ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/OpenCL/OpenCL.pm
(Generate patch)

Comparing OpenCL/OpenCL.pm (file contents):
Revision 1.9 by root, Thu Nov 17 01:36:52 2011 UTC vs.
Revision 1.77 by root, Sat May 5 12:54:03 2012 UTC

18vendors) - usually there is only one. 18vendors) - usually there is only one.
19 19
20Each platform gives you access to a number of OpenCL::Device objects, e.g. 20Each platform gives you access to a number of OpenCL::Device objects, e.g.
21your graphics card. 21your graphics card.
22 22
23From a platform and some devices, you create an OpenCL::Context, which is 23From a platform and some device(s), you create an OpenCL::Context, which is
24a very central object in OpenCL: Once you have a context you can create 24a very central object in OpenCL: Once you have a context you can create
25most other objects: 25most other objects:
26 26
27OpenCL::Program objects, which store source code and, after building 27OpenCL::Program objects, which store source code and, after building for a
28("compiling and linking"), also binary programs. For each kernel function 28specific device ("compiling and linking"), also binary programs. For each
29in a program you can then create an OpenCL::Kernel object which represents 29kernel function in a program you can then create an OpenCL::Kernel object
30basically a function call with argument values. 30which represents basically a function call with argument values.
31 31
32OpenCL::Memory objects of various flavours: OpenCL::Buffers objects (flat 32OpenCL::Memory objects of various flavours: OpenCL::Buffer objects (flat
33memory areas, think array) and OpenCL::Image objects (think 2d or 3d 33memory areas, think arrays or structs) and OpenCL::Image objects (think 2d
34array) for bulk data and input and output for kernels. 34or 3d array) for bulk data and input and output for kernels.
35 35
36OpenCL::Sampler objects, which are kind of like texture filter modes in 36OpenCL::Sampler objects, which are kind of like texture filter modes in
37OpenGL. 37OpenGL.
38 38
39OpenCL::Queue objects - command queues, which allow you to submit memory 39OpenCL::Queue objects - command queues, which allow you to submit memory
43 43
44OpenCL::Event objects are used to signal when something is complete. 44OpenCL::Event objects are used to signal when something is complete.
45 45
46=head2 HELPFUL RESOURCES 46=head2 HELPFUL RESOURCES
47 47
48The OpenCL spec used to develop this module (1.2 spec was available, but 48The OpenCL specs used to develop this module - download these and keept
49no implementation was available to me :). 49hema round, they are required reference material:
50 50
51 http://www.khronos.org/registry/cl/specs/opencl-1.1.pdf 51 http://www.khronos.org/registry/cl/specs/opencl-1.1.pdf
52 http://www.khronos.org/registry/cl/specs/opencl-1.2.pdf
53 http://www.khronos.org/registry/cl/specs/opencl-1.2-extensions.pdf
52 54
53OpenCL manpages: 55OpenCL manpages:
54 56
55 http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/ 57 http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/
58 http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/
59
60If you are into UML class diagrams, the following diagram might help - if
61not, it will be mildly confusing (also, the class hierarchy of this module
62is much more fine-grained):
63
64 http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/classDiagram.html
65
66Here's a tutorial from AMD (very AMD-centric, too), not sure how useful it
67is, but at least it's free of charge:
68
69 http://developer.amd.com/zones/OpenCLZone/courses/Documents/Introduction_to_OpenCL_Programming%20Training_Guide%20%28201005%29.pdf
70
71And here's NVIDIA's OpenCL Best Practises Guide:
72
73 http://developer.download.nvidia.com/compute/cuda/3_2/toolkit/docs/OpenCL_Best_Practices_Guide.pdf
56 74
57=head1 BASIC WORKFLOW 75=head1 BASIC WORKFLOW
58 76
59To get something done, you basically have to do this once: 77To get something done, you basically have to do this once (refer to the
78examples below for actual code, this is just a high-level description):
60 79
61Find some platform (e.g. the first one) and some device (e.g. the first 80Find some platform (e.g. the first one) and some device(s) (e.g. the first
62device you can find), and create a context from those. 81device of the platform), and create a context from those.
63 82
64Create a command queue from your context, and program objects from your 83Create program objects from your OpenCL source code, then build (compile)
65OpenCL source code, build the programs. 84the programs for each device you want to run them on.
66 85
67Create kernel objects for all kernels you want to use. 86Create kernel objects for all kernels you want to use (surprisingly, these
87are not device-specific).
68 88
69Then, to execute stuff, you repeat this: 89Then, to execute stuff, you repeat these steps, possibly resuing or
90sharing some buffers:
70 91
71Create some input and output buffers from your context. Initialise the 92Create some input and output buffers from your context. Set these as
72input buffers with data. Set these as arguments to your kernel. 93arguments to your kernel.
94
95Enqueue buffer writes to initialise your input buffers (when not
96initialised at creation time).
73 97
74Enqueue the kernel execution. 98Enqueue the kernel execution.
75 99
76Enqueue buffer reads for your output buffer to read results. 100Enqueue buffer reads for your output buffer to read results.
77 101
78The next section shows how this can be done. 102=head1 OPENCL 1.1 VS. OPENCL 1.2
103
104This module supports both OpenCL version 1.1 and 1.2, although the OpenCL
1051.2 interface hasn't been tested much for lack of availability of an
106actual implementation.
107
108Every function or method in this manual page that interfaces to a
109particular OpenCL function has a link to the its C manual page.
110
111If the link contains a F<1.1>, then this function is an OpenCL 1.1
112function. Most but not all also exist in OpenCL 1.2, and this module
113tries to emulate the missing ones for you, when told to do so at
114compiletime. You cna check whether a function was removed in OpenCL 1.2 by
115replacing the F<1.1> component in the URL by F<1.2>.
116
117If the link contains a F<1.2>, then this is a OpenCL 1.2-only
118function. Even if the module was compiled with OpenCL 1.2 header files
119and has an 1.2 OpenCL library, calling such a function on a platform that
120doesn't implement 1.2 causes undefined behaviour, usually a crash (But
121this is not guaranteed).
122
123You can find out whether this module was compiled to prefer 1.1
124functionality by ooking at C<OpenCL::PREFER_1_1> - if it is true, then
1251.1 functions generally are implemented using 1.1 OpenCL functions. If it
126is false, then 1.1 functions missing from 1.2 are emulated by calling 1.2
127fucntions.
128
129This is a somewhat sorry state of affairs, but the Khronos group choose to
130make every release of OpenCL source and binary incompatible with previous
131releases.
79 132
80=head1 EXAMPLES 133=head1 EXAMPLES
81 134
82=head2 Enumerate all devices and get contexts for them. 135=head2 Enumerate all devices and get contexts for them.
83 136
137Best run this once to get a feel for the platforms and devices in your
138system.
139
84 for my $platform (OpenCL::platforms) { 140 for my $platform (OpenCL::platforms) {
85 warn $platform->info (OpenCL::PLATFORM_NAME); 141 printf "platform: %s\n", $platform->name;
86 warn $platform->info (OpenCL::PLATFORM_EXTENSIONS); 142 printf "extensions: %s\n", $platform->extensions;
87 for my $device ($platform->devices) { 143 for my $device ($platform->devices) {
88 warn $device->info (OpenCL::DEVICE_NAME); 144 printf "+ device: %s\n", $device->name;
89 my $ctx = $device->context_simple; 145 my $ctx = $platform->context (undef, [$device]);
90 # do stuff 146 # do stuff
91 } 147 }
92 } 148 }
93 149
94=head2 Get a useful context and a command queue. 150=head2 Get a useful context and a command queue.
95 151
96 my $dev = ((OpenCL::platforms)[0]->devices)[0]; 152This is a useful boilerplate for any OpenCL program that only wants to use
97 my $ctx = $dev->context_simple; 153one device,
98 my $queue = $ctx->queue ($dev); 154
155 my ($platform) = OpenCL::platforms; # find first platform
156 my ($dev) = $platform->devices; # find first device of platform
157 my $ctx = $platform->context (undef, [$dev]); # create context out of those
158 my $queue = $ctx->queue ($dev); # create a command queue for the device
99 159
100=head2 Print all supported image formats of a context. 160=head2 Print all supported image formats of a context.
101 161
162Best run this once for your context, to see whats available and how to
163gather information.
164
102 for my $type (OpenCL::MEM_OBJECT_IMAGE2D, OpenCL::MEM_OBJECT_IMAGE3D) { 165 for my $type (OpenCL::MEM_OBJECT_IMAGE2D, OpenCL::MEM_OBJECT_IMAGE3D) {
103 say "supported image formats for ", OpenCL::enum2str $type; 166 print "supported image formats for ", OpenCL::enum2str $type, "\n";
104 167
105 for my $f ($ctx->supported_image_formats (0, $type)) { 168 for my $f ($ctx->supported_image_formats (0, $type)) {
106 printf " %-10s %-20s\n", OpenCL::enum2str $f->[0], OpenCL::enum2str $f->[1]; 169 printf " %-10s %-20s\n", OpenCL::enum2str $f->[0], OpenCL::enum2str $f->[1];
107 } 170 }
108 } 171 }
110=head2 Create a buffer with some predefined data, read it back synchronously, 173=head2 Create a buffer with some predefined data, read it back synchronously,
111then asynchronously. 174then asynchronously.
112 175
113 my $buf = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, "helmut"); 176 my $buf = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, "helmut");
114 177
115 $queue->enqueue_read_buffer ($buf, 1, 1, 3, my $data); 178 $queue->read_buffer ($buf, 1, 1, 3, my $data);
116 warn $data; 179 print "$data\n";
117 180
118 my $ev = $queue->enqueue_read_buffer ($buf, 0, 1, 3, my $data); 181 my $ev = $queue->read_buffer ($buf, 0, 1, 3, my $data);
119 $ev->wait; 182 $ev->wait;
120 warn $data; 183 print "$data\n"; # prints "elm"
121 184
122=head2 Create and build a program, then create a kernel out of one of its 185=head2 Create and build a program, then create a kernel out of one of its
123functions. 186functions.
124 187
125 my $src = ' 188 my $src = '
126 __kernel void 189 kernel void
127 squareit (__global float *input, __global float *output) 190 squareit (global float *input, global float *output)
128 { 191 {
129 size_t id = get_global_id (0); 192 $id = get_global_id (0);
130 output [id] = input [id] * input [id]; 193 output [id] = input [id] * input [id];
131 } 194 }
132 '; 195 ';
133 196
134 my $prog = $ctx->program_with_source ($src); 197 my $prog = $ctx->build_program ($src);
135
136 eval { $prog->build ($dev); 1 }
137 or die $prog->build_info ($dev, OpenCL::PROGRAM_BUILD_LOG);
138
139 my $kernel = $prog->kernel ("squareit"); 198 my $kernel = $prog->kernel ("squareit");
140 199
141=head2 Create some input and output float buffers, then call squareit on them. 200=head2 Create some input and output float buffers, then call the
201'squareit' kernel on them.
142 202
143 my $input = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, pack "f*", 1, 2, 3, 4.5); 203 my $input = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, pack "f*", 1, 2, 3, 4.5);
144 my $output = $ctx->buffer (0, OpenCL::SIZEOF_FLOAT * 5); 204 my $output = $ctx->buffer (0, OpenCL::SIZEOF_FLOAT * 5);
145 205
146 # set buffer 206 # set buffer
147 $kernel->set_buffer (0, $input); 207 $kernel->set_buffer (0, $input);
148 $kernel->set_buffer (1, $output); 208 $kernel->set_buffer (1, $output);
149 209
150 # execute it for all 4 numbers 210 # execute it for all 4 numbers
151 $queue->enqueue_nd_range_kernel ($kernel, undef, [4], undef); 211 $queue->nd_range_kernel ($kernel, undef, [4], undef);
152 212
153 # enqueue a synchronous read 213 # enqueue a synchronous read
154 $queue->enqueue_read_buffer ($output, 1, 0, OpenCL::SIZEOF_FLOAT * 4, my $data); 214 $queue->read_buffer ($output, 1, 0, OpenCL::SIZEOF_FLOAT * 4, my $data);
155 215
156 # print the results: 216 # print the results:
157 say join ", ", unpack "f*", $data; 217 printf "%s\n", join ", ", unpack "f*", $data;
158 218
159=head2 The same enqueue operations as before, but assuming an out-of-order queue, 219=head2 The same enqueue operations as before, but assuming an out-of-order queue,
160showing off barriers. 220showing off barriers.
161 221
162 # execute it for all 4 numbers 222 # execute it for all 4 numbers
163 $queue->enqueue_nd_range_kernel ($kernel, undef, [4], undef); 223 $queue->nd_range_kernel ($kernel, undef, [4], undef);
164 224
165 # enqueue a barrier to ensure in-order execution 225 # enqueue a barrier to ensure in-order execution
166 $queue->enqueue_barrier; 226 $queue->barrier;
167 227
168 # enqueue an async read 228 # enqueue an async read
169 $queue->enqueue_read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data); 229 $queue->read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data);
170 230
171 # wait for all requests to finish 231 # wait for all requests to finish
172 $queue->finish; 232 $queue->finish;
173 233
174=head2 The same enqueue operations as before, but assuming an out-of-order queue, 234=head2 The same enqueue operations as before, but assuming an out-of-order queue,
175showing off event objects and wait lists. 235showing off event objects and wait lists.
176 236
177 # execute it for all 4 numbers 237 # execute it for all 4 numbers
178 my $ev = $queue->enqueue_nd_range_kernel ($kernel, undef, [4], undef); 238 my $ev = $queue->nd_range_kernel ($kernel, undef, [4], undef);
179 239
180 # enqueue an async read 240 # enqueue an async read
181 $ev = $queue->enqueue_read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data, $ev); 241 $ev = $queue->read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data, $ev);
182 242
183 # wait for the last event to complete 243 # wait for the last event to complete
184 $ev->wait; 244 $ev->wait;
185 245
246=head2 Use the OpenGL module to share a texture between OpenCL and OpenGL and draw some julia
247set flight effect.
248
249This is quite a long example to get you going - you can download it from
250L<http://cvs.schmorp.de/OpenCL/examples/juliaflight>.
251
252 use OpenGL ":all";
253 use OpenCL;
254
255 my $S = $ARGV[0] || 256; # window/texture size, smaller is faster
256
257 # open a window and create a gl texture
258 OpenGL::glpOpenWindow width => $S, height => $S;
259 my $texid = glGenTextures_p 1;
260 glBindTexture GL_TEXTURE_2D, $texid;
261 glTexImage2D_c GL_TEXTURE_2D, 0, GL_RGBA8, $S, $S, 0, GL_RGBA, GL_UNSIGNED_BYTE, 0;
262
263 # find and use the first opencl device that let's us get a shared opengl context
264 my $platform;
265 my $dev;
266 my $ctx;
267
268 for (OpenCL::platforms) {
269 $platform = $_;
270 for ($platform->devices) {
271 $dev = $_;
272 $ctx = $platform->context ([OpenCL::GLX_DISPLAY_KHR, undef, OpenCL::GL_CONTEXT_KHR, undef], [$dev])
273 and last;
274 }
275 }
276
277 $ctx
278 or die "cannot find suitable OpenCL device\n";
279
280 my $queue = $ctx->queue ($dev);
281
282 # now attach an opencl image2d object to the opengl texture
283 my $tex = $ctx->gl_texture2d (OpenCL::MEM_WRITE_ONLY, GL_TEXTURE_2D, 0, $texid);
284
285 # now the boring opencl code
286 my $src = <<EOF;
287 kernel void
288 juliatunnel (write_only image2d_t img, float time)
289 {
290 int2 xy = (int2)(get_global_id (0), get_global_id (1));
291 float2 p = convert_float2 (xy) / $S.f * 2.f - 1.f;
292
293 float2 m = (float2)(1.f, p.y) / fabs (p.x); // tunnel
294 m.x = fabs (fmod (m.x + time * 0.05f, 4.f) - 2.f);
295
296 float2 z = m;
297 float2 c = (float2)(sin (time * 0.01133f), cos (time * 0.02521f));
298
299 for (int i = 0; i < 25 && dot (z, z) < 4.f; ++i) // standard julia
300 z = (float2)(z.x * z.x - z.y * z.y, 2.f * z.x * z.y) + c;
301
302 float3 colour = (float3)(z.x, z.y, atan2 (z.y, z.x));
303 write_imagef (img, xy, (float4)(colour * p.x * p.x, 1.));
304 }
305 EOF
306
307 my $prog = $ctx->build_program ($src);
308 my $kernel = $prog->kernel ("juliatunnel");
309
310 # program compiled, kernel ready, now draw and loop
311
312 for (my $time; ; ++$time) {
313 # acquire objects from opengl
314 $queue->acquire_gl_objects ([$tex]);
315
316 # configure and run our kernel
317 $kernel->setf ("mf", $tex, $time*2); # mf = memory object, float
318 $queue->nd_range_kernel ($kernel, undef, [$S, $S], undef);
319
320 # release objects to opengl again
321 $queue->release_gl_objects ([$tex]);
322
323 # wait
324 $queue->finish;
325
326 # now draw the texture, the defaults should be all right
327 glTexParameterf GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST;
328
329 glEnable GL_TEXTURE_2D;
330 glBegin GL_QUADS;
331 glTexCoord2f 0, 1; glVertex3i -1, -1, -1;
332 glTexCoord2f 0, 0; glVertex3i 1, -1, -1;
333 glTexCoord2f 1, 0; glVertex3i 1, 1, -1;
334 glTexCoord2f 1, 1; glVertex3i -1, 1, -1;
335 glEnd;
336
337 glXSwapBuffers;
338
339 select undef, undef, undef, 1/60;
340 }
341
342=head2 How to modify the previous example to not rely on GL sharing.
343
344For those poor souls with only a sucky CPU OpenCL implementation, you
345currently have to read the image into some perl scalar, and then modify a
346texture or use glDrawPixels or so).
347
348First, when you don't need gl sharing, you can create the context much simpler:
349
350 $ctx = $platform->context (undef, [$dev])
351
352To use a texture, you would modify the above example by creating an
353OpenCL::Image manually instead of deriving it from a texture:
354
355 my $tex = $ctx->image2d (OpenCL::MEM_WRITE_ONLY, OpenCL::RGBA, OpenCL::UNORM_INT8, $S, $S);
356
357And in the draw loop, intead of acquire_gl_objects/release_gl_objects, you
358would read the image2d after the kernel has written it:
359
360 $queue->read_image ($tex, 0, 0, 0, 0, $S, $S, 1, 0, 0, my $data);
361
362And then you would upload the pixel data to the texture (or use glDrawPixels):
363
364 glTexSubImage2D_s GL_TEXTURE_2D, 0, 0, 0, $S, $S, GL_RGBA, GL_UNSIGNED_BYTE, $data;
365
366The fully modified example can be found at
367L<http://cvs.schmorp.de/OpenCL/examples/juliaflight-nosharing>.
368
369=head2 Julia sets look soooo 80ies.
370
371Then colour them differently, e.g. using orbit traps! Replace the loop and
372colour calculation from the previous examples by this:
373
374 float2 dm = (float2)(1.f, 1.f);
375
376 for (int i = 0; i < 25; ++i)
377 {
378 z = (float2)(z.x * z.x - z.y * z.y, 2.f * z.x * z.y) + c;
379 dm = fmin (dm, (float2)(fabs (dot (z, z) - 1.f), fabs (z.x - 1.f)));
380 }
381
382 float3 colour = (float3)(dm.x * dm.y, dm.x * dm.y, dm.x);
383
384Also try C<-10.f> instead of C<-1.f>.
385
186=head1 DOCUMENTATION 386=head1 DOCUMENTATION
187 387
188=head2 BASIC CONVENTIONS 388=head2 BASIC CONVENTIONS
189 389
190This is not a 1:1 C-style translation of OpenCL to Perl - instead I 390This is not a one-to-one C-style translation of OpenCL to Perl - instead
191attempted to make the interface as type-safe as possible and introducing 391I attempted to make the interface as type-safe as possible by introducing
192object syntax where it makes sense. There are a number of important 392object syntax where it makes sense. There are a number of important
193differences between the OpenCL C API and this module: 393differences between the OpenCL C API and this module:
194 394
195=over 4 395=over 4
196 396
197=item * Object lifetime managament is automatic - there is no need 397=item * Object lifetime managament is automatic - there is no need
198to free objects explicitly (C<clReleaseXXX>), the release function 398to free objects explicitly (C<clReleaseXXX>), the release function
199is called automatically once all Perl references to it go away. 399is called automatically once all Perl references to it go away.
200 400
201=item * OpenCL uses CamelCase for function names (C<clGetPlatformInfo>), 401=item * OpenCL uses CamelCase for function names
402(e.g. C<clGetPlatformIDs>, C<clGetPlatformInfo>), while this module
202while this module uses underscores as word separator and often leaves out 403uses underscores as word separator and often leaves out prefixes
203prefixes (C<< $platform->info >>). 404(C<OpenCL::platforms>, C<< $platform->info >>).
204 405
205=item * OpenCL often specifies fixed vector function arguments as short 406=item * OpenCL often specifies fixed vector function arguments as short
206arrays (C<size_t origin[3]>), while this module explicitly expects the 407arrays (C<size_t origin[3]>), while this module explicitly expects the
207components as separate arguments- 408components as separate arguments (C<$orig_x, $orig_y, $orig_z>) in
409function calls.
208 410
209=item * Where possible, one of the pitch values is calculated from the 411=item * Structures are often specified by flattening out their components
210perl scalar length and need not be specified. 412as with short vectors, and returned as arrayrefs.
211 413
212=item * When enqueuing commands, the wait list is specified by adding 414=item * When enqueuing commands, the wait list is specified by adding
213extra arguments to the function - anywhere a C<$wait_events...> argument 415extra arguments to the function - anywhere a C<$wait_events...> argument
214is documented this can be any number of event objects. 416is documented this can be any number of event objects. As an extsnion
417implemented by this module, C<undef> values will be ignored in the event
418list.
215 419
216=item * When enqueuing commands, if the enqueue method is called in void 420=item * When enqueuing commands, if the enqueue method is called in void
217context, no event is created. In all other contexts an event is returned 421context, no event is created. In all other contexts an event is returned
218by the method. 422by the method.
219 423
239 ulong IV - Q 443 ulong IV - Q
240 float NV float f 444 float NV float f
241 half IV ushort S 445 half IV ushort S
242 double NV double d 446 double NV double d
243 447
448=head2 GLX SUPPORT
449
450Due to the sad state that OpenGL support is in in Perl (mostly the OpenGL
451module, which has little to no documentation and has little to no support
452for glX), this module, as a special extension, treats context creation
453properties C<OpenCL::GLX_DISPLAY_KHR> and C<OpenCL::GL_CONTEXT_KHR>
454specially: If either or both of these are C<undef>, then the OpenCL
455module tries to dynamically resolve C<glXGetCurrentDisplay> and
456C<glXGetCurrentContext>, call these functions and use their return values
457instead.
458
459For this to work, the OpenGL library must be loaded, a GLX context must
460have been created and be made current, and C<dlsym> must be available and
461capable of finding the function via C<RTLD_DEFAULT>.
462
463=head2 EVENT SYSTEM
464
465OpenCL can generate a number of (potentially) asynchronous events, for
466example, after compiling a program, to signal a context-related error or,
467perhaps most important, to signal completion of queued jobs (by setting
468callbacks on OpenCL::Event objects).
469
470The OpenCL module converts all these callbacks into events - you can
471still register callbacks, but they are not executed when your OpenCL
472implementation calls the actual callback, but only later. Therefore, none
473of the limitations of OpenCL callbacks apply to the perl implementation:
474it is perfectly safe to make blocking operations from event callbacks, and
475enqueued operations don't need to be flushed.
476
477To facilitate this, this module maintains an event queue - each
478time an asynchronous event happens, it is queued, and perl will be
479interrupted. This is implemented via the L<Async::Interrupt> module. In
480addition, this module has L<AnyEvent> support, so it can seamlessly
481integrate itself into many event loops.
482
483Since L<Async::Interrupt> is a bit hard to understand, here are some case examples:
484
485=head3 Don't use callbacks.
486
487When your program never uses any callbacks, then there will never be any
488notifications you need to take care of, and therefore no need to worry
489about all this.
490
491You can achieve a great deal by explicitly waiting for events, or using
492barriers and flush calls. In many programs, there is no need at all to
493tinker with asynchronous events.
494
495=head3 Use AnyEvent
496
497This module automatically registers a watcher that invokes all outstanding
498event callbacks when AnyEvent is initialised (and block asynchronous
499interruptions). Using this mode of operations is the safest and most
500recommended one.
501
502To use this, simply use AnyEvent and this module normally, make sure you
503have an event loop running:
504
505 use Gtk2 -init;
506 use AnyEvent;
507
508 # initialise AnyEvent, by creating a watcher, or:
509 AnyEvent::detect;
510
511 my $e = $queue->marker;
512 $e->cb (sub {
513 warn "opencl is finished\n";
514 })
515
516 main Gtk2;
517
518Note that this module will not initialise AnyEvent for you. Before
519AnyEvent is initialised, the module will asynchronously interrupt perl
520instead. To avoid any surprises, it's best to explicitly initialise
521AnyEvent.
522
523You can temporarily enable asynchronous interruptions (see next paragraph)
524by calling C<$OpenCL::INTERRUPT->unblock> and disable them again by
525calling C<$OpenCL::INTERRUPT->block>.
526
527=head3 Let yourself be interrupted at any time
528
529This mode is the default unless AnyEvent is loaded and initialised. In
530this mode, OpenCL asynchronously interrupts a running perl program. The
531emphasis is on both I<asynchronously> and I<running> here.
532
533Asynchronously means that perl might execute your callbacks at any
534time. For example, in the following code (I<THAT YOU SHOULD NOT COPY>),
535the C<until> loop following the marker call will be interrupted by the
536callback:
537
538 my $e = $queue->marker;
539 my $flag;
540 $e->cb (sub { $flag = 1 });
541 1 until $flag;
542 # $flag is now 1
543
544The reason why you shouldn't blindly copy the above code is that
545busy waiting is a really really bad thing, and really really bad for
546performance.
547
548While at first this asynchronous business might look exciting, it can be
549really hard, because you need to be prepared for the callback code to be
550executed at any time, which limits the amount of things the callback code
551can do safely.
552
553This can be mitigated somewhat by using C<<
554$OpenCL::INTERRUPT->scope_block >> (see the L<Async::Interrupt>
555documentation for details).
556
557The other problem is that your program must be actively I<running> to be
558interrupted. When you calculate stuff, your program is running. When you
559hang in some C functions or other block execution (by calling C<sleep>,
560C<select>, running an event loop and so on), your program is waiting, not
561running.
562
563One way around that would be to attach a read watcher to your event loop,
564listening for events on C<< $OpenCL::INTERRUPT->pipe_fileno >>, using a
565dummy callback (C<sub { }>) to temporarily execute some perl code.
566
567That is then awfully close to using the built-in AnyEvent support above,
568though, so consider that one instead.
569
570=head3 Be creative
571
572OpenCL exports the L<Async::Interrupt> object it uses in the global
573variable C<$OpenCL::INTERRUPT>. You can configure it in any way you like.
574
575So if you want to feel like a real pro, err, wait, if you feel no risk
576menas no fun, you can experiment by implementing your own mode of
577operations.
578
579=cut
580
581package OpenCL;
582
583use common::sense;
584use Carp ();
585use Async::Interrupt ();
586
587our $POLL_FUNC; # set by XS
588
589BEGIN {
590 our $VERSION = '0.99';
591
592 require XSLoader;
593 XSLoader::load (__PACKAGE__, $VERSION);
594
595 @OpenCL::Platform::ISA =
596 @OpenCL::Device::ISA =
597 @OpenCL::Context::ISA =
598 @OpenCL::Queue::ISA =
599 @OpenCL::Memory::ISA =
600 @OpenCL::Sampler::ISA =
601 @OpenCL::Program::ISA =
602 @OpenCL::Kernel::ISA =
603 @OpenCL::Event::ISA = OpenCL::Object::;
604
605 @OpenCL::SubDevice::ISA = OpenCL::Device::;
606
607 @OpenCL::Buffer::ISA =
608 @OpenCL::Image::ISA = OpenCL::Memory::;
609
610 @OpenCL::BufferObj::ISA = OpenCL::Buffer::;
611
612 @OpenCL::Image2D::ISA =
613 @OpenCL::Image3D::ISA =
614 @OpenCL::Image2DArray::ISA =
615 @OpenCL::Image1D::ISA =
616 @OpenCL::Image1DArray::ISA =
617 @OpenCL::Image1DBuffer::ISA = OpenCL::Image::;
618
619 @OpenCL::UserEvent::ISA = OpenCL::Event::;
620
621 @OpenCL::MappedBuffer::ISA =
622 @OpenCL::MappedImage::ISA = OpenCL::Mapped::;
623}
624
244=head2 THE OpenCL PACKAGE 625=head2 THE OpenCL PACKAGE
245 626
246=over 4 627=over 4
247 628
248=item $int = OpenCL::errno 629=item $int = OpenCL::errno
249 630
250The last error returned by a function - it's only changed on errors. 631The last error returned by a function - it's only valid after an error occured
632and before calling another OpenCL function.
251 633
252=item $str = OpenCL::err2str $errval 634=item $str = OpenCL::err2str [$errval]
253 635
254Comverts an error value into a human readable string. 636Converts an error value into a human readable string. IF no error value is
637given, then the last error will be used (as returned by OpenCL::errno).
255 638
256=item $str = OpenCL::enum2str $enum 639=item $str = OpenCL::enum2str $enum
257 640
258Converts most enum values (inof parameter names, image format constants, 641Converts most enum values (of parameter names, image format constants,
259object types, addressing and filter modes, command types etc.) into a 642object types, addressing and filter modes, command types etc.) into a
260human readbale string. When confronted with some random integer it can be 643human readable string. When confronted with some random integer it can be
261very helpful to pass it through this function to maybe get some readable 644very helpful to pass it through this function to maybe get some readable
262string out of it. 645string out of it.
263 646
264=item @platforms = OpenCL::platforms 647=item @platforms = OpenCL::platforms
265 648
266Returns all available OpenCL::Platform objects. 649Returns all available OpenCL::Platform objects.
267 650
268L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html> 651L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html>
269 652
270=item $ctx = OpenCL::context_from_type_simple $type = OpenCL::DEVICE_TYPE_DEFAULT 653=item $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr
271 654
272Tries to create a context from a default device and platform - never worked for me. 655Tries to create a context from a default device and platform type - never worked for me.
656Consider using C<< $platform->context_from_type >> instead.
657
658type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, OpenCL::DEVICE_TYPE_GPU,
659OpenCL::DEVICE_TYPE_ACCELERATOR, OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL.
273 660
274L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html> 661L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html>
275 662
663=item $ctx = OpenCL::context $properties, \@devices, $callback->($err, $pvt) = $print_stderr)
664
665Create a new OpenCL::Context object using the given device object(s).
666Consider using C<< $platform->context >> instead.
667
668L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html>
669
276=item OpenCL::wait_for_events $wait_events... 670=item OpenCL::wait_for_events $wait_events...
277 671
278Waits for all events to complete. 672Waits for all events to complete.
279 673
280L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html> 674L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html>
281 675
676=item OpenCL::poll
677
678Checks if there are any outstanding events (see L<EVENT SYSTEM>) and
679invokes their callbacks.
680
681=item $OpenCL::INTERRUPT
682
683The L<Async::Interrupt> object used to signal asynchronous events (see
684L<EVENT SYSTEM>).
685
686=cut
687
688our $INTERRUPT = new Async::Interrupt c_cb => [$POLL_FUNC, 0];
689
690&_eq_initialise ($INTERRUPT->signal_func);
691
692=item $OpenCL::WATCHER
693
694The L<AnyEvent> watcher object used to watch for asynchronous events (see
695L<EVENT SYSTEM>). This variable is C<undef> until L<AnyEvent> has been
696loaded I<and> initialised (e.g. by calling C<AnyEvent::detect>).
697
698=cut
699
700our $WATCHER;
701
702sub _init_anyevent {
703 $INTERRUPT->block;
704 $WATCHER = AE::io ($INTERRUPT->pipe_fileno, 0, sub { $INTERRUPT->handle });
705}
706
707if (defined $AnyEvent::MODEL) {
708 _init_anyevent;
709} else {
710 push @AnyEvent::post_detect, \&_init_anyevent;
711}
712
282=back 713=back
283 714
715=head2 THE OpenCL::Object CLASS
716
717This is the base class for all objects in the OpenCL module. The only
718method it implements is the C<id> method, which is only useful if you want
719to interface to OpenCL on the C level.
720
721=over 4
722
723=item $iv = $obj->id
724
725OpenCL objects are represented by pointers or integers on the C level. If
726you want to interface to an OpenCL object directly on the C level, then
727you need this value, which is returned by this method. You should use an
728C<IV> type in your code and cast that to the correct type.
729
730=cut
731
732sub OpenCL::Object::id {
733 ref $_[0] eq "SCALAR"
734 ? ${ $_[0] }
735 : $_[0][0]
736}
737
738=back
739
284=head2 THE OpenCL::Platform CLASS 740=head2 THE OpenCL::Platform CLASS
285 741
286=over 4 742=over 4
287 743
744=item @devices = $platform->devices ($type = OpenCL::DEVICE_TYPE_ALL)
745
746Returns a list of matching OpenCL::Device objects.
747
748=item $ctx = $platform->context_from_type ($properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr)
749
750Tries to create a context. Never worked for me, and you need devices explicitly anyway.
751
752L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html>
753
754=item $ctx = $platform->context ($properties, \@devices, $callback->($err, $pvt) = $print_stderr)
755
756Create a new OpenCL::Context object using the given device object(s)- a
757CL_CONTEXT_PLATFORM property is supplied automatically.
758
759L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html>
760
288=item $packed_value = $platform->info ($name) 761=item $packed_value = $platform->info ($name)
289 762
290Calls C<clGetPlatformInfo> and returns the packed, raw value - for 763Calls C<clGetPlatformInfo> and returns the packed, raw value - for
291strings, this will be the string, for other values you probably need to 764strings, this will be the string (possibly including terminating \0), for
292use the correct C<unpack>. This might get improved in the future. Hopefully. 765other values you probably need to use the correct C<unpack>.
766
767It's best to avoid this method and use one of the following convenience
768wrappers.
293 769
294L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformInfo.html> 770L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformInfo.html>
295 771
296=item @devices = $platform->devices ($type = OpenCL::DEVICE_TYPE_ALL) 772=item $platform->unload_compiler
297 773
298Returns a list of matching OpenCL::Device objects. 774Attempts to unload the compiler for this platform, for endless
775profit. Does nothing on OpenCL 1.1.
299 776
300=item $ctx = $platform->context_from_type_simple ($type = OpenCL::DEVICE_TYPE_DEFAULT)
301
302Tries to create a context. Never worked for me.
303
304L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html> 777L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clUnloadPlatformCompiler.html>
778
779=for gengetinfo begin platform
780
781=item $string = $platform->profile
782
783Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_PROFILE> and returns the result.
784
785=item $string = $platform->version
786
787Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_VERSION> and returns the result.
788
789=item $string = $platform->name
790
791Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_NAME> and returns the result.
792
793=item $string = $platform->vendor
794
795Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_VENDOR> and returns the result.
796
797=item $string = $platform->extensions
798
799Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_EXTENSIONS> and returns the result.
800
801=for gengetinfo end platform
305 802
306=back 803=back
307 804
308=head2 THE OpenCL::Device CLASS 805=head2 THE OpenCL::Device CLASS
309 806
313 810
314See C<< $platform->info >> for details. 811See C<< $platform->info >> for details.
315 812
316L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetDeviceInfo.html> 813L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetDeviceInfo.html>
317 814
318=item $ctx = $device->context_simple 815=item @devices = $device->sub_devices (\@properties)
319 816
320Convenience function to create a new OpenCL::Context object. 817Creates OpencL::SubDevice objects by partitioning an existing device.
321 818
322L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html> 819L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateSubDevices.html>
820
821=for gengetinfo begin device
822
823=item $device_type = $device->type
824
825Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_TYPE> and returns the result.
826
827=item $uint = $device->vendor_id
828
829Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_VENDOR_ID> and returns the result.
830
831=item $uint = $device->max_compute_units
832
833Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_COMPUTE_UNITS> and returns the result.
834
835=item $uint = $device->max_work_item_dimensions
836
837Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WORK_ITEM_DIMENSIONS> and returns the result.
838
839=item $int = $device->max_work_group_size
840
841Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WORK_GROUP_SIZE> and returns the result.
842
843=item @ints = $device->max_work_item_sizes
844
845Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WORK_ITEM_SIZES> and returns the result.
846
847=item $uint = $device->preferred_vector_width_char
848
849Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_CHAR> and returns the result.
850
851=item $uint = $device->preferred_vector_width_short
852
853Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_SHORT> and returns the result.
854
855=item $uint = $device->preferred_vector_width_int
856
857Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_INT> and returns the result.
858
859=item $uint = $device->preferred_vector_width_long
860
861Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_LONG> and returns the result.
862
863=item $uint = $device->preferred_vector_width_float
864
865Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT> and returns the result.
866
867=item $uint = $device->preferred_vector_width_double
868
869Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE> and returns the result.
870
871=item $uint = $device->max_clock_frequency
872
873Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_CLOCK_FREQUENCY> and returns the result.
874
875=item $bitfield = $device->address_bits
876
877Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_ADDRESS_BITS> and returns the result.
878
879=item $uint = $device->max_read_image_args
880
881Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_READ_IMAGE_ARGS> and returns the result.
882
883=item $uint = $device->max_write_image_args
884
885Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WRITE_IMAGE_ARGS> and returns the result.
886
887=item $ulong = $device->max_mem_alloc_size
888
889Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_MEM_ALLOC_SIZE> and returns the result.
890
891=item $int = $device->image2d_max_width
892
893Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE2D_MAX_WIDTH> and returns the result.
894
895=item $int = $device->image2d_max_height
896
897Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE2D_MAX_HEIGHT> and returns the result.
898
899=item $int = $device->image3d_max_width
900
901Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE3D_MAX_WIDTH> and returns the result.
902
903=item $int = $device->image3d_max_height
904
905Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE3D_MAX_HEIGHT> and returns the result.
906
907=item $int = $device->image3d_max_depth
908
909Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE3D_MAX_DEPTH> and returns the result.
910
911=item $uint = $device->image_support
912
913Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE_SUPPORT> and returns the result.
914
915=item $int = $device->max_parameter_size
916
917Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_PARAMETER_SIZE> and returns the result.
918
919=item $uint = $device->max_samplers
920
921Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_SAMPLERS> and returns the result.
922
923=item $uint = $device->mem_base_addr_align
924
925Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MEM_BASE_ADDR_ALIGN> and returns the result.
926
927=item $uint = $device->min_data_type_align_size
928
929Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MIN_DATA_TYPE_ALIGN_SIZE> and returns the result.
930
931=item $device_fp_config = $device->single_fp_config
932
933Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_SINGLE_FP_CONFIG> and returns the result.
934
935=item $device_mem_cache_type = $device->global_mem_cache_type
936
937Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_CACHE_TYPE> and returns the result.
938
939=item $uint = $device->global_mem_cacheline_size
940
941Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_CACHELINE_SIZE> and returns the result.
942
943=item $ulong = $device->global_mem_cache_size
944
945Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_CACHE_SIZE> and returns the result.
946
947=item $ulong = $device->global_mem_size
948
949Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_SIZE> and returns the result.
950
951=item $ulong = $device->max_constant_buffer_size
952
953Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_CONSTANT_BUFFER_SIZE> and returns the result.
954
955=item $uint = $device->max_constant_args
956
957Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_CONSTANT_ARGS> and returns the result.
958
959=item $device_local_mem_type = $device->local_mem_type
960
961Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_LOCAL_MEM_TYPE> and returns the result.
962
963=item $ulong = $device->local_mem_size
964
965Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_LOCAL_MEM_SIZE> and returns the result.
966
967=item $boolean = $device->error_correction_support
968
969Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_ERROR_CORRECTION_SUPPORT> and returns the result.
970
971=item $int = $device->profiling_timer_resolution
972
973Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PROFILING_TIMER_RESOLUTION> and returns the result.
974
975=item $boolean = $device->endian_little
976
977Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_ENDIAN_LITTLE> and returns the result.
978
979=item $boolean = $device->available
980
981Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_AVAILABLE> and returns the result.
982
983=item $boolean = $device->compiler_available
984
985Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_COMPILER_AVAILABLE> and returns the result.
986
987=item $device_exec_capabilities = $device->execution_capabilities
988
989Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_EXECUTION_CAPABILITIES> and returns the result.
990
991=item $command_queue_properties = $device->properties
992
993Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_QUEUE_PROPERTIES> and returns the result.
994
995=item $ = $device->platform
996
997Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PLATFORM> and returns the result.
998
999=item $string = $device->name
1000
1001Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NAME> and returns the result.
1002
1003=item $string = $device->vendor
1004
1005Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_VENDOR> and returns the result.
1006
1007=item $string = $device->driver_version
1008
1009Calls C<clGetDeviceInfo> with C<OpenCL::DRIVER_VERSION> and returns the result.
1010
1011=item $string = $device->profile
1012
1013Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PROFILE> and returns the result.
1014
1015=item $string = $device->version
1016
1017Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_VERSION> and returns the result.
1018
1019=item $string = $device->extensions
1020
1021Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_EXTENSIONS> and returns the result.
1022
1023=item $uint = $device->preferred_vector_width_half
1024
1025Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_HALF> and returns the result.
1026
1027=item $uint = $device->native_vector_width_char
1028
1029Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_CHAR> and returns the result.
1030
1031=item $uint = $device->native_vector_width_short
1032
1033Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_SHORT> and returns the result.
1034
1035=item $uint = $device->native_vector_width_int
1036
1037Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_INT> and returns the result.
1038
1039=item $uint = $device->native_vector_width_long
1040
1041Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_LONG> and returns the result.
1042
1043=item $uint = $device->native_vector_width_float
1044
1045Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_FLOAT> and returns the result.
1046
1047=item $uint = $device->native_vector_width_double
1048
1049Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE> and returns the result.
1050
1051=item $uint = $device->native_vector_width_half
1052
1053Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_HALF> and returns the result.
1054
1055=item $device_fp_config = $device->double_fp_config
1056
1057Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_DOUBLE_FP_CONFIG> and returns the result.
1058
1059=item $device_fp_config = $device->half_fp_config
1060
1061Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_HALF_FP_CONFIG> and returns the result.
1062
1063=item $boolean = $device->host_unified_memory
1064
1065Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_HOST_UNIFIED_MEMORY> and returns the result.
1066
1067=item $device = $device->parent_device_ext
1068
1069Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PARENT_DEVICE_EXT> and returns the result.
1070
1071=item @device_partition_property_exts = $device->partition_types_ext
1072
1073Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PARTITION_TYPES_EXT> and returns the result.
1074
1075=item @device_partition_property_exts = $device->affinity_domains_ext
1076
1077Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_AFFINITY_DOMAINS_EXT> and returns the result.
1078
1079=item $uint = $device->reference_count_ext
1080
1081Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_REFERENCE_COUNT_EXT> and returns the result.
1082
1083=item @device_partition_property_exts = $device->partition_style_ext
1084
1085Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PARTITION_STYLE_EXT> and returns the result.
1086
1087=for gengetinfo end device
323 1088
324=back 1089=back
325 1090
326=head2 THE OpenCL::Context CLASS 1091=head2 THE OpenCL::Context CLASS
327 1092
1093An OpenCL::Context is basically a container, or manager, for a number of
1094devices of a platform. It is used to create all sorts of secondary objects
1095such as buffers, queues, programs and so on.
1096
1097All context creation functions and methods take a list of properties
1098(type-value pairs). All property values can be specified as integers -
1099some additionally support other types:
1100
328=over 4 1101=over 4
329 1102
330=item $packed_value = $ctx->info ($name) 1103=item OpenCL::CONTEXT_PLATFORM
331 1104
332See C<< $platform->info >> for details. 1105Also accepts OpenCL::Platform objects.
333 1106
334L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetContextInfo.html> 1107=item OpenCL::GLX_DISPLAY_KHR
1108
1109Also accepts C<undef>, in which case a deep and troubling hack is engaged
1110to find the current glx display (see L<GLX SUPPORT>).
1111
1112=item OpenCL::GL_CONTEXT_KHR
1113
1114Also accepts C<undef>, in which case a deep and troubling hack is engaged
1115to find the current glx context (see L<GLX SUPPORT>).
1116
1117=back
1118
1119=over 4
1120
1121=item $prog = $ctx->build_program ($program, $options = "")
1122
1123This convenience function tries to build the program on all devices in
1124the context. If the build fails, then the function will C<croak> with the
1125build log. Otherwise ti returns the program object.
1126
1127The C<$program> can either be a C<OpenCL::Program> object or a string
1128containing the program. In the latter case, a program objetc will be
1129created automatically.
1130
1131=cut
1132
1133sub OpenCL::Context::build_program {
1134 my ($self, $prog, $options) = @_;
1135
1136 $prog = $self->program_with_source ($prog)
1137 unless ref $prog;
1138
1139 eval { $prog->build (undef, $options); 1 }
1140 or errno == BUILD_PROGRAM_FAILURE
1141 or errno == INVALID_BINARY # workaround nvidia bug
1142 or Carp::croak "OpenCL::Context->build_program: " . err2str;
1143
1144 # we check status for all devices
1145 for my $dev ($self->devices) {
1146 $prog->build_status ($dev) == BUILD_SUCCESS
1147 or Carp::croak "Building OpenCL program for device '" . $dev->name . "' failed:\n"
1148 . $prog->build_log ($dev);
1149 }
1150
1151 $prog
1152}
335 1153
336=item $queue = $ctx->queue ($device, $properties) 1154=item $queue = $ctx->queue ($device, $properties)
337 1155
338Create a new OpenCL::Queue object from the context and the given device. 1156Create a new OpenCL::Queue object from the context and the given device.
339 1157
340L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateCommandQueue.html> 1158L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateCommandQueue.html>
341 1159
1160Example: create an out-of-order queue.
1161
1162 $queue = $ctx->queue ($device, OpenCL::QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE);
1163
342=item $ev = $ctx->user_event 1164=item $ev = $ctx->user_event
343 1165
344Creates a new OpenCL::UserEvent object. 1166Creates a new OpenCL::UserEvent object.
345 1167
346L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateUserEvent.html> 1168L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateUserEvent.html>
347 1169
348=item $buf = $ctx->buffer ($flags, $len) 1170=item $buf = $ctx->buffer ($flags, $len)
349 1171
350Creates a new OpenCL::Buffer object with the given flags and octet-size. 1172Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object with the
1173given flags and octet-size.
1174
1175flags: OpenCL::MEM_READ_WRITE, OpenCL::MEM_WRITE_ONLY, OpenCL::MEM_READ_ONLY,
1176OpenCL::MEM_USE_HOST_PTR, OpenCL::MEM_ALLOC_HOST_PTR, OpenCL::MEM_COPY_HOST_PTR,
1177OpenCL::MEM_HOST_WRITE_ONLY, OpenCL::MEM_HOST_READ_ONLY, OpenCL::MEM_HOST_NO_ACCESS.
351 1178
352L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateBuffer.html> 1179L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateBuffer.html>
353 1180
354=item $buf = $ctx->buffer_sv ($flags, $data) 1181=item $buf = $ctx->buffer_sv ($flags, $data)
355 1182
356Creates a new OpenCL::Buffer object and initialise it with the given data values. 1183Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object and
1184initialise it with the given data values.
357 1185
1186=item $img = $ctx->image ($self, $flags, $channel_order, $channel_type, $type, $width, $height, $depth = 0, $array_size = 0, $row_pitch = 0, $slice_pitch = 0, $num_mip_level = 0, $num_samples = 0, $*data = &PL_sv_undef)
1187
1188Creates a new OpenCL::Image object and optionally initialises it with
1189the given data values.
1190
1191channel_order: OpenCL::R, OpenCL::A, OpenCL::RG, OpenCL::RA, OpenCL::RGB,
1192OpenCL::RGBA, OpenCL::BGRA, OpenCL::ARGB, OpenCL::INTENSITY, OpenCL::LUMINANCE,
1193OpenCL::Rx, OpenCL::RGx, OpenCL::RGBx.
1194
1195channel_type: OpenCL::SNORM_INT8, OpenCL::SNORM_INT16, OpenCL::UNORM_INT8,
1196OpenCL::UNORM_INT16, OpenCL::UNORM_SHORT_565, OpenCL::UNORM_SHORT_555,
1197OpenCL::UNORM_INT_101010, OpenCL::SIGNED_INT8, OpenCL::SIGNED_INT16,
1198OpenCL::SIGNED_INT32, OpenCL::UNSIGNED_INT8, OpenCL::UNSIGNED_INT16,
1199OpenCL::UNSIGNED_INT32, OpenCL::HALF_FLOAT, OpenCL::FLOAT.
1200
1201
1202type: OpenCL::MEM_OBJECT_BUFFER, OpenCL::MEM_OBJECT_IMAGE2D,
1203OpenCL::MEM_OBJECT_IMAGE3D, OpenCL::MEM_OBJECT_IMAGE2D_ARRAY,
1204OpenCL::MEM_OBJECT_IMAGE1D, OpenCL::MEM_OBJECT_IMAGE1D_ARRAY,
1205OpenCL::MEM_OBJECT_IMAGE1D_BUFFER.
1206
1207L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateImage.html>
1208
358=item $img = $ctx->image2d ($flags, $channel_order, $channel_type, $width, $height, $data) 1209=item $img = $ctx->image2d ($flags, $channel_order, $channel_type, $width, $height, $row_pitch = 0, $data = undef)
359 1210
360Creates a new OpenCL::Image2D object and optionally initialises it with the given data values. 1211Creates a new OpenCL::Image2D object and optionally initialises it with
1212the given data values.
361 1213
362L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateImage2D.html> 1214L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateImage2D.html>
363 1215
364=item $img = $ctx->image3d ($flags, $channel_order, $channel_type, $width, $height, $depth, $slice_pitch, $data) 1216=item $img = $ctx->image3d ($flags, $channel_order, $channel_type, $width, $height, $depth, $row_pitch = 0, $slice_pitch = 0, $data = undef)
365 1217
366Creates a new OpenCL::Image3D object and optionally initialises it with the given data values. 1218Creates a new OpenCL::Image3D object and optionally initialises it with
1219the given data values.
367 1220
368L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateImage3D.html> 1221L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateImage3D.html>
1222
1223=item $buffer = $ctx->gl_buffer ($flags, $bufobj)
1224
1225Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object that refers to the given
1226OpenGL buffer object.
1227
1228flags: OpenCL::MEM_READ_WRITE, OpenCL::MEM_READ_ONLY, OpenCL::MEM_WRITE_ONLY.
1229
1230http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLBuffer.html
1231
1232=item $img = $ctx->gl_texture ($flags, $target, $miplevel, $texture)
1233
1234Creates a new OpenCL::Image object that refers to the given OpenGL
1235texture object or buffer.
1236
1237target: GL_TEXTURE_1D, GL_TEXTURE_1D_ARRAY, GL_TEXTURE_BUFFER,
1238GL_TEXTURE_2D, GL_TEXTURE_2D_ARRAY, GL_TEXTURE_3D,
1239GL_TEXTURE_CUBE_MAP_POSITIVE_X, GL_TEXTURE_CUBE_MAP_POSITIVE_Y,
1240GL_TEXTURE_CUBE_MAP_POSITIVE_Z, GL_TEXTURE_CUBE_MAP_NEGATIVE_X,
1241GL_TEXTURE_CUBE_MAP_NEGATIVE_Y, GL_TEXTURE_CUBE_MAP_NEGATIVE_Z,
1242GL_TEXTURE_RECTANGLE/GL_TEXTURE_RECTANGLE_ARB.
1243
1244http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateFromGLTexture.html
1245
1246=item $img = $ctx->gl_texture2d ($flags, $target, $miplevel, $texture)
1247
1248Creates a new OpenCL::Image2D object that refers to the given OpenGL
12492D texture object.
1250
1251http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture2D.html
1252
1253=item $img = $ctx->gl_texture3d ($flags, $target, $miplevel, $texture)
1254
1255Creates a new OpenCL::Image3D object that refers to the given OpenGL
12563D texture object.
1257
1258http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture3D.html
1259
1260=item $ctx->gl_renderbuffer ($flags, $renderbuffer)
1261
1262Creates a new OpenCL::Image2D object that refers to the given OpenGL
1263render buffer.
1264
1265http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLRenderbuffer.html
369 1266
370=item @formats = $ctx->supported_image_formats ($flags, $image_type) 1267=item @formats = $ctx->supported_image_formats ($flags, $image_type)
371 1268
372Returns a list of matching image formats - each format is an arrayref with 1269Returns a list of matching image formats - each format is an arrayref with
373two values, $channel_order and $channel_type, in it. 1270two values, $channel_order and $channel_type, in it.
376 1273
377=item $sampler = $ctx->sampler ($normalized_coords, $addressing_mode, $filter_mode) 1274=item $sampler = $ctx->sampler ($normalized_coords, $addressing_mode, $filter_mode)
378 1275
379Creates a new OpenCL::Sampler object. 1276Creates a new OpenCL::Sampler object.
380 1277
1278addressing_mode: OpenCL::ADDRESS_NONE, OpenCL::ADDRESS_CLAMP_TO_EDGE,
1279OpenCL::ADDRESS_CLAMP, OpenCL::ADDRESS_REPEAT, OpenCL::ADDRESS_MIRRORED_REPEAT.
1280
1281filter_mode: OpenCL::FILTER_NEAREST, OpenCL::FILTER_LINEAR.
1282
381L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateSampler.html> 1283L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateSampler.html>
382 1284
383=item $program = $ctx->program_with_source ($string) 1285=item $program = $ctx->program_with_source ($string)
384 1286
385Creates a new OpenCL::Program object from the given source code. 1287Creates a new OpenCL::Program object from the given source code.
386 1288
387L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateProgramWithSource.html> 1289L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateProgramWithSource.html>
388 1290
1291=item ($program, \@status) = $ctx->program_with_binary (\@devices, \@binaries)
1292
1293Creates a new OpenCL::Program object from the given binaries.
1294
1295L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateProgramWithBinary.html>
1296
1297Example: clone an existing program object that contains a successfully
1298compiled program, no matter how useless this is.
1299
1300 my $clone = $ctx->program_with_binary ([$prog->devices], [$prog->binaries]);
1301
1302=item $program = $ctx->program_with_built_in_kernels (\@devices, $kernel_names)
1303
1304Creates a new OpenCL::Program object from the given built-in kernel names.
1305
1306L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateProgramWithBuiltInKernels.html>
1307
1308=item $program = $ctx->link_program (\@devices, $options, \@programs, $cb->($program) = undef)
1309
1310Links all (already compiled) program objects specified in C<@programs>
1311together and returns a new OpenCL::Program object with the result.
1312
1313L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clLinkProgram.html>
1314
1315=item $packed_value = $ctx->info ($name)
1316
1317See C<< $platform->info >> for details.
1318
1319L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetContextInfo.html>
1320
1321=for gengetinfo begin context
1322
1323=item $uint = $context->reference_count
1324
1325Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_REFERENCE_COUNT> and returns the result.
1326
1327=item @devices = $context->devices
1328
1329Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_DEVICES> and returns the result.
1330
1331=item @property_ints = $context->properties
1332
1333Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_PROPERTIES> and returns the result.
1334
1335=item $uint = $context->num_devices
1336
1337Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_NUM_DEVICES> and returns the result.
1338
1339=for gengetinfo end context
1340
389=back 1341=back
390 1342
391=head2 THE OpenCL::Queue CLASS 1343=head2 THE OpenCL::Queue CLASS
392 1344
393An OpenCL::Queue represents an execution queue for OpenCL. You execute 1345An OpenCL::Queue represents an execution queue for OpenCL. You execute
394requests by calling their respective C<enqueue_xxx> method and waitinf for 1346requests by calling their respective method and waiting for it to complete
395it to complete in some way. 1347in some way.
396 1348
397All the enqueue methods return an event object that can be used to wait 1349Most methods that enqueue some request return an event object that can
398for completion, unless the method is called in void context, in which case 1350be used to wait for completion (optionally using a callback), unless
399no event object is created. 1351the method is called in void context, in which case no event object is
1352created.
400 1353
401They also allow you to specify any number of other event objects that this 1354They also allow you to specify any number of other event objects that this
402request has to wait for before it starts executing, by simply passing the 1355request has to wait for before it starts executing, by simply passing the
403event objects as extra parameters to the enqueue methods. 1356event objects as extra parameters to the enqueue methods. To simplify
1357program design, this module ignores any C<undef> values in the list of
1358events. This makes it possible to code operations such as this, without
1359having to put a valid event object into C<$event> first:
1360
1361 $event = $queue->xxx (..., $event);
404 1362
405Queues execute in-order by default, without any parallelism, so in most 1363Queues execute in-order by default, without any parallelism, so in most
406cases (i.e. you use only one queue) it's not necessary to wait for or 1364cases (i.e. you use only one queue) it's not necessary to wait for or
407create event objects. 1365create event objects, althoguh an our of order queue is often a bit
1366faster.
408 1367
409=over 4 1368=over 4
410 1369
411=item $packed_value = $ctx->info ($name)
412
413See C<< $platform->info >> for details.
414
415L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetCommandQueueInfo.html>
416
417=item $ev = $queue->enqueue_read_buffer ($buffer, $blocking, $offset, $len, $data, $wait_events...) 1370=item $ev = $queue->read_buffer ($buffer, $blocking, $offset, $len, $data, $wait_events...)
418 1371
419Reads data from buffer into the given string. 1372Reads data from buffer into the given string.
420 1373
421L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBuffer.html> 1374L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBuffer.html>
422 1375
423=item $ev = $queue->enqueue_write_buffer ($buffer, $blocking, $offset, $data, $wait_events...) 1376=item $ev = $queue->write_buffer ($buffer, $blocking, $offset, $data, $wait_events...)
424 1377
425Writes data to buffer from the given string. 1378Writes data to buffer from the given string.
426 1379
427L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBuffer.html> 1380L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBuffer.html>
428 1381
429=item $ev = $queue->enqueue_copy_buffer ($src, $dst, $src_offset, $dst_offset, $len, $wait_events...) 1382=item $ev = $queue->copy_buffer ($src, $dst, $src_offset, $dst_offset, $len, $wait_events...)
430 1383
431L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBuffer.html> 1384L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBuffer.html>
432 1385
1386=item $ev = $queue->read_buffer_rect (OpenCL::Memory buf, cl_bool blocking, $buf_x, $buf_y, $buf_z, $host_x, $host_y, $host_z, $width, $height, $depth, $buf_row_pitch, $buf_slice_pitch, $host_row_pitch, $host_slice_pitch, $data, $wait_events...)
1387
1388http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html
1389
1390=item $ev = $queue->write_buffer_rect (OpenCL::Memory buf, cl_bool blocking, $buf_x, $buf_y, $buf_z, $host_x, $host_y, $host_z, $width, $height, $depth, $buf_row_pitch, $buf_slice_pitch, $host_row_pitch, $host_slice_pitch, $data, $wait_events...)
1391
1392http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html
1393
1394=item $ev = $queue->copy_buffer_to_image ($src_buffer, $dst_image, $src_offset, $dst_x, $dst_y, $dst_z, $width, $height, $depth, $wait_events...)
1395
1396L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>
1397
433=item $ev = $queue->enqueue_read_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...) 1398=item $ev = $queue->read_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...)
1399
1400C<$row_pitch> (and C<$slice_pitch>) can be C<0>, in which case the OpenCL
1401module uses the image width (and height) to supply default values.
434 1402
435L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadImage.html> 1403L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadImage.html>
436 1404
437=item $ev = $queue->enqueue_write_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $data, $wait_events...) 1405=item $ev = $queue->write_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...)
438 1406
1407C<$row_pitch> (and C<$slice_pitch>) can be C<0>, in which case the OpenCL
1408module uses the image width (and height) to supply default values.
439L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteImage.html> 1409L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteImage.html>
440 1410
1411=item $ev = $queue->copy_image ($src_image, $dst_image, $src_x, $src_y, $src_z, $dst_x, $dst_y, $dst_z, $width, $height, $depth, $wait_events...)
1412
1413L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyImage.html>
1414
1415=item $ev = $queue->copy_image_to_buffer ($src_image, $dst_image, $src_x, $src_y, $src_z, $width, $height, $depth, $dst_offset, $wait_events...)
1416
1417L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyImageToBuffer.html>
1418
441=item $ev = $queue->enqueue_copy_buffer_rect ($src, $dst, $src_x, $src_y, $src_z, $dst_x, $dst_y, $dst_z, $width, $height, $depth, $src_row_pitch, $src_slice_pitch, 4dst_row_pitch, $dst_slice_pitch, $ait_event...) 1419=item $ev = $queue->copy_buffer_rect ($src, $dst, $src_x, $src_y, $src_z, $dst_x, $dst_y, $dst_z, $width, $height, $depth, $src_row_pitch, $src_slice_pitch, $dst_row_pitch, $dst_slice_pitch, $wait_event...)
442 1420
443Yeah. 1421Yeah.
444 1422
445L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferRect.html>
446
447=item $ev = $queue->enqueue_copy_buffer_to_image (OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
448
449L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>. 1423L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>.
450 1424
451=item $ev = $queue->enqueue_copy_image (OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1425=item $ev = $queue->fill_buffer ($mem, $pattern, $offset, $size, ...)
452 1426
1427Fills the given buffer object with repeated applications of C<$pattern>,
1428starting at C<$offset> for C<$size> octets.
1429
1430L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueFillBuffer.html>
1431
1432=item $ev = $queue->fill_image ($img, $r, $g, $b, $a, $x, $y, $z, $width, $height, $depth, ...)
1433
1434Fills the given image area with the given rgba colour components. The
1435components are normally floating point values between C<0> and C<1>,
1436except when the image channel data type is a signe dor unsigned
1437unnormalised format, in which case the range is determined by the format.
1438
453L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyImage.html> 1439L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueFillImage.html>
454 1440
455=item $ev = $queue->enqueue_copy_image_to_buffer (OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
456
457L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyImageToBuffer.html>
458
459=item $ev = $queue->enqueue_task ($kernel, $wait_events...) 1441=item $ev = $queue->task ($kernel, $wait_events...)
460 1442
461L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueTask.html> 1443L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueTask.html>
462 1444
463=item $ev = $queue->enqueue_nd_range_kernel ($kernel, @$global_work_offset, @$global_work_size, @$local_work_size, $wait_events...) 1445=item $ev = $queue->nd_range_kernel ($kernel, \@global_work_offset, \@global_work_size, \@local_work_size, $wait_events...)
464 1446
465Enqueues a kernel execution. 1447Enqueues a kernel execution.
466 1448
467@$global_work_size must be specified as a reference to an array of 1449\@global_work_size must be specified as a reference to an array of
468integers specifying the work sizes (element counts). 1450integers specifying the work sizes (element counts).
469 1451
470@$global_work_offset must be either C<undef> (in which case all offsets 1452\@global_work_offset must be either C<undef> (in which case all offsets
471are C<0>), or a reference to an array of work offsets, with the same number 1453are C<0>), or a reference to an array of work offsets, with the same number
472of elements as @$global_work_size. 1454of elements as \@global_work_size.
473 1455
474@$local_work_size must be either C<undef> (in which case the 1456\@local_work_size must be either C<undef> (in which case the
475implementation is supposed to choose good local work sizes), or a 1457implementation is supposed to choose good local work sizes), or a
476reference to an array of local work sizes, with the same number of 1458reference to an array of local work sizes, with the same number of
477elements as @$global_work_size. 1459elements as \@global_work_size.
478 1460
479L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueNDRangeKernel.html> 1461L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueNDRangeKernel.html>
480 1462
481=item $ev = $queue->enqueue_marker 1463=item $ev = $queue->migrate_mem_objects (\@mem_objects, $flags, $wait_events...)
482 1464
1465Migrates a number of OpenCL::Memory objects to or from the device.
1466
1467flags: OpenCL::MIGRATE_MEM_OBJECT_HOST, OpenCL::MIGRATE_MEM_OBJECT_CONTENT_UNDEFINED
1468
1469L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueMigrateMemObjects.html>
1470
1471=item $ev = $queue->acquire_gl_objects ([object, ...], $wait_events...)
1472
1473Enqueues a list (an array-ref of OpenCL::Memory objects) to be acquired
1474for subsequent OpenCL usage.
1475
483L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueMarker.html> 1476L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueAcquireGLObjects.html>
484 1477
1478=item $ev = $queue->release_gl_objects ([object, ...], $wait_events...)
1479
1480Enqueues a list (an array-ref of OpenCL::Memory objects) to be released
1481for subsequent OpenGL usage.
1482
1483L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReleaseGLObjects.html>
1484
485=item $ev = $queue->enqueue_wait_for_events ($wait_events...) 1485=item $ev = $queue->wait_for_events ($wait_events...)
486 1486
487L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWaitForEvents.html> 1487L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWaitForEvents.html>
488 1488
489=item $queue->enqueue_barrier 1489=item $ev = $queue->marker ($wait_events...)
490 1490
1491L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueMarkerWithWaitList.html>
1492
1493=item $ev = $queue->barrier ($wait_events...)
1494
491L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueBarrier.html> 1495L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueBarrierWithWaitList.html>
492 1496
493=item $queue->flush 1497=item $queue->flush
494 1498
495L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFlush.html> 1499L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFlush.html>
496 1500
497=item $queue->finish 1501=item $queue->finish
498 1502
499L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFinish.html> 1503L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFinish.html>
500 1504
1505=item $packed_value = $queue->info ($name)
1506
1507See C<< $platform->info >> for details.
1508
1509L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetCommandQueueInfo.html>
1510
1511=for gengetinfo begin command_queue
1512
1513=item $ctx = $command_queue->context
1514
1515Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_CONTEXT> and returns the result.
1516
1517=item $device = $command_queue->device
1518
1519Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_DEVICE> and returns the result.
1520
1521=item $uint = $command_queue->reference_count
1522
1523Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_REFERENCE_COUNT> and returns the result.
1524
1525=item $command_queue_properties = $command_queue->properties
1526
1527Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_PROPERTIES> and returns the result.
1528
1529=for gengetinfo end command_queue
1530
501=back 1531=back
502 1532
1533=head3 MEMORY MAPPED BUFFERS
1534
1535OpenCL allows you to map buffers and images to host memory (read: perl
1536scalars). This is done much like reading or copying a buffer, by enqueuing
1537a map or unmap operation on the command queue.
1538
1539The map operations return an C<OpenCL::Mapped> object - see L<THE
1540OpenCL::Mapped CLASS> section for details on what to do with these
1541objects.
1542
1543The object will be unmapped automatically when the mapped object is
1544destroyed (you can use a barrier to make sure the unmap has finished,
1545before using the buffer in a kernel), but you can also enqueue an unmap
1546operation manually.
1547
1548=over 4
1549
1550=item $mapped_buffer = $queue->map_buffer ($buf, $blocking=1, $map_flags=OpenCL::MAP_READ|OpenCL::MAP_WRITE, $offset=0, $size=undef, $wait_events...)
1551
1552Maps the given buffer into host memory and returns an
1553C<OpenCL::MappedBuffer> object. If C<$size> is specified as undef, then
1554the map will extend to the end of the buffer.
1555
1556map_flags: OpenCL::MAP_READ, OpenCL::MAP_WRITE, OpenCL::MAP_WRITE_INVALIDATE_REGION.
1557
1558L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueMapBuffer.html>
1559
1560Example: map the buffer $buf fully and replace the first 4 bytes by "abcd", then unmap.
1561
1562 {
1563 my $mapped = $queue->map_buffer ($buf, 1, OpenCL::MAP_WRITE);
1564 substr $$mapped, 0, 4, "abcd";
1565 } # asynchronously unmap because $mapped is destroyed
1566
1567=item $mapped_image = $queue->map_image ($img, $blocking=1, $map_flags=OpenCL::MAP_READ|OpenCL::MAP_WRITE, $x=0, $y=0, $z=0, $width=undef, $height=undef, $depth=undef, $wait_events...)
1568
1569Maps the given image area into host memory and return an
1570C<OpenCL::MappedImage> object.
1571
1572If any of C<$width>, C<$height> and/or C<$depth> are C<undef> then they
1573will be replaced by the maximum possible value.
1574
1575L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueMapImage.html>
1576
1577Example: map an image (with OpenCL::UNSIGNED_INT8 channel type) and set
1578the first channel of the leftmost column to 5, then explicitly unmap
1579it. You are not necessarily meant to do it this way, this example just
1580shows you the accessors to use :)
1581
1582 my $mapped = $queue->map_image ($image, 1, OpenCL::MAP_WRITE);
1583
1584 $mapped->set ($_ * $mapped->row_pitch, pack "C", 5)
1585 for 0..$image->height;
1586
1587 $mapped->unmap;.
1588 $mapped->wait; # only needed for out of order queues normally
1589
1590=item $ev = $queue->unmap ($mapped, $wait_events...)
1591
1592Unmaps the data from host memory. You must not call any methods that
1593modify the data, or modify the data scalar directly, after calling this
1594method.
1595
1596The mapped event object will always be passed as part of the
1597$wait_events. The mapped event object will be replaced by the new event
1598object that this request creates.
1599
1600=back
1601
503=head2 THE OpenCL::Memory CLASS 1602=head2 THE OpenCL::Memory CLASS
504 1603
505This the superclass of all memory objects - OpenCL::Buffer, OpenCL::Image, 1604This the superclass of all memory objects - OpenCL::Buffer, OpenCL::Image,
506OpenCL::Image2D and OpenCL::Image3D. The subclasses of this class 1605OpenCL::Image2D and OpenCL::Image3D.
507currently only exist to allow type-checking.
508 1606
509=over 4 1607=over 4
510 1608
511=item $packed_value = $memory->info ($name) 1609=item $packed_value = $memory->info ($name)
512 1610
513See C<< $platform->info >> for details. 1611See C<< $platform->info >> for details.
514 1612
515L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetMemObjectInfo.html> 1613L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetMemObjectInfo.html>
516 1614
1615=for gengetinfo begin mem
1616
1617=item $mem_object_type = $mem->type
1618
1619Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_TYPE> and returns the result.
1620
1621=item $mem_flags = $mem->flags
1622
1623Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_FLAGS> and returns the result.
1624
1625=item $int = $mem->size
1626
1627Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_SIZE> and returns the result.
1628
1629=item $ptr_value = $mem->host_ptr
1630
1631Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_HOST_PTR> and returns the result.
1632
1633=item $uint = $mem->map_count
1634
1635Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_MAP_COUNT> and returns the result.
1636
1637=item $uint = $mem->reference_count
1638
1639Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_REFERENCE_COUNT> and returns the result.
1640
1641=item $ctx = $mem->context
1642
1643Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_CONTEXT> and returns the result.
1644
1645=item $mem = $mem->associated_memobject
1646
1647Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_ASSOCIATED_MEMOBJECT> and returns the result.
1648
1649=item $int = $mem->offset
1650
1651Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_OFFSET> and returns the result.
1652
1653=for gengetinfo end mem
1654
1655=item ($type, $name) = $mem->gl_object_info
1656
1657Returns the OpenGL object type (e.g. OpenCL::GL_OBJECT_TEXTURE2D) and the
1658object "name" (e.g. the texture name) used to create this memory object.
1659
1660L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetGLObjectInfo.html>
1661
517=back 1662=back
518 1663
1664=head2 THE OpenCL::Buffer CLASS
1665
1666This is a subclass of OpenCL::Memory, and the superclass of
1667OpenCL::BufferObj. Its purpose is simply to distinguish between buffers
1668and sub-buffers.
1669
1670=head2 THE OpenCL::BufferObj CLASS
1671
1672This is a subclass of OpenCL::Buffer and thus OpenCL::Memory. It exists
1673because one cna create sub buffers of OpenLC::BufferObj objects, but not
1674sub buffers from these sub buffers.
1675
1676=over 4
1677
1678=item $subbuf = $buf_obj->sub_buffer_region ($flags, $origin, $size)
1679
1680Creates an OpenCL::Buffer objects from this buffer and returns it. The
1681C<buffer_create_type> is assumed to be C<CL_BUFFER_CREATE_TYPE_REGION>.
1682
1683L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateSubBuffer.html>
1684
1685=back
1686
1687=head2 THE OpenCL::Image CLASS
1688
1689This is the superclass of all image objects - OpenCL::Image1D,
1690OpenCL::Image1DArray, OpenCL::Image1DBuffer, OpenCL::Image2D,
1691OpenCL::Image2DArray and OpenCL::Image3D.
1692
1693=over 4
1694
1695=item $packed_value = $image->image_info ($name)
1696
1697See C<< $platform->info >> for details.
1698
1699The reason this method is not called C<info> is that there already is an
1700C<< ->info >> method inherited from C<OpenCL::Memory>.
1701
1702L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetImageInfo.html>
1703
1704=item ($channel_order, $channel_data_type) = $image->format
1705
1706Returns the channel order and type used to create the image by calling
1707C<clGetImageInfo> with C<CL_IMAGE_FORMAT>.
1708
1709=for gengetinfo begin image
1710
1711=item $int = $image->element_size
1712
1713Calls C<clGetImageInfo> with C<OpenCL::IMAGE_ELEMENT_SIZE> and returns the result.
1714
1715=item $int = $image->row_pitch
1716
1717Calls C<clGetImageInfo> with C<OpenCL::IMAGE_ROW_PITCH> and returns the result.
1718
1719=item $int = $image->slice_pitch
1720
1721Calls C<clGetImageInfo> with C<OpenCL::IMAGE_SLICE_PITCH> and returns the result.
1722
1723=item $int = $image->width
1724
1725Calls C<clGetImageInfo> with C<OpenCL::IMAGE_WIDTH> and returns the result.
1726
1727=item $int = $image->height
1728
1729Calls C<clGetImageInfo> with C<OpenCL::IMAGE_HEIGHT> and returns the result.
1730
1731=item $int = $image->depth
1732
1733Calls C<clGetImageInfo> with C<OpenCL::IMAGE_DEPTH> and returns the result.
1734
1735=for gengetinfo end image
1736
1737=for gengetinfo begin gl_texture
1738
1739=item $GLenum = $gl_texture->target
1740
1741Calls C<clGetGLTextureInfo> with C<OpenCL::GL_TEXTURE_TARGET> and returns the result.
1742
1743=item $GLint = $gl_texture->gl_mipmap_level
1744
1745Calls C<clGetGLTextureInfo> with C<OpenCL::GL_MIPMAP_LEVEL> and returns the result.
1746
1747=for gengetinfo end gl_texture
1748
1749=back
1750
519=head2 THE OpenCL::Sampler CLASS 1751=head2 THE OpenCL::Sampler CLASS
520 1752
521=over 4 1753=over 4
522 1754
523=item $packed_value = $sampler->info ($name) 1755=item $packed_value = $sampler->info ($name)
524 1756
525See C<< $platform->info >> for details. 1757See C<< $platform->info >> for details.
526 1758
527L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetSamplerInfo.html> 1759L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetSamplerInfo.html>
528 1760
1761=for gengetinfo begin sampler
1762
1763=item $uint = $sampler->reference_count
1764
1765Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_REFERENCE_COUNT> and returns the result.
1766
1767=item $ctx = $sampler->context
1768
1769Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_CONTEXT> and returns the result.
1770
1771=item $addressing_mode = $sampler->normalized_coords
1772
1773Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_NORMALIZED_COORDS> and returns the result.
1774
1775=item $filter_mode = $sampler->addressing_mode
1776
1777Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_ADDRESSING_MODE> and returns the result.
1778
1779=item $boolean = $sampler->filter_mode
1780
1781Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_FILTER_MODE> and returns the result.
1782
1783=for gengetinfo end sampler
1784
529=back 1785=back
530 1786
531=head2 THE OpenCL::Program CLASS 1787=head2 THE OpenCL::Program CLASS
532 1788
533=over 4 1789=over 4
534 1790
535=item $packed_value = $program->info ($name) 1791=item $program->build (\@devices = undef, $options = "", $cb->($program) = undef)
536 1792
537See C<< $platform->info >> for details.
538
539L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProgramInfo.html>
540
541=item $program->build ($device, $options = "")
542
543Tries to build the program with the givne options. 1793Tries to build the program with the given options. See also the
1794C<$ctx->build> convenience function.
1795
1796If a callback is specified, then it will be called when compilation is
1797finished. Note that many OpenCL implementations block your program while
1798compiling whether you use a callback or not. See C<build_async> if you
1799want to make sure the build is done in the background.
1800
1801Note that some OpenCL implementations act up badly, and don't call the
1802callback in some error cases (but call it in others). This implementation
1803assumes the callback will always be called, and leaks memory if this is
1804not so. So best make sure you don't pass in invalid values.
1805
1806Some implementations fail with C<OpenCL::INVALID_BINARY> when the
1807compilation state is successful but some later stage fails.
1808
1809options: C<-D name>, C<-D name=definition>, C<-I dir>,
1810C<-cl-single-precision-constant>, C<-cl-denorms-are-zero>,
1811C<-cl-fp32-correctly-rounded-divide-sqrt>, C<-cl-opt-disable>,
1812C<-cl-mad-enable>, C<-cl-no-signed-zeros>, C<-cl-unsafe-math-optimizations>,
1813C<-cl-finite-math-only>, C<-cl-fast-relaxed-math>,
1814C<-w>, C<-Werror>, C<-cl-std=CL1.1/CL1.2>, C<-cl-kernel-arg-info>,
1815C<-create-library>, C<-enable-link-options>.
1816
1817build_status: OpenCL::BUILD_SUCCESS, OpenCL::BUILD_NONE,
1818OpenCL::BUILD_ERROR, OpenCL::BUILD_IN_PROGRESS.
544 1819
545L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html> 1820L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html>
1821
1822=item $program->build_async (\@devices = undef, $options = "", $cb->($program) = undef)
1823
1824Similar to C<< ->build >>, except it starts a thread, and never fails (you
1825need to check the compilation status form the callback, or by polling).
1826
1827=item $program->compile (\@devices = undef, $options = "", \%headers = undef, $cb->($program) = undef)
1828
1829Compiles the given program for the given devices (or all devices if
1830undef). If C<$headers> is given, it must be a hashref with include name =>
1831OpenCL::Program pairs.
1832
1833L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCompileProgram.html>
546 1834
547=item $packed_value = $program->build_info ($device, $name) 1835=item $packed_value = $program->build_info ($device, $name)
548 1836
549Similar to C<< $platform->info >>, but returns build info for a previous 1837Similar to C<< $platform->info >>, but returns build info for a previous
550build attempt for the given device. 1838build attempt for the given device.
556Creates an OpenCL::Kernel object out of the named C<__kernel> function in 1844Creates an OpenCL::Kernel object out of the named C<__kernel> function in
557the program. 1845the program.
558 1846
559L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernel.html> 1847L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernel.html>
560 1848
1849=item @kernels = $program->kernels_in_program
1850
1851Returns all kernels successfully compiled for all devices in program.
1852
1853http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernelsInProgram.html
1854
1855=for gengetinfo begin program_build
1856
1857=item $build_status = $program->build_status ($device)
1858
1859Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BUILD_STATUS> and returns the result.
1860
1861=item $string = $program->build_options ($device)
1862
1863Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BUILD_OPTIONS> and returns the result.
1864
1865=item $string = $program->build_log ($device)
1866
1867Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BUILD_LOG> and returns the result.
1868
1869=for gengetinfo end program_build
1870
1871=item $packed_value = $program->info ($name)
1872
1873See C<< $platform->info >> for details.
1874
1875L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProgramInfo.html>
1876
1877=for gengetinfo begin program
1878
1879=item $uint = $program->reference_count
1880
1881Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_REFERENCE_COUNT> and returns the result.
1882
1883=item $ctx = $program->context
1884
1885Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_CONTEXT> and returns the result.
1886
1887=item $uint = $program->num_devices
1888
1889Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_NUM_DEVICES> and returns the result.
1890
1891=item @devices = $program->devices
1892
1893Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_DEVICES> and returns the result.
1894
1895=item $string = $program->source
1896
1897Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_SOURCE> and returns the result.
1898
1899=item @ints = $program->binary_sizes
1900
1901Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_BINARY_SIZES> and returns the result.
1902
1903=for gengetinfo end program
1904
1905=item @blobs = $program->binaries
1906
1907Returns a string for the compiled binary for every device associated with
1908the program, empty strings indicate missing programs, and an empty result
1909means no program binaries are available.
1910
1911These "binaries" are often, in fact, informative low-level assembly
1912sources.
1913
1914L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProgramInfo.html>
1915
561=back 1916=back
562 1917
563=head2 THE OpenCL::Kernel CLASS 1918=head2 THE OpenCL::Kernel CLASS
564 1919
565=over 4 1920=over 4
568 1923
569See C<< $platform->info >> for details. 1924See C<< $platform->info >> for details.
570 1925
571L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelInfo.html> 1926L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelInfo.html>
572 1927
1928=for gengetinfo begin kernel
1929
1930=item $string = $kernel->function_name
1931
1932Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_FUNCTION_NAME> and returns the result.
1933
1934=item $uint = $kernel->num_args
1935
1936Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_NUM_ARGS> and returns the result.
1937
1938=item $uint = $kernel->reference_count
1939
1940Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_REFERENCE_COUNT> and returns the result.
1941
1942=item $ctx = $kernel->context
1943
1944Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_CONTEXT> and returns the result.
1945
1946=item $program = $kernel->program
1947
1948Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_PROGRAM> and returns the result.
1949
1950=for gengetinfo end kernel
1951
1952=item $packed_value = $kernel->work_group_info ($device, $name)
1953
1954See C<< $platform->info >> for details.
1955
1956L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelWorkGroupInfo.html>
1957
1958=for gengetinfo begin kernel_work_group
1959
1960=item $int = $kernel->work_group_size ($device)
1961
1962Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_WORK_GROUP_SIZE> and returns the result.
1963
1964=item @ints = $kernel->compile_work_group_size ($device)
1965
1966Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_COMPILE_WORK_GROUP_SIZE> and returns the result.
1967
1968=item $ulong = $kernel->local_mem_size ($device)
1969
1970Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_LOCAL_MEM_SIZE> and returns the result.
1971
1972=item $int = $kernel->preferred_work_group_size_multiple ($device)
1973
1974Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE> and returns the result.
1975
1976=item $ulong = $kernel->private_mem_size ($device)
1977
1978Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_PRIVATE_MEM_SIZE> and returns the result.
1979
1980=for gengetinfo end kernel_work_group
1981
1982=item $packed_value = $kernel->arg_info ($idx, $name)
1983
1984See C<< $platform->info >> for details.
1985
1986L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clGetKernelArgInfo.html>
1987
1988=for gengetinfo begin kernel_arg
1989
1990=item $kernel_arg_address_qualifier = $kernel->arg_address_qualifier ($idx)
1991
1992Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_ADDRESS_QUALIFIER> and returns the result.
1993
1994=item $kernel_arg_access_qualifier = $kernel->arg_access_qualifier ($idx)
1995
1996Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_ACCESS_QUALIFIER> and returns the result.
1997
1998=item $string = $kernel->arg_type_name ($idx)
1999
2000Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_TYPE_NAME> and returns the result.
2001
2002=item $kernel_arg_type_qualifier = $kernel->arg_type_qualifier ($idx)
2003
2004Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_TYPE_QUALIFIER> and returns the result.
2005
2006=item $string = $kernel->arg_name ($idx)
2007
2008Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_NAME> and returns the result.
2009
2010=for gengetinfo end kernel_arg
2011
2012=item $kernel->setf ($format, ...)
2013
2014Sets the arguments of a kernel. Since OpenCL 1.1 doesn't have a generic
2015way to set arguments (and with OpenCL 1.2 it might be rather slow), you
2016need to specify a format argument, much as with C<printf>, to tell OpenCL
2017what type of argument it is.
2018
2019The format arguments are single letters:
2020
2021 c char
2022 C unsigned char
2023 s short
2024 S unsigned short
2025 i int
2026 I unsigned int
2027 l long
2028 L unsigned long
2029
2030 h half float (0..65535)
2031 f float
2032 d double
2033
2034 z local (octet size)
2035
2036 m memory object (buffer or image)
2037 a sampler
2038 e event
2039
2040Space characters in the format string are ignored.
2041
2042Example: set the arguments for a kernel that expects an int, two floats, a buffer and an image.
2043
2044 $kernel->setf ("i ff mm", 5, 0.5, 3, $buffer, $image);
2045
573=item $kernel->set_TYPE ($index, $value) 2046=item $kernel->set_TYPE ($index, $value)
574 2047
2048=item $kernel->set_char ($index, $value)
2049
2050=item $kernel->set_uchar ($index, $value)
2051
2052=item $kernel->set_short ($index, $value)
2053
2054=item $kernel->set_ushort ($index, $value)
2055
2056=item $kernel->set_int ($index, $value)
2057
2058=item $kernel->set_uint ($index, $value)
2059
2060=item $kernel->set_long ($index, $value)
2061
2062=item $kernel->set_ulong ($index, $value)
2063
2064=item $kernel->set_half ($index, $value)
2065
2066=item $kernel->set_float ($index, $value)
2067
2068=item $kernel->set_double ($index, $value)
2069
2070=item $kernel->set_memory ($index, $value)
2071
2072=item $kernel->set_buffer ($index, $value)
2073
2074=item $kernel->set_image ($index, $value)
2075
2076=item $kernel->set_sampler ($index, $value)
2077
2078=item $kernel->set_local ($index, $value)
2079
2080=item $kernel->set_event ($index, $value)
2081
575This is a family of methods to set the kernel argument with the number C<$index> to the give C<$value>. 2082This is a family of methods to set the kernel argument with the number
576 2083C<$index> to the give C<$value>.
577TYPE is one of C<char>, C<uchar>, C<short>, C<ushort>, C<int>, C<uint>,
578C<long>, C<ulong>, C<half>, C<float>, C<double>, C<memory>, C<buffer>,
579C<image2d>, C<image3d>, C<sampler> or C<event>.
580 2084
581Chars and integers (including the half type) are specified as integers, 2085Chars and integers (including the half type) are specified as integers,
582float and double as floating point values, memory/buffer/image2d/image3d 2086float and double as floating point values, memory/buffer/image must be
583must be an object of that type or C<undef>, and sampler and event must be 2087an object of that type or C<undef>, local-memory arguments are set by
584objects of that type. 2088specifying the size, and sampler and event must be objects of that type.
2089
2090Note that C<set_memory> works for all memory objects (all types of buffers
2091and images) - the main purpose of the more specific C<set_TYPE> functions
2092is type checking.
2093
2094Setting an argument for a kernel does NOT keep a reference to the object -
2095for example, if you set an argument to some image object, free the image,
2096and call the kernel, you will run into undefined behaviour.
585 2097
586L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetKernelArg.html> 2098L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetKernelArg.html>
587 2099
588=back 2100=back
589 2101
592This is the superclass for all event objects (including OpenCL::UserEvent 2104This is the superclass for all event objects (including OpenCL::UserEvent
593objects). 2105objects).
594 2106
595=over 4 2107=over 4
596 2108
2109=item $ev->wait
2110
2111Waits for the event to complete.
2112
2113L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html>
2114
2115=item $ev->cb ($exec_callback_type, $callback->($event, $event_command_exec_status))
2116
2117Adds a callback to the callback stack for the given event type. There is
2118no way to remove a callback again.
2119
2120L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetEventCallback.html>
2121
597=item $packed_value = $ev->info ($name) 2122=item $packed_value = $ev->info ($name)
598 2123
599See C<< $platform->info >> for details. 2124See C<< $platform->info >> for details.
600 2125
601L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetEventInfo.html> 2126L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetEventInfo.html>
602 2127
603=item $ev->wait 2128=for gengetinfo begin event
604 2129
605Waits for the event to complete. 2130=item $queue = $event->command_queue
606 2131
2132Calls C<clGetEventInfo> with C<OpenCL::EVENT_COMMAND_QUEUE> and returns the result.
2133
2134=item $command_type = $event->command_type
2135
2136Calls C<clGetEventInfo> with C<OpenCL::EVENT_COMMAND_TYPE> and returns the result.
2137
2138=item $uint = $event->reference_count
2139
2140Calls C<clGetEventInfo> with C<OpenCL::EVENT_REFERENCE_COUNT> and returns the result.
2141
2142=item $uint = $event->command_execution_status
2143
2144Calls C<clGetEventInfo> with C<OpenCL::EVENT_COMMAND_EXECUTION_STATUS> and returns the result.
2145
2146=item $ctx = $event->context
2147
2148Calls C<clGetEventInfo> with C<OpenCL::EVENT_CONTEXT> and returns the result.
2149
2150=for gengetinfo end event
2151
2152=item $packed_value = $ev->profiling_info ($name)
2153
2154See C<< $platform->info >> for details.
2155
2156The reason this method is not called C<info> is that there already is an
2157C<< ->info >> method.
2158
607L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html> 2159L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProfilingInfo.html>
2160
2161=for gengetinfo begin profiling
2162
2163=item $ulong = $event->profiling_command_queued
2164
2165Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_QUEUED> and returns the result.
2166
2167=item $ulong = $event->profiling_command_submit
2168
2169Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_SUBMIT> and returns the result.
2170
2171=item $ulong = $event->profiling_command_start
2172
2173Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_START> and returns the result.
2174
2175=item $ulong = $event->profiling_command_end
2176
2177Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_END> and returns the result.
2178
2179=for gengetinfo end profiling
608 2180
609=back 2181=back
610 2182
611=head2 THE OpenCL::UserEvent CLASS 2183=head2 THE OpenCL::UserEvent CLASS
612 2184
614 2186
615=over 4 2187=over 4
616 2188
617=item $ev->set_status ($execution_status) 2189=item $ev->set_status ($execution_status)
618 2190
2191Sets the execution status of the user event. Can only be called once,
2192either with OpenCL::COMPLETE or a negative number as status.
2193
2194execution_status: OpenCL::COMPLETE or a negative integer.
2195
619L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetUserEventStatus.html> 2196L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetUserEventStatus.html>
620 2197
621=back 2198=back
622 2199
2200=head2 THE OpenCL::Mapped CLASS
2201
2202This class represents objects mapped into host memory. They are
2203represented by a blessed string scalar. The string data is the mapped
2204memory area, that is, if you read or write it, then the mapped object is
2205accessed directly.
2206
2207You must only ever use operations that modify the string in-place - for
2208example, a C<substr> that doesn't change the length, or maybe a regex that
2209doesn't change the length. Any other operation might cause the data to be
2210copied.
2211
2212When the object is destroyed it will enqueue an implicit unmap operation
2213on the queue that was used to create it.
2214
2215Keep in mind that you I<need> to unmap (or destroy) mapped objects before
2216OpenCL sees the changes, even if some implementations don't need this
2217sometimes.
2218
2219Example, replace the first two floats in the mapped buffer by 1 and 2.
2220
2221 my $mapped = $queue->map_buffer ($buf, ...
2222 $mapped->event->wait; # make sure it's there
2223
2224 # now replace first 8 bytes by new data, which is exactly 8 bytes long
2225 # we blindly assume device endianness to equal host endianness
2226 # (and of course, we assume iee 754 single precision floats :)
2227 substr $$mapped, 0, 8, pack "f*", 1, 2;
2228
2229=over 4
2230
2231=item $ev = $mapped->unmap ($wait_events...)
2232
2233Unmaps the mapped memory object, using the queue originally used to create
2234it, quite similarly to C<< $queue->unmap ($mapped, ...) >>.
2235
2236=item $bool = $mapped->mapped
2237
2238Returns whether the object is still mapped - true before an C<unmap> is
2239enqueued, false afterwards.
2240
2241=item $ev = $mapped->event
2242
2243Return the event object associated with the mapped object. Initially, this
2244will be the event object created when mapping the object, and after an
2245unmap, this will be the event object that the unmap operation created.
2246
2247=item $mapped->wait
2248
2249Same as C<< $mapped->event->wait >> - makes sure no operations on this
2250mapped object are outstanding.
2251
2252=item $bytes = $mapped->size
2253
2254Returns the size of the mapped area, in bytes. Same as C<length $$mapped>.
2255
2256=item $ptr = $mapped->ptr
2257
2258Returns the raw memory address of the mapped area.
2259
2260=item $mapped->set ($offset, $data)
2261
2262Replaces the data at the given C<$offset> in the memory area by the new
2263C<$data>. This method is safer than direct manipulation of C<$mapped>
2264because it does bounds-checking, but also slower.
2265
2266=item $data = $mapped->get ($offset, $length)
2267
2268Returns (without copying) a scalar representing the data at the given
2269C<$offset> and C<$length> in the mapped memory area. This is the same as
2270the following substr, except much slower;
2271
2272 $data = substr $$mapped, $offset, $length
2273
623=cut 2274=cut
624 2275
625package OpenCL; 2276sub OpenCL::Mapped::get {
626 2277 substr ${$_[0]}, $_[1], $_[2]
627use common::sense;
628
629BEGIN {
630 our $VERSION = '0.03';
631
632 require XSLoader;
633 XSLoader::load (__PACKAGE__, $VERSION);
634
635 @OpenCL::Buffer::ISA =
636 @OpenCL::Image::ISA = OpenCL::Memory::;
637
638 @OpenCL::Image2D::ISA =
639 @OpenCL::Image3D::ISA = OpenCL::Image::;
640
641 @OpenCL::UserEvent::ISA = OpenCL::Event::;
642} 2278}
2279
2280=back
2281
2282=head2 THE OpenCL::MappedBuffer CLASS
2283
2284This is a subclass of OpenCL::Mapped, representing mapped buffers.
2285
2286=head2 THE OpenCL::MappedImage CLASS
2287
2288This is a subclass of OpenCL::Mapped, representing mapped images.
2289
2290=over 4
2291
2292=item $bytes = $mapped->row_pitch
2293
2294=item $bytes = $mapped->slice_pitch
2295
2296Return the row or slice pitch of the image that has been mapped.
2297
2298=back
2299
2300
2301=cut
643 2302
6441; 23031;
645 2304
646=head1 AUTHOR 2305=head1 AUTHOR
647 2306

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines