… | |
… | |
96 | initialised at creation time). |
96 | initialised at creation time). |
97 | |
97 | |
98 | Enqueue the kernel execution. |
98 | Enqueue the kernel execution. |
99 | |
99 | |
100 | Enqueue buffer reads for your output buffer to read results. |
100 | Enqueue buffer reads for your output buffer to read results. |
|
|
101 | |
|
|
102 | =head1 OPENCL 1.1 VS. OPENCL 1.2 |
|
|
103 | |
|
|
104 | This module supports both OpenCL version 1.1 and 1.2, although the OpenCL |
|
|
105 | 1.2 interface hasn't been tested much for lack of availability of an |
|
|
106 | actual implementation. |
|
|
107 | |
|
|
108 | Every function or method in this manual page that interfaces to a |
|
|
109 | particular OpenCL function has a link to the its C manual page. |
|
|
110 | |
|
|
111 | If the link contains a F<1.1>, then this function is an OpenCL 1.1 |
|
|
112 | function. Most but not all also exist in OpenCL 1.2, and this module |
|
|
113 | tries to emulate the missing ones for you, when told to do so at |
|
|
114 | compiletime. You cna check whether a function was removed in OpenCL 1.2 by |
|
|
115 | replacing the F<1.1> component in the URL by F<1.2>. |
|
|
116 | |
|
|
117 | If the link contains a F<1.2>, then this is a OpenCL 1.2-only |
|
|
118 | function. Even if the module was compiled with OpenCL 1.2 header files |
|
|
119 | and has an 1.2 OpenCL library, calling such a function on a platform that |
|
|
120 | doesn't implement 1.2 causes undefined behaviour, usually a crash (But |
|
|
121 | this is not guaranteed). |
|
|
122 | |
|
|
123 | You can find out whether this module was compiled to prefer 1.1 |
|
|
124 | functionality by ooking at C<OpenCL::PREFER_1_1> - if it is true, then |
|
|
125 | 1.1 functions generally are implemented using 1.1 OpenCL functions. If it |
|
|
126 | is false, then 1.1 functions missing from 1.2 are emulated by calling 1.2 |
|
|
127 | fucntions. |
|
|
128 | |
|
|
129 | This is a somewhat sorry state of affairs, but the Khronos group choose to |
|
|
130 | make every release of OpenCL source and binary incompatible with previous |
|
|
131 | releases. |
101 | |
132 | |
102 | =head1 EXAMPLES |
133 | =head1 EXAMPLES |
103 | |
134 | |
104 | =head2 Enumerate all devices and get contexts for them. |
135 | =head2 Enumerate all devices and get contexts for them. |
105 | |
136 | |
… | |
… | |
434 | OpenCL can generate a number of (potentially) asynchronous events, for |
465 | OpenCL can generate a number of (potentially) asynchronous events, for |
435 | example, after compiling a program, to signal a context-related error or, |
466 | example, after compiling a program, to signal a context-related error or, |
436 | perhaps most important, to signal completion of queued jobs (by setting |
467 | perhaps most important, to signal completion of queued jobs (by setting |
437 | callbacks on OpenCL::Event objects). |
468 | callbacks on OpenCL::Event objects). |
438 | |
469 | |
|
|
470 | The OpenCL module converts all these callbacks into events - you can |
|
|
471 | still register callbacks, but they are not executed when your OpenCL |
|
|
472 | implementation calls the actual callback, but only later. Therefore, none |
|
|
473 | of the limitations of OpenCL callbacks apply to the perl implementation: |
|
|
474 | it is perfectly safe to make blocking operations from event callbacks, and |
|
|
475 | enqueued operations don't need to be flushed. |
|
|
476 | |
439 | To facilitate this, this module maintains an event queue - each |
477 | To facilitate this, this module maintains an event queue - each |
440 | time an asynchronous event happens, it is queued, and perl will be |
478 | time an asynchronous event happens, it is queued, and perl will be |
441 | interrupted. This is implemented via the L<Async::Interrupt> module. In |
479 | interrupted. This is implemented via the L<Async::Interrupt> module. In |
442 | addition, this module has L<AnyEvent> support, so it can seamlessly |
480 | addition, this module has L<AnyEvent> support, so it can seamlessly |
443 | integrate itself into many event loops. |
481 | integrate itself into many event loops. |
444 | |
482 | |
445 | Since this module is a bit hard to understand, here are some case examples: |
483 | Since L<Async::Interrupt> is a bit hard to understand, here are some case examples: |
446 | |
484 | |
447 | =head3 Don't use callbacks. |
485 | =head3 Don't use callbacks. |
448 | |
486 | |
449 | When your program never uses any callbacks, then there will never be any |
487 | When your program never uses any callbacks, then there will never be any |
450 | notifications you need to take care of, and therefore no need to worry |
488 | notifications you need to take care of, and therefore no need to worry |
… | |
… | |
613 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html> |
651 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetPlatformIDs.html> |
614 | |
652 | |
615 | =item $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr |
653 | =item $ctx = OpenCL::context_from_type $properties, $type = OpenCL::DEVICE_TYPE_DEFAULT, $callback->($err, $pvt) = $print_stderr |
616 | |
654 | |
617 | Tries to create a context from a default device and platform type - never worked for me. |
655 | Tries to create a context from a default device and platform type - never worked for me. |
|
|
656 | Consider using C<< $platform->context_from_type >> instead. |
618 | |
657 | |
619 | type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, OpenCL::DEVICE_TYPE_GPU, |
658 | type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, OpenCL::DEVICE_TYPE_GPU, |
620 | OpenCL::DEVICE_TYPE_ACCELERATOR, OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL. |
659 | OpenCL::DEVICE_TYPE_ACCELERATOR, OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL. |
621 | |
660 | |
622 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html> |
661 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContextFromType.html> |
623 | |
662 | |
624 | =item $ctx = OpenCL::context $properties, \@devices, $callback->($err, $pvt) = $print_stderr) |
663 | =item $ctx = OpenCL::context $properties, \@devices, $callback->($err, $pvt) = $print_stderr) |
625 | |
664 | |
626 | Create a new OpenCL::Context object using the given device object(s). This |
665 | Create a new OpenCL::Context object using the given device object(s). |
627 | function isn't implemented yet, use C<< $platform->context >> instead. |
666 | Consider using C<< $platform->context >> instead. |
628 | |
667 | |
629 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html> |
668 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clCreateContext.html> |
630 | |
669 | |
631 | =item OpenCL::wait_for_events $wait_events... |
670 | =item OpenCL::wait_for_events $wait_events... |
632 | |
671 | |
… | |
… | |
768 | =over 4 |
807 | =over 4 |
769 | |
808 | |
770 | =item $packed_value = $device->info ($name) |
809 | =item $packed_value = $device->info ($name) |
771 | |
810 | |
772 | See C<< $platform->info >> for details. |
811 | See C<< $platform->info >> for details. |
|
|
812 | |
|
|
813 | type: OpenCL::DEVICE_TYPE_DEFAULT, OpenCL::DEVICE_TYPE_CPU, |
|
|
814 | OpenCL::DEVICE_TYPE_GPU, OpenCL::DEVICE_TYPE_ACCELERATOR, |
|
|
815 | OpenCL::DEVICE_TYPE_CUSTOM, OpenCL::DEVICE_TYPE_ALL. |
|
|
816 | |
|
|
817 | fp_config: OpenCL::FP_DENORM, OpenCL::FP_INF_NAN, OpenCL::FP_ROUND_TO_NEAREST, |
|
|
818 | OpenCL::FP_ROUND_TO_ZERO, OpenCL::FP_ROUND_TO_INF, OpenCL::FP_FMA, |
|
|
819 | OpenCL::FP_SOFT_FLOAT, OpenCL::FP_CORRECTLY_ROUNDED_DIVIDE_SQRT. |
|
|
820 | |
|
|
821 | mem_cache_type: OpenCL::NONE, OpenCL::READ_ONLY_CACHE, OpenCL::READ_WRITE_CACHE. |
|
|
822 | |
|
|
823 | local_mem_type: OpenCL::LOCAL, OpenCL::GLOBAL. |
|
|
824 | |
|
|
825 | exec_capabilities: OpenCL::EXEC_KERNEL, OpenCL::EXEC_NATIVE_KERNEL. |
|
|
826 | |
|
|
827 | command_queue_properties: OpenCL::QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE, |
|
|
828 | OpenCL::QUEUE_PROFILING_ENABLE. |
|
|
829 | |
|
|
830 | partition_properties: OpenCL::DEVICE_PARTITION_EQUALLY, |
|
|
831 | OpenCL::DEVICE_PARTITION_BY_COUNTS, OpenCL::DEVICE_PARTITION_BY_COUNTS_LIST_END, |
|
|
832 | OpenCL::DEVICE_PARTITION_BY_AFFINITY_DOMAIN. |
|
|
833 | |
|
|
834 | affinity_domain: OpenCL::DEVICE_AFFINITY_DOMAIN_NUMA, |
|
|
835 | OpenCL::DEVICE_AFFINITY_DOMAIN_L4_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L3_CACHE, |
|
|
836 | OpenCL::DEVICE_AFFINITY_DOMAIN_L2_CACHE, OpenCL::DEVICE_AFFINITY_DOMAIN_L1_CACHE, |
|
|
837 | OpenCL::DEVICE_AFFINITY_DOMAIN_NEXT_PARTITIONABLE. |
773 | |
838 | |
774 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetDeviceInfo.html> |
839 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetDeviceInfo.html> |
775 | |
840 | |
776 | =item @devices = $device->sub_devices (\@properties) |
841 | =item @devices = $device->sub_devices (\@properties) |
777 | |
842 | |
… | |
… | |
1157 | OpenCL::UNORM_INT16, OpenCL::UNORM_SHORT_565, OpenCL::UNORM_SHORT_555, |
1222 | OpenCL::UNORM_INT16, OpenCL::UNORM_SHORT_565, OpenCL::UNORM_SHORT_555, |
1158 | OpenCL::UNORM_INT_101010, OpenCL::SIGNED_INT8, OpenCL::SIGNED_INT16, |
1223 | OpenCL::UNORM_INT_101010, OpenCL::SIGNED_INT8, OpenCL::SIGNED_INT16, |
1159 | OpenCL::SIGNED_INT32, OpenCL::UNSIGNED_INT8, OpenCL::UNSIGNED_INT16, |
1224 | OpenCL::SIGNED_INT32, OpenCL::UNSIGNED_INT8, OpenCL::UNSIGNED_INT16, |
1160 | OpenCL::UNSIGNED_INT32, OpenCL::HALF_FLOAT, OpenCL::FLOAT. |
1225 | OpenCL::UNSIGNED_INT32, OpenCL::HALF_FLOAT, OpenCL::FLOAT. |
1161 | |
1226 | |
1162 | |
|
|
1163 | type: OpenCL::MEM_OBJECT_BUFFER, OpenCL::MEM_OBJECT_IMAGE2D, |
1227 | type: OpenCL::MEM_OBJECT_BUFFER, OpenCL::MEM_OBJECT_IMAGE2D, |
1164 | OpenCL::MEM_OBJECT_IMAGE3D, OpenCL::MEM_OBJECT_IMAGE2D_ARRAY, |
1228 | OpenCL::MEM_OBJECT_IMAGE3D, OpenCL::MEM_OBJECT_IMAGE2D_ARRAY, |
1165 | OpenCL::MEM_OBJECT_IMAGE1D, OpenCL::MEM_OBJECT_IMAGE1D_ARRAY, |
1229 | OpenCL::MEM_OBJECT_IMAGE1D, OpenCL::MEM_OBJECT_IMAGE1D_ARRAY, |
1166 | OpenCL::MEM_OBJECT_IMAGE1D_BUFFER. |
1230 | OpenCL::MEM_OBJECT_IMAGE1D_BUFFER. |
1167 | |
1231 | |
… | |
… | |
1263 | =item $program = $ctx->program_with_built_in_kernels (\@devices, $kernel_names) |
1327 | =item $program = $ctx->program_with_built_in_kernels (\@devices, $kernel_names) |
1264 | |
1328 | |
1265 | Creates a new OpenCL::Program object from the given built-in kernel names. |
1329 | Creates a new OpenCL::Program object from the given built-in kernel names. |
1266 | |
1330 | |
1267 | L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateProgramWithBuiltInKernels.html> |
1331 | L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCreateProgramWithBuiltInKernels.html> |
|
|
1332 | |
|
|
1333 | =item $program = $ctx->link_program (\@devices, $options, \@programs, $cb->($program) = undef) |
|
|
1334 | |
|
|
1335 | Links all (already compiled) program objects specified in C<@programs> |
|
|
1336 | together and returns a new OpenCL::Program object with the result. |
|
|
1337 | |
|
|
1338 | L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clLinkProgram.html> |
1268 | |
1339 | |
1269 | =item $packed_value = $ctx->info ($name) |
1340 | =item $packed_value = $ctx->info ($name) |
1270 | |
1341 | |
1271 | See C<< $platform->info >> for details. |
1342 | See C<< $platform->info >> for details. |
1272 | |
1343 | |
… | |
… | |
1766 | C<-cl-mad-enable>, C<-cl-no-signed-zeros>, C<-cl-unsafe-math-optimizations>, |
1837 | C<-cl-mad-enable>, C<-cl-no-signed-zeros>, C<-cl-unsafe-math-optimizations>, |
1767 | C<-cl-finite-math-only>, C<-cl-fast-relaxed-math>, |
1838 | C<-cl-finite-math-only>, C<-cl-fast-relaxed-math>, |
1768 | C<-w>, C<-Werror>, C<-cl-std=CL1.1/CL1.2>, C<-cl-kernel-arg-info>, |
1839 | C<-w>, C<-Werror>, C<-cl-std=CL1.1/CL1.2>, C<-cl-kernel-arg-info>, |
1769 | C<-create-library>, C<-enable-link-options>. |
1840 | C<-create-library>, C<-enable-link-options>. |
1770 | |
1841 | |
|
|
1842 | build_status: OpenCL::BUILD_SUCCESS, OpenCL::BUILD_NONE, |
|
|
1843 | OpenCL::BUILD_ERROR, OpenCL::BUILD_IN_PROGRESS. |
|
|
1844 | |
1771 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html> |
1845 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clBuildProgram.html> |
1772 | |
1846 | |
1773 | =item $program->build_async (\@devices = undef, $options = "", $cb->($program) = undef) |
1847 | =item $program->build_async (\@devices = undef, $options = "", $cb->($program) = undef) |
1774 | |
1848 | |
1775 | Similar to C<< ->build >>, except it starts a thread, and never fails (you |
1849 | Similar to C<< ->build >>, except it starts a thread, and never fails (you |
1776 | need to check the compilation status form the callback, or by polling). |
1850 | need to check the compilation status form the callback, or by polling). |
1777 | |
1851 | |
1778 | build_status: OpenCL::BUILD_SUCCESS, OpenCL::BUILD_NONE, |
1852 | =item $program->compile (\@devices = undef, $options = "", \%headers = undef, $cb->($program) = undef) |
1779 | OpenCL::BUILD_ERROR, OpenCL::BUILD_IN_PROGRESS. |
1853 | |
|
|
1854 | Compiles the given program for the given devices (or all devices if |
|
|
1855 | undef). If C<$headers> is given, it must be a hashref with include name => |
|
|
1856 | OpenCL::Program pairs. |
|
|
1857 | |
|
|
1858 | L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCompileProgram.html> |
1780 | |
1859 | |
1781 | =item $packed_value = $program->build_info ($device, $name) |
1860 | =item $packed_value = $program->build_info ($device, $name) |
1782 | |
1861 | |
1783 | Similar to C<< $platform->info >>, but returns build info for a previous |
1862 | Similar to C<< $platform->info >>, but returns build info for a previous |
1784 | build attempt for the given device. |
1863 | build attempt for the given device. |
|
|
1864 | |
|
|
1865 | binary_type: OpenCL::PROGRAM_BINARY_TYPE_NONE, |
|
|
1866 | OpenCL::PROGRAM_BINARY_TYPE_COMPILED_OBJECT, |
|
|
1867 | OpenCL::PROGRAM_BINARY_TYPE_LIBRARY, |
|
|
1868 | OpenCL::PROGRAM_BINARY_TYPE_EXECUTABLE. |
1785 | |
1869 | |
1786 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetBuildInfo.html> |
1870 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetBuildInfo.html> |
1787 | |
1871 | |
1788 | =item $kernel = $program->kernel ($function_name) |
1872 | =item $kernel = $program->kernel ($function_name) |
1789 | |
1873 | |
… | |
… | |
1897 | |
1981 | |
1898 | =item $packed_value = $kernel->work_group_info ($device, $name) |
1982 | =item $packed_value = $kernel->work_group_info ($device, $name) |
1899 | |
1983 | |
1900 | See C<< $platform->info >> for details. |
1984 | See C<< $platform->info >> for details. |
1901 | |
1985 | |
1902 | The reason this method is not called C<info> is that there already is an |
|
|
1903 | C<< ->info >> method. |
|
|
1904 | |
|
|
1905 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelWorkGroupInfo.html> |
1986 | L<http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetKernelWorkGroupInfo.html> |
1906 | |
1987 | |
1907 | =for gengetinfo begin kernel_work_group |
1988 | =for gengetinfo begin kernel_work_group |
1908 | |
1989 | |
1909 | =item $int = $kernel->work_group_size ($device) |
1990 | =item $int = $kernel->work_group_size ($device) |
… | |
… | |
1926 | |
2007 | |
1927 | Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_PRIVATE_MEM_SIZE> and returns the result. |
2008 | Calls C<clGetKernelWorkGroupInfo> with C<OpenCL::KERNEL_PRIVATE_MEM_SIZE> and returns the result. |
1928 | |
2009 | |
1929 | =for gengetinfo end kernel_work_group |
2010 | =for gengetinfo end kernel_work_group |
1930 | |
2011 | |
|
|
2012 | =item $packed_value = $kernel->arg_info ($idx, $name) |
|
|
2013 | |
|
|
2014 | See C<< $platform->info >> for details. |
|
|
2015 | |
|
|
2016 | L<http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clGetKernelArgInfo.html> |
|
|
2017 | |
1931 | =for gengetinfo begin kernel_arg_info |
2018 | =for gengetinfo begin kernel_arg |
1932 | |
2019 | |
|
|
2020 | =item $kernel_arg_address_qualifier = $kernel->arg_address_qualifier ($idx) |
|
|
2021 | |
|
|
2022 | Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_ADDRESS_QUALIFIER> and returns the result. |
|
|
2023 | |
|
|
2024 | =item $kernel_arg_access_qualifier = $kernel->arg_access_qualifier ($idx) |
|
|
2025 | |
|
|
2026 | Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_ACCESS_QUALIFIER> and returns the result. |
|
|
2027 | |
|
|
2028 | =item $string = $kernel->arg_type_name ($idx) |
|
|
2029 | |
|
|
2030 | Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_TYPE_NAME> and returns the result. |
|
|
2031 | |
|
|
2032 | =item $kernel_arg_type_qualifier = $kernel->arg_type_qualifier ($idx) |
|
|
2033 | |
|
|
2034 | Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_TYPE_QUALIFIER> and returns the result. |
|
|
2035 | |
|
|
2036 | =item $string = $kernel->arg_name ($idx) |
|
|
2037 | |
|
|
2038 | Calls C<clGetKernelArgInfo> with C<OpenCL::KERNEL_ARG_NAME> and returns the result. |
|
|
2039 | |
1933 | =for gengetinfo end kernel_arg_info |
2040 | =for gengetinfo end kernel_arg |
1934 | |
2041 | |
1935 | =item $kernel->setf ($format, ...) |
2042 | =item $kernel->setf ($format, ...) |
1936 | |
2043 | |
1937 | Sets the arguments of a kernel. Since OpenCL 1.1 doesn't have a generic |
2044 | Sets the arguments of a kernel. Since OpenCL 1.1 doesn't have a generic |
1938 | way to set arguments (and with OpenCL 1.2 it might be rather slow), you |
2045 | way to set arguments (and with OpenCL 1.2 it might be rather slow), you |