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

Comparing OpenCL/OpenCL.pm (file contents):
Revision 1.45 by root, Sat Apr 21 17:50:27 2012 UTC vs.
Revision 1.57 by root, Tue Apr 24 23:58:34 2012 UTC

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:
49no implementation was available to me :).
50 49
51 http://www.khronos.org/registry/cl/specs/opencl-1.1.pdf 50 http://www.khronos.org/registry/cl/specs/opencl-1.1.pdf
51 http://www.khronos.org/registry/cl/specs/opencl-1.2.pdf
52 http://www.khronos.org/registry/cl/specs/opencl-1.2-extensions.pdf
52 53
53OpenCL manpages: 54OpenCL manpages:
54 55
55 http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/ 56 http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/
57 http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/
56 58
57If you are into UML class diagrams, the following diagram might help - if 59If you are into UML class diagrams, the following diagram might help - if
58not, it will be mildly cobfusing: 60not, it will be mildly confusing (also, the class hierarchy of this module
61is much more fine-grained):
59 62
60 http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/classDiagram.html 63 http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/classDiagram.html
61 64
62Here's a tutorial from AMD (very AMD-centric, too), not sure how useful it 65Here's a tutorial from AMD (very AMD-centric, too), not sure how useful it
63is, but at least it's free of charge: 66is, but at least it's free of charge:
64 67
65 http://developer.amd.com/zones/OpenCLZone/courses/Documents/Introduction_to_OpenCL_Programming%20Training_Guide%20%28201005%29.pdf 68 http://developer.amd.com/zones/OpenCLZone/courses/Documents/Introduction_to_OpenCL_Programming%20Training_Guide%20%28201005%29.pdf
157 $id = get_global_id (0); 160 $id = get_global_id (0);
158 output [id] = input [id] * input [id]; 161 output [id] = input [id] * input [id];
159 } 162 }
160 '; 163 ';
161 164
162 my $prog = $ctx->program_with_source ($src); 165 my $prog = $ctx->build_program ($src);
163
164 # build croaks on compile errors, so catch it and print the compile errors
165 eval { $prog->build ($dev, "-cl-fast-relaxed-math"); 1 }
166 or die $prog->build_log;
167
168 my $kernel = $prog->kernel ("squareit"); 166 my $kernel = $prog->kernel ("squareit");
169 167
170=head2 Create some input and output float buffers, then call the 168=head2 Create some input and output float buffers, then call the
171'squareit' kernel on them. 169'squareit' kernel on them.
172 170
267 265
268 float3 colour = (float3)(z.x, z.y, z.x * z.y); 266 float3 colour = (float3)(z.x, z.y, z.x * z.y);
269 write_imagef (img, (int2)(get_global_id (0), get_global_id (1)), (float4)(colour * p.x * p.x, 1.)); 267 write_imagef (img, (int2)(get_global_id (0), get_global_id (1)), (float4)(colour * p.x * p.x, 1.));
270 } 268 }
271 EOF 269 EOF
270
272 my $prog = $ctx->program_with_source ($src); 271 my $prog = $ctx->build_program ($src);
273 eval { $prog->build ($dev); 1 }
274 or die $prog->build_log ($dev);
275
276 my $kernel = $prog->kernel ("juliatunnel"); 272 my $kernel = $prog->kernel ("juliatunnel");
277 273
278 # program compiled, kernel ready, now draw and loop 274 # program compiled, kernel ready, now draw and loop
279 275
280 for (my $time; ; ++$time) { 276 for (my $time; ; ++$time) {
383 379
384For this to work, the OpenGL library must be loaded, a GLX context must 380For this to work, the OpenGL library must be loaded, a GLX context must
385have been created and be made current, and C<dlsym> must be available and 381have been created and be made current, and C<dlsym> must be available and
386capable of finding the function via C<RTLD_DEFAULT>. 382capable of finding the function via C<RTLD_DEFAULT>.
387 383
384=head2 EVENT SYSTEM
385
386OpenCL can generate a number of (potentially) asynchronous events, for
387example, after compiling a program, to signal a context-related error or,
388perhaps most important, to signal completion of queued jobs (by setting
389callbacks on OpenCL::Event objects).
390
391To facilitate this, this module maintains an event queue - each
392time an asynchronous event happens, it is queued, and perl will be
393interrupted. This is implemented via the L<Async::Interrupt> module. In
394addition, this module has L<AnyEvent> support, so it can seamlessly
395integrate itself into many event loops.
396
397Since this module is a bit hard to understand, here are some case examples:
398
399=head3 Don't use callbacks.
400
401When your program never uses any callbacks, then there will never be any
402notifications you need to take care of, and therefore no need to worry
403about all this.
404
405You can achieve a great deal by explicitly waiting for events, or using
406barriers and flush calls. In many programs, there is no need at all to
407tinker with asynchronous events.
408
409=head3 Use AnyEvent
410
411This module automatically registers a watcher that invokes all outstanding
412event callbacks when AnyEvent is initialised (and block asynchronous
413interruptions). Using this mode of operations is the safest and most
414recommended one.
415
416To use this, simply use AnyEvent and this module normally, make sure you
417have an event loop running:
418
419 use Gtk2 -init;
420 use AnyEvent;
421
422 # initialise AnyEvent, by creating a watcher, or:
423 AnyEvent::detect;
424
425 my $e = $queue->enqueue_marker;
426 $e->cb (sub {
427 warn "opencl is finished\n";
428 })
429
430 main Gtk2;
431
432Note that this module will not initialise AnyEvent for you. Before
433AnyEvent is initialised, the module will asynchronously interrupt perl
434instead. To avoid any surprises, it's best to explicitly initialise
435AnyEvent.
436
437You can temporarily enable asynchronous interruptions (see next paragraph)
438by calling C<$OpenCL::INTERRUPT->unblock> and disable them again by
439calling C<$OpenCL::INTERRUPT->block>.
440
441=head3 Let yourself be interrupted at any time
442
443This mode is the default unless AnyEvent is loaded and initialised. In
444this mode, OpenCL asynchronously interrupts a running perl program. The
445emphasis is on both I<asynchronously> and I<running> here.
446
447Asynchronously means that perl might execute your callbacks at any
448time. For example, in the following code (I<THAT YOU SHOULD NOT COPY>),
449the C<until> loop following the marker call will be interrupted by the
450callback:
451
452 my $e = $queue->enqueue_marker;
453 my $flag;
454 $e->cb (sub { $flag = 1 });
455 1 until $flag;
456 # $flag is now 1
457
458The reason why you shouldn't blindly copy the above code is that
459busy waiting is a really really bad thing, and really really bad for
460performance.
461
462While at first this asynchronous business might look exciting, it can be
463really hard, because you need to be prepared for the callback code to be
464executed at any time, which limits the amount of things the callback code
465can do safely.
466
467This can be mitigated somewhat by using C<<
468$OpenCL::INTERRUPT->scope_block >> (see the L<Async::Interrupt>
469documentation for details).
470
471The other problem is that your program must be actively I<running> to be
472interrupted. When you calculate stuff, your program is running. When you
473hang in some C functions or other block execution (by calling C<sleep>,
474C<select>, running an event loop and so on), your program is waiting, not
475running.
476
477One way around that would be to attach a read watcher to your event loop,
478listening for events on C<< $OpenCL::INTERRUPT->pipe_fileno >>, using a
479dummy callback (C<sub { }>) to temporarily execute some perl code.
480
481That is then awfully close to using the built-in AnyEvent support above,
482though, so consider that one instead.
483
484=head3 Be creative
485
486OpenCL exports the L<Async::Interrupt> object it uses in the global
487variable C<$OpenCL::INTERRUPT>. You can configure it in any way you like.
488
489So if you want to feel like a real pro, err, wait, if you feel no risk
490menas no fun, you can experiment by implementing your own mode of
491operations.
492
493=cut
494
495package OpenCL;
496
497use common::sense;
498use Async::Interrupt ();
499
500our $POLL_FUNC; # set by XS
501
502BEGIN {
503 our $VERSION = '0.97';
504
505 require XSLoader;
506 XSLoader::load (__PACKAGE__, $VERSION);
507
508 @OpenCL::Platform::ISA =
509 @OpenCL::Device::ISA =
510 @OpenCL::Context::ISA =
511 @OpenCL::Queue::ISA =
512 @OpenCL::Memory::ISA =
513 @OpenCL::Sampler::ISA =
514 @OpenCL::Program::ISA =
515 @OpenCL::Kernel::ISA =
516 @OpenCL::Event::ISA = OpenCL::Object::;
517
518 @OpenCL::Buffer::ISA =
519 @OpenCL::Image::ISA = OpenCL::Memory::;
520
521 @OpenCL::BufferObj::ISA = OpenCL::Buffer::;
522
523 @OpenCL::Image2D::ISA =
524 @OpenCL::Image3D::ISA =
525 @OpenCL::Image2DArray::ISA =
526 @OpenCL::Image1D::ISA =
527 @OpenCL::Image1DArray::ISA =
528 @OpenCL::Image1DBuffer::ISA = OpenCL::Image::;
529
530 @OpenCL::UserEvent::ISA = OpenCL::Event::;
531}
532
388=head2 THE OpenCL PACKAGE 533=head2 THE OpenCL PACKAGE
389 534
390=over 4 535=over 4
391 536
392=item $int = OpenCL::errno 537=item $int = OpenCL::errno
410 555
411Returns all available OpenCL::Platform objects. 556Returns all available OpenCL::Platform objects.
412 557
413L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html> 558L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html>
414 559
415=item $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $notify = undef 560=item $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr
416 561
417Tries to create a context from a default device and platform - never worked for me. 562Tries to create a context from a default device and platform type - never worked for me.
418 563
419L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html> 564L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html>
420 565
566=item $ctx = OpenCL::context $properties, \@devices, $callback->($err, $pvt) = $print_stderr)
567
568Create a new OpenCL::Context object using the given device object(s). This
569function isn't implemented yet, use C<< $platform->context >> instead.
570
571L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html>
572
421=item OpenCL::wait_for_events $wait_events... 573=item OpenCL::wait_for_events $wait_events...
422 574
423Waits for all events to complete. 575Waits for all events to complete.
424 576
425L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html> 577L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html>
426 578
579=item OpenCL::poll
580
581Checks if there are any outstanding events (see L<EVENT SYSTEM>) and
582invokes their callbacks.
583
584=item $OpenCL::INTERRUPT
585
586The L<Async::Interrupt> object used to signal asynchronous events (see
587L<EVENT SYSTEM>).
588
589=cut
590
591our $INTERRUPT = new Async::Interrupt c_cb => [$POLL_FUNC, 0];
592
593&_eq_initialise ($INTERRUPT->signal_func);
594
595=item $OpenCL::WATCHER
596
597The L<AnyEvent> watcher object used to watch for asynchronous events (see
598L<EVENT SYSTEM>). This variable is C<undef> until L<AnyEvent> has been
599loaded I<and> initialised (e.g. by calling C<AnyEvent::detect>).
600
601=cut
602
603our $WATCHER;
604
605sub _init_anyevent {
606 $INTERRUPT->block;
607 $WATCHER = AE::io ($INTERRUPT->pipe_fileno, 0, sub { $INTERRUPT->handle });
608}
609
610if (defined $AnyEvent::MODEL) {
611 _init_anyevent;
612} else {
613 push @AnyEvent::post_detect, \&_init_anyevent;
614}
615
427=back 616=back
428 617
618=head2 THE OpenCL::Object CLASS
619
620This is the base class for all objects in the OpenCL module. The only
621method it implements is the C<id> method, which is only useful if you want
622to interface to OpenCL on the C level.
623
624=over 4
625
626=item $iv = $obj->id
627
628OpenCL objects are represented by pointers or integers on the C level. If
629you want to interface to an OpenCL object directly on the C level, then
630you need this value, which is returned by this method. You should use an
631C<IV> type in your code and cast that to the correct type.
632
633=cut
634
635sub OpenCL::Object::id {
636 ref $_[0] eq "SCALAR"
637 ? ${ $_[0] }
638 : $_[0][0]
639}
640
641=back
642
429=head2 THE OpenCL::Platform CLASS 643=head2 THE OpenCL::Platform CLASS
430 644
431=over 4 645=over 4
432 646
433=item @devices = $platform->devices ($type = OpenCL::DEVICE_TYPE_ALL) 647=item @devices = $platform->devices ($type = OpenCL::DEVICE_TYPE_ALL)
434 648
435Returns a list of matching OpenCL::Device objects. 649Returns a list of matching OpenCL::Device objects.
436 650
437=item $ctx = $platform->context_from_type ($properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $notify = undef) 651=item $ctx = $platform->context_from_type ($properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr)
438 652
439Tries to create a context. Never worked for me, and you need devices explicitly anyway. 653Tries to create a context. Never worked for me, and you need devices explicitly anyway.
440 654
441L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html> 655L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html>
442 656
443=item $ctx = $platform->context ($properties = undef, @$devices, $notify = undef) 657=item $ctx = $platform->context ($properties, \@devices, $callback->($err, $pvt) = $print_stderr)
444 658
445Create a new OpenCL::Context object using the given device object(s)- a 659Create a new OpenCL::Context object using the given device object(s)- a
446CL_CONTEXT_PLATFORM property is supplied automatically. 660CL_CONTEXT_PLATFORM property is supplied automatically.
447 661
448L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html> 662L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html>
456It's best to avoid this method and use one of the following convenience 670It's best to avoid this method and use one of the following convenience
457wrappers. 671wrappers.
458 672
459L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformInfo.html> 673L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformInfo.html>
460 674
675=item $platform->unload_compiler
676
677Attempts to unload the compiler for this platform, for endless
678profit. Does nothing on OpenCL 1.1.
679
680L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clUnloadPlatformCompiler.html>
681
461=for gengetinfo begin platform 682=for gengetinfo begin platform
462 683
463=item $string = $platform->profile 684=item $string = $platform->profile
464 685
465Calls C<clGetPlatformInfo> with C<CL_PLATFORM_PROFILE> and returns the result. 686Calls C<clGetPlatformInfo> with C<CL_PLATFORM_PROFILE> and returns the result.
752 973
753Calls C<clGetDeviceInfo> with C<CL_DEVICE_AFFINITY_DOMAINS_EXT> and returns the result. 974Calls C<clGetDeviceInfo> with C<CL_DEVICE_AFFINITY_DOMAINS_EXT> and returns the result.
754 975
755=item $uint = $device->reference_count_ext 976=item $uint = $device->reference_count_ext
756 977
757Calls C<clGetDeviceInfo> with C<CL_DEVICE_REFERENCE_COUNT_EXT > and returns the result. 978Calls C<clGetDeviceInfo> with C<CL_DEVICE_REFERENCE_COUNT_EXT> and returns the result.
758 979
759=item @device_partition_property_exts = $device->partition_style_ext 980=item @device_partition_property_exts = $device->partition_style_ext
760 981
761Calls C<clGetDeviceInfo> with C<CL_DEVICE_PARTITION_STYLE_EXT> and returns the result. 982Calls C<clGetDeviceInfo> with C<CL_DEVICE_PARTITION_STYLE_EXT> and returns the result.
762 983
765=back 986=back
766 987
767=head2 THE OpenCL::Context CLASS 988=head2 THE OpenCL::Context CLASS
768 989
769=over 4 990=over 4
991
992=item $prog = $ctx->build_program ($program, $options = "")
993
994This convenience function tries to build the program on all devices in
995the context. If the build fails, then the function will C<croak> with the
996build log. Otherwise ti returns the program object.
997
998The C<$program> can either be a C<OpenCL::Program> object or a string
999containing the program. In the latter case, a program objetc will be
1000created automatically.
1001
1002=cut
1003
1004sub OpenCL::Context::build_program {
1005 my ($self, $prog, $options) = @_;
1006
1007 require Carp;
1008
1009 $prog = $self->program_with_source ($prog)
1010 unless ref $prog;
1011
1012 # we build separately per device so we instantly know which one failed
1013 for my $dev ($self->devices) {
1014 eval { $prog->build ([$dev], $options); 1 }
1015 or Carp::croak ("Building OpenCL program for device '" . $dev->name . "' failed:\n"
1016 . $prog->build_log ($dev));
1017 }
1018
1019 $prog
1020}
770 1021
771=item $queue = $ctx->queue ($device, $properties) 1022=item $queue = $ctx->queue ($device, $properties)
772 1023
773Create a new OpenCL::Queue object from the context and the given device. 1024Create a new OpenCL::Queue object from the context and the given device.
774 1025
794=item $buf = $ctx->buffer_sv ($flags, $data) 1045=item $buf = $ctx->buffer_sv ($flags, $data)
795 1046
796Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object and 1047Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object and
797initialise it with the given data values. 1048initialise it with the given data values.
798 1049
1050=item $img = $ctx->image ($self, $flags, $channel_order, $channel_type, $type, $width, $height, $depth, $array_size = 0, $row_pitch = 0, $slice_pitch = 0, $num_mip_level = 0, $num_samples = 0, $*data = &PL_sv_undef)
1051
1052Creates a new OpenCL::Image object and optionally initialises it with
1053the given data values.
1054
1055L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateImage.html>
1056
799=item $img = $ctx->image2d ($flags, $channel_order, $channel_type, $width, $height, $row_pitch = 0, $data = undef) 1057=item $img = $ctx->image2d ($flags, $channel_order, $channel_type, $width, $height, $row_pitch = 0, $data = undef)
800 1058
801Creates a new OpenCL::Image2D object and optionally initialises it with 1059Creates a new OpenCL::Image2D object and optionally initialises it with
802the given data values. 1060the given data values.
803 1061
815Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object that refers to the given 1073Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object that refers to the given
816OpenGL buffer object. 1074OpenGL buffer object.
817 1075
818http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLBuffer.html 1076http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLBuffer.html
819 1077
1078=item $img = $ctx->gl_texture ($flags, $target, $miplevel, $texture)
1079
1080Creates a new OpenCL::Image object that refers to the given OpenGL
1081texture object or buffer.
1082
1083http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateFromGLTexture.html
1084
820=item $ctx->gl_texture2d ($flags, $target, $miplevel, $texture) 1085=item $img = $ctx->gl_texture2d ($flags, $target, $miplevel, $texture)
821 1086
822Creates a new OpenCL::Image2D object that refers to the given OpenGL 1087Creates a new OpenCL::Image2D object that refers to the given OpenGL
8232D texture object. 10882D texture object.
824 1089
825http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture2D.html 1090http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture2D.html
826 1091
827=item $ctx->gl_texture3d ($flags, $target, $miplevel, $texture) 1092=item $img = $ctx->gl_texture3d ($flags, $target, $miplevel, $texture)
828 1093
829Creates a new OpenCL::Image3D object that refers to the given OpenGL 1094Creates a new OpenCL::Image3D object that refers to the given OpenGL
8303D texture object. 10953D texture object.
831 1096
832http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture3D.html 1097http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture3D.html
959 1224
960Yeah. 1225Yeah.
961 1226
962L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>. 1227L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>.
963 1228
1229=item $ev = $queue->enqueue_fill_buffer ($mem, $pattern, $offset, $size, ...)
1230
1231Fills the given buffer object with repeated applications of C<$pattern>,
1232starting at C<$offset> for C<$size> octets.
1233
1234L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueFillBuffer.html>
1235
1236=item $ev = $queue->enqueue_fill_image ($img, $r, $g, $b, $a, $x, $y, $z, $width, $height, $depth, ...)
1237
1238Fills the given image area with the given rgba colour components. The
1239components are normally floating point values between C<0> and C<1>,
1240except when the image channel data type is a signe dor unsigned
1241unnormalised format, in which case the range is determined by the format.
1242
1243L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueFillImage.html>
1244
964=item $ev = $queue->enqueue_task ($kernel, $wait_events...) 1245=item $ev = $queue->enqueue_task ($kernel, $wait_events...)
965 1246
966L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueTask.html> 1247L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueTask.html>
967 1248
968=item $ev = $queue->enqueue_nd_range_kernel ($kernel, @$global_work_offset, @$global_work_size, @$local_work_size, $wait_events...) 1249=item $ev = $queue->enqueue_nd_range_kernel ($kernel, \@global_work_offset, \@global_work_size, \@local_work_size, $wait_events...)
969 1250
970Enqueues a kernel execution. 1251Enqueues a kernel execution.
971 1252
972@$global_work_size must be specified as a reference to an array of 1253\@global_work_size must be specified as a reference to an array of
973integers specifying the work sizes (element counts). 1254integers specifying the work sizes (element counts).
974 1255
975@$global_work_offset must be either C<undef> (in which case all offsets 1256\@global_work_offset must be either C<undef> (in which case all offsets
976are C<0>), or a reference to an array of work offsets, with the same number 1257are C<0>), or a reference to an array of work offsets, with the same number
977of elements as @$global_work_size. 1258of elements as \@global_work_size.
978 1259
979@$local_work_size must be either C<undef> (in which case the 1260\@local_work_size must be either C<undef> (in which case the
980implementation is supposed to choose good local work sizes), or a 1261implementation is supposed to choose good local work sizes), or a
981reference to an array of local work sizes, with the same number of 1262reference to an array of local work sizes, with the same number of
982elements as @$global_work_size. 1263elements as \@global_work_size.
983 1264
984L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueNDRangeKernel.html> 1265L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueNDRangeKernel.html>
985
986=item $ev = $queue->enqueue_marker ($wait_events...)
987
988L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueMarker.html>
989 1266
990=item $ev = $queue->enqueue_acquire_gl_objects ([object, ...], $wait_events...) 1267=item $ev = $queue->enqueue_acquire_gl_objects ([object, ...], $wait_events...)
991 1268
992Enqueues a list (an array-ref of OpenCL::Memory objects) to be acquired 1269Enqueues a list (an array-ref of OpenCL::Memory objects) to be acquired
993for subsequent OpenCL usage. 1270for subsequent OpenCL usage.
1003 1280
1004=item $ev = $queue->enqueue_wait_for_events ($wait_events...) 1281=item $ev = $queue->enqueue_wait_for_events ($wait_events...)
1005 1282
1006L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWaitForEvents.html> 1283L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWaitForEvents.html>
1007 1284
1008=item $queue->enqueue_barrier 1285=item $ev = $queue->enqueue_marker ($wait_events...)
1009 1286
1287L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueMarkerWithWaitList.html>
1288
1289=item $ev = $queue->enqueue_barrier ($wait_events...)
1290
1010L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueBarrier.html> 1291L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueBarrierWithWaitList.html>
1011 1292
1012=item $queue->flush 1293=item $queue->flush
1013 1294
1014L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFlush.html> 1295L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFlush.html>
1015 1296
1130 1411
1131=back 1412=back
1132 1413
1133=head2 THE OpenCL::Image CLASS 1414=head2 THE OpenCL::Image CLASS
1134 1415
1135This is the superclass of all image objects - OpenCL::Image2D and OpenCL::Image3D. 1416This is the superclass of all image objects - OpenCL::Image1D,
1417OpenCL::Image1DArray, OpenCL::Image1DBuffer, OpenCL::Image2D,
1418OpenCL::Image2DArray and OpenCL::Image3D.
1136 1419
1137=over 4 1420=over 4
1138 1421
1139=item $packed_value = $ev->image_info ($name) 1422=item $packed_value = $image->image_info ($name)
1140 1423
1141See C<< $platform->info >> for details. 1424See C<< $platform->info >> for details.
1142 1425
1143The reason this method is not called C<info> is that there already is an 1426The reason this method is not called C<info> is that there already is an
1144C<< ->info >> method inherited from C<OpenCL::Memory>. 1427C<< ->info >> method inherited from C<OpenCL::Memory>.
1145 1428
1146L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetImageInfo.html> 1429L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetImageInfo.html>
1147 1430
1431=item ($channel_order, $channel_data_type) = $image->format
1432
1433Returns the channel order and type used to create the image by calling
1434C<clGetImageInfo> with C<CL_IMAGE_FORMAT>.
1435
1148=for gengetinfo begin image 1436=for gengetinfo begin image
1149 1437
1150=item $int = $image->element_size 1438=item $int = $image->element_size
1151 1439
1152Calls C<clGetImageInfo> with C<CL_IMAGE_ELEMENT_SIZE> and returns the result. 1440Calls C<clGetImageInfo> with C<CL_IMAGE_ELEMENT_SIZE> and returns the result.
1225 1513
1226=head2 THE OpenCL::Program CLASS 1514=head2 THE OpenCL::Program CLASS
1227 1515
1228=over 4 1516=over 4
1229 1517
1230=item $program->build ($device, $options = "") 1518=item $program->build (\@devices = undef, $options = "", $cb->($program) = undef)
1231 1519
1232Tries to build the program with the givne options. 1520Tries to build the program with the given options. See also the
1521C<$ctx->build> convenience function.
1522
1523If a callback is specified, then it will be called when compilation is
1524finished. Note that many OpenCL implementations block your program while
1525compiling whether you use a callback or not. See C<build_async> if you
1526want to make sure the build is done in the background.
1527
1528Note that some OpenCL implementations atc up badly, and don't call the
1529callback in some error cases (but call it in others). This implementation
1530assumes the callback will always be called, and leaks memory if this is
1531not so. So best make sure you don't pass in invalid values.
1233 1532
1234L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html> 1533L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html>
1534
1535=item $program->build_async (\@devices = undef, $options = "", $cb->($program) = undef)
1536
1537Similar to C<< ->build >>, except it starts a thread, and never fails (you
1538need to check the compilation status form the callback, or by polling).
1235 1539
1236=item $packed_value = $program->build_info ($device, $name) 1540=item $packed_value = $program->build_info ($device, $name)
1237 1541
1238Similar to C<< $platform->info >>, but returns build info for a previous 1542Similar to C<< $platform->info >>, but returns build info for a previous
1239build attempt for the given device. 1543build attempt for the given device.
1244 1548
1245Creates an OpenCL::Kernel object out of the named C<__kernel> function in 1549Creates an OpenCL::Kernel object out of the named C<__kernel> function in
1246the program. 1550the program.
1247 1551
1248L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernel.html> 1552L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernel.html>
1553
1554=item @kernels = $program->kernels_in_program
1555
1556Returns all kernels successfully compiled for all devices in program.
1557
1558http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernelsInProgram.html
1249 1559
1250=for gengetinfo begin program_build 1560=for gengetinfo begin program_build
1251 1561
1252=item $build_status = $program->build_status ($device) 1562=item $build_status = $program->build_status ($device)
1253 1563
1389float and double as floating point values, memory/buffer/image2d/image3d 1699float and double as floating point values, memory/buffer/image2d/image3d
1390must be an object of that type or C<undef>, local-memory arguments are 1700must be an object of that type or C<undef>, local-memory arguments are
1391set by specifying the size, and sampler and event must be objects of that 1701set by specifying the size, and sampler and event must be objects of that
1392type. 1702type.
1393 1703
1704Setting an argument for a kernel does NOT keep a reference to the object -
1705for example, if you set an argument to some image object, free the image,
1706and call the kernel, you will run into undefined behaviour.
1707
1394L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetKernelArg.html> 1708L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetKernelArg.html>
1395 1709
1396=back 1710=back
1397 1711
1398=head2 THE OpenCL::Event CLASS 1712=head2 THE OpenCL::Event CLASS
1406 1720
1407Waits for the event to complete. 1721Waits for the event to complete.
1408 1722
1409L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html> 1723L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html>
1410 1724
1725=item $ev->cb ($exec_callback_type, $callback->($event, $event_command_exec_status))
1726
1727Adds a callback to the callback stack for the given event type. There is
1728no way to remove a callback again.
1729
1730L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetEventCallback.html>
1731
1411=item $packed_value = $ev->info ($name) 1732=item $packed_value = $ev->info ($name)
1412 1733
1413See C<< $platform->info >> for details. 1734See C<< $platform->info >> for details.
1414 1735
1415L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetEventInfo.html> 1736L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetEventInfo.html>
1475 1796
1476=over 4 1797=over 4
1477 1798
1478=item $ev->set_status ($execution_status) 1799=item $ev->set_status ($execution_status)
1479 1800
1801Sets the execution status of the user event. Can only be called once,
1802either with OpenCL::COMPLETE or a negative number as status.
1803
1480L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetUserEventStatus.html> 1804L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetUserEventStatus.html>
1481 1805
1482=back 1806=back
1483 1807
1484=cut 1808=cut
1485
1486package OpenCL;
1487
1488use common::sense;
1489
1490BEGIN {
1491 our $VERSION = '0.96';
1492
1493 require XSLoader;
1494 XSLoader::load (__PACKAGE__, $VERSION);
1495
1496 @OpenCL::Buffer::ISA =
1497 @OpenCL::Image::ISA = OpenCL::Memory::;
1498
1499 @OpenCL::BufferObj::ISA = OpenCL::Buffer::;
1500
1501 @OpenCL::Image2D::ISA =
1502 @OpenCL::Image3D::ISA = OpenCL::Image::;
1503
1504 @OpenCL::UserEvent::ISA = OpenCL::Event::;
1505}
1506 1809
15071; 18101;
1508 1811
1509=head1 AUTHOR 1812=head1 AUTHOR
1510 1813

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines