ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/OpenCL/OpenCL.pm

Comparing OpenCL/OpenCL.pm (file contents):
Revision 1.2 by root, Tue Nov 15 09:24:40 2011 UTC vs.
Revision 1.80 by root, Sat May 5 13:55:59 2012 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	OpenCL - bindings to, well, OpenCL	3	OpenCL - Open Computing Language Bindings
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use OpenCL;	7	use OpenCL;
8		8
9	=head1 DESCRIPTION	9	=head1 DESCRIPTION
10		10
11	This is an early release which is not useful yet.	11	This is an early release which might be useful, but hasn't seen much testing.
12		12
		13	=head2 OpenCL FROM 10000 FEET HEIGHT
		14
		15	Here is a high level overview of OpenCL:
		16
		17	First you need to find one or more OpenCL::Platforms (kind of like
		18	vendors) - usually there is only one.
		19
		20	Each platform gives you access to a number of OpenCL::Device objects, e.g.
		21	your graphics card.
		22
		23	From a platform and some device(s), you create an OpenCL::Context, which is
		24	a very central object in OpenCL: Once you have a context you can create
		25	most other objects:
		26
		27	OpenCL::Program objects, which store source code and, after building for a
		28	specific device ("compiling and linking"), also binary programs. For each
		29	kernel function in a program you can then create an OpenCL::Kernel object
		30	which represents basically a function call with argument values.
		31
		32	OpenCL::Memory objects of various flavours: OpenCL::Buffer objects (flat
		33	memory areas, think arrays or structs) and OpenCL::Image objects (think 2D
		34	or 3D array) for bulk data and input and output for kernels.
		35
		36	OpenCL::Sampler objects, which are kind of like texture filter modes in
		37	OpenGL.
		38
		39	OpenCL::Queue objects - command queues, which allow you to submit memory
		40	reads, writes and copies, as well as kernel calls to your devices. They
		41	also offer a variety of methods to synchronise request execution, for
		42	example with barriers or OpenCL::Event objects.
		43
		44	OpenCL::Event objects are used to signal when something is complete.
		45
		46	=head2 HELPFUL RESOURCES
		47
		48	The OpenCL specs used to develop this module - download these and keept
		49	hema round, they are required reference material:
		50
		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
		54
		55	OpenCL manpages:
		56
		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
		60	If you are into UML class diagrams, the following diagram might help - if
		61	not, it will be mildly confusing (also, the class hierarchy of this module
		62	is much more fine-grained):
		63
		64	http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/classDiagram.html
		65
		66	Here's a tutorial from AMD (very AMD-centric, too), not sure how useful it
		67	is, 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
		71	And 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
		74
		75	=head1 BASIC WORKFLOW
		76
		77	To get something done, you basically have to do this once (refer to the
		78	examples below for actual code, this is just a high-level description):
		79
		80	Find some platform (e.g. the first one) and some device(s) (e.g. the first
		81	device of the platform), and create a context from those.
		82
		83	Create program objects from your OpenCL source code, then build (compile)
		84	the programs for each device you want to run them on.
		85
		86	Create kernel objects for all kernels you want to use (surprisingly, these
		87	are not device-specific).
		88
		89	Then, to execute stuff, you repeat these steps, possibly resuing or
		90	sharing some buffers:
		91
		92	Create some input and output buffers from your context. Set these as
		93	arguments to your kernel.
		94
		95	Enqueue buffer writes to initialise your input buffers (when not
		96	initialised at creation time).
		97
		98	Enqueue the kernel execution.
		99
		100	Enqueue buffer reads for your output buffer to read results.
		101
		102	=head1 EXAMPLES
		103
13	Enumerate all devices and get contexts for them;	104	=head2 Enumerate all devices and get contexts for them.
		105
		106	Best run this once to get a feel for the platforms and devices in your
		107	system.
14		108
15	for my $platform (OpenCL::platforms) {	109	for my $platform (OpenCL::platforms) {
16	warn $platform->info (OpenCL::PLATFORM_NAME);	110	printf "platform: %s\n", $platform->name;
17	warn $platform->info (OpenCL::PLATFORM_EXTENSIONS);	111	printf "extensions: %s\n", $platform->extensions;
18	for my $device ($platform->devices) {	112	for my $device ($platform->devices) {
19	warn $device->info (OpenCL::DEVICE_NAME);	113	printf "+ device: %s\n", $device->name;
20	my $ctx = $device->context_simple;	114	my $ctx = $platform->context (undef, [$device]);
21	# do stuff	115	# do stuff
22	}	116	}
23	}	117	}
24		118
25	Get a useful context and a command queue:	119	=head2 Get a useful context and a command queue.
26		120
27	my $dev = ((OpenCL::platforms)[0]->devices)[0];	121	This is a useful boilerplate for any OpenCL program that only wants to use
28	my $ctx = $dev->context_simple;	122	one device,
		123
		124	my ($platform) = OpenCL::platforms; # find first platform
		125	my ($dev) = $platform->devices; # find first device of platform
		126	my $ctx = $platform->context (undef, [$dev]); # create context out of those
		127	my $queue = $ctx->queue ($dev); # create a command queue for the device
		128
		129	=head2 Print all supported image formats of a context.
		130
		131	Best run this once for your context, to see whats available and how to
		132	gather information.
		133
		134	for my $type (OpenCL::MEM_OBJECT_IMAGE2D, OpenCL::MEM_OBJECT_IMAGE3D) {
		135	print "supported image formats for ", OpenCL::enum2str $type, "\n";
		136
		137	for my $f ($ctx->supported_image_formats (0, $type)) {
		138	printf " %-10s %-20s\n", OpenCL::enum2str $f->[0], OpenCL::enum2str $f->[1];
		139	}
		140	}
		141
		142	=head2 Create a buffer with some predefined data, read it back synchronously,
		143	then asynchronously.
		144
		145	my $buf = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, "helmut");
		146
		147	$queue->read_buffer ($buf, 1, 1, 3, my $data);
		148	print "$data\n";
		149
		150	my $ev = $queue->read_buffer ($buf, 0, 1, 3, my $data);
		151	$ev->wait;
		152	print "$data\n"; # prints "elm"
		153
		154	=head2 Create and build a program, then create a kernel out of one of its
		155	functions.
		156
		157	my $src = '
		158	kernel void
		159	squareit (global float *input, global float *output)
		160	{
		161	$id = get_global_id (0);
		162	output [id] = input [id] * input [id];
		163	}
		164	';
		165
		166	my $prog = $ctx->build_program ($src);
		167	my $kernel = $prog->kernel ("squareit");
		168
		169	=head2 Create some input and output float buffers, then call the
		170	'squareit' kernel on them.
		171
		172	my $input = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, pack "f*", 1, 2, 3, 4.5);
		173	my $output = $ctx->buffer (0, OpenCL::SIZEOF_FLOAT * 5);
		174
		175	# set buffer
		176	$kernel->set_buffer (0, $input);
		177	$kernel->set_buffer (1, $output);
		178
		179	# execute it for all 4 numbers
		180	$queue->nd_range_kernel ($kernel, undef, [4], undef);
		181
		182	# enqueue a synchronous read
		183	$queue->read_buffer ($output, 1, 0, OpenCL::SIZEOF_FLOAT * 4, my $data);
		184
		185	# print the results:
		186	printf "%s\n", join ", ", unpack "f*", $data;
		187
		188	=head2 The same enqueue operations as before, but assuming an out-of-order queue,
		189	showing off barriers.
		190
		191	# execute it for all 4 numbers
		192	$queue->nd_range_kernel ($kernel, undef, [4], undef);
		193
		194	# enqueue a barrier to ensure in-order execution
		195	$queue->barrier;
		196
		197	# enqueue an async read
		198	$queue->read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data);
		199
		200	# wait for all requests to finish
		201	$queue->finish;
		202
		203	=head2 The same enqueue operations as before, but assuming an out-of-order queue,
		204	showing off event objects and wait lists.
		205
		206	# execute it for all 4 numbers
		207	my $ev = $queue->nd_range_kernel ($kernel, undef, [4], undef);
		208
		209	# enqueue an async read
		210	$ev = $queue->read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data, $ev);
		211
		212	# wait for the last event to complete
		213	$ev->wait;
		214
		215	=head2 Use the OpenGL module to share a texture between OpenCL and OpenGL and draw some julia
		216	set flight effect.
		217
		218	This is quite a long example to get you going - you can download it from
		219	L<http://cvs.schmorp.de/OpenCL/examples/juliaflight>.
		220
		221	use OpenGL ":all";
		222	use OpenCL;
		223
		224	my $S = $ARGV[0] || 256; # window/texture size, smaller is faster
		225
		226	# open a window and create a gl texture
		227	OpenGL::glpOpenWindow width => $S, height => $S;
		228	my $texid = glGenTextures_p 1;
		229	glBindTexture GL_TEXTURE_2D, $texid;
		230	glTexImage2D_c GL_TEXTURE_2D, 0, GL_RGBA8, $S, $S, 0, GL_RGBA, GL_UNSIGNED_BYTE, 0;
		231
		232	# find and use the first opencl device that let's us get a shared opengl context
		233	my $platform;
		234	my $dev;
		235	my $ctx;
		236
		237	for (OpenCL::platforms) {
		238	$platform = $_;
		239	for ($platform->devices) {
		240	$dev = $_;
		241	$ctx = $platform->context ([OpenCL::GLX_DISPLAY_KHR, undef, OpenCL::GL_CONTEXT_KHR, undef], [$dev])
		242	and last;
		243	}
		244	}
		245
		246	$ctx
		247	or die "cannot find suitable OpenCL device\n";
		248
29	my $queue = $ctx->command_queue_simple ($dev);	249	my $queue = $ctx->queue ($dev);
		250
		251	# now attach an opencl image2d object to the opengl texture
		252	my $tex = $ctx->gl_texture2d (OpenCL::MEM_WRITE_ONLY, GL_TEXTURE_2D, 0, $texid);
		253
		254	# now the boring opencl code
		255	my $src = <<EOF;
		256	kernel void
		257	juliatunnel (write_only image2d_t img, float time)
		258	{
		259	int2 xy = (int2)(get_global_id (0), get_global_id (1));
		260	float2 p = convert_float2 (xy) / $S.f * 2.f - 1.f;
		261
		262	float2 m = (float2)(1.f, p.y) / fabs (p.x); // tunnel
		263	m.x = fabs (fmod (m.x + time * 0.05f, 4.f) - 2.f);
		264
		265	float2 z = m;
		266	float2 c = (float2)(sin (time * 0.01133f), cos (time * 0.02521f));
		267
		268	for (int i = 0; i < 25 && dot (z, z) < 4.f; ++i) // standard julia
		269	z = (float2)(z.x * z.x - z.y * z.y, 2.f * z.x * z.y) + c;
		270
		271	float3 colour = (float3)(z.x, z.y, atan2 (z.y, z.x));
		272	write_imagef (img, xy, (float4)(colour * p.x * p.x, 1.));
		273	}
		274	EOF
		275
		276	my $prog = $ctx->build_program ($src);
		277	my $kernel = $prog->kernel ("juliatunnel");
		278
		279	# program compiled, kernel ready, now draw and loop
		280
		281	for (my $time; ; ++$time) {
		282	# acquire objects from opengl
		283	$queue->acquire_gl_objects ([$tex]);
		284
		285	# configure and run our kernel
		286	$kernel->setf ("mf", $tex, $time*2); # mf = memory object, float
		287	$queue->nd_range_kernel ($kernel, undef, [$S, $S], undef);
		288
		289	# release objects to opengl again
		290	$queue->release_gl_objects ([$tex]);
		291
		292	# wait
		293	$queue->finish;
		294
		295	# now draw the texture, the defaults should be all right
		296	glTexParameterf GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST;
		297
		298	glEnable GL_TEXTURE_2D;
		299	glBegin GL_QUADS;
		300	glTexCoord2f 0, 1; glVertex3i -1, -1, -1;
		301	glTexCoord2f 0, 0; glVertex3i 1, -1, -1;
		302	glTexCoord2f 1, 0; glVertex3i 1, 1, -1;
		303	glTexCoord2f 1, 1; glVertex3i -1, 1, -1;
		304	glEnd;
		305
		306	glXSwapBuffers;
		307
		308	select undef, undef, undef, 1/60;
		309	}
		310
		311	=head2 How to modify the previous example to not rely on GL sharing.
		312
		313	For those poor souls with only a sucky CPU OpenCL implementation, you
		314	currently have to read the image into some perl scalar, and then modify a
		315	texture or use glDrawPixels or so).
		316
		317	First, when you don't need gl sharing, you can create the context much simpler:
		318
		319	$ctx = $platform->context (undef, [$dev])
		320
		321	To use a texture, you would modify the above example by creating an
		322	OpenCL::Image manually instead of deriving it from a texture:
		323
		324	my $tex = $ctx->image2d (OpenCL::MEM_WRITE_ONLY, OpenCL::RGBA, OpenCL::UNORM_INT8, $S, $S);
		325
		326	And in the draw loop, intead of acquire_gl_objects/release_gl_objects, you
		327	would read the image2d after the kernel has written it:
		328
		329	$queue->read_image ($tex, 0, 0, 0, 0, $S, $S, 1, 0, 0, my $data);
		330
		331	And then you would upload the pixel data to the texture (or use glDrawPixels):
		332
		333	glTexSubImage2D_s GL_TEXTURE_2D, 0, 0, 0, $S, $S, GL_RGBA, GL_UNSIGNED_BYTE, $data;
		334
		335	The fully modified example can be found at
		336	L<http://cvs.schmorp.de/OpenCL/examples/juliaflight-nosharing>.
		337
		338	=head2 Julia sets look soooo 80ies.
		339
		340	Then colour them differently, e.g. using orbit traps! Replace the loop and
		341	colour calculation from the previous examples by this:
		342
		343	float2 dm = (float2)(1.f, 1.f);
		344
		345	for (int i = 0; i < 25; ++i)
		346	{
		347	z = (float2)(z.x * z.x - z.y * z.y, 2.f * z.x * z.y) + c;
		348	dm = fmin (dm, (float2)(fabs (dot (z, z) - 1.f), fabs (z.x - 1.f)));
		349	}
		350
		351	float3 colour = (float3)(dm.x * dm.y, dm.x * dm.y, dm.x);
		352
		353	Also try C<-10.f> instead of C<-1.f>.
		354
		355	=head1 DOCUMENTATION
		356
		357	=head2 BASIC CONVENTIONS
		358
		359	This is not a one-to-one C-style translation of OpenCL to Perl - instead
		360	I attempted to make the interface as type-safe as possible by introducing
		361	object syntax where it makes sense. There are a number of important
		362	differences between the OpenCL C API and this module:
30		363
31	=over 4	364	=over 4
32		365
		366	=item * Object lifetime managament is automatic - there is no need
		367	to free objects explicitly (C<clReleaseXXX>), the release function
		368	is called automatically once all Perl references to it go away.
		369
		370	=item * OpenCL uses CamelCase for function names
		371	(e.g. C<clGetPlatformIDs>, C<clGetPlatformInfo>), while this module
		372	uses underscores as word separator and often leaves out prefixes
		373	(C<OpenCL::platforms>, C<< $platform->info >>).
		374
		375	=item * OpenCL often specifies fixed vector function arguments as short
		376	arrays (C<size_t origin[3]>), while this module explicitly expects the
		377	components as separate arguments (C<$orig_x, $orig_y, $orig_z>) in
		378	function calls.
		379
		380	=item * Structures are often specified by flattening out their components
		381	as with short vectors, and returned as arrayrefs.
		382
		383	=item * When enqueuing commands, the wait list is specified by adding
		384	extra arguments to the function - anywhere a C<$wait_events...> argument
		385	is documented this can be any number of event objects. As an extsnion
		386	implemented by this module, C<undef> values will be ignored in the event
		387	list.
		388
		389	=item * When enqueuing commands, if the enqueue method is called in void
		390	context, no event is created. In all other contexts an event is returned
		391	by the method.
		392
		393	=item * This module expects all functions to return C<OpenCL::SUCCESS>. If any
		394	other status is returned the function will throw an exception, so you
		395	don't normally have to to any error checking.
		396
		397	=back
		398
		399	=head2 CONSTANTS
		400
		401	All C<CL_xxx> constants that this module supports are always available
		402	in the C<OpenCL> namespace as C<OpenCL::xxx> (i.e. without the C<CL_>
		403	prefix). Constants which are not defined in the heaer files used during
		404	compilation, or otherwise are not available, will have the value C<-1>.
		405
		406	=head2 OPENCL 1.1 VS. OPENCL 1.2
		407
		408	This module supports both OpenCL version 1.1 and 1.2, although the OpenCL
		409	1.2 interface hasn't been tested much for lack of availability of an
		410	actual implementation.
		411
		412	Every function or method in this manual page that interfaces to a
		413	particular OpenCL function has a link to the its C manual page.
		414
		415	If the link contains a F<1.1>, then this function is an OpenCL 1.1
		416	function. Most but not all also exist in OpenCL 1.2, and this module
		417	tries to emulate the missing ones for you, when told to do so at
		418	compiletime. You cna check whether a function was removed in OpenCL 1.2 by
		419	replacing the F<1.1> component in the URL by F<1.2>.
		420
		421	If the link contains a F<1.2>, then this is a OpenCL 1.2-only
		422	function. Even if the module was compiled with OpenCL 1.2 header files
		423	and has an 1.2 OpenCL library, calling such a function on a platform that
		424	doesn't implement 1.2 causes undefined behaviour, usually a crash (But
		425	this is not guaranteed).
		426
		427	You can find out whether this module was compiled to prefer 1.1
		428	functionality by ooking at C<OpenCL::PREFER_1_1> - if it is true, then
		429	1.1 functions generally are implemented using 1.1 OpenCL functions. If it
		430	is false, then 1.1 functions missing from 1.2 are emulated by calling 1.2
		431	fucntions.
		432
		433	This is a somewhat sorry state of affairs, but the Khronos group choose to
		434	make every release of OpenCL source and binary incompatible with previous
		435	releases.
		436
		437	=head2 PERL AND OPENCL TYPES
		438
		439	This handy(?) table lists OpenCL types and their perl, PDL and pack/unpack
		440	format equivalents:
		441
		442	OpenCL perl PDL pack/unpack
		443	char IV - c
		444	uchar IV byte C
		445	short IV short s
		446	ushort IV ushort S
		447	int IV long? l
		448	uint IV - L
		449	long IV longlong q
		450	ulong IV - Q
		451	float NV float f
		452	half IV ushort S
		453	double NV double d
		454
		455	=head2 GLX SUPPORT
		456
		457	Due to the sad state that OpenGL support is in in Perl (mostly the OpenGL
		458	module, which has little to no documentation and has little to no support
		459	for glX), this module, as a special extension, treats context creation
		460	properties C<OpenCL::GLX_DISPLAY_KHR> and C<OpenCL::GL_CONTEXT_KHR>
		461	specially: If either or both of these are C<undef>, then the OpenCL
		462	module tries to dynamically resolve C<glXGetCurrentDisplay> and
		463	C<glXGetCurrentContext>, call these functions and use their return values
		464	instead.
		465
		466	For this to work, the OpenGL library must be loaded, a GLX context must
		467	have been created and be made current, and C<dlsym> must be available and
		468	capable of finding the function via C<RTLD_DEFAULT>.
		469
		470	=head2 EVENT SYSTEM
		471
		472	OpenCL can generate a number of (potentially) asynchronous events, for
		473	example, after compiling a program, to signal a context-related error or,
		474	perhaps most important, to signal completion of queued jobs (by setting
		475	callbacks on OpenCL::Event objects).
		476
		477	The OpenCL module converts all these callbacks into events - you can
		478	still register callbacks, but they are not executed when your OpenCL
		479	implementation calls the actual callback, but only later. Therefore, none
		480	of the limitations of OpenCL callbacks apply to the perl implementation:
		481	it is perfectly safe to make blocking operations from event callbacks, and
		482	enqueued operations don't need to be flushed.
		483
		484	To facilitate this, this module maintains an event queue - each
		485	time an asynchronous event happens, it is queued, and perl will be
		486	interrupted. This is implemented via the L<Async::Interrupt> module. In
		487	addition, this module has L<AnyEvent> support, so it can seamlessly
		488	integrate itself into many event loops.
		489
		490	Since L<Async::Interrupt> is a bit hard to understand, here are some case examples:
		491
		492	=head3 Don't use callbacks.
		493
		494	When your program never uses any callbacks, then there will never be any
		495	notifications you need to take care of, and therefore no need to worry
		496	about all this.
		497
		498	You can achieve a great deal by explicitly waiting for events, or using
		499	barriers and flush calls. In many programs, there is no need at all to
		500	tinker with asynchronous events.
		501
		502	=head3 Use AnyEvent
		503
		504	This module automatically registers a watcher that invokes all outstanding
		505	event callbacks when AnyEvent is initialised (and block asynchronous
		506	interruptions). Using this mode of operations is the safest and most
		507	recommended one.
		508
		509	To use this, simply use AnyEvent and this module normally, make sure you
		510	have an event loop running:
		511
		512	use Gtk2 -init;
		513	use AnyEvent;
		514
		515	# initialise AnyEvent, by creating a watcher, or:
		516	AnyEvent::detect;
		517
		518	my $e = $queue->marker;
		519	$e->cb (sub {
		520	warn "opencl is finished\n";
		521	})
		522
		523	main Gtk2;
		524
		525	Note that this module will not initialise AnyEvent for you. Before
		526	AnyEvent is initialised, the module will asynchronously interrupt perl
		527	instead. To avoid any surprises, it's best to explicitly initialise
		528	AnyEvent.
		529
		530	You can temporarily enable asynchronous interruptions (see next paragraph)
		531	by calling C<$OpenCL::INTERRUPT->unblock> and disable them again by
		532	calling C<$OpenCL::INTERRUPT->block>.
		533
		534	=head3 Let yourself be interrupted at any time
		535
		536	This mode is the default unless AnyEvent is loaded and initialised. In
		537	this mode, OpenCL asynchronously interrupts a running perl program. The
		538	emphasis is on both I<asynchronously> and I<running> here.
		539
		540	Asynchronously means that perl might execute your callbacks at any
		541	time. For example, in the following code (I<THAT YOU SHOULD NOT COPY>),
		542	the C<until> loop following the marker call will be interrupted by the
		543	callback:
		544
		545	my $e = $queue->marker;
		546	my $flag;
		547	$e->cb (sub { $flag = 1 });
		548	1 until $flag;
		549	# $flag is now 1
		550
		551	The reason why you shouldn't blindly copy the above code is that
		552	busy waiting is a really really bad thing, and really really bad for
		553	performance.
		554
		555	While at first this asynchronous business might look exciting, it can be
		556	really hard, because you need to be prepared for the callback code to be
		557	executed at any time, which limits the amount of things the callback code
		558	can do safely.
		559
		560	This can be mitigated somewhat by using C<<
		561	$OpenCL::INTERRUPT->scope_block >> (see the L<Async::Interrupt>
		562	documentation for details).
		563
		564	The other problem is that your program must be actively I<running> to be
		565	interrupted. When you calculate stuff, your program is running. When you
		566	hang in some C functions or other block execution (by calling C<sleep>,
		567	C<select>, running an event loop and so on), your program is waiting, not
		568	running.
		569
		570	One way around that would be to attach a read watcher to your event loop,
		571	listening for events on C<< $OpenCL::INTERRUPT->pipe_fileno >>, using a
		572	dummy callback (C<sub { }>) to temporarily execute some perl code.
		573
		574	That is then awfully close to using the built-in AnyEvent support above,
		575	though, so consider that one instead.
		576
		577	=head3 Be creative
		578
		579	OpenCL exports the L<Async::Interrupt> object it uses in the global
		580	variable C<$OpenCL::INTERRUPT>. You can configure it in any way you like.
		581
		582	So if you want to feel like a real pro, err, wait, if you feel no risk
		583	menas no fun, you can experiment by implementing your own mode of
		584	operations.
		585
33	=cut	586	=cut
34		587
35	package OpenCL;	588	package OpenCL;
36		589
37	use common::sense;	590	use common::sense;
		591	use Carp ();
		592	use Async::Interrupt ();
		593
		594	our $POLL_FUNC; # set by XS
38		595
39	BEGIN {	596	BEGIN {
40	our $VERSION = '0.01';	597	our $VERSION = '0.99';
41		598
42	require XSLoader;	599	require XSLoader;
43	XSLoader::load (__PACKAGE__, $VERSION);	600	XSLoader::load (__PACKAGE__, $VERSION);
		601
		602	@OpenCL::Platform::ISA =
		603	@OpenCL::Device::ISA =
		604	@OpenCL::Context::ISA =
		605	@OpenCL::Queue::ISA =
		606	@OpenCL::Memory::ISA =
		607	@OpenCL::Sampler::ISA =
		608	@OpenCL::Program::ISA =
		609	@OpenCL::Kernel::ISA =
		610	@OpenCL::Event::ISA = OpenCL::Object::;
		611
		612	@OpenCL::SubDevice::ISA = OpenCL::Device::;
		613
		614	@OpenCL::Buffer::ISA =
		615	@OpenCL::Image::ISA = OpenCL::Memory::;
		616
		617	@OpenCL::BufferObj::ISA = OpenCL::Buffer::;
		618
		619	@OpenCL::Image2D::ISA =
		620	@OpenCL::Image3D::ISA =
		621	@OpenCL::Image2DArray::ISA =
		622	@OpenCL::Image1D::ISA =
		623	@OpenCL::Image1DArray::ISA =
		624	@OpenCL::Image1DBuffer::ISA = OpenCL::Image::;
		625
		626	@OpenCL::UserEvent::ISA = OpenCL::Event::;
		627
		628	@OpenCL::MappedBuffer::ISA =
		629	@OpenCL::MappedImage::ISA = OpenCL::Mapped::;
44	}	630	}
45		631
		632	=head2 THE OpenCL PACKAGE
		633
		634	=over 4
		635
		636	=item $int = OpenCL::errno
		637
		638	The last error returned by a function - it's only valid after an error occured
		639	and before calling another OpenCL function.
		640
		641	=item $str = OpenCL::err2str [$errval]
		642
		643	Converts an error value into a human readable string. IF no error value is
		644	given, then the last error will be used (as returned by OpenCL::errno).
		645
		646	=item $str = OpenCL::enum2str $enum
		647
		648	Converts most enum values (of parameter names, image format constants,
		649	object types, addressing and filter modes, command types etc.) into a
		650	human readable string. When confronted with some random integer it can be
		651	very helpful to pass it through this function to maybe get some readable
		652	string out of it.
		653
		654	=item @platforms = OpenCL::platforms
		655
		656	Returns all available OpenCL::Platform objects.
		657
		658	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html>
		659
		660	=item $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr
		661
		662	Tries to create a context from a default device and platform type - never worked for me.
		663	Consider using C<< $platform->context_from_type >> instead.
		664
		665	type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, OpenCL::DEVICE_TYPE_GPU,
		666	OpenCL::DEVICE_TYPE_ACCELERATOR, OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL.
		667
		668	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html>
		669
		670	=item $ctx = OpenCL::context $properties, \@devices, $callback->($err, $pvt) = $print_stderr)
		671
		672	Create a new OpenCL::Context object using the given device object(s).
		673	Consider using C<< $platform->context >> instead.
		674
		675	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html>
		676
		677	=item OpenCL::wait_for_events $wait_events...
		678
		679	Waits for all events to complete.
		680
		681	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html>
		682
		683	=item OpenCL::poll
		684
		685	Checks if there are any outstanding events (see L<EVENT SYSTEM>) and
		686	invokes their callbacks.
		687
		688	=item $OpenCL::INTERRUPT
		689
		690	The L<Async::Interrupt> object used to signal asynchronous events (see
		691	L<EVENT SYSTEM>).
		692
		693	=cut
		694
		695	our $INTERRUPT = new Async::Interrupt c_cb => [$POLL_FUNC, 0];
		696
		697	&_eq_initialise ($INTERRUPT->signal_func);
		698
		699	=item $OpenCL::WATCHER
		700
		701	The L<AnyEvent> watcher object used to watch for asynchronous events (see
		702	L<EVENT SYSTEM>). This variable is C<undef> until L<AnyEvent> has been
		703	loaded I<and> initialised (e.g. by calling C<AnyEvent::detect>).
		704
		705	=cut
		706
		707	our $WATCHER;
		708
		709	sub _init_anyevent {
		710	$INTERRUPT->block;
		711	$WATCHER = AE::io ($INTERRUPT->pipe_fileno, 0, sub { $INTERRUPT->handle });
		712	}
		713
		714	if (defined $AnyEvent::MODEL) {
		715	_init_anyevent;
		716	} else {
		717	push @AnyEvent::post_detect, \&_init_anyevent;
		718	}
		719
		720	=back
		721
		722	=head2 THE OpenCL::Object CLASS
		723
		724	This is the base class for all objects in the OpenCL module. The only
		725	method it implements is the C<id> method, which is only useful if you want
		726	to interface to OpenCL on the C level.
		727
		728	=over 4
		729
		730	=item $iv = $obj->id
		731
		732	OpenCL objects are represented by pointers or integers on the C level. If
		733	you want to interface to an OpenCL object directly on the C level, then
		734	you need this value, which is returned by this method. You should use an
		735	C<IV> type in your code and cast that to the correct type.
		736
		737	=cut
		738
		739	sub OpenCL::Object::id {
		740	ref $_[0] eq "SCALAR"
		741	? ${ $_[0] }
		742	: $_[0][0]
		743	}
		744
		745	=back
		746
		747	=head2 THE OpenCL::Platform CLASS
		748
		749	=over 4
		750
		751	=item @devices = $platform->devices ($type = OpenCL::DEVICE_TYPE_ALL)
		752
		753	Returns a list of matching OpenCL::Device objects.
		754
		755	=item $ctx = $platform->context_from_type ($properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr)
		756
		757	Tries to create a context. Never worked for me, and you need devices explicitly anyway.
		758
		759	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html>
		760
		761	=item $ctx = $platform->context ($properties, \@devices, $callback->($err, $pvt) = $print_stderr)
		762
		763	Create a new OpenCL::Context object using the given device object(s)- a
		764	OpenCL::CONTEXT_PLATFORM property is supplied automatically.
		765
		766	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html>
		767
		768	=item $packed_value = $platform->info ($name)
		769
		770	Calls C<clGetPlatformInfo> and returns the packed, raw value - for
		771	strings, this will be the string (possibly including terminating \0), for
		772	other values you probably need to use the correct C<unpack>.
		773
		774	It's best to avoid this method and use one of the following convenience
		775	wrappers.
		776
		777	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformInfo.html>
		778
		779	=item $platform->unload_compiler
		780
		781	Attempts to unload the compiler for this platform, for endless
		782	profit. Does nothing on OpenCL 1.1.
		783
		784	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clUnloadPlatformCompiler.html>
		785
		786	=for gengetinfo begin platform
		787
		788	=item $string = $platform->profile
		789
		790	Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_PROFILE> and returns the result.
		791
		792	=item $string = $platform->version
		793
		794	Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_VERSION> and returns the result.
		795
		796	=item $string = $platform->name
		797
		798	Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_NAME> and returns the result.
		799
		800	=item $string = $platform->vendor
		801
		802	Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_VENDOR> and returns the result.
		803
		804	=item $string = $platform->extensions
		805
		806	Calls C<clGetPlatformInfo> with C<OpenCL::PLATFORM_EXTENSIONS> and returns the result.
		807
		808	=for gengetinfo end platform
		809
		810	=back
		811
		812	=head2 THE OpenCL::Device CLASS
		813
		814	=over 4
		815
		816	=item $packed_value = $device->info ($name)
		817
		818	See C<< $platform->info >> for details.
		819
		820	type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU,
		821	OpenCL::DEVICE_TYPE_GPU, OpenCL::DEVICE_TYPE_ACCELERATOR,
		822	OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL.
		823
		824	fp_config: OpenCL::FP_DENORM, OpenCL::FP_INF_NAN, OpenCL::FP_ROUND_TO_NEAREST,
		825	OpenCL::FP_ROUND_TO_ZERO, OpenCL::FP_ROUND_TO_INF, OpenCL::FP_FMA,
		826	OpenCL::FP_SOFT_FLOAT, OpenCL::FP_CORRECTLY_ROUNDED_DIVIDE_SQRT.
		827
		828	mem_cache_type: OpenCL::NONE, OpenCL::READ_ONLY_CACHE, OpenCL::READ_WRITE_CACHE.
		829
		830	local_mem_type: OpenCL::LOCAL, OpenCL::GLOBAL.
		831
		832	exec_capabilities: OpenCL::EXEC_KERNEL, OpenCL::EXEC_NATIVE_KERNEL.
		833
		834	command_queue_properties: OpenCL::QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE,
		835	OpenCL::QUEUE_PROFILING_ENABLE.
		836
		837	partition_properties: OpenCL::DEVICE_PARTITION_EQUALLY,
		838	OpenCL::DEVICE_PARTITION_BY_COUNTS, OpenCL::DEVICE_PARTITION_BY_COUNTS_LIST_END,
		839	OpenCL::DEVICE_PARTITION_BY_AFFINITY_DOMAIN.
		840
		841	affinity_domain: OpenCL::DEVICE_AFFINITY_DOMAIN_NUMA,
		842	OpenCL::DEVICE_AFFINITY_DOMAIN_L4_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L3_CACHE,
		843	OpenCL::DEVICE_AFFINITY_DOMAIN_L2_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L1_CACHE,
		844	OpenCL::DEVICE_AFFINITY_DOMAIN_NEXT_PARTITIONABLE.
		845
		846	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetDeviceInfo.html>
		847
		848	=item @devices = $device->sub_devices (\@properties)
		849
		850	Creates OpencL::SubDevice objects by partitioning an existing device.
		851
		852	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateSubDevices.html>
		853
		854	=for gengetinfo begin device
		855
		856	=item $device_type = $device->type
		857
		858	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_TYPE> and returns the result.
		859
		860	=item $uint = $device->vendor_id
		861
		862	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_VENDOR_ID> and returns the result.
		863
		864	=item $uint = $device->max_compute_units
		865
		866	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_COMPUTE_UNITS> and returns the result.
		867
		868	=item $uint = $device->max_work_item_dimensions
		869
		870	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WORK_ITEM_DIMENSIONS> and returns the result.
		871
		872	=item $int = $device->max_work_group_size
		873
		874	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WORK_GROUP_SIZE> and returns the result.
		875
		876	=item @ints = $device->max_work_item_sizes
		877
		878	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WORK_ITEM_SIZES> and returns the result.
		879
		880	=item $uint = $device->preferred_vector_width_char
		881
		882	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_CHAR> and returns the result.
		883
		884	=item $uint = $device->preferred_vector_width_short
		885
		886	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_SHORT> and returns the result.
		887
		888	=item $uint = $device->preferred_vector_width_int
		889
		890	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_INT> and returns the result.
		891
		892	=item $uint = $device->preferred_vector_width_long
		893
		894	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_LONG> and returns the result.
		895
		896	=item $uint = $device->preferred_vector_width_float
		897
		898	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT> and returns the result.
		899
		900	=item $uint = $device->preferred_vector_width_double
		901
		902	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE> and returns the result.
		903
		904	=item $uint = $device->max_clock_frequency
		905
		906	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_CLOCK_FREQUENCY> and returns the result.
		907
		908	=item $bitfield = $device->address_bits
		909
		910	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_ADDRESS_BITS> and returns the result.
		911
		912	=item $uint = $device->max_read_image_args
		913
		914	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_READ_IMAGE_ARGS> and returns the result.
		915
		916	=item $uint = $device->max_write_image_args
		917
		918	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_WRITE_IMAGE_ARGS> and returns the result.
		919
		920	=item $ulong = $device->max_mem_alloc_size
		921
		922	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_MEM_ALLOC_SIZE> and returns the result.
		923
		924	=item $int = $device->image2d_max_width
		925
		926	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE2D_MAX_WIDTH> and returns the result.
		927
		928	=item $int = $device->image2d_max_height
		929
		930	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE2D_MAX_HEIGHT> and returns the result.
		931
		932	=item $int = $device->image3d_max_width
		933
		934	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE3D_MAX_WIDTH> and returns the result.
		935
		936	=item $int = $device->image3d_max_height
		937
		938	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE3D_MAX_HEIGHT> and returns the result.
		939
		940	=item $int = $device->image3d_max_depth
		941
		942	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE3D_MAX_DEPTH> and returns the result.
		943
		944	=item $uint = $device->image_support
		945
		946	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_IMAGE_SUPPORT> and returns the result.
		947
		948	=item $int = $device->max_parameter_size
		949
		950	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_PARAMETER_SIZE> and returns the result.
		951
		952	=item $uint = $device->max_samplers
		953
		954	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_SAMPLERS> and returns the result.
		955
		956	=item $uint = $device->mem_base_addr_align
		957
		958	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MEM_BASE_ADDR_ALIGN> and returns the result.
		959
		960	=item $uint = $device->min_data_type_align_size
		961
		962	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MIN_DATA_TYPE_ALIGN_SIZE> and returns the result.
		963
		964	=item $device_fp_config = $device->single_fp_config
		965
		966	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_SINGLE_FP_CONFIG> and returns the result.
		967
		968	=item $device_mem_cache_type = $device->global_mem_cache_type
		969
		970	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_CACHE_TYPE> and returns the result.
		971
		972	=item $uint = $device->global_mem_cacheline_size
		973
		974	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_CACHELINE_SIZE> and returns the result.
		975
		976	=item $ulong = $device->global_mem_cache_size
		977
		978	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_CACHE_SIZE> and returns the result.
		979
		980	=item $ulong = $device->global_mem_size
		981
		982	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_GLOBAL_MEM_SIZE> and returns the result.
		983
		984	=item $ulong = $device->max_constant_buffer_size
		985
		986	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_CONSTANT_BUFFER_SIZE> and returns the result.
		987
		988	=item $uint = $device->max_constant_args
		989
		990	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_MAX_CONSTANT_ARGS> and returns the result.
		991
		992	=item $device_local_mem_type = $device->local_mem_type
		993
		994	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_LOCAL_MEM_TYPE> and returns the result.
		995
		996	=item $ulong = $device->local_mem_size
		997
		998	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_LOCAL_MEM_SIZE> and returns the result.
		999
		1000	=item $boolean = $device->error_correction_support
		1001
		1002	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_ERROR_CORRECTION_SUPPORT> and returns the result.
		1003
		1004	=item $int = $device->profiling_timer_resolution
		1005
		1006	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PROFILING_TIMER_RESOLUTION> and returns the result.
		1007
		1008	=item $boolean = $device->endian_little
		1009
		1010	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_ENDIAN_LITTLE> and returns the result.
		1011
		1012	=item $boolean = $device->available
		1013
		1014	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_AVAILABLE> and returns the result.
		1015
		1016	=item $boolean = $device->compiler_available
		1017
		1018	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_COMPILER_AVAILABLE> and returns the result.
		1019
		1020	=item $device_exec_capabilities = $device->execution_capabilities
		1021
		1022	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_EXECUTION_CAPABILITIES> and returns the result.
		1023
		1024	=item $command_queue_properties = $device->properties
		1025
		1026	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_QUEUE_PROPERTIES> and returns the result.
		1027
		1028	=item $ = $device->platform
		1029
		1030	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PLATFORM> and returns the result.
		1031
		1032	=item $string = $device->name
		1033
		1034	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NAME> and returns the result.
		1035
		1036	=item $string = $device->vendor
		1037
		1038	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_VENDOR> and returns the result.
		1039
		1040	=item $string = $device->driver_version
		1041
		1042	Calls C<clGetDeviceInfo> with C<OpenCL::DRIVER_VERSION> and returns the result.
		1043
		1044	=item $string = $device->profile
		1045
		1046	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PROFILE> and returns the result.
		1047
		1048	=item $string = $device->version
		1049
		1050	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_VERSION> and returns the result.
		1051
		1052	=item $string = $device->extensions
		1053
		1054	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_EXTENSIONS> and returns the result.
		1055
		1056	=item $uint = $device->preferred_vector_width_half
		1057
		1058	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_HALF> and returns the result.
		1059
		1060	=item $uint = $device->native_vector_width_char
		1061
		1062	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_CHAR> and returns the result.
		1063
		1064	=item $uint = $device->native_vector_width_short
		1065
		1066	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_SHORT> and returns the result.
		1067
		1068	=item $uint = $device->native_vector_width_int
		1069
		1070	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_INT> and returns the result.
		1071
		1072	=item $uint = $device->native_vector_width_long
		1073
		1074	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_LONG> and returns the result.
		1075
		1076	=item $uint = $device->native_vector_width_float
		1077
		1078	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_FLOAT> and returns the result.
		1079
		1080	=item $uint = $device->native_vector_width_double
		1081
		1082	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE> and returns the result.
		1083
		1084	=item $uint = $device->native_vector_width_half
		1085
		1086	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_HALF> and returns the result.
		1087
		1088	=item $device_fp_config = $device->double_fp_config
		1089
		1090	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_DOUBLE_FP_CONFIG> and returns the result.
		1091
		1092	=item $device_fp_config = $device->half_fp_config
		1093
		1094	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_HALF_FP_CONFIG> and returns the result.
		1095
		1096	=item $boolean = $device->host_unified_memory
		1097
		1098	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_HOST_UNIFIED_MEMORY> and returns the result.
		1099
		1100	=item $device = $device->parent_device_ext
		1101
		1102	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PARENT_DEVICE_EXT> and returns the result.
		1103
		1104	=item @device_partition_property_exts = $device->partition_types_ext
		1105
		1106	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PARTITION_TYPES_EXT> and returns the result.
		1107
		1108	=item @device_partition_property_exts = $device->affinity_domains_ext
		1109
		1110	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_AFFINITY_DOMAINS_EXT> and returns the result.
		1111
		1112	=item $uint = $device->reference_count_ext
		1113
		1114	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_REFERENCE_COUNT_EXT> and returns the result.
		1115
		1116	=item @device_partition_property_exts = $device->partition_style_ext
		1117
		1118	Calls C<clGetDeviceInfo> with C<OpenCL::DEVICE_PARTITION_STYLE_EXT> and returns the result.
		1119
		1120	=for gengetinfo end device
		1121
		1122	=back
		1123
		1124	=head2 THE OpenCL::Context CLASS
		1125
		1126	An OpenCL::Context is basically a container, or manager, for a number of
		1127	devices of a platform. It is used to create all sorts of secondary objects
		1128	such as buffers, queues, programs and so on.
		1129
		1130	All context creation functions and methods take a list of properties
		1131	(type-value pairs). All property values can be specified as integers -
		1132	some additionally support other types:
		1133
		1134	=over 4
		1135
		1136	=item OpenCL::CONTEXT_PLATFORM
		1137
		1138	Also accepts OpenCL::Platform objects.
		1139
		1140	=item OpenCL::GLX_DISPLAY_KHR
		1141
		1142	Also accepts C<undef>, in which case a deep and troubling hack is engaged
		1143	to find the current glx display (see L<GLX SUPPORT>).
		1144
		1145	=item OpenCL::GL_CONTEXT_KHR
		1146
		1147	Also accepts C<undef>, in which case a deep and troubling hack is engaged
		1148	to find the current glx context (see L<GLX SUPPORT>).
		1149
		1150	=back
		1151
		1152	=over 4
		1153
		1154	=item $prog = $ctx->build_program ($program, $options = "")
		1155
		1156	This convenience function tries to build the program on all devices in
		1157	the context. If the build fails, then the function will C<croak> with the
		1158	build log. Otherwise ti returns the program object.
		1159
		1160	The C<$program> can either be a C<OpenCL::Program> object or a string
		1161	containing the program. In the latter case, a program objetc will be
		1162	created automatically.
		1163
		1164	=cut
		1165
		1166	sub OpenCL::Context::build_program {
		1167	my ($self, $prog, $options) = @_;
		1168
		1169	$prog = $self->program_with_source ($prog)
		1170	unless ref $prog;
		1171
		1172	eval { $prog->build (undef, $options); 1 }
		1173	or errno == BUILD_PROGRAM_FAILURE
		1174	or errno == INVALID_BINARY # workaround nvidia bug
		1175	or Carp::croak "OpenCL::Context->build_program: " . err2str;
		1176
		1177	# we check status for all devices
		1178	for my $dev ($self->devices) {
		1179	$prog->build_status ($dev) == BUILD_SUCCESS
		1180	or Carp::croak "Building OpenCL program for device '" . $dev->name . "' failed:\n"
		1181	. $prog->build_log ($dev);
		1182	}
		1183
		1184	$prog
		1185	}
		1186
		1187	=item $queue = $ctx->queue ($device, $properties)
		1188
		1189	Create a new OpenCL::Queue object from the context and the given device.
		1190
		1191	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateCommandQueue.html>
		1192
		1193	Example: create an out-of-order queue.
		1194
		1195	$queue = $ctx->queue ($device, OpenCL::QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE);
		1196
		1197	=item $ev = $ctx->user_event
		1198
		1199	Creates a new OpenCL::UserEvent object.
		1200
		1201	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateUserEvent.html>
		1202
		1203	=item $buf = $ctx->buffer ($flags, $len)
		1204
		1205	Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object with the
		1206	given flags and octet-size.
		1207
		1208	flags: OpenCL::MEM_READ_WRITE, OpenCL::MEM_WRITE_ONLY, OpenCL::MEM_READ_ONLY,
		1209	OpenCL::MEM_USE_HOST_PTR, OpenCL::MEM_ALLOC_HOST_PTR, OpenCL::MEM_COPY_HOST_PTR,
		1210	OpenCL::MEM_HOST_WRITE_ONLY, OpenCL::MEM_HOST_READ_ONLY, OpenCL::MEM_HOST_NO_ACCESS.
		1211
		1212	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateBuffer.html>
		1213
		1214	=item $buf = $ctx->buffer_sv ($flags, $data)
		1215
		1216	Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object and
		1217	initialise it with the given data values.
		1218
		1219	=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)
		1220
		1221	Creates a new OpenCL::Image object and optionally initialises it with
		1222	the given data values.
		1223
		1224	channel_order: OpenCL::R, OpenCL::A, OpenCL::RG, OpenCL::RA, OpenCL::RGB,
		1225	OpenCL::RGBA, OpenCL::BGRA, OpenCL::ARGB, OpenCL::INTENSITY, OpenCL::LUMINANCE,
		1226	OpenCL::Rx, OpenCL::RGx, OpenCL::RGBx.
		1227
		1228	channel_type: OpenCL::SNORM_INT8, OpenCL::SNORM_INT16, OpenCL::UNORM_INT8,
		1229	OpenCL::UNORM_INT16, OpenCL::UNORM_SHORT_565, OpenCL::UNORM_SHORT_555,
		1230	OpenCL::UNORM_INT_101010, OpenCL::SIGNED_INT8, OpenCL::SIGNED_INT16,
		1231	OpenCL::SIGNED_INT32, OpenCL::UNSIGNED_INT8, OpenCL::UNSIGNED_INT16,
		1232	OpenCL::UNSIGNED_INT32, OpenCL::HALF_FLOAT, OpenCL::FLOAT.
		1233
		1234	type: OpenCL::MEM_OBJECT_BUFFER, OpenCL::MEM_OBJECT_IMAGE2D,
		1235	OpenCL::MEM_OBJECT_IMAGE3D, OpenCL::MEM_OBJECT_IMAGE2D_ARRAY,
		1236	OpenCL::MEM_OBJECT_IMAGE1D, OpenCL::MEM_OBJECT_IMAGE1D_ARRAY,
		1237	OpenCL::MEM_OBJECT_IMAGE1D_BUFFER.
		1238
		1239	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateImage.html>
		1240
		1241	=item $img = $ctx->image2d ($flags, $channel_order, $channel_type, $width, $height, $row_pitch = 0, $data = undef)
		1242
		1243	Creates a new OpenCL::Image2D object and optionally initialises it with
		1244	the given data values.
		1245
		1246	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateImage2D.html>
		1247
		1248	=item $img = $ctx->image3d ($flags, $channel_order, $channel_type, $width, $height, $depth, $row_pitch = 0, $slice_pitch = 0, $data = undef)
		1249
		1250	Creates a new OpenCL::Image3D object and optionally initialises it with
		1251	the given data values.
		1252
		1253	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateImage3D.html>
		1254
		1255	=item $buffer = $ctx->gl_buffer ($flags, $bufobj)
		1256
		1257	Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object that refers to the given
		1258	OpenGL buffer object.
		1259
		1260	flags: OpenCL::MEM_READ_WRITE, OpenCL::MEM_READ_ONLY, OpenCL::MEM_WRITE_ONLY.
		1261
		1262	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLBuffer.html
		1263
		1264	=item $img = $ctx->gl_texture ($flags, $target, $miplevel, $texture)
		1265
		1266	Creates a new OpenCL::Image object that refers to the given OpenGL
		1267	texture object or buffer.
		1268
		1269	target: GL_TEXTURE_1D, GL_TEXTURE_1D_ARRAY, GL_TEXTURE_BUFFER,
		1270	GL_TEXTURE_2D, GL_TEXTURE_2D_ARRAY, GL_TEXTURE_3D,
		1271	GL_TEXTURE_CUBE_MAP_POSITIVE_X, GL_TEXTURE_CUBE_MAP_POSITIVE_Y,
		1272	GL_TEXTURE_CUBE_MAP_POSITIVE_Z, GL_TEXTURE_CUBE_MAP_NEGATIVE_X,
		1273	GL_TEXTURE_CUBE_MAP_NEGATIVE_Y, GL_TEXTURE_CUBE_MAP_NEGATIVE_Z,
		1274	GL_TEXTURE_RECTANGLE/GL_TEXTURE_RECTANGLE_ARB.
		1275
		1276	http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateFromGLTexture.html
		1277
		1278	=item $img = $ctx->gl_texture2d ($flags, $target, $miplevel, $texture)
		1279
		1280	Creates a new OpenCL::Image2D object that refers to the given OpenGL
		1281	2D texture object.
		1282
		1283	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture2D.html
		1284
		1285	=item $img = $ctx->gl_texture3d ($flags, $target, $miplevel, $texture)
		1286
		1287	Creates a new OpenCL::Image3D object that refers to the given OpenGL
		1288	3D texture object.
		1289
		1290	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLTexture3D.html
		1291
		1292	=item $ctx->gl_renderbuffer ($flags, $renderbuffer)
		1293
		1294	Creates a new OpenCL::Image2D object that refers to the given OpenGL
		1295	render buffer.
		1296
		1297	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFromGLRenderbuffer.html
		1298
		1299	=item @formats = $ctx->supported_image_formats ($flags, $image_type)
		1300
		1301	Returns a list of matching image formats - each format is an arrayref with
		1302	two values, $channel_order and $channel_type, in it.
		1303
		1304	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetSupportedImageFormats.html>
		1305
		1306	=item $sampler = $ctx->sampler ($normalized_coords, $addressing_mode, $filter_mode)
		1307
		1308	Creates a new OpenCL::Sampler object.
		1309
		1310	addressing_mode: OpenCL::ADDRESS_NONE, OpenCL::ADDRESS_CLAMP_TO_EDGE,
		1311	OpenCL::ADDRESS_CLAMP, OpenCL::ADDRESS_REPEAT, OpenCL::ADDRESS_MIRRORED_REPEAT.
		1312
		1313	filter_mode: OpenCL::FILTER_NEAREST, OpenCL::FILTER_LINEAR.
		1314
		1315	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateSampler.html>
		1316
		1317	=item $program = $ctx->program_with_source ($string)
		1318
		1319	Creates a new OpenCL::Program object from the given source code.
		1320
		1321	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateProgramWithSource.html>
		1322
		1323	=item ($program, \@status) = $ctx->program_with_binary (\@devices, \@binaries)
		1324
		1325	Creates a new OpenCL::Program object from the given binaries.
		1326
		1327	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateProgramWithBinary.html>
		1328
		1329	Example: clone an existing program object that contains a successfully
		1330	compiled program, no matter how useless this is.
		1331
		1332	my $clone = $ctx->program_with_binary ([$prog->devices], [$prog->binaries]);
		1333
		1334	=item $program = $ctx->program_with_built_in_kernels (\@devices, $kernel_names)
		1335
		1336	Creates a new OpenCL::Program object from the given built-in kernel names.
		1337
		1338	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateProgramWithBuiltInKernels.html>
		1339
		1340	=item $program = $ctx->link_program (\@devices, $options, \@programs, $cb->($program) = undef)
		1341
		1342	Links all (already compiled) program objects specified in C<@programs>
		1343	together and returns a new OpenCL::Program object with the result.
		1344
		1345	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clLinkProgram.html>
		1346
		1347	=item $packed_value = $ctx->info ($name)
		1348
		1349	See C<< $platform->info >> for details.
		1350
		1351	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetContextInfo.html>
		1352
		1353	=for gengetinfo begin context
		1354
		1355	=item $uint = $context->reference_count
		1356
		1357	Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_REFERENCE_COUNT> and returns the result.
		1358
		1359	=item @devices = $context->devices
		1360
		1361	Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_DEVICES> and returns the result.
		1362
		1363	=item @property_ints = $context->properties
		1364
		1365	Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_PROPERTIES> and returns the result.
		1366
		1367	=item $uint = $context->num_devices
		1368
		1369	Calls C<clGetContextInfo> with C<OpenCL::CONTEXT_NUM_DEVICES> and returns the result.
		1370
		1371	=for gengetinfo end context
		1372
		1373	=back
		1374
		1375	=head2 THE OpenCL::Queue CLASS
		1376
		1377	An OpenCL::Queue represents an execution queue for OpenCL. You execute
		1378	requests by calling their respective method and waiting for it to complete
		1379	in some way.
		1380
		1381	Most methods that enqueue some request return an event object that can
		1382	be used to wait for completion (optionally using a callback), unless
		1383	the method is called in void context, in which case no event object is
		1384	created.
		1385
		1386	They also allow you to specify any number of other event objects that this
		1387	request has to wait for before it starts executing, by simply passing the
		1388	event objects as extra parameters to the enqueue methods. To simplify
		1389	program design, this module ignores any C<undef> values in the list of
		1390	events. This makes it possible to code operations such as this, without
		1391	having to put a valid event object into C<$event> first:
		1392
		1393	$event = $queue->xxx (..., $event);
		1394
		1395	Queues execute in-order by default, without any parallelism, so in most
		1396	cases (i.e. you use only one queue) it's not necessary to wait for or
		1397	create event objects, althoguh an our of order queue is often a bit
		1398	faster.
		1399
		1400	=over 4
		1401
		1402	=item $ev = $queue->read_buffer ($buffer, $blocking, $offset, $len, $data, $wait_events...)
		1403
		1404	Reads data from buffer into the given string.
		1405
		1406	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBuffer.html>
		1407
		1408	=item $ev = $queue->write_buffer ($buffer, $blocking, $offset, $data, $wait_events...)
		1409
		1410	Writes data to buffer from the given string.
		1411
		1412	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBuffer.html>
		1413
		1414	=item $ev = $queue->copy_buffer ($src, $dst, $src_offset, $dst_offset, $len, $wait_events...)
		1415
		1416	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBuffer.html>
		1417
		1418	=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...)
		1419
		1420	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html
		1421
		1422	=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...)
		1423
		1424	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html
		1425
		1426	=item $ev = $queue->copy_buffer_to_image ($src_buffer, $dst_image, $src_offset, $dst_x, $dst_y, $dst_z, $width, $height, $depth, $wait_events...)
		1427
		1428	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>
		1429
		1430	=item $ev = $queue->read_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...)
		1431
		1432	C<$row_pitch> (and C<$slice_pitch>) can be C<0>, in which case the OpenCL
		1433	module uses the image width (and height) to supply default values.
		1434
		1435	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadImage.html>
		1436
		1437	=item $ev = $queue->write_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...)
		1438
		1439	C<$row_pitch> (and C<$slice_pitch>) can be C<0>, in which case the OpenCL
		1440	module uses the image width (and height) to supply default values.
		1441	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteImage.html>
		1442
		1443	=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...)
		1444
		1445	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyImage.html>
		1446
		1447	=item $ev = $queue->copy_image_to_buffer ($src_image, $dst_image, $src_x, $src_y, $src_z, $width, $height, $depth, $dst_offset, $wait_events...)
		1448
		1449	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyImageToBuffer.html>
		1450
		1451	=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...)
		1452
		1453	Yeah.
		1454
		1455	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueCopyBufferToImage.html>.
		1456
		1457	=item $ev = $queue->fill_buffer ($mem, $pattern, $offset, $size, ...)
		1458
		1459	Fills the given buffer object with repeated applications of C<$pattern>,
		1460	starting at C<$offset> for C<$size> octets.
		1461
		1462	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueFillBuffer.html>
		1463
		1464	=item $ev = $queue->fill_image ($img, $r, $g, $b, $a, $x, $y, $z, $width, $height, $depth, ...)
		1465
		1466	Fills the given image area with the given rgba colour components. The
		1467	components are normally floating point values between C<0> and C<1>,
		1468	except when the image channel data type is a signe dor unsigned
		1469	unnormalised format, in which case the range is determined by the format.
		1470
		1471	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueFillImage.html>
		1472
		1473	=item $ev = $queue->task ($kernel, $wait_events...)
		1474
		1475	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueTask.html>
		1476
		1477	=item $ev = $queue->nd_range_kernel ($kernel, \@global_work_offset, \@global_work_size, \@local_work_size, $wait_events...)
		1478
		1479	Enqueues a kernel execution.
		1480
		1481	\@global_work_size must be specified as a reference to an array of
		1482	integers specifying the work sizes (element counts).
		1483
		1484	\@global_work_offset must be either C<undef> (in which case all offsets
		1485	are C<0>), or a reference to an array of work offsets, with the same number
		1486	of elements as \@global_work_size.
		1487
		1488	\@local_work_size must be either C<undef> (in which case the
		1489	implementation is supposed to choose good local work sizes), or a
		1490	reference to an array of local work sizes, with the same number of
		1491	elements as \@global_work_size.
		1492
		1493	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueNDRangeKernel.html>
		1494
		1495	=item $ev = $queue->migrate_mem_objects (\@mem_objects, $flags, $wait_events...)
		1496
		1497	Migrates a number of OpenCL::Memory objects to or from the device.
		1498
		1499	flags: OpenCL::MIGRATE_MEM_OBJECT_HOST, OpenCL::MIGRATE_MEM_OBJECT_CONTENT_UNDEFINED
		1500
		1501	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueMigrateMemObjects.html>
		1502
		1503	=item $ev = $queue->acquire_gl_objects ([object, ...], $wait_events...)
		1504
		1505	Enqueues a list (an array-ref of OpenCL::Memory objects) to be acquired
		1506	for subsequent OpenCL usage.
		1507
		1508	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueAcquireGLObjects.html>
		1509
		1510	=item $ev = $queue->release_gl_objects ([object, ...], $wait_events...)
		1511
		1512	Enqueues a list (an array-ref of OpenCL::Memory objects) to be released
		1513	for subsequent OpenGL usage.
		1514
		1515	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReleaseGLObjects.html>
		1516
		1517	=item $ev = $queue->wait_for_events ($wait_events...)
		1518
		1519	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWaitForEvents.html>
		1520
		1521	=item $ev = $queue->marker ($wait_events...)
		1522
		1523	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueMarkerWithWaitList.html>
		1524
		1525	=item $ev = $queue->barrier ($wait_events...)
		1526
		1527	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueBarrierWithWaitList.html>
		1528
		1529	=item $queue->flush
		1530
		1531	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFlush.html>
		1532
		1533	=item $queue->finish
		1534
		1535	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clFinish.html>
		1536
		1537	=item $packed_value = $queue->info ($name)
		1538
		1539	See C<< $platform->info >> for details.
		1540
		1541	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetCommandQueueInfo.html>
		1542
		1543	=for gengetinfo begin command_queue
		1544
		1545	=item $ctx = $command_queue->context
		1546
		1547	Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_CONTEXT> and returns the result.
		1548
		1549	=item $device = $command_queue->device
		1550
		1551	Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_DEVICE> and returns the result.
		1552
		1553	=item $uint = $command_queue->reference_count
		1554
		1555	Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_REFERENCE_COUNT> and returns the result.
		1556
		1557	=item $command_queue_properties = $command_queue->properties
		1558
		1559	Calls C<clGetCommandQueueInfo> with C<OpenCL::QUEUE_PROPERTIES> and returns the result.
		1560
		1561	=for gengetinfo end command_queue
		1562
		1563	=back
		1564
		1565	=head3 MEMORY MAPPED BUFFERS
		1566
		1567	OpenCL allows you to map buffers and images to host memory (read: perl
		1568	scalars). This is done much like reading or copying a buffer, by enqueuing
		1569	a map or unmap operation on the command queue.
		1570
		1571	The map operations return an C<OpenCL::Mapped> object - see L<THE
		1572	OpenCL::Mapped CLASS> section for details on what to do with these
		1573	objects.
		1574
		1575	The object will be unmapped automatically when the mapped object is
		1576	destroyed (you can use a barrier to make sure the unmap has finished,
		1577	before using the buffer in a kernel), but you can also enqueue an unmap
		1578	operation manually.
		1579
		1580	=over 4
		1581
		1582	=item $mapped_buffer = $queue->map_buffer ($buf, $blocking=1, $map_flags=OpenCL::MAP_READ|OpenCL::MAP_WRITE, $offset=0, $size=undef, $wait_events...)
		1583
		1584	Maps the given buffer into host memory and returns an
		1585	C<OpenCL::MappedBuffer> object. If C<$size> is specified as undef, then
		1586	the map will extend to the end of the buffer.
		1587
		1588	map_flags: OpenCL::MAP_READ, OpenCL::MAP_WRITE, OpenCL::MAP_WRITE_INVALIDATE_REGION.
		1589
		1590	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueMapBuffer.html>
		1591
		1592	Example: map the buffer $buf fully and replace the first 4 bytes by "abcd", then unmap.
		1593
		1594	{
		1595	my $mapped = $queue->map_buffer ($buf, 1, OpenCL::MAP_WRITE);
		1596	substr $$mapped, 0, 4, "abcd";
		1597	} # asynchronously unmap because $mapped is destroyed
		1598
		1599	=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...)
		1600
		1601	Maps the given image area into host memory and return an
		1602	C<OpenCL::MappedImage> object.
		1603
		1604	If any of C<$width>, C<$height> and/or C<$depth> are C<undef> then they
		1605	will be replaced by the maximum possible value.
		1606
		1607	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueMapImage.html>
		1608
		1609	Example: map an image (with OpenCL::UNSIGNED_INT8 channel type) and set
		1610	the first channel of the leftmost column to 5, then explicitly unmap
		1611	it. You are not necessarily meant to do it this way, this example just
		1612	shows you the accessors to use :)
		1613
		1614	my $mapped = $queue->map_image ($image, 1, OpenCL::MAP_WRITE);
		1615
		1616	$mapped->set ($_ * $mapped->row_pitch, pack "C", 5)
		1617	for 0..$image->height;
		1618
		1619	$mapped->unmap;.
		1620	$mapped->wait; # only needed for out of order queues normally
		1621
		1622	=item $ev = $queue->unmap ($mapped, $wait_events...)
		1623
		1624	Unmaps the data from host memory. You must not call any methods that
		1625	modify the data, or modify the data scalar directly, after calling this
		1626	method.
		1627
		1628	The mapped event object will always be passed as part of the
		1629	$wait_events. The mapped event object will be replaced by the new event
		1630	object that this request creates.
		1631
		1632	=back
		1633
		1634	=head2 THE OpenCL::Memory CLASS
		1635
		1636	This the superclass of all memory objects - OpenCL::Buffer, OpenCL::Image,
		1637	OpenCL::Image2D and OpenCL::Image3D.
		1638
		1639	=over 4
		1640
		1641	=item $packed_value = $memory->info ($name)
		1642
		1643	See C<< $platform->info >> for details.
		1644
		1645	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetMemObjectInfo.html>
		1646
		1647	=for gengetinfo begin mem
		1648
		1649	=item $mem_object_type = $mem->type
		1650
		1651	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_TYPE> and returns the result.
		1652
		1653	=item $mem_flags = $mem->flags
		1654
		1655	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_FLAGS> and returns the result.
		1656
		1657	=item $int = $mem->size
		1658
		1659	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_SIZE> and returns the result.
		1660
		1661	=item $ptr_value = $mem->host_ptr
		1662
		1663	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_HOST_PTR> and returns the result.
		1664
		1665	=item $uint = $mem->map_count
		1666
		1667	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_MAP_COUNT> and returns the result.
		1668
		1669	=item $uint = $mem->reference_count
		1670
		1671	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_REFERENCE_COUNT> and returns the result.
		1672
		1673	=item $ctx = $mem->context
		1674
		1675	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_CONTEXT> and returns the result.
		1676
		1677	=item $mem = $mem->associated_memobject
		1678
		1679	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_ASSOCIATED_MEMOBJECT> and returns the result.
		1680
		1681	=item $int = $mem->offset
		1682
		1683	Calls C<clGetMemObjectInfo> with C<OpenCL::MEM_OFFSET> and returns the result.
		1684
		1685	=for gengetinfo end mem
		1686
		1687	=item ($type, $name) = $mem->gl_object_info
		1688
		1689	Returns the OpenGL object type (e.g. OpenCL::GL_OBJECT_TEXTURE2D) and the
		1690	object "name" (e.g. the texture name) used to create this memory object.
		1691
		1692	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetGLObjectInfo.html>
		1693
		1694	=back
		1695
		1696	=head2 THE OpenCL::Buffer CLASS
		1697
		1698	This is a subclass of OpenCL::Memory, and the superclass of
		1699	OpenCL::BufferObj. Its purpose is simply to distinguish between buffers
		1700	and sub-buffers.
		1701
		1702	=head2 THE OpenCL::BufferObj CLASS
		1703
		1704	This is a subclass of OpenCL::Buffer and thus OpenCL::Memory. It exists
		1705	because one cna create sub buffers of OpenLC::BufferObj objects, but not
		1706	sub buffers from these sub buffers.
		1707
		1708	=over 4
		1709
		1710	=item $subbuf = $buf_obj->sub_buffer_region ($flags, $origin, $size)
		1711
		1712	Creates an OpenCL::Buffer objects from this buffer and returns it. The
		1713	C<buffer_create_type> is assumed to be C<OpenCL::BUFFER_CREATE_TYPE_REGION>.
		1714
		1715	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateSubBuffer.html>
		1716
		1717	=back
		1718
		1719	=head2 THE OpenCL::Image CLASS
		1720
		1721	This is the superclass of all image objects - OpenCL::Image1D,
		1722	OpenCL::Image1DArray, OpenCL::Image1DBuffer, OpenCL::Image2D,
		1723	OpenCL::Image2DArray and OpenCL::Image3D.
		1724
		1725	=over 4
		1726
		1727	=item $packed_value = $image->image_info ($name)
		1728
		1729	See C<< $platform->info >> for details.
		1730
		1731	The reason this method is not called C<info> is that there already is an
		1732	C<< ->info >> method inherited from C<OpenCL::Memory>.
		1733
		1734	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetImageInfo.html>
		1735
		1736	=item ($channel_order, $channel_data_type) = $image->format
		1737
		1738	Returns the channel order and type used to create the image by calling
		1739	C<clGetImageInfo> with C<OpenCL::IMAGE_FORMAT>.
		1740
		1741	=for gengetinfo begin image
		1742
		1743	=item $int = $image->element_size
		1744
		1745	Calls C<clGetImageInfo> with C<OpenCL::IMAGE_ELEMENT_SIZE> and returns the result.
		1746
		1747	=item $int = $image->row_pitch
		1748
		1749	Calls C<clGetImageInfo> with C<OpenCL::IMAGE_ROW_PITCH> and returns the result.
		1750
		1751	=item $int = $image->slice_pitch
		1752
		1753	Calls C<clGetImageInfo> with C<OpenCL::IMAGE_SLICE_PITCH> and returns the result.
		1754
		1755	=item $int = $image->width
		1756
		1757	Calls C<clGetImageInfo> with C<OpenCL::IMAGE_WIDTH> and returns the result.
		1758
		1759	=item $int = $image->height
		1760
		1761	Calls C<clGetImageInfo> with C<OpenCL::IMAGE_HEIGHT> and returns the result.
		1762
		1763	=item $int = $image->depth
		1764
		1765	Calls C<clGetImageInfo> with C<OpenCL::IMAGE_DEPTH> and returns the result.
		1766
		1767	=for gengetinfo end image
		1768
		1769	=for gengetinfo begin gl_texture
		1770
		1771	=item $GLenum = $gl_texture->target
		1772
		1773	Calls C<clGetGLTextureInfo> with C<OpenCL::GL_TEXTURE_TARGET> and returns the result.
		1774
		1775	=item $GLint = $gl_texture->gl_mipmap_level
		1776
		1777	Calls C<clGetGLTextureInfo> with C<OpenCL::GL_MIPMAP_LEVEL> and returns the result.
		1778
		1779	=for gengetinfo end gl_texture
		1780
		1781	=back
		1782
		1783	=head2 THE OpenCL::Sampler CLASS
		1784
		1785	=over 4
		1786
		1787	=item $packed_value = $sampler->info ($name)
		1788
		1789	See C<< $platform->info >> for details.
		1790
		1791	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetSamplerInfo.html>
		1792
		1793	=for gengetinfo begin sampler
		1794
		1795	=item $uint = $sampler->reference_count
		1796
		1797	Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_REFERENCE_COUNT> and returns the result.
		1798
		1799	=item $ctx = $sampler->context
		1800
		1801	Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_CONTEXT> and returns the result.
		1802
		1803	=item $addressing_mode = $sampler->normalized_coords
		1804
		1805	Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_NORMALIZED_COORDS> and returns the result.
		1806
		1807	=item $filter_mode = $sampler->addressing_mode
		1808
		1809	Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_ADDRESSING_MODE> and returns the result.
		1810
		1811	=item $boolean = $sampler->filter_mode
		1812
		1813	Calls C<clGetSamplerInfo> with C<OpenCL::SAMPLER_FILTER_MODE> and returns the result.
		1814
		1815	=for gengetinfo end sampler
		1816
		1817	=back
		1818
		1819	=head2 THE OpenCL::Program CLASS
		1820
		1821	=over 4
		1822
		1823	=item $program->build (\@devices = undef, $options = "", $cb->($program) = undef)
		1824
		1825	Tries to build the program with the given options. See also the
		1826	C<$ctx->build> convenience function.
		1827
		1828	If a callback is specified, then it will be called when compilation is
		1829	finished. Note that many OpenCL implementations block your program while
		1830	compiling whether you use a callback or not. See C<build_async> if you
		1831	want to make sure the build is done in the background.
		1832
		1833	Note that some OpenCL implementations act up badly, and don't call the
		1834	callback in some error cases (but call it in others). This implementation
		1835	assumes the callback will always be called, and leaks memory if this is
		1836	not so. So best make sure you don't pass in invalid values.
		1837
		1838	Some implementations fail with C<OpenCL::INVALID_BINARY> when the
		1839	compilation state is successful but some later stage fails.
		1840
		1841	options: C<-D name>, C<-D name=definition>, C<-I dir>,
		1842	C<-cl-single-precision-constant>, C<-cl-denorms-are-zero>,
		1843	C<-cl-fp32-correctly-rounded-divide-sqrt>, C<-cl-opt-disable>,
		1844	C<-cl-mad-enable>, C<-cl-no-signed-zeros>, C<-cl-unsafe-math-optimizations>,
		1845	C<-cl-finite-math-only>, C<-cl-fast-relaxed-math>,
		1846	C<-w>, C<-Werror>, C<-cl-std=CL1.1/CL1.2>, C<-cl-kernel-arg-info>,
		1847	C<-create-library>, C<-enable-link-options>.
		1848
		1849	build_status: OpenCL::BUILD_SUCCESS, OpenCL::BUILD_NONE,
		1850	OpenCL::BUILD_ERROR, OpenCL::BUILD_IN_PROGRESS.
		1851
		1852	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html>
		1853
		1854	=item $program->build_async (\@devices = undef, $options = "", $cb->($program) = undef)
		1855
		1856	Similar to C<< ->build >>, except it starts a thread, and never fails (you
		1857	need to check the compilation status form the callback, or by polling).
		1858
		1859	=item $program->compile (\@devices = undef, $options = "", \%headers = undef, $cb->($program) = undef)
		1860
		1861	Compiles the given program for the given devices (or all devices if
		1862	undef). If C<$headers> is given, it must be a hashref with include name =>
		1863	OpenCL::Program pairs.
		1864
		1865	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCompileProgram.html>
		1866
		1867	=item $packed_value = $program->build_info ($device, $name)
		1868
		1869	Similar to C<< $platform->info >>, but returns build info for a previous
		1870	build attempt for the given device.
		1871
		1872	binary_type: OpenCL::PROGRAM_BINARY_TYPE_NONE,
		1873	OpenCL::PROGRAM_BINARY_TYPE_COMPILED_OBJECT,
		1874	OpenCL::PROGRAM_BINARY_TYPE_LIBRARY,
		1875	OpenCL::PROGRAM_BINARY_TYPE_EXECUTABLE.
		1876
		1877	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetBuildInfo.html>
		1878
		1879	=item $kernel = $program->kernel ($function_name)
		1880
		1881	Creates an OpenCL::Kernel object out of the named C<__kernel> function in
		1882	the program.
		1883
		1884	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernel.html>
		1885
		1886	=item @kernels = $program->kernels_in_program
		1887
		1888	Returns all kernels successfully compiled for all devices in program.
		1889
		1890	http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKernelsInProgram.html
		1891
		1892	=for gengetinfo begin program_build
		1893
		1894	=item $build_status = $program->build_status ($device)
		1895
		1896	Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BUILD_STATUS> and returns the result.
		1897
		1898	=item $string = $program->build_options ($device)
		1899
		1900	Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BUILD_OPTIONS> and returns the result.
		1901
		1902	=item $string = $program->build_log ($device)
		1903
		1904	Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BUILD_LOG> and returns the result.
		1905
		1906	=item $binary_type = $program->binary_type ($device)
		1907
		1908	Calls C<clGetProgramBuildInfo> with C<OpenCL::PROGRAM_BINARY_TYPE> and returns the result.
		1909
		1910	=for gengetinfo end program_build
		1911
		1912	=item $packed_value = $program->info ($name)
		1913
		1914	See C<< $platform->info >> for details.
		1915
		1916	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProgramInfo.html>
		1917
		1918	=for gengetinfo begin program
		1919
		1920	=item $uint = $program->reference_count
		1921
		1922	Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_REFERENCE_COUNT> and returns the result.
		1923
		1924	=item $ctx = $program->context
		1925
		1926	Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_CONTEXT> and returns the result.
		1927
		1928	=item $uint = $program->num_devices
		1929
		1930	Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_NUM_DEVICES> and returns the result.
		1931
		1932	=item @devices = $program->devices
		1933
		1934	Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_DEVICES> and returns the result.
		1935
		1936	=item $string = $program->source
		1937
		1938	Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_SOURCE> and returns the result.
		1939
		1940	=item @ints = $program->binary_sizes
		1941
		1942	Calls C<clGetProgramInfo> with C<OpenCL::PROGRAM_BINARY_SIZES> and returns the result.
		1943
		1944	=for gengetinfo end program
		1945
		1946	=item @blobs = $program->binaries
		1947
		1948	Returns a string for the compiled binary for every device associated with
		1949	the program, empty strings indicate missing programs, and an empty result
		1950	means no program binaries are available.
		1951
		1952	These "binaries" are often, in fact, informative low-level assembly
		1953	sources.
		1954
		1955	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProgramInfo.html>
		1956
		1957	=back
		1958
		1959	=head2 THE OpenCL::Kernel CLASS
		1960
		1961	=over 4
		1962
		1963	=item $packed_value = $kernel->info ($name)
		1964
		1965	See C<< $platform->info >> for details.
		1966
		1967	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelInfo.html>
		1968
		1969	=for gengetinfo begin kernel
		1970
		1971	=item $string = $kernel->function_name
		1972
		1973	Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_FUNCTION_NAME> and returns the result.
		1974
		1975	=item $uint = $kernel->num_args
		1976
		1977	Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_NUM_ARGS> and returns the result.
		1978
		1979	=item $uint = $kernel->reference_count
		1980
		1981	Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_REFERENCE_COUNT> and returns the result.
		1982
		1983	=item $ctx = $kernel->context
		1984
		1985	Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_CONTEXT> and returns the result.
		1986
		1987	=item $program = $kernel->program
		1988
		1989	Calls C<clGetKernelInfo> with C<OpenCL::KERNEL_PROGRAM> and returns the result.
		1990
		1991	=for gengetinfo end kernel
		1992
		1993	=item $packed_value = $kernel->work_group_info ($device, $name)
		1994
		1995	See C<< $platform->info >> for details.
		1996
		1997	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelWorkGroupInfo.html>
		1998
		1999	=for gengetinfo begin kernel_work_group
		2000
		2001	=item $int = $kernel->work_group_size ($device)
		2002
		2003	Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_WORK_GROUP_SIZE> and returns the result.
		2004
		2005	=item @ints = $kernel->compile_work_group_size ($device)
		2006
		2007	Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_COMPILE_WORK_GROUP_SIZE> and returns the result.
		2008
		2009	=item $ulong = $kernel->local_mem_size ($device)
		2010
		2011	Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_LOCAL_MEM_SIZE> and returns the result.
		2012
		2013	=item $int = $kernel->preferred_work_group_size_multiple ($device)
		2014
		2015	Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE> and returns the result.
		2016
		2017	=item $ulong = $kernel->private_mem_size ($device)
		2018
		2019	Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_PRIVATE_MEM_SIZE> and returns the result.
		2020
		2021	=for gengetinfo end kernel_work_group
		2022
		2023	=item $packed_value = $kernel->arg_info ($idx, $name)
		2024
		2025	See C<< $platform->info >> for details.
		2026
		2027	L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clGetKernelArgInfo.html>
		2028
		2029	=for gengetinfo begin kernel_arg
		2030
		2031	=item $kernel_arg_address_qualifier = $kernel->arg_address_qualifier ($idx)
		2032
		2033	Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_ADDRESS_QUALIFIER> and returns the result.
		2034
		2035	=item $kernel_arg_access_qualifier = $kernel->arg_access_qualifier ($idx)
		2036
		2037	Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_ACCESS_QUALIFIER> and returns the result.
		2038
		2039	=item $string = $kernel->arg_type_name ($idx)
		2040
		2041	Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_TYPE_NAME> and returns the result.
		2042
		2043	=item $kernel_arg_type_qualifier = $kernel->arg_type_qualifier ($idx)
		2044
		2045	Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_TYPE_QUALIFIER> and returns the result.
		2046
		2047	=item $string = $kernel->arg_name ($idx)
		2048
		2049	Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_NAME> and returns the result.
		2050
		2051	=for gengetinfo end kernel_arg
		2052
		2053	=item $kernel->setf ($format, ...)
		2054
		2055	Sets the arguments of a kernel. Since OpenCL 1.1 doesn't have a generic
		2056	way to set arguments (and with OpenCL 1.2 it might be rather slow), you
		2057	need to specify a format argument, much as with C<printf>, to tell OpenCL
		2058	what type of argument it is.
		2059
		2060	The format arguments are single letters:
		2061
		2062	c char
		2063	C unsigned char
		2064	s short
		2065	S unsigned short
		2066	i int
		2067	I unsigned int
		2068	l long
		2069	L unsigned long
		2070
		2071	h half float (0..65535)
		2072	f float
		2073	d double
		2074
		2075	z local (octet size)
		2076
		2077	m memory object (buffer or image)
		2078	a sampler
		2079	e event
		2080
		2081	Space characters in the format string are ignored.
		2082
		2083	Example: set the arguments for a kernel that expects an int, two floats, a buffer and an image.
		2084
		2085	$kernel->setf ("i ff mm", 5, 0.5, 3, $buffer, $image);
		2086
		2087	=item $kernel->set_TYPE ($index, $value)
		2088
		2089	=item $kernel->set_char ($index, $value)
		2090
		2091	=item $kernel->set_uchar ($index, $value)
		2092
		2093	=item $kernel->set_short ($index, $value)
		2094
		2095	=item $kernel->set_ushort ($index, $value)
		2096
		2097	=item $kernel->set_int ($index, $value)
		2098
		2099	=item $kernel->set_uint ($index, $value)
		2100
		2101	=item $kernel->set_long ($index, $value)
		2102
		2103	=item $kernel->set_ulong ($index, $value)
		2104
		2105	=item $kernel->set_half ($index, $value)
		2106
		2107	=item $kernel->set_float ($index, $value)
		2108
		2109	=item $kernel->set_double ($index, $value)
		2110
		2111	=item $kernel->set_memory ($index, $value)
		2112
		2113	=item $kernel->set_buffer ($index, $value)
		2114
		2115	=item $kernel->set_image ($index, $value)
		2116
		2117	=item $kernel->set_sampler ($index, $value)
		2118
		2119	=item $kernel->set_local ($index, $value)
		2120
		2121	=item $kernel->set_event ($index, $value)
		2122
		2123	This is a family of methods to set the kernel argument with the number
		2124	C<$index> to the give C<$value>.
		2125
		2126	Chars and integers (including the half type) are specified as integers,
		2127	float and double as floating point values, memory/buffer/image must be
		2128	an object of that type or C<undef>, local-memory arguments are set by
		2129	specifying the size, and sampler and event must be objects of that type.
		2130
		2131	Note that C<set_memory> works for all memory objects (all types of buffers
		2132	and images) - the main purpose of the more specific C<set_TYPE> functions
		2133	is type checking.
		2134
		2135	Setting an argument for a kernel does NOT keep a reference to the object -
		2136	for example, if you set an argument to some image object, free the image,
		2137	and call the kernel, you will run into undefined behaviour.
		2138
		2139	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetKernelArg.html>
		2140
		2141	=back
		2142
		2143	=head2 THE OpenCL::Event CLASS
		2144
		2145	This is the superclass for all event objects (including OpenCL::UserEvent
		2146	objects).
		2147
		2148	=over 4
		2149
		2150	=item $ev->wait
		2151
		2152	Waits for the event to complete.
		2153
		2154	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clWaitForEvents.html>
		2155
		2156	=item $ev->cb ($exec_callback_type, $callback->($event, $event_command_exec_status))
		2157
		2158	Adds a callback to the callback stack for the given event type. There is
		2159	no way to remove a callback again.
		2160
		2161	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetEventCallback.html>
		2162
		2163	=item $packed_value = $ev->info ($name)
		2164
		2165	See C<< $platform->info >> for details.
		2166
		2167	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetEventInfo.html>
		2168
		2169	=for gengetinfo begin event
		2170
		2171	=item $queue = $event->command_queue
		2172
		2173	Calls C<clGetEventInfo> with C<OpenCL::EVENT_COMMAND_QUEUE> and returns the result.
		2174
		2175	=item $command_type = $event->command_type
		2176
		2177	Calls C<clGetEventInfo> with C<OpenCL::EVENT_COMMAND_TYPE> and returns the result.
		2178
		2179	=item $uint = $event->reference_count
		2180
		2181	Calls C<clGetEventInfo> with C<OpenCL::EVENT_REFERENCE_COUNT> and returns the result.
		2182
		2183	=item $uint = $event->command_execution_status
		2184
		2185	Calls C<clGetEventInfo> with C<OpenCL::EVENT_COMMAND_EXECUTION_STATUS> and returns the result.
		2186
		2187	=item $ctx = $event->context
		2188
		2189	Calls C<clGetEventInfo> with C<OpenCL::EVENT_CONTEXT> and returns the result.
		2190
		2191	=for gengetinfo end event
		2192
		2193	=item $packed_value = $ev->profiling_info ($name)
		2194
		2195	See C<< $platform->info >> for details.
		2196
		2197	The reason this method is not called C<info> is that there already is an
		2198	C<< ->info >> method.
		2199
		2200	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetProfilingInfo.html>
		2201
		2202	=for gengetinfo begin profiling
		2203
		2204	=item $ulong = $event->profiling_command_queued
		2205
		2206	Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_QUEUED> and returns the result.
		2207
		2208	=item $ulong = $event->profiling_command_submit
		2209
		2210	Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_SUBMIT> and returns the result.
		2211
		2212	=item $ulong = $event->profiling_command_start
		2213
		2214	Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_START> and returns the result.
		2215
		2216	=item $ulong = $event->profiling_command_end
		2217
		2218	Calls C<clGetEventProfilingInfo> with C<OpenCL::PROFILING_COMMAND_END> and returns the result.
		2219
		2220	=for gengetinfo end profiling
		2221
		2222	=back
		2223
		2224	=head2 THE OpenCL::UserEvent CLASS
		2225
		2226	This is a subclass of OpenCL::Event.
		2227
		2228	=over 4
		2229
		2230	=item $ev->set_status ($execution_status)
		2231
		2232	Sets the execution status of the user event. Can only be called once,
		2233	either with OpenCL::COMPLETE or a negative number as status.
		2234
		2235	execution_status: OpenCL::COMPLETE or a negative integer.
		2236
		2237	L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetUserEventStatus.html>
		2238
		2239	=back
		2240
		2241	=head2 THE OpenCL::Mapped CLASS
		2242
		2243	This class represents objects mapped into host memory. They are
		2244	represented by a blessed string scalar. The string data is the mapped
		2245	memory area, that is, if you read or write it, then the mapped object is
		2246	accessed directly.
		2247
		2248	You must only ever use operations that modify the string in-place - for
		2249	example, a C<substr> that doesn't change the length, or maybe a regex that
		2250	doesn't change the length. Any other operation might cause the data to be
		2251	copied.
		2252
		2253	When the object is destroyed it will enqueue an implicit unmap operation
		2254	on the queue that was used to create it.
		2255
		2256	Keep in mind that you I<need> to unmap (or destroy) mapped objects before
		2257	OpenCL sees the changes, even if some implementations don't need this
		2258	sometimes.
		2259
		2260	Example, replace the first two floats in the mapped buffer by 1 and 2.
		2261
		2262	my $mapped = $queue->map_buffer ($buf, ...
		2263	$mapped->event->wait; # make sure it's there
		2264
		2265	# now replace first 8 bytes by new data, which is exactly 8 bytes long
		2266	# we blindly assume device endianness to equal host endianness
		2267	# (and of course, we assume iee 754 single precision floats :)
		2268	substr $$mapped, 0, 8, pack "f*", 1, 2;
		2269
		2270	=over 4
		2271
		2272	=item $ev = $mapped->unmap ($wait_events...)
		2273
		2274	Unmaps the mapped memory object, using the queue originally used to create
		2275	it, quite similarly to C<< $queue->unmap ($mapped, ...) >>.
		2276
		2277	=item $bool = $mapped->mapped
		2278
		2279	Returns whether the object is still mapped - true before an C<unmap> is
		2280	enqueued, false afterwards.
		2281
		2282	=item $ev = $mapped->event
		2283
		2284	Return the event object associated with the mapped object. Initially, this
		2285	will be the event object created when mapping the object, and after an
		2286	unmap, this will be the event object that the unmap operation created.
		2287
		2288	=item $mapped->wait
		2289
		2290	Same as C<< $mapped->event->wait >> - makes sure no operations on this
		2291	mapped object are outstanding.
		2292
		2293	=item $bytes = $mapped->size
		2294
		2295	Returns the size of the mapped area, in bytes. Same as C<length $$mapped>.
		2296
		2297	=item $ptr = $mapped->ptr
		2298
		2299	Returns the raw memory address of the mapped area.
		2300
		2301	=item $mapped->set ($offset, $data)
		2302
		2303	Replaces the data at the given C<$offset> in the memory area by the new
		2304	C<$data>. This method is safer than direct manipulation of C<$mapped>
		2305	because it does bounds-checking, but also slower.
		2306
		2307	=item $data = $mapped->get ($offset, $length)
		2308
		2309	Returns (without copying) a scalar representing the data at the given
		2310	C<$offset> and C<$length> in the mapped memory area. This is the same as
		2311	the following substr, except much slower;
		2312
		2313	$data = substr $$mapped, $offset, $length
		2314
		2315	=cut
		2316
		2317	sub OpenCL::Mapped::get {
		2318	substr ${$_[0]}, $_[1], $_[2]
		2319	}
		2320
		2321	=back
		2322
		2323	=head2 THE OpenCL::MappedBuffer CLASS
		2324
		2325	This is a subclass of OpenCL::Mapped, representing mapped buffers.
		2326
		2327	=head2 THE OpenCL::MappedImage CLASS
		2328
		2329	This is a subclass of OpenCL::Mapped, representing mapped images.
		2330
		2331	=over 4
		2332
		2333	=item $bytes = $mapped->row_pitch
		2334
		2335	=item $bytes = $mapped->slice_pitch
		2336
		2337	Return the row or slice pitch of the image that has been mapped.
		2338
		2339	=back
		2340
		2341
		2342	=cut
		2343
46	1;	2344	1;
47
48	=back
49		2345
50	=head1 AUTHOR	2346	=head1 AUTHOR
51		2347
52	Marc Lehmann <schmorp@schmorp.de>	2348	Marc Lehmann <schmorp@schmorp.de>
53	http://home.schmorp.de/	2349	http://home.schmorp.de/

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines