_t mNumThreads;
285
RunThread(cl_kernel kernel,cl_command_queue queue,cl_mem stream0,cl_mem stream1,size_t numThreads)286 RunThread( cl_kernel kernel, cl_command_queue queue, cl_mem stream0, cl_mem stream1, size_t numThreads )
287 : mKernel( kernel ), mQueue( queue ), mStream0( stream0 ), mStream1( stream1 ), mNumThreads( numThreads )
288 {
289 }
290
SetRunData(cl_int rowIdx,cl_event fenceEvent)291 void SetRunData( cl_int rowIdx, cl_event fenceEvent )
292 {
293 mRowIdx = rowIdx;
294 mFenceEvent = fenceEvent;
295 }
296
IRun(void)297 virtual void * IRun( void )
298 {
299 cl_int error = run_cl_kernel( mKernel, mQueue, mStream0, mStream1, mRowIdx, mFenceEvent, mNumThreads );
300 return (void *)(intptr_t)error;
301 }
302 };
303
304
test_fence_sync_single(cl_device_id device,cl_context context,cl_command_queue queue,bool separateThreads,GLint rend_vs,GLint read_vs,cl_device_id rend_device)305 int test_fence_sync_single( cl_device_id device, cl_context context, cl_command_queue queue, bool separateThreads, GLint rend_vs, GLint read_vs, cl_device_id rend_device )
306 {
307 int error;
308 const int framebufferSize = 512;
309 cl_platform_id platform_id = NULL;
310
311 if( !is_extension_available( device, "cl_khr_gl_event" ) )
312 {
313 log_info( "NOTE: cl_khr_gl_event extension not present on this device; skipping fence sync test\n" );
314 return 0;
315 }
316
317 error = clGetDeviceInfo(device,
318 CL_DEVICE_PLATFORM,
319 sizeof(platform_id),
320 &platform_id,
321 NULL);
322 if(error)
323 {
324 return error;
325 }
326
327 clCreateEventFromGLsyncKHR_ptr = \
328 (clCreateEventFromGLsyncKHR_fn)clGetExtensionFunctionAddressForPlatform(platform_id,"clCreateEventFromGLsyncKHR");
329 if( clCreateEventFromGLsyncKHR_ptr == NULL )
330 {
331 log_error( "ERROR: Unable to run fence_sync test (clCreateEventFromGLsyncKHR function not discovered!)\n" );
332 clCreateEventFromGLsyncKHR_ptr = \
333 (clCreateEventFromGLsyncKHR_fn)clGetExtensionFunctionAddressForPlatform(platform_id, "clCreateEventFromGLsyncAPPLE");
334 return -1;
335 }
336
337 #ifdef USING_ARB_sync
338 char *gl_version_str = (char*)glGetString( GL_VERSION );
339 float glCoreVersion;
340 sscanf(gl_version_str, "%f", &glCoreVersion);
341 if( glCoreVersion < 3.0f )
342 {
343 log_info( "OpenGL version %f does not support fence/sync! Skipping test.\n", glCoreVersion );
344 return 0;
345 }
346
347 #ifdef __APPLE__
348 CGLContextObj currCtx = CGLGetCurrentContext();
349 CGLPixelFormatObj pixFmt = CGLGetPixelFormat(currCtx);
350 GLint val, screen;
351 CGLGetVirtualScreen(currCtx, &screen);
352 CGLDescribePixelFormat(pixFmt, screen, kCGLPFAOpenGLProfile, &val);
353 if(val != kCGLOGLPVersion_3_2_Core)
354 {
355 log_error( "OpenGL context was not created with OpenGL version >= 3.0 profile even though platform supports it"
356 "OpenGL profile %f does not support fence/sync! Skipping test.\n", glCoreVersion );
357 return -1;
358 }
359 #else
360 // Need platform specific way to query if current GL context was created with 3.x profile
361 log_error( "ERROR: not implemented\n\n" );
362 return -1;
363 #endif
364
365 InitSyncFns();
366 #endif
367
368 #ifdef __APPLE__
369 CGLSetVirtualScreen(CGLGetCurrentContext(), rend_vs);
370 #else
371 // Need platform specific way to set device with id rend_vs the current
372 // rendering target
373 log_error( "ERROR: not implemented\n\n" );
374 return -1;
375 #endif
376
377 GLint posLoc, colLoc;
378 GLuint shaderprogram = createShaderProgram(&posLoc, &colLoc);
379 if(!shaderprogram)
380 {
381 log_error("Failed to create shader program\n");
382 return -1;
383 }
384
385 float l = 0.0f; float r = framebufferSize;
386 float b = 0.0f; float t = framebufferSize;
387
388 float projMatrix[16] = { 2.0f/(r-l), 0.0f, 0.0f, 0.0f,
389 0.0f, 2.0f/(t-b), 0.0f, 0.0f,
390 0.0f, 0.0f, -1.0f, 0.0f,
391 -(r+l)/(r-l), -(t+b)/(t-b), 0.0f, 1.0f
392 };
393
394 glUseProgram(shaderprogram);
395 GLuint projMatLoc = glGetUniformLocation(shaderprogram, "projMatrix");
396 glUniformMatrix4fv(projMatLoc, 1, 0, projMatrix);
397 glUseProgram(0);
398
399 // Note: the framebuffer is just the target to verify our results against, so we don't
400 // really care to go through all the possible formats in this case
401 glFramebufferWrapper glFramebuffer;
402 glRenderbufferWrapper glRenderbuffer;
403 error = CreateGLRenderbufferRaw( framebufferSize, 128, GL_COLOR_ATTACHMENT0_EXT,
404 GL_RGBA, GL_UNSIGNED_BYTE,
405 &glFramebuffer, &glRenderbuffer );
406 if( error != 0 )
407 return error;
408
409 // GLuint vao;
410 // glGenVertexArrays(1, &vao);
411 // glBindVertexArray(vao);
412
413 glBufferWrapper vtxBuffer, colorBuffer;
414 glGenBuffers( 1, &vtxBuffer );
415 glGenBuffers( 1, &colorBuffer );
416
417 const int numHorizVertices = ( framebufferSize * 64 ) + 1;
418
419 glBindBuffer( GL_ARRAY_BUFFER, vtxBuffer );
420 glBufferData( GL_ARRAY_BUFFER, sizeof( GLfloat ) * numHorizVertices * 2 * 4, NULL, GL_STATIC_DRAW );
421
422 glBindBuffer( GL_ARRAY_BUFFER, colorBuffer );
423 glBufferData( GL_ARRAY_BUFFER, sizeof( GLfloat ) * numHorizVertices * 2 * 4, NULL, GL_STATIC_DRAW );
424
425 clProgramWrapper program;
426 clKernelWrapper kernel;
427 clMemWrapper streams[ 2 ];
428
429 if( create_single_kernel_helper( context, &program, &kernel, 1, updateBuffersKernel, "update" ) )
430 return -1;
431
432
433 streams[ 0 ] = (*clCreateFromGLBuffer_ptr)( context, CL_MEM_READ_WRITE, vtxBuffer, &error );
434 test_error( error, "Unable to create CL buffer from GL vertex buffer" );
435
436 streams[ 1 ] = (*clCreateFromGLBuffer_ptr)( context, CL_MEM_READ_WRITE, colorBuffer, &error );
437 test_error( error, "Unable to create CL buffer from GL color buffer" );
438
439 error = clSetKernelArg( kernel, 0, sizeof( streams[ 0 ] ), &streams[ 0 ] );
440 test_error( error, "Unable to set kernel arguments" );
441
442 error = clSetKernelArg( kernel, 1, sizeof( streams[ 1 ] ), &streams[ 1 ] );
443 test_error( error, "Unable to set kernel arguments" );
444
445 cl_int horizWrap = (cl_int)framebufferSize;
446 error = clSetKernelArg( kernel, 2, sizeof( horizWrap ), &horizWrap );
447 test_error( error, "Unable to set kernel arguments" );
448
449 glViewport( 0, 0, framebufferSize, framebufferSize );
450 glClearColor( 0, 0, 0, 0 );
451 glClear( GL_COLOR_BUFFER_BIT );
452 glClear( GL_DEPTH_BUFFER_BIT );
453 glDisable( GL_DEPTH_TEST );
454 glEnable( GL_BLEND );
455 glBlendFunc( GL_ONE, GL_ONE );
456
457 clEventWrapper fenceEvent;
458 GLsync glFence = 0;
459
460 // Do a loop through 8 different horizontal stripes against the framebuffer
461 RunThread thread( kernel, queue, streams[ 0 ], streams[ 1 ], (size_t)numHorizVertices );
462
463 for( int i = 0; i < 8; i++ )
464 {
465 // if current rendering device is not the compute device and
466 // separateThreads == false which means compute is going on same
467 // thread and we are using implicit synchronization (no GLSync obj used)
468 // then glFlush by clEnqueueAcquireGLObject is not sufficient ... we need
469 // to wait for rendering to finish on other device before CL can start
470 // writing to CL/GL shared mem objects. When separateThreads is true i.e.
471 // we are using GLSync obj to synchronize then we dont need to call glFinish
472 // here since CL should wait for rendering on other device before this
473 // GLSync object to finish before it starts writing to shared mem object.
474 // Also rend_device == compute_device no need to call glFinish
475 if(rend_device != device && !separateThreads)
476 glFinish();
477
478 if( separateThreads )
479 {
480 thread.SetRunData( (cl_int)i, fenceEvent );
481 thread.Start();
482
483 error = (cl_int)(size_t)thread.Join();
484 }
485 else
486 {
487 error = run_cl_kernel( kernel, queue, streams[ 0 ], streams[ 1 ], (cl_int)i, fenceEvent, (size_t)numHorizVertices );
488 }
489 test_error( error, "Unable to run CL kernel" );
490
491 glUseProgram(shaderprogram);
492 glEnableVertexAttribArray(posLoc);
493 glEnableVertexAttribArray(colLoc);
494 glBindBuffer( GL_ARRAY_BUFFER, vtxBuffer );
495 glVertexAttribPointer(posLoc, 4, GL_FLOAT, GL_FALSE, 4*sizeof(GLfloat), 0);
496 glBindBuffer( GL_ARRAY_BUFFER, colorBuffer );
497 glVertexAttribPointer(colLoc, 4, GL_FLOAT, GL_FALSE, 4*sizeof(GLfloat), 0);
498 glBindBuffer( GL_ARRAY_BUFFER, 0 );
499
500 glDrawArrays( GL_TRIANGLE_STRIP, 0, numHorizVertices * 2 );
501
502 glDisableVertexAttribArray(posLoc);
503 glDisableVertexAttribArray(colLoc);
504 glUseProgram(0);
505
506 if( separateThreads )
507 {
508 // If we're on the same thread, then we're testing implicit syncing, so we
509 // don't need the actual fence code
510 if( fenceEvent != NULL )
511 {
512 clReleaseEvent( fenceEvent );
513 glDeleteSyncFunc( glFence );
514 }
515
516 glFence = glFenceSyncFunc( GL_SYNC_GPU_COMMANDS_COMPLETE, 0 );
517 fenceEvent = clCreateEventFromGLsyncKHR_ptr( context, glFence, &error );
518 test_error( error, "Unable to create CL event from GL fence" );
519
520 // in case of explicit synchronization, we just wait for the sync object to complete
521 // in clEnqueueAcquireGLObject but we dont flush. Its application's responsibility
522 // to flush on the context on which glSync is created
523 glFlush();
524 }
525 }
526
527 if( glFence != 0 )
528 // Don't need the final release for fenceEvent, because the wrapper will take care of that
529 glDeleteSyncFunc( glFence );
530
531 #ifdef __APPLE__
532 CGLSetVirtualScreen(CGLGetCurrentContext(), read_vs);
533 #else
534 // Need platform specific code to set the current rendering device (OpenGL target)
535 // to device with id read_vs so that next glReadPixels get submitted to that device
536 log_error( "ERROR: not implemented\n\n" );
537 return -1;
538 #endif
539 // Grab the contents of the final framebuffer
540 BufferOwningPtr<char> resultData( ReadGLRenderbuffer( glFramebuffer, glRenderbuffer,
541 GL_COLOR_ATTACHMENT0_EXT,
542 GL_RGBA8_OES, GL_UNSIGNED_BYTE, GL_RGBA, GL_UNSIGNED_BYTE, kUChar,
543 framebufferSize, 128 ) );
544
545 // Check the contents now. We should end up with solid color bands 32 pixels high and the
546 // full width of the framebuffer, at values (128,128,128) due to the additive blending
547 for( int i = 0; i < 8; i++ )
548 {
549 for( int y = 0; y < 4; y++ )
550 {
551 // Note: coverage will be double because the 63-0 triangle overwrites again at the end of the pass
552 cl_uchar valA = ( ( ( i + 1 ) & 1 ) ) * numHorizVertices * 2 / framebufferSize;
553 cl_uchar valB = ( ( ( i + 1 ) & 2 ) >> 1 ) * numHorizVertices * 2 / framebufferSize;
554 cl_uchar valC = ( ( ( i + 1 ) & 4 ) >> 2 ) * numHorizVertices * 2 / framebufferSize;
555
556 cl_uchar *row = (cl_uchar *)&resultData[ ( i * 16 + y ) * framebufferSize * 4 ];
557 for( int x = 0; x < ( framebufferSize - 1 ) - 1; x++ )
558 {
559 if( ( row[ x * 4 ] != valA ) || ( row[ x * 4 + 1 ] != valB ) ||
560 ( row[ x * 4 + 2 ] != valC ) )
561 {
562 log_error( "ERROR: Output framebuffer did not validate!\n" );
563 DumpGLBuffer( GL_UNSIGNED_BYTE, framebufferSize, 128, resultData );
564 log_error( "RUNS:\n" );
565 uint32_t *p = (uint32_t *)(char *)resultData;
566 size_t a = 0;
567 for( size_t t = 1; t < framebufferSize * framebufferSize; t++ )
568 {
569 if( p[ a ] != 0 )
570 {
571 if( p[ t ] == 0 )
572 {
573 log_error(
574 "RUN: %zu to %zu (%d,%d to %d,%d) 0x%08x\n",
575 a, t - 1, (int)(a % framebufferSize),
576 (int)(a / framebufferSize),
577 (int)((t - 1) % framebufferSize),
578 (int)((t - 1) / framebufferSize), p[a]);
579 a = t;
580 }
581 }
582 else
583 {
584 if( p[ t ] != 0 )
585 {
586 a = t;
587 }
588 }
589
590 }
591 return -1;
592 }
593 }
594 }
595 }
596
597 glDeleteBuffers( 1, &vtxBuffer );
598 glDeleteBuffers( 1, &colorBuffer );
599 destroyShaderProgram(shaderprogram);
600 // glDeleteVertexArrays(1, &vao);
601 return 0;
602 }
603
test_fence_sync(cl_device_id device,cl_context context,cl_command_queue queue,int numElements)604 int test_fence_sync( cl_device_id device, cl_context context, cl_command_queue queue, int numElements )
605 {
606 GLint vs_count = 0;
607 cl_device_id *device_list = NULL;
608
609 if( !is_extension_available( device, "cl_khr_gl_event" ) )
610 {
611 log_info( "NOTE: cl_khr_gl_event extension not present on this device; skipping fence sync test\n" );
612 return 0;
613 }
614 #ifdef __APPLE__
615 CGLContextObj ctx = CGLGetCurrentContext();
616 CGLPixelFormatObj pix = CGLGetPixelFormat(ctx);
617 CGLError err = CGLDescribePixelFormat(pix, 0, kCGLPFAVirtualScreenCount, &vs_count);
618
619 device_list = (cl_device_id *) malloc(sizeof(cl_device_id)*vs_count);
620 clGetGLContextInfoAPPLE(context, ctx, CL_CGL_DEVICES_FOR_SUPPORTED_VIRTUAL_SCREENS_APPLE, sizeof(cl_device_id)*vs_count, device_list, NULL);
621 #else
622 // Need platform specific way of getting devices from CL context to which OpenGL can render
623 // If not available it can be replaced with clGetContextInfo with CL_CONTEXT_DEVICES
624 log_error( "ERROR: not implemented\n\n" );
625 return -1;
626 #endif
627
628 GLint rend_vs, read_vs;
629 int error = 0;
630 int any_failed = 0;
631
632 // Loop through all the devices capable to OpenGL rendering
633 // and set them as current rendering target
634 for(rend_vs = 0; rend_vs < vs_count; rend_vs++)
635 {
636 // Loop through all the devices and set them as current
637 // compute target
638 for(read_vs = 0; read_vs < vs_count; read_vs++)
639 {
640 cl_device_id rend_device = device_list[rend_vs], read_device = device_list[read_vs];
641 char rend_name[200], read_name[200];
642
643 clGetDeviceInfo(rend_device, CL_DEVICE_NAME, sizeof(rend_name), rend_name, NULL);
644 clGetDeviceInfo(read_device, CL_DEVICE_NAME, sizeof(read_name), read_name, NULL);
645
646 log_info("Rendering on: %s, read back on: %s\n", rend_name, read_name);
647 error = test_fence_sync_single( device, context, queue, false, rend_vs, read_vs, rend_device );
648 any_failed |= error;
649 if( error != 0 )
650 log_error( "ERROR: Implicit syncing with GL sync events failed!\n\n" );
651 else
652 log_info("Implicit syncing Passed\n");
653
654 error = test_fence_sync_single( device, context, queue, true, rend_vs, read_vs, rend_device );
655 any_failed |= error;
656 if( error != 0 )
657 log_error( "ERROR: Explicit syncing with GL sync events failed!\n\n" );
658 else
659 log_info("Explicit syncing Passed\n");
660 }
661 }
662
663 free(device_list);
664
665 return any_failed;
666 }
667