NAME OpenCL - Open Computing Language Bindings SYNOPSIS use OpenCL; DESCRIPTION This is an early release which might be useful, but hasn't seen much testing. OpenCL FROM 10000 FEET HEIGHT Here is a high level overview of OpenCL: First you need to find one or more OpenCL::Platforms (kind of like vendors) - usually there is only one. Each platform gives you access to a number of OpenCL::Device objects, e.g. your graphics card. From a platform and some device(s), you create an OpenCL::Context, which is a very central object in OpenCL: Once you have a context you can create most other objects: OpenCL::Program objects, which store source code and, after building for a specific device ("compiling and linking"), also binary programs. For each kernel function in a program you can then create an OpenCL::Kernel object which represents basically a function call with argument values. OpenCL::Memory objects of various flavours: OpenCL::Buffer objects (flat memory areas, think arrays or structs) and OpenCL::Image objects (think 2D or 3D array) for bulk data and input and output for kernels. OpenCL::Sampler objects, which are kind of like texture filter modes in OpenGL. OpenCL::Queue objects - command queues, which allow you to submit memory reads, writes and copies, as well as kernel calls to your devices. They also offer a variety of methods to synchronise request execution, for example with barriers or OpenCL::Event objects. OpenCL::Event objects are used to signal when something is complete. HELPFUL RESOURCES The OpenCL specs used to develop this module - download these and keept hema round, they are required reference material: http://www.khronos.org/registry/cl/specs/opencl-1.1.pdf http://www.khronos.org/registry/cl/specs/opencl-1.2.pdf http://www.khronos.org/registry/cl/specs/opencl-1.2-extensions.pdf OpenCL manpages: http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/ http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/ If you are into UML class diagrams, the following diagram might help - if not, it will be mildly confusing (also, the class hierarchy of this module is much more fine-grained): http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/classDiagram.html Here's a tutorial from AMD (very AMD-centric, too), not sure how useful it is, but at least it's free of charge: http://developer.amd.com/zones/OpenCLZone/courses/Documents/Introduction_to_OpenCL_Programming%20Training_Guide%20%28201005%29.pdf And here's NVIDIA's OpenCL Best Practises Guide: http://developer.download.nvidia.com/compute/cuda/3_2/toolkit/docs/OpenCL_Best_Practices_Guide.pdf BASIC WORKFLOW To get something done, you basically have to do this once (refer to the examples below for actual code, this is just a high-level description): Find some platform (e.g. the first one) and some device(s) (e.g. the first device of the platform), and create a context from those. Create program objects from your OpenCL source code, then build (compile) the programs for each device you want to run them on. Create kernel objects for all kernels you want to use (surprisingly, these are not device-specific). Then, to execute stuff, you repeat these steps, possibly resuing or sharing some buffers: Create some input and output buffers from your context. Set these as arguments to your kernel. Enqueue buffer writes to initialise your input buffers (when not initialised at creation time). Enqueue the kernel execution. Enqueue buffer reads for your output buffer to read results. EXAMPLES Enumerate all devices and get contexts for them. Best run this once to get a feel for the platforms and devices in your system. for my $platform (OpenCL::platforms) { printf "platform: %s\n", $platform->name; printf "extensions: %s\n", $platform->extensions; for my $device ($platform->devices) { printf "+ device: %s\n", $device->name; my $ctx = $platform->context (undef, [$device]); # do stuff } } Get a useful context and a command queue. This is a useful boilerplate for any OpenCL program that only wants to use one device, my ($platform) = OpenCL::platforms; # find first platform my ($dev) = $platform->devices; # find first device of platform my $ctx = $platform->context (undef, [$dev]); # create context out of those my $queue = $ctx->queue ($dev); # create a command queue for the device Print all supported image formats of a context. Best run this once for your context, to see whats available and how to gather information. for my $type (OpenCL::MEM_OBJECT_IMAGE2D, OpenCL::MEM_OBJECT_IMAGE3D) { print "supported image formats for ", OpenCL::enum2str $type, "\n"; for my $f ($ctx->supported_image_formats (0, $type)) { printf " %-10s %-20s\n", OpenCL::enum2str $f->[0], OpenCL::enum2str $f->[1]; } } Create a buffer with some predefined data, read it back synchronously, then asynchronously. my $buf = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, "helmut"); $queue->read_buffer ($buf, 1, 1, 3, my $data); print "$data\n"; my $ev = $queue->read_buffer ($buf, 0, 1, 3, my $data); $ev->wait; print "$data\n"; # prints "elm" Create and build a program, then create a kernel out of one of its functions. my $src = ' kernel void squareit (global float *input, global float *output) { $id = get_global_id (0); output [id] = input [id] * input [id]; } '; my $prog = $ctx->build_program ($src); my $kernel = $prog->kernel ("squareit"); Create some input and output float buffers, then call the 'squareit' kernel on them. my $input = $ctx->buffer_sv (OpenCL::MEM_COPY_HOST_PTR, pack "f*", 1, 2, 3, 4.5); my $output = $ctx->buffer (0, OpenCL::SIZEOF_FLOAT * 5); # set buffer $kernel->set_buffer (0, $input); $kernel->set_buffer (1, $output); # execute it for all 4 numbers $queue->nd_range_kernel ($kernel, undef, [4], undef); # enqueue a synchronous read $queue->read_buffer ($output, 1, 0, OpenCL::SIZEOF_FLOAT * 4, my $data); # print the results: printf "%s\n", join ", ", unpack "f*", $data; The same enqueue operations as before, but assuming an out-of-order queue, showing off barriers. # execute it for all 4 numbers $queue->nd_range_kernel ($kernel, undef, [4], undef); # enqueue a barrier to ensure in-order execution $queue->barrier; # enqueue an async read $queue->read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data); # wait for all requests to finish $queue->finish; The same enqueue operations as before, but assuming an out-of-order queue, showing off event objects and wait lists. # execute it for all 4 numbers my $ev = $queue->nd_range_kernel ($kernel, undef, [4], undef); # enqueue an async read $ev = $queue->read_buffer ($output, 0, 0, OpenCL::SIZEOF_FLOAT * 4, my $data, $ev); # wait for the last event to complete $ev->wait; Use the OpenGL module to share a texture between OpenCL and OpenGL and draw some julia set flight effect. This is quite a long example to get you going - you can also download it from . use OpenGL ":all"; use OpenCL; my $S = $ARGV[0] || 256; # window/texture size, smaller is faster # open a window and create a gl texture OpenGL::glpOpenWindow width => $S, height => $S; my $texid = glGenTextures_p 1; glBindTexture GL_TEXTURE_2D, $texid; glTexImage2D_c GL_TEXTURE_2D, 0, GL_RGBA8, $S, $S, 0, GL_RGBA, GL_UNSIGNED_BYTE, 0; # find and use the first opencl device that let's us get a shared opengl context my $platform; my $dev; my $ctx; for (OpenCL::platforms) { $platform = $_; for ($platform->devices) { $dev = $_; $ctx = $platform->context ([OpenCL::GLX_DISPLAY_KHR, undef, OpenCL::GL_CONTEXT_KHR, undef], [$dev]) and last; } } $ctx or die "cannot find suitable OpenCL device\n"; my $queue = $ctx->queue ($dev); # now attach an opencl image2d object to the opengl texture my $tex = $ctx->gl_texture2d (OpenCL::MEM_WRITE_ONLY, GL_TEXTURE_2D, 0, $texid); # now the boring opencl code my $src = <build_program ($src); my $kernel = $prog->kernel ("juliatunnel"); # program compiled, kernel ready, now draw and loop for (my $time; ; ++$time) { # acquire objects from opengl $queue->acquire_gl_objects ([$tex]); # configure and run our kernel $kernel->setf ("mf", $tex, $time*2); # mf = memory object, float $queue->nd_range_kernel ($kernel, undef, [$S, $S], undef); # release objects to opengl again $queue->release_gl_objects ([$tex]); # wait $queue->finish; # now draw the texture, the defaults should be all right glTexParameterf GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST; glEnable GL_TEXTURE_2D; glBegin GL_QUADS; glTexCoord2f 0, 1; glVertex3i -1, -1, -1; glTexCoord2f 0, 0; glVertex3i 1, -1, -1; glTexCoord2f 1, 0; glVertex3i 1, 1, -1; glTexCoord2f 1, 1; glVertex3i -1, 1, -1; glEnd; glXSwapBuffers; select undef, undef, undef, 1/60; } How to modify the previous example to not rely on GL sharing. For those poor souls with only a sucky CPU OpenCL implementation, you currently have to read the image into some perl scalar, and then modify a texture or use glDrawPixels or so). First, when you don't need gl sharing, you can create the context much simpler: $ctx = $platform->context (undef, [$dev]) To use a texture, you would modify the above example by creating an OpenCL::Image manually instead of deriving it from a texture: my $tex = $ctx->image2d (OpenCL::MEM_WRITE_ONLY, OpenCL::RGBA, OpenCL::UNORM_INT8, $S, $S); And in the draw loop, intead of acquire_gl_objects/release_gl_objects, you would read the image2d after the kernel has written it: $queue->read_image ($tex, 0, 0, 0, 0, $S, $S, 1, 0, 0, my $data); And then you would upload the pixel data to the texture (or use glDrawPixels): glTexSubImage2D_s GL_TEXTURE_2D, 0, 0, 0, $S, $S, GL_RGBA, GL_UNSIGNED_BYTE, $data; The fully modified example can be found at . Julia sets look soooo 80ies. Then colour them differently, e.g. using orbit traps! Replace the loop and colour calculation from the previous examples by this: float2 dm = (float2)(1.f, 1.f); for (int i = 0; i < 25; ++i) { z = (float2)(z.x * z.x - z.y * z.y, 2.f * z.x * z.y) + c; dm = fmin (dm, (float2)(fabs (dot (z, z) - 1.f), fabs (z.x - 1.f))); } float3 colour = (float3)(dm.x * dm.y, dm.x * dm.y, dm.x); Also try "-10.f" instead of "-1.f". DOCUMENTATION BASIC CONVENTIONS This is not a one-to-one C-style translation of OpenCL to Perl - instead I attempted to make the interface as type-safe as possible by introducing object syntax where it makes sense. There are a number of important differences between the OpenCL C API and this module: * Object lifetime managament is automatic - there is no need to free objects explicitly ("clReleaseXXX"), the release function is called automatically once all Perl references to it go away. * OpenCL uses CamelCase for function names (e.g. "clGetPlatformIDs", "clGetPlatformInfo"), while this module uses underscores as word separator and often leaves out prefixes ("OpenCL::platforms", "$platform->info"). * OpenCL often specifies fixed vector function arguments as short arrays ("size_t origin[3]"), while this module explicitly expects the components as separate arguments ("$orig_x, $orig_y, $orig_z") in function calls. * Structures are often specified by flattening out their components as with short vectors, and returned as arrayrefs. * When enqueuing commands, the wait list is specified by adding extra arguments to the function - anywhere a "$wait_events..." argument is documented this can be any number of event objects. As an extsnion implemented by this module, "undef" values will be ignored in the event list. * When enqueuing commands, if the enqueue method is called in void context, no event is created. In all other contexts an event is returned by the method. * This module expects all functions to return "OpenCL::SUCCESS". If any other status is returned the function will throw an exception, so you don't normally have to to any error checking. CONSTANTS All "CL_xxx" constants that this module supports are always available in the "OpenCL" namespace as "OpenCL::xxx" (i.e. without the "CL_" prefix). Constants which are not defined in the header files used during compilation, or otherwise are not available, will have the value 0 (in some cases, this will make them indistinguishable from real constants, sorry). The latest version of this module knows and exports the constants listed in . OPENCL 1.1 VS. OPENCL 1.2 This module supports both OpenCL version 1.1 and 1.2, although the OpenCL 1.2 interface hasn't been tested much for lack of availability of an actual implementation. Every function or method in this manual page that interfaces to a particular OpenCL function has a link to the its C manual page. If the link contains a 1.1, then this function is an OpenCL 1.1 function. Most but not all also exist in OpenCL 1.2, and this module tries to emulate the missing ones for you, when told to do so at compiletime. You can check whether a function was removed in OpenCL 1.2 by replacing the 1.1 component in the URL by 1.2. If the link contains a 1.2, then this is a OpenCL 1.2-only function. Even if the module was compiled with OpenCL 1.2 header files and has an 1.2 OpenCL library, calling such a function on a platform that doesn't implement 1.2 causes undefined behaviour, usually a crash (But this is not guaranteed). You can find out whether this module was compiled to prefer 1.1 functionality by ooking at "OpenCL::PREFER_1_1" - if it is true, then 1.1 functions generally are implemented using 1.1 OpenCL functions. If it is false, then 1.1 functions missing from 1.2 are emulated by calling 1.2 fucntions. This is a somewhat sorry state of affairs, but the Khronos group choose to make every release of OpenCL source and binary incompatible with previous releases. PERL AND OPENCL TYPES This handy(?) table lists OpenCL types and their perl, PDL and pack/unpack format equivalents: OpenCL perl PDL pack/unpack char IV - c uchar IV byte C short IV short s ushort IV ushort S int IV long? l uint IV - L long IV longlong q ulong IV - Q float NV float f half IV ushort S double NV double d GLX SUPPORT Due to the sad state that OpenGL support is in in Perl (mostly the OpenGL module, which has little to no documentation and has little to no support for glX), this module, as a special extension, treats context creation properties "OpenCL::GLX_DISPLAY_KHR" and "OpenCL::GL_CONTEXT_KHR" specially: If either or both of these are "undef", then the OpenCL module tries to dynamically resolve "glXGetCurrentDisplay" and "glXGetCurrentContext", call these functions and use their return values instead. For this to work, the OpenGL library must be loaded, a GLX context must have been created and be made current, and "dlsym" must be available and capable of finding the function via "RTLD_DEFAULT". EVENT SYSTEM OpenCL can generate a number of (potentially) asynchronous events, for example, after compiling a program, to signal a context-related error or, perhaps most important, to signal completion of queued jobs (by setting callbacks on OpenCL::Event objects). The OpenCL module converts all these callbacks into events - you can still register callbacks, but they are not executed when your OpenCL implementation calls the actual callback, but only later. Therefore, none of the limitations of OpenCL callbacks apply to the perl implementation: it is perfectly safe to make blocking operations from event callbacks, and enqueued operations don't need to be flushed. To facilitate this, this module maintains an event queue - each time an asynchronous event happens, it is queued, and perl will be interrupted. This is implemented via the Async::Interrupt module. In addition, this module has AnyEvent support, so it can seamlessly integrate itself into many event loops. Since Async::Interrupt is a bit hard to understand, here are some case examples: Don't use callbacks. When your program never uses any callbacks, then there will never be any notifications you need to take care of, and therefore no need to worry about all this. You can achieve a great deal by explicitly waiting for events, or using barriers and flush calls. In many programs, there is no need at all to tinker with asynchronous events. Use AnyEvent This module automatically registers a watcher that invokes all outstanding event callbacks when AnyEvent is initialised (and block asynchronous interruptions). Using this mode of operations is the safest and most recommended one. To use this, simply use AnyEvent and this module normally, make sure you have an event loop running: use Gtk2 -init; use AnyEvent; # initialise AnyEvent, by creating a watcher, or: AnyEvent::detect; my $e = $queue->marker; $e->cb (sub { warn "opencl is finished\n"; }) main Gtk2; Note that this module will not initialise AnyEvent for you. Before AnyEvent is initialised, the module will asynchronously interrupt perl instead. To avoid any surprises, it's best to explicitly initialise AnyEvent. You can temporarily enable asynchronous interruptions (see next paragraph) by calling "$OpenCL::INTERRUPT-"unblock> and disable them again by calling "$OpenCL::INTERRUPT-"block>. Let yourself be interrupted at any time This mode is the default unless AnyEvent is loaded and initialised. In this mode, OpenCL asynchronously interrupts a running perl program. The emphasis is on both *asynchronously* and *running* here. Asynchronously means that perl might execute your callbacks at any time. For example, in the following code (*THAT YOU SHOULD NOT COPY*), the "until" loop following the marker call will be interrupted by the callback: my $e = $queue->marker; my $flag; $e->cb (sub { $flag = 1 }); 1 until $flag; # $flag is now 1 The reason why you shouldn't blindly copy the above code is that busy waiting is a really really bad thing, and really really bad for performance. While at first this asynchronous business might look exciting, it can be really hard, because you need to be prepared for the callback code to be executed at any time, which limits the amount of things the callback code can do safely. This can be mitigated somewhat by using "$OpenCL::INTERRUPT->scope_block" (see the Async::Interrupt documentation for details). The other problem is that your program must be actively *running* to be interrupted. When you calculate stuff, your program is running. When you hang in some C functions or other block execution (by calling "sleep", "select", running an event loop and so on), your program is waiting, not running. One way around that would be to attach a read watcher to your event loop, listening for events on "$OpenCL::INTERRUPT->pipe_fileno", using a dummy callback ("sub { }") to temporarily execute some perl code. That is then awfully close to using the built-in AnyEvent support above, though, so consider that one instead. Be creative OpenCL exports the Async::Interrupt object it uses in the global variable $OpenCL::INTERRUPT. You can configure it in any way you like. So if you want to feel like a real pro, err, wait, if you feel no risk menas no fun, you can experiment by implementing your own mode of operations. THE OpenCL PACKAGE $int = OpenCL::errno The last error returned by a function - it's only valid after an error occured and before calling another OpenCL function. $str = OpenCL::err2str [$errval] Converts an error value into a human readable string. If no error value is given, then the last error will be used (as returned by OpenCL::errno). The latest version of this module knows the error constants listed in . $str = OpenCL::enum2str $enum Converts most enum values (of parameter names, image format constants, object types, addressing and filter modes, command types etc.) into a human readable string. When confronted with some random integer it can be very helpful to pass it through this function to maybe get some readable string out of it. The latest version of this module knows the enumaration constants listed in . @platforms = OpenCL::platforms Returns all available OpenCL::Platform objects. $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr Tries to create a context from a default device and platform type - never worked for me. Consider using "$platform->context_from_type" instead. type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, OpenCL::DEVICE_TYPE_GPU, OpenCL::DEVICE_TYPE_ACCELERATOR, OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL. $ctx = OpenCL::context $properties, \@devices, $callback->($err, $pvt) = $print_stderr) Create a new OpenCL::Context object using the given device object(s). Consider using "$platform->context" instead. OpenCL::wait_for_events $wait_events... Waits for all events to complete. OpenCL::poll Checks if there are any outstanding events (see "EVENT SYSTEM") and invokes their callbacks. $OpenCL::INTERRUPT The Async::Interrupt object used to signal asynchronous events (see "EVENT SYSTEM"). $OpenCL::WATCHER The AnyEvent watcher object used to watch for asynchronous events (see "EVENT SYSTEM"). This variable is "undef" until AnyEvent has been loaded *and* initialised (e.g. by calling "AnyEvent::detect"). THE OpenCL::Object CLASS This is the base class for all objects in the OpenCL module. The only method it implements is the "id" method, which is only useful if you want to interface to OpenCL on the C level. $iv = $obj->id OpenCL objects are represented by pointers or integers on the C level. If you want to interface to an OpenCL object directly on the C level, then you need this value, which is returned by this method. You should use an "IV" type in your code and cast that to the correct type. THE OpenCL::Platform CLASS @devices = $platform->devices ($type = OpenCL::DEVICE_TYPE_ALL) Returns a list of matching OpenCL::Device objects. $ctx = $platform->context_from_type ($properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr) Tries to create a context. Never worked for me, and you need devices explicitly anyway. $ctx = $platform->context ($properties, \@devices, $callback->($err, $pvt) = $print_stderr) Create a new OpenCL::Context object using the given device object(s)- a OpenCL::CONTEXT_PLATFORM property is supplied automatically. $packed_value = $platform->info ($name) Calls "clGetPlatformInfo" and returns the packed, raw value - for strings, this will be the string (possibly including terminating \0), for other values you probably need to use the correct "unpack". It's best to avoid this method and use one of the following convenience wrappers. $platform->unload_compiler Attempts to unload the compiler for this platform, for endless profit. Does nothing on OpenCL 1.1. $string = $platform->profile Calls "clGetPlatformInfo" with "OpenCL::PLATFORM_PROFILE" and returns the result. $string = $platform->version Calls "clGetPlatformInfo" with "OpenCL::PLATFORM_VERSION" and returns the result. $string = $platform->name Calls "clGetPlatformInfo" with "OpenCL::PLATFORM_NAME" and returns the result. $string = $platform->vendor Calls "clGetPlatformInfo" with "OpenCL::PLATFORM_VENDOR" and returns the result. $string = $platform->extensions Calls "clGetPlatformInfo" with "OpenCL::PLATFORM_EXTENSIONS" and returns the result. THE OpenCL::Device CLASS $packed_value = $device->info ($name) See "$platform->info" for details. type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, OpenCL::DEVICE_TYPE_GPU, OpenCL::DEVICE_TYPE_ACCELERATOR, OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL. fp_config: OpenCL::FP_DENORM, OpenCL::FP_INF_NAN, OpenCL::FP_ROUND_TO_NEAREST, OpenCL::FP_ROUND_TO_ZERO, OpenCL::FP_ROUND_TO_INF, OpenCL::FP_FMA, OpenCL::FP_SOFT_FLOAT, OpenCL::FP_CORRECTLY_ROUNDED_DIVIDE_SQRT. mem_cache_type: OpenCL::NONE, OpenCL::READ_ONLY_CACHE, OpenCL::READ_WRITE_CACHE. local_mem_type: OpenCL::LOCAL, OpenCL::GLOBAL. exec_capabilities: OpenCL::EXEC_KERNEL, OpenCL::EXEC_NATIVE_KERNEL. command_queue_properties: OpenCL::QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE, OpenCL::QUEUE_PROFILING_ENABLE. partition_properties: OpenCL::DEVICE_PARTITION_EQUALLY, OpenCL::DEVICE_PARTITION_BY_COUNTS, OpenCL::DEVICE_PARTITION_BY_COUNTS_LIST_END, OpenCL::DEVICE_PARTITION_BY_AFFINITY_DOMAIN. affinity_domain: OpenCL::DEVICE_AFFINITY_DOMAIN_NUMA, OpenCL::DEVICE_AFFINITY_DOMAIN_L4_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L3_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L2_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L1_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_NEXT_PARTITIONABLE. @devices = $device->sub_devices (\@properties) Creates OpencL::SubDevice objects by partitioning an existing device. $device_type = $device->type Calls "clGetDeviceInfo" with "OpenCL::DEVICE_TYPE" and returns the result. $uint = $device->vendor_id Calls "clGetDeviceInfo" with "OpenCL::DEVICE_VENDOR_ID" and returns the result. $uint = $device->max_compute_units Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_COMPUTE_UNITS" and returns the result. $uint = $device->max_work_item_dimensions Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_WORK_ITEM_DIMENSIONS" and returns the result. $int = $device->max_work_group_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_WORK_GROUP_SIZE" and returns the result. @ints = $device->max_work_item_sizes Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_WORK_ITEM_SIZES" and returns the result. $uint = $device->preferred_vector_width_char Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_CHAR" and returns the result. $uint = $device->preferred_vector_width_short Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_SHORT" and returns the result. $uint = $device->preferred_vector_width_int Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_INT" and returns the result. $uint = $device->preferred_vector_width_long Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_LONG" and returns the result. $uint = $device->preferred_vector_width_float Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT" and returns the result. $uint = $device->preferred_vector_width_double Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE" and returns the result. $uint = $device->max_clock_frequency Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_CLOCK_FREQUENCY" and returns the result. $bitfield = $device->address_bits Calls "clGetDeviceInfo" with "OpenCL::DEVICE_ADDRESS_BITS" and returns the result. $uint = $device->max_read_image_args Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_READ_IMAGE_ARGS" and returns the result. $uint = $device->max_write_image_args Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_WRITE_IMAGE_ARGS" and returns the result. $ulong = $device->max_mem_alloc_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_MEM_ALLOC_SIZE" and returns the result. $int = $device->image2d_max_width Calls "clGetDeviceInfo" with "OpenCL::DEVICE_IMAGE2D_MAX_WIDTH" and returns the result. $int = $device->image2d_max_height Calls "clGetDeviceInfo" with "OpenCL::DEVICE_IMAGE2D_MAX_HEIGHT" and returns the result. $int = $device->image3d_max_width Calls "clGetDeviceInfo" with "OpenCL::DEVICE_IMAGE3D_MAX_WIDTH" and returns the result. $int = $device->image3d_max_height Calls "clGetDeviceInfo" with "OpenCL::DEVICE_IMAGE3D_MAX_HEIGHT" and returns the result. $int = $device->image3d_max_depth Calls "clGetDeviceInfo" with "OpenCL::DEVICE_IMAGE3D_MAX_DEPTH" and returns the result. $uint = $device->image_support Calls "clGetDeviceInfo" with "OpenCL::DEVICE_IMAGE_SUPPORT" and returns the result. $int = $device->max_parameter_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_PARAMETER_SIZE" and returns the result. $uint = $device->max_samplers Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_SAMPLERS" and returns the result. $uint = $device->mem_base_addr_align Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MEM_BASE_ADDR_ALIGN" and returns the result. $uint = $device->min_data_type_align_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MIN_DATA_TYPE_ALIGN_SIZE" and returns the result. $device_fp_config = $device->single_fp_config Calls "clGetDeviceInfo" with "OpenCL::DEVICE_SINGLE_FP_CONFIG" and returns the result. $device_mem_cache_type = $device->global_mem_cache_type Calls "clGetDeviceInfo" with "OpenCL::DEVICE_GLOBAL_MEM_CACHE_TYPE" and returns the result. $uint = $device->global_mem_cacheline_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_GLOBAL_MEM_CACHELINE_SIZE" and returns the result. $ulong = $device->global_mem_cache_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_GLOBAL_MEM_CACHE_SIZE" and returns the result. $ulong = $device->global_mem_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_GLOBAL_MEM_SIZE" and returns the result. $ulong = $device->max_constant_buffer_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_CONSTANT_BUFFER_SIZE" and returns the result. $uint = $device->max_constant_args Calls "clGetDeviceInfo" with "OpenCL::DEVICE_MAX_CONSTANT_ARGS" and returns the result. $device_local_mem_type = $device->local_mem_type Calls "clGetDeviceInfo" with "OpenCL::DEVICE_LOCAL_MEM_TYPE" and returns the result. $ulong = $device->local_mem_size Calls "clGetDeviceInfo" with "OpenCL::DEVICE_LOCAL_MEM_SIZE" and returns the result. $boolean = $device->error_correction_support Calls "clGetDeviceInfo" with "OpenCL::DEVICE_ERROR_CORRECTION_SUPPORT" and returns the result. $int = $device->profiling_timer_resolution Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PROFILING_TIMER_RESOLUTION" and returns the result. $boolean = $device->endian_little Calls "clGetDeviceInfo" with "OpenCL::DEVICE_ENDIAN_LITTLE" and returns the result. $boolean = $device->available Calls "clGetDeviceInfo" with "OpenCL::DEVICE_AVAILABLE" and returns the result. $boolean = $device->compiler_available Calls "clGetDeviceInfo" with "OpenCL::DEVICE_COMPILER_AVAILABLE" and returns the result. $device_exec_capabilities = $device->execution_capabilities Calls "clGetDeviceInfo" with "OpenCL::DEVICE_EXECUTION_CAPABILITIES" and returns the result. $command_queue_properties = $device->properties Calls "clGetDeviceInfo" with "OpenCL::DEVICE_QUEUE_PROPERTIES" and returns the result. $ = $device->platform Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PLATFORM" and returns the result. $string = $device->name Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NAME" and returns the result. $string = $device->vendor Calls "clGetDeviceInfo" with "OpenCL::DEVICE_VENDOR" and returns the result. $string = $device->driver_version Calls "clGetDeviceInfo" with "OpenCL::DRIVER_VERSION" and returns the result. $string = $device->profile Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PROFILE" and returns the result. $string = $device->version Calls "clGetDeviceInfo" with "OpenCL::DEVICE_VERSION" and returns the result. $string = $device->extensions Calls "clGetDeviceInfo" with "OpenCL::DEVICE_EXTENSIONS" and returns the result. $uint = $device->preferred_vector_width_half Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PREFERRED_VECTOR_WIDTH_HALF" and returns the result. $uint = $device->native_vector_width_char Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_CHAR" and returns the result. $uint = $device->native_vector_width_short Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_SHORT" and returns the result. $uint = $device->native_vector_width_int Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_INT" and returns the result. $uint = $device->native_vector_width_long Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_LONG" and returns the result. $uint = $device->native_vector_width_float Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_FLOAT" and returns the result. $uint = $device->native_vector_width_double Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE" and returns the result. $uint = $device->native_vector_width_half Calls "clGetDeviceInfo" with "OpenCL::DEVICE_NATIVE_VECTOR_WIDTH_HALF" and returns the result. $device_fp_config = $device->double_fp_config Calls "clGetDeviceInfo" with "OpenCL::DEVICE_DOUBLE_FP_CONFIG" and returns the result. $device_fp_config = $device->half_fp_config Calls "clGetDeviceInfo" with "OpenCL::DEVICE_HALF_FP_CONFIG" and returns the result. $boolean = $device->host_unified_memory Calls "clGetDeviceInfo" with "OpenCL::DEVICE_HOST_UNIFIED_MEMORY" and returns the result. $device = $device->parent_device_ext Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PARENT_DEVICE_EXT" and returns the result. @device_partition_property_exts = $device->partition_types_ext Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PARTITION_TYPES_EXT" and returns the result. @device_partition_property_exts = $device->affinity_domains_ext Calls "clGetDeviceInfo" with "OpenCL::DEVICE_AFFINITY_DOMAINS_EXT" and returns the result. $uint = $device->reference_count_ext Calls "clGetDeviceInfo" with "OpenCL::DEVICE_REFERENCE_COUNT_EXT" and returns the result. @device_partition_property_exts = $device->partition_style_ext Calls "clGetDeviceInfo" with "OpenCL::DEVICE_PARTITION_STYLE_EXT" and returns the result. THE OpenCL::Context CLASS An OpenCL::Context is basically a container, or manager, for a number of devices of a platform. It is used to create all sorts of secondary objects such as buffers, queues, programs and so on. All context creation functions and methods take a list of properties (type-value pairs). All property values can be specified as integers - some additionally support other types: OpenCL::CONTEXT_PLATFORM Also accepts OpenCL::Platform objects. OpenCL::GLX_DISPLAY_KHR Also accepts "undef", in which case a deep and troubling hack is engaged to find the current glx display (see "GLX SUPPORT"). OpenCL::GL_CONTEXT_KHR Also accepts "undef", in which case a deep and troubling hack is engaged to find the current glx context (see "GLX SUPPORT"). $prog = $ctx->build_program ($program, $options = "") This convenience function tries to build the program on all devices in the context. If the build fails, then the function will "croak" with the build log. Otherwise ti returns the program object. The $program can either be a "OpenCL::Program" object or a string containing the program. In the latter case, a program objetc will be created automatically. $queue = $ctx->queue ($device, $properties) Create a new OpenCL::Queue object from the context and the given device. Example: create an out-of-order queue. $queue = $ctx->queue ($device, OpenCL::QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE); $ev = $ctx->user_event Creates a new OpenCL::UserEvent object. $buf = $ctx->buffer ($flags, $len) Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object with the given flags and octet-size. flags: OpenCL::MEM_READ_WRITE, OpenCL::MEM_WRITE_ONLY, OpenCL::MEM_READ_ONLY, OpenCL::MEM_USE_HOST_PTR, OpenCL::MEM_ALLOC_HOST_PTR, OpenCL::MEM_COPY_HOST_PTR, OpenCL::MEM_HOST_WRITE_ONLY, OpenCL::MEM_HOST_READ_ONLY, OpenCL::MEM_HOST_NO_ACCESS. $buf = $ctx->buffer_sv ($flags, $data) Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object and initialise it with the given data values. $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) Creates a new OpenCL::Image object and optionally initialises it with the given data values. channel_order: OpenCL::R, OpenCL::A, OpenCL::RG, OpenCL::RA, OpenCL::RGB, OpenCL::RGBA, OpenCL::BGRA, OpenCL::ARGB, OpenCL::INTENSITY, OpenCL::LUMINANCE, OpenCL::Rx, OpenCL::RGx, OpenCL::RGBx. channel_type: OpenCL::SNORM_INT8, OpenCL::SNORM_INT16, OpenCL::UNORM_INT8, OpenCL::UNORM_INT16, OpenCL::UNORM_SHORT_565, OpenCL::UNORM_SHORT_555, OpenCL::UNORM_INT_101010, OpenCL::SIGNED_INT8, OpenCL::SIGNED_INT16, OpenCL::SIGNED_INT32, OpenCL::UNSIGNED_INT8, OpenCL::UNSIGNED_INT16, OpenCL::UNSIGNED_INT32, OpenCL::HALF_FLOAT, OpenCL::FLOAT. type: OpenCL::MEM_OBJECT_BUFFER, OpenCL::MEM_OBJECT_IMAGE2D, OpenCL::MEM_OBJECT_IMAGE3D, OpenCL::MEM_OBJECT_IMAGE2D_ARRAY, OpenCL::MEM_OBJECT_IMAGE1D, OpenCL::MEM_OBJECT_IMAGE1D_ARRAY, OpenCL::MEM_OBJECT_IMAGE1D_BUFFER. $img = $ctx->image2d ($flags, $channel_order, $channel_type, $width, $height, $row_pitch = 0, $data = undef) Creates a new OpenCL::Image2D object and optionally initialises it with the given data values. $img = $ctx->image3d ($flags, $channel_order, $channel_type, $width, $height, $depth, $row_pitch = 0, $slice_pitch = 0, $data = undef) Creates a new OpenCL::Image3D object and optionally initialises it with the given data values. $buffer = $ctx->gl_buffer ($flags, $bufobj) Creates a new OpenCL::Buffer (actually OpenCL::BufferObj) object that refers to the given OpenGL buffer object. flags: OpenCL::MEM_READ_WRITE, OpenCL::MEM_READ_ONLY, OpenCL::MEM_WRITE_ONLY. http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFr omGLBuffer.html $img = $ctx->gl_texture ($flags, $target, $miplevel, $texture) Creates a new OpenCL::Image object that refers to the given OpenGL texture object or buffer. target: GL_TEXTURE_1D, GL_TEXTURE_1D_ARRAY, GL_TEXTURE_BUFFER, GL_TEXTURE_2D, GL_TEXTURE_2D_ARRAY, GL_TEXTURE_3D, GL_TEXTURE_CUBE_MAP_POSITIVE_X, GL_TEXTURE_CUBE_MAP_POSITIVE_Y, GL_TEXTURE_CUBE_MAP_POSITIVE_Z, GL_TEXTURE_CUBE_MAP_NEGATIVE_X, GL_TEXTURE_CUBE_MAP_NEGATIVE_Y, GL_TEXTURE_CUBE_MAP_NEGATIVE_Z, GL_TEXTURE_RECTANGLE/GL_TEXTURE_RECTANGLE_ARB. http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateFr omGLTexture.html $img = $ctx->gl_texture2d ($flags, $target, $miplevel, $texture) Creates a new OpenCL::Image2D object that refers to the given OpenGL 2D texture object. http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFr omGLTexture2D.html $img = $ctx->gl_texture3d ($flags, $target, $miplevel, $texture) Creates a new OpenCL::Image3D object that refers to the given OpenGL 3D texture object. http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFr omGLTexture3D.html $ctx->gl_renderbuffer ($flags, $renderbuffer) Creates a new OpenCL::Image2D object that refers to the given OpenGL render buffer. http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateFr omGLRenderbuffer.html @formats = $ctx->supported_image_formats ($flags, $image_type) Returns a list of matching image formats - each format is an arrayref with two values, $channel_order and $channel_type, in it. $sampler = $ctx->sampler ($normalized_coords, $addressing_mode, $filter_mode) Creates a new OpenCL::Sampler object. addressing_mode: OpenCL::ADDRESS_NONE, OpenCL::ADDRESS_CLAMP_TO_EDGE, OpenCL::ADDRESS_CLAMP, OpenCL::ADDRESS_REPEAT, OpenCL::ADDRESS_MIRRORED_REPEAT. filter_mode: OpenCL::FILTER_NEAREST, OpenCL::FILTER_LINEAR. $program = $ctx->program_with_source ($string) Creates a new OpenCL::Program object from the given source code. ($program, \@status) = $ctx->program_with_binary (\@devices, \@binaries) Creates a new OpenCL::Program object from the given binaries. Example: clone an existing program object that contains a successfully compiled program, no matter how useless this is. my $clone = $ctx->program_with_binary ([$prog->devices], [$prog->binaries]); $program = $ctx->program_with_built_in_kernels (\@devices, $kernel_names) Creates a new OpenCL::Program object from the given built-in kernel names. $program = $ctx->link_program (\@devices, $options, \@programs, $cb->($program) = undef) Links all (already compiled) program objects specified in @programs together and returns a new OpenCL::Program object with the result. $packed_value = $ctx->info ($name) See "$platform->info" for details. $uint = $context->reference_count Calls "clGetContextInfo" with "OpenCL::CONTEXT_REFERENCE_COUNT" and returns the result. @devices = $context->devices Calls "clGetContextInfo" with "OpenCL::CONTEXT_DEVICES" and returns the result. @property_ints = $context->properties Calls "clGetContextInfo" with "OpenCL::CONTEXT_PROPERTIES" and returns the result. $uint = $context->num_devices Calls "clGetContextInfo" with "OpenCL::CONTEXT_NUM_DEVICES" and returns the result. THE OpenCL::Queue CLASS An OpenCL::Queue represents an execution queue for OpenCL. You execute requests by calling their respective method and waiting for it to complete in some way. Most methods that enqueue some request return an event object that can be used to wait for completion (optionally using a callback), unless the method is called in void context, in which case no event object is created. They also allow you to specify any number of other event objects that this request has to wait for before it starts executing, by simply passing the event objects as extra parameters to the enqueue methods. To simplify program design, this module ignores any "undef" values in the list of events. This makes it possible to code operations such as this, without having to put a valid event object into $event first: $event = $queue->xxx (..., $event); Queues execute in-order by default, without any parallelism, so in most cases (i.e. you use only one queue) it's not necessary to wait for or create event objects, althoguh an our of order queue is often a bit faster. $ev = $queue->read_buffer ($buffer, $blocking, $offset, $len, $data, $wait_events...) Reads data from buffer into the given string. $ev = $queue->write_buffer ($buffer, $blocking, $offset, $data, $wait_events...) Writes data to buffer from the given string. $ev = $queue->copy_buffer ($src, $dst, $src_offset, $dst_offset, $len, $wait_events...) $eue->read_buffer_rect ($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...) http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueR eadBufferRect.html $ev = $queue->write_buffer_rect ($buf, $blocking, $buf_y, $host_x, $host_z, $height, $buf_row_pitch, $host_row_pitch, $data, $wait_events...) http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueW riteBufferRect.html $ev = $queue->copy_buffer_to_image ($src_buffer, $dst_image, $src_offset, $dst_x, $dst_y, $dst_z, $width, $height, $depth, $wait_events...) $ev = $queue->read_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...) $row_pitch (and $slice_pitch) can be 0, in which case the OpenCL module uses the image width (and height) to supply default values. $ev = $queue->write_image ($src, $blocking, $x, $y, $z, $width, $height, $depth, $row_pitch, $slice_pitch, $data, $wait_events...) $row_pitch (and $slice_pitch) can be 0, in which case the OpenCL module uses the image width (and height) to supply default values. $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...) $ev = $queue->copy_image_to_buffer ($src_image, $dst_image, $src_x, $src_y, $src_z, $width, $height, $depth, $dst_offset, $wait_events...) $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...) Yeah. . $ev = $queue->fill_buffer ($mem, $pattern, $offset, $size, ...) Fills the given buffer object with repeated applications of $pattern, starting at $offset for $size octets. $ev = $queue->fill_image ($img, $r, $g, $b, $a, $x, $y, $z, $width, $height, $depth, ...) Fills the given image area with the given rgba colour components. The components are normally floating point values between 0 and 1, except when the image channel data type is a signe dor unsigned unnormalised format, in which case the range is determined by the format. $ev = $queue->task ($kernel, $wait_events...) $ev = $queue->nd_range_kernel ($kernel, \@global_work_offset, \@global_work_size, \@local_work_size, $wait_events...) Enqueues a kernel execution. \@global_work_size must be specified as a reference to an array of integers specifying the work sizes (element counts). \@global_work_offset must be either "undef" (in which case all offsets are 0), or a reference to an array of work offsets, with the same number of elements as \@global_work_size. \@local_work_size must be either "undef" (in which case the implementation is supposed to choose good local work sizes), or a reference to an array of local work sizes, with the same number of elements as \@global_work_size. $ev = $queue->migrate_mem_objects (\@mem_objects, $flags, $wait_events...) Migrates a number of OpenCL::Memory objects to or from the device. flags: OpenCL::MIGRATE_MEM_OBJECT_HOST, OpenCL::MIGRATE_MEM_OBJECT_CONTENT_UNDEFINED $ev = $queue->acquire_gl_objects ([object, ...], $wait_events...) Enqueues a list (an array-ref of OpenCL::Memory objects) to be acquired for subsequent OpenCL usage. $ev = $queue->release_gl_objects ([object, ...], $wait_events...) Enqueues a list (an array-ref of OpenCL::Memory objects) to be released for subsequent OpenGL usage. $ev = $queue->wait_for_events ($wait_events...) $ev = $queue->marker ($wait_events...) $ev = $queue->barrier ($wait_events...) $queue->flush $queue->finish $packed_value = $queue->info ($name) See "$platform->info" for details. $ctx = $command_queue->context Calls "clGetCommandQueueInfo" with "OpenCL::QUEUE_CONTEXT" and returns the result. $device = $command_queue->device Calls "clGetCommandQueueInfo" with "OpenCL::QUEUE_DEVICE" and returns the result. $uint = $command_queue->reference_count Calls "clGetCommandQueueInfo" with "OpenCL::QUEUE_REFERENCE_COUNT" and returns the result. $command_queue_properties = $command_queue->properties Calls "clGetCommandQueueInfo" with "OpenCL::QUEUE_PROPERTIES" and returns the result. MEMORY MAPPED BUFFERS OpenCL allows you to map buffers and images to host memory (read: perl scalars). This is done much like reading or copying a buffer, by enqueuing a map or unmap operation on the command queue. The map operations return an "OpenCL::Mapped" object - see "THE OpenCL::Mapped CLASS" section for details on what to do with these objects. The object will be unmapped automatically when the mapped object is destroyed (you can use a barrier to make sure the unmap has finished, before using the buffer in a kernel), but you can also enqueue an unmap operation manually. $mapped_buffer = $queue->map_buffer ($buf, $blocking=1, $map_flags=OpenCL::MAP_READ|OpenCL::MAP_WRITE, $offset=0, $size=undef, $wait_events...) Maps the given buffer into host memory and returns an "OpenCL::MappedBuffer" object. If $size is specified as undef, then the map will extend to the end of the buffer. map_flags: OpenCL::MAP_READ, OpenCL::MAP_WRITE, OpenCL::MAP_WRITE_INVALIDATE_REGION. Example: map the buffer $buf fully and replace the first 4 bytes by "abcd", then unmap. { my $mapped = $queue->map_buffer ($buf, 1, OpenCL::MAP_WRITE); substr $$mapped, 0, 4, "abcd"; } # asynchronously unmap because $mapped is destroyed $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...) Maps the given image area into host memory and return an "OpenCL::MappedImage" object. If any of $width, $height and/or $depth are "undef" then they will be replaced by the maximum possible value. Example: map an image (with OpenCL::UNSIGNED_INT8 channel type) and set the first channel of the leftmost column to 5, then explicitly unmap it. You are not necessarily meant to do it this way, this example just shows you the accessors to use :) my $mapped = $queue->map_image ($image, 1, OpenCL::MAP_WRITE); $mapped->write ($_ * $mapped->row_pitch, pack "C", 5) for 0 .. $mapped->height - 1; $mapped->unmap;. $mapped->wait; # only needed for out of order queues normally $ev = $queue->unmap ($mapped, $wait_events...) Unmaps the data from host memory. You must not call any methods that modify the data, or modify the data scalar directly, after calling this method. The mapped event object will always be passed as part of the $wait_events. The mapped event object will be replaced by the new event object that this request creates. THE OpenCL::Memory CLASS This the superclass of all memory objects - OpenCL::Buffer, OpenCL::Image, OpenCL::Image2D and OpenCL::Image3D. $packed_value = $memory->info ($name) See "$platform->info" for details. $memory->destructor_callback ($cb->()) Sets a callback that will be invoked after the memory object is destructed. $mem_object_type = $mem->type Calls "clGetMemObjectInfo" with "OpenCL::MEM_TYPE" and returns the result. $mem_flags = $mem->flags Calls "clGetMemObjectInfo" with "OpenCL::MEM_FLAGS" and returns the result. $int = $mem->size Calls "clGetMemObjectInfo" with "OpenCL::MEM_SIZE" and returns the result. $ptr_value = $mem->host_ptr Calls "clGetMemObjectInfo" with "OpenCL::MEM_HOST_PTR" and returns the result. $uint = $mem->map_count Calls "clGetMemObjectInfo" with "OpenCL::MEM_MAP_COUNT" and returns the result. $uint = $mem->reference_count Calls "clGetMemObjectInfo" with "OpenCL::MEM_REFERENCE_COUNT" and returns the result. $ctx = $mem->context Calls "clGetMemObjectInfo" with "OpenCL::MEM_CONTEXT" and returns the result. $mem = $mem->associated_memobject Calls "clGetMemObjectInfo" with "OpenCL::MEM_ASSOCIATED_MEMOBJECT" and returns the result. $int = $mem->offset Calls "clGetMemObjectInfo" with "OpenCL::MEM_OFFSET" and returns the result. ($type, $name) = $mem->gl_object_info Returns the OpenGL object type (e.g. OpenCL::GL_OBJECT_TEXTURE2D) and the object "name" (e.g. the texture name) used to create this memory object. THE OpenCL::Buffer CLASS This is a subclass of OpenCL::Memory, and the superclass of OpenCL::BufferObj. Its purpose is simply to distinguish between buffers and sub-buffers. THE OpenCL::BufferObj CLASS This is a subclass of OpenCL::Buffer and thus OpenCL::Memory. It exists because one cna create sub buffers of OpenLC::BufferObj objects, but not sub buffers from these sub buffers. $subbuf = $buf_obj->sub_buffer_region ($flags, $origin, $size) Creates an OpenCL::Buffer objects from this buffer and returns it. The "buffer_create_type" is assumed to be "OpenCL::BUFFER_CREATE_TYPE_REGION". THE OpenCL::Image CLASS This is the superclass of all image objects - OpenCL::Image1D, OpenCL::Image1DArray, OpenCL::Image1DBuffer, OpenCL::Image2D, OpenCL::Image2DArray and OpenCL::Image3D. $packed_value = $image->image_info ($name) See "$platform->info" for details. The reason this method is not called "info" is that there already is an "->info" method inherited from "OpenCL::Memory". ($channel_order, $channel_data_type) = $image->format Returns the channel order and type used to create the image by calling "clGetImageInfo" with "OpenCL::IMAGE_FORMAT". $int = $image->element_size Calls "clGetImageInfo" with "OpenCL::IMAGE_ELEMENT_SIZE" and returns the result. $int = $image->row_pitch Calls "clGetImageInfo" with "OpenCL::IMAGE_ROW_PITCH" and returns the result. $int = $image->slice_pitch Calls "clGetImageInfo" with "OpenCL::IMAGE_SLICE_PITCH" and returns the result. $int = $image->width Calls "clGetImageInfo" with "OpenCL::IMAGE_WIDTH" and returns the result. $int = $image->height Calls "clGetImageInfo" with "OpenCL::IMAGE_HEIGHT" and returns the result. $int = $image->depth Calls "clGetImageInfo" with "OpenCL::IMAGE_DEPTH" and returns the result. $GLenum = $gl_texture->target Calls "clGetGLTextureInfo" with "OpenCL::GL_TEXTURE_TARGET" and returns the result. $GLint = $gl_texture->gl_mipmap_level Calls "clGetGLTextureInfo" with "OpenCL::GL_MIPMAP_LEVEL" and returns the result. THE OpenCL::Sampler CLASS $packed_value = $sampler->info ($name) See "$platform->info" for details. $uint = $sampler->reference_count Calls "clGetSamplerInfo" with "OpenCL::SAMPLER_REFERENCE_COUNT" and returns the result. $ctx = $sampler->context Calls "clGetSamplerInfo" with "OpenCL::SAMPLER_CONTEXT" and returns the result. $addressing_mode = $sampler->normalized_coords Calls "clGetSamplerInfo" with "OpenCL::SAMPLER_NORMALIZED_COORDS" and returns the result. $filter_mode = $sampler->addressing_mode Calls "clGetSamplerInfo" with "OpenCL::SAMPLER_ADDRESSING_MODE" and returns the result. $boolean = $sampler->filter_mode Calls "clGetSamplerInfo" with "OpenCL::SAMPLER_FILTER_MODE" and returns the result. THE OpenCL::Program CLASS $program->build (\@devices = undef, $options = "", $cb->($program) = undef) Tries to build the program with the given options. See also the "$ctx-"build> convenience function. If a callback is specified, then it will be called when compilation is finished. Note that many OpenCL implementations block your program while compiling whether you use a callback or not. See "build_async" if you want to make sure the build is done in the background. Note that some OpenCL implementations act up badly, and don't call the callback in some error cases (but call it in others). This implementation assumes the callback will always be called, and leaks memory if this is not so. So best make sure you don't pass in invalid values. Some implementations fail with "OpenCL::INVALID_BINARY" when the compilation state is successful but some later stage fails. options: "-D name", "-D name=definition", "-I dir", "-cl-single-precision-constant", "-cl-denorms-are-zero", "-cl-fp32-correctly-rounded-divide-sqrt", "-cl-opt-disable", "-cl-mad-enable", "-cl-no-signed-zeros", "-cl-unsafe-math-optimizations", "-cl-finite-math-only", "-cl-fast-relaxed-math", "-w", "-Werror", "-cl-std=CL1.1/CL1.2", "-cl-kernel-arg-info", "-create-library", "-enable-link-options". build_status: OpenCL::BUILD_SUCCESS, OpenCL::BUILD_NONE, OpenCL::BUILD_ERROR, OpenCL::BUILD_IN_PROGRESS. $program->build_async (\@devices = undef, $options = "", $cb->($program) = undef) Similar to "->build", except it starts a thread, and never fails (you need to check the compilation status form the callback, or by polling). $program->compile (\@devices = undef, $options = "", \%headers = undef, $cb->($program) = undef) Compiles the given program for the given devices (or all devices if undef). If $headers is given, it must be a hashref with include name => OpenCL::Program pairs. $packed_value = $program->build_info ($device, $name) Similar to "$platform->info", but returns build info for a previous build attempt for the given device. binary_type: OpenCL::PROGRAM_BINARY_TYPE_NONE, OpenCL::PROGRAM_BINARY_TYPE_COMPILED_OBJECT, OpenCL::PROGRAM_BINARY_TYPE_LIBRARY, OpenCL::PROGRAM_BINARY_TYPE_EXECUTABLE. $kernel = $program->kernel ($function_name) Creates an OpenCL::Kernel object out of the named "__kernel" function in the program. @kernels = $program->kernels_in_program Returns all kernels successfully compiled for all devices in program. http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateKe rnelsInProgram.html $build_status = $program->build_status ($device) Calls "clGetProgramBuildInfo" with "OpenCL::PROGRAM_BUILD_STATUS" and returns the result. $string = $program->build_options ($device) Calls "clGetProgramBuildInfo" with "OpenCL::PROGRAM_BUILD_OPTIONS" and returns the result. $string = $program->build_log ($device) Calls "clGetProgramBuildInfo" with "OpenCL::PROGRAM_BUILD_LOG" and returns the result. $binary_type = $program->binary_type ($device) Calls "clGetProgramBuildInfo" with "OpenCL::PROGRAM_BINARY_TYPE" and returns the result. $packed_value = $program->info ($name) See "$platform->info" for details. $uint = $program->reference_count Calls "clGetProgramInfo" with "OpenCL::PROGRAM_REFERENCE_COUNT" and returns the result. $ctx = $program->context Calls "clGetProgramInfo" with "OpenCL::PROGRAM_CONTEXT" and returns the result. $uint = $program->num_devices Calls "clGetProgramInfo" with "OpenCL::PROGRAM_NUM_DEVICES" and returns the result. @devices = $program->devices Calls "clGetProgramInfo" with "OpenCL::PROGRAM_DEVICES" and returns the result. $string = $program->source Calls "clGetProgramInfo" with "OpenCL::PROGRAM_SOURCE" and returns the result. @ints = $program->binary_sizes Calls "clGetProgramInfo" with "OpenCL::PROGRAM_BINARY_SIZES" and returns the result. @blobs = $program->binaries Returns a string for the compiled binary for every device associated with the program, empty strings indicate missing programs, and an empty result means no program binaries are available. These "binaries" are often, in fact, informative low-level assembly sources. THE OpenCL::Kernel CLASS $packed_value = $kernel->info ($name) See "$platform->info" for details. $string = $kernel->function_name Calls "clGetKernelInfo" with "OpenCL::KERNEL_FUNCTION_NAME" and returns the result. $uint = $kernel->num_args Calls "clGetKernelInfo" with "OpenCL::KERNEL_NUM_ARGS" and returns the result. $uint = $kernel->reference_count Calls "clGetKernelInfo" with "OpenCL::KERNEL_REFERENCE_COUNT" and returns the result. $ctx = $kernel->context Calls "clGetKernelInfo" with "OpenCL::KERNEL_CONTEXT" and returns the result. $program = $kernel->program Calls "clGetKernelInfo" with "OpenCL::KERNEL_PROGRAM" and returns the result. $packed_value = $kernel->work_group_info ($device, $name) See "$platform->info" for details. $int = $kernel->work_group_size ($device) Calls "clGetKernelWorkGroupInfo" with "OpenCL::KERNEL_WORK_GROUP_SIZE" and returns the result. @ints = $kernel->compile_work_group_size ($device) Calls "clGetKernelWorkGroupInfo" with "OpenCL::KERNEL_COMPILE_WORK_GROUP_SIZE" and returns the result. $ulong = $kernel->local_mem_size ($device) Calls "clGetKernelWorkGroupInfo" with "OpenCL::KERNEL_LOCAL_MEM_SIZE" and returns the result. $int = $kernel->preferred_work_group_size_multiple ($device) Calls "clGetKernelWorkGroupInfo" with "OpenCL::KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE" and returns the result. $ulong = $kernel->private_mem_size ($device) Calls "clGetKernelWorkGroupInfo" with "OpenCL::KERNEL_PRIVATE_MEM_SIZE" and returns the result. $packed_value = $kernel->arg_info ($idx, $name) See "$platform->info" for details. $kernel_arg_address_qualifier = $kernel->arg_address_qualifier ($idx) Calls "clGetKernelArgInfo" with "OpenCL::KERNEL_ARG_ADDRESS_QUALIFIER" and returns the result. $kernel_arg_access_qualifier = $kernel->arg_access_qualifier ($idx) Calls "clGetKernelArgInfo" with "OpenCL::KERNEL_ARG_ACCESS_QUALIFIER" and returns the result. $string = $kernel->arg_type_name ($idx) Calls "clGetKernelArgInfo" with "OpenCL::KERNEL_ARG_TYPE_NAME" and returns the result. $kernel_arg_type_qualifier = $kernel->arg_type_qualifier ($idx) Calls "clGetKernelArgInfo" with "OpenCL::KERNEL_ARG_TYPE_QUALIFIER" and returns the result. $string = $kernel->arg_name ($idx) Calls "clGetKernelArgInfo" with "OpenCL::KERNEL_ARG_NAME" and returns the result. $kernel->setf ($format, ...) Sets the arguments of a kernel. Since OpenCL 1.1 doesn't have a generic way to set arguments (and with OpenCL 1.2 it might be rather slow), you need to specify a format argument, much as with "printf", to tell OpenCL what type of argument it is. The format arguments are single letters: c char C unsigned char s short S unsigned short i int I unsigned int l long L unsigned long h half float (0..65535) f float d double z local (octet size) m memory object (buffer or image) a sampler e event Space characters in the format string are ignored. Example: set the arguments for a kernel that expects an int, two floats, a buffer and an image. $kernel->setf ("i ff mm", 5, 0.5, 3, $buffer, $image); $kernel->set_TYPE ($index, $value) $kernel->set_char ($index, $value) $kernel->set_uchar ($index, $value) $kernel->set_short ($index, $value) $kernel->set_ushort ($index, $value) $kernel->set_int ($index, $value) $kernel->set_uint ($index, $value) $kernel->set_long ($index, $value) $kernel->set_ulong ($index, $value) $kernel->set_half ($index, $value) $kernel->set_float ($index, $value) $kernel->set_double ($index, $value) $kernel->set_memory ($index, $value) $kernel->set_buffer ($index, $value) $kernel->set_image ($index, $value) $kernel->set_sampler ($index, $value) $kernel->set_local ($index, $value) $kernel->set_event ($index, $value) This is a family of methods to set the kernel argument with the number $index to the give $value. Chars and integers (including the half type) are specified as integers, float and double as floating point values, memory/buffer/image must be an object of that type or "undef", local-memory arguments are set by specifying the size, and sampler and event must be objects of that type. Note that "set_memory" works for all memory objects (all types of buffers and images) - the main purpose of the more specific "set_TYPE" functions is type checking. Setting an argument for a kernel does NOT keep a reference to the object - for example, if you set an argument to some image object, free the image, and call the kernel, you will run into undefined behaviour. THE OpenCL::Event CLASS This is the superclass for all event objects (including OpenCL::UserEvent objects). $ev->wait Waits for the event to complete. $ev->cb ($exec_callback_type, $callback->($event, $event_command_exec_status)) Adds a callback to the callback stack for the given event type. There is no way to remove a callback again. $packed_value = $ev->info ($name) See "$platform->info" for details. $queue = $event->command_queue Calls "clGetEventInfo" with "OpenCL::EVENT_COMMAND_QUEUE" and returns the result. $command_type = $event->command_type Calls "clGetEventInfo" with "OpenCL::EVENT_COMMAND_TYPE" and returns the result. $uint = $event->reference_count Calls "clGetEventInfo" with "OpenCL::EVENT_REFERENCE_COUNT" and returns the result. $uint = $event->command_execution_status Calls "clGetEventInfo" with "OpenCL::EVENT_COMMAND_EXECUTION_STATUS" and returns the result. $ctx = $event->context Calls "clGetEventInfo" with "OpenCL::EVENT_CONTEXT" and returns the result. $packed_value = $ev->profiling_info ($name) See "$platform->info" for details. The reason this method is not called "info" is that there already is an "->info" method. $ulong = $event->profiling_command_queued Calls "clGetEventProfilingInfo" with "OpenCL::PROFILING_COMMAND_QUEUED" and returns the result. $ulong = $event->profiling_command_submit Calls "clGetEventProfilingInfo" with "OpenCL::PROFILING_COMMAND_SUBMIT" and returns the result. $ulong = $event->profiling_command_start Calls "clGetEventProfilingInfo" with "OpenCL::PROFILING_COMMAND_START" and returns the result. $ulong = $event->profiling_command_end Calls "clGetEventProfilingInfo" with "OpenCL::PROFILING_COMMAND_END" and returns the result. THE OpenCL::UserEvent CLASS This is a subclass of OpenCL::Event. $ev->set_status ($execution_status) Sets the execution status of the user event. Can only be called once, either with OpenCL::COMPLETE or a negative number as status. execution_status: OpenCL::COMPLETE or a negative integer. THE OpenCL::Mapped CLASS This class represents objects mapped into host memory. They are represented by a blessed string scalar. The string data is the mapped memory area, that is, if you read or write it, then the mapped object is accessed directly. You must only ever use operations that modify the string in-place - for example, a "substr" that doesn't change the length, or maybe a regex that doesn't change the length. Any other operation might cause the data to be copied. When the object is destroyed it will enqueue an implicit unmap operation on the queue that was used to create it. Keep in mind that you *need* to unmap (or destroy) mapped objects before OpenCL sees the changes, even if some implementations don't need this sometimes. Example, replace the first two floats in the mapped buffer by 1 and 2. my $mapped = $queue->map_buffer ($buf, ... $mapped->event->wait; # make sure it's there # now replace first 8 bytes by new data, which is exactly 8 bytes long # we blindly assume device endianness to equal host endianness # (and of course, we assume iee 754 single precision floats :) substr $$mapped, 0, 8, pack "f*", 1, 2; $ev = $mapped->unmap ($wait_events...) Unmaps the mapped memory object, using the queue originally used to create it, quite similarly to "$queue->unmap ($mapped, ...)". $bool = $mapped->mapped Returns whether the object is still mapped - true before an "unmap" is enqueued, false afterwards. $ev = $mapped->event Return the event object associated with the mapped object. Initially, this will be the event object created when mapping the object, and after an unmap, this will be the event object that the unmap operation created. $mapped->wait Same as "$mapped->event->wait" - makes sure no operations on this mapped object are outstanding. $bytes = $mapped->size Returns the size of the mapped area, in bytes. Same as "length $$mapped". $ptr = $mapped->ptr Returns the raw memory address of the mapped area. $mapped->set ($offset, $data) Replaces the data at the given $offset in the memory area by the new $data. This method is safer than direct manipulation of $mapped because it does bounds-checking, but also slower. $data = $mapped->get ($offset, $length) Returns (without copying) a scalar representing the data at the given $offset and $length in the mapped memory area. This is the same as the following substr, except much slower; $data = substr $$mapped, $offset, $length THE OpenCL::MappedBuffer CLASS This is a subclass of OpenCL::Mapped, representing mapped buffers. THE OpenCL::MappedImage CLASS This is a subclass of OpenCL::Mapped, representing mapped images. $pixels = $mapped->width $pixels = $mapped->height $pixels = $mapped->depth Return the width/height/depth of the mapped image region, in pixels. $bytes = $mapped->row_pitch $bytes = $mapped->slice_pitch Return the row or slice pitch of the image that has been mapped. $bytes = $mapped->element_size Return the size of a single pixel. $data = $mapped->get_row ($count, $x=0, $y=0, $z=0) Return $count pixels from the given coordinates. The pixel data must be completely contained within a single row. If $count is "undef", then all the remaining pixels in that row are returned. $mapped->set_row ($data, $x=0, $y=0, $z=0) Write the given pixel data at the given coordinate. The pixel data must be completely contained within a single row. AUTHOR Marc Lehmann http://home.schmorp.de/