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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.50 by root, Tue Apr 24 14:57:06 2012 UTC vs.
Revision 1.59 by root, Sun Apr 29 19:38:05 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h"
7#include "schmorp.h"
4 8
5#ifdef I_DLFCN 9#ifdef I_DLFCN
6 #include <dlfcn.h> 10 #include <dlfcn.h>
7#endif 11#endif
8 12
189 if (res) \ 193 if (res) \
190 FAIL (name); 194 FAIL (name);
191 195
192/*****************************************************************************/ 196/*****************************************************************************/
193 197
194#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
195#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
196
197static void *
198SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
199{
200 if (SvROK (sv) && sv_derived_from (sv, pkg))
201 return (void *)SvIV (SvRV (sv));
202
203 croak ("%s: %s is not of type %s", func, svname, pkg);
204}
205
206/*****************************************************************************/
207
208static cl_context_properties * 198static cl_context_properties *
209SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
210{ 200{
211 if (!sv || !SvOK (sv)) 201 if (!sv || !SvOK (sv))
212 if (extra) 202 if (extra)
270 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
271} 261}
272 262
273/*****************************************************************************/ 263/*****************************************************************************/
274 264
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr))
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{
274 if (SvROK (sv) && sv_derived_from (sv, pkg))
275 return (void *)SvIV (SvRV (sv));
276
277 croak ("%s: %s is not of type %s", func, svname, pkg);
278}
279
280/*****************************************************************************/
281/* callback stuff */
282
283/* default context callback, log to stderr */
284static void CL_CALLBACK
285context_default_notify (const char *msg, const void *info, size_t cb, void *data)
286{
287 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
288}
289
290typedef struct
291{
292 int free_cb;
293 void (*push)(void *data1, void *data2, void *data3);
294} eq_vtbl;
295
296typedef struct eq_item
297{
298 struct eq_item *next;
299 eq_vtbl *vtbl;
300 SV *cb;
301 void *data1, *data2, *data3;
302} eq_item;
303
304static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail;
308
309static void
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{
312 eq_item *item = malloc (sizeof (eq_item));
313
314 item->next = 0;
315 item->vtbl = vtbl;
316 item->cb = cb;
317 item->data1 = data1;
318 item->data2 = data2;
319 item->data3 = data3;
320
321 X_LOCK (eq_lock);
322
323 *(eq_head ? &eq_tail->next : &eq_head) = item;
324 eq_tail = item;
325
326 X_UNLOCK (eq_lock);
327
328 eq_signal_func (eq_signal_arg, 0);
329}
330
331static eq_item *
332eq_dec (void)
333{
334 eq_item *res;
335
336 X_LOCK (eq_lock);
337
338 res = eq_head;
339 if (res)
340 eq_head = res->next;
341
342 X_UNLOCK (eq_lock);
343
344 return res;
345}
346
347static void
348eq_poll (void)
349{
350 eq_item *item;
351
352 while ((item = eq_dec ()))
353 {
354 ENTER;
355 SAVETMPS;
356
357 dSP;
358 PUSHMARK (SP);
359 EXTEND (SP, 2);
360
361 if (item->vtbl->free_cb)
362 sv_2mortal (item->cb);
363
364 PUTBACK;
365 item->vtbl->push (item->data1, item->data2, item->data3);
366
367 SV *cb = item->cb;
368 free (item);
369
370 call_sv (cb, G_DISCARD | G_VOID);
371
372 FREETMPS;
373 LEAVE;
374 }
375}
376
377static void
378eq_poll_interrupt (pTHX_ void *c_arg, int value)
379{
380 eq_poll ();
381}
382
383/*****************************************************************************/
384/* context notify */
385
386static void
387eq_context_push (void *data1, void *data2, void *data3)
388{
389 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
392 PUTBACK;
393
394 free (data1);
395 free (data2);
396}
397
398static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
399
400static void CL_CALLBACK
401eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
402{
403 void *pvt_copy = malloc (cb);
404 memcpy (pvt_copy, pvt, cb);
405 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
406}
407
408#define CONTEXT_NOTIFY_CALLBACK \
409 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
410 void *user_data = 0; \
411 \
412 if (SvOK (notify)) \
413 { \
414 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \
416 }
417
418static SV *
419new_clobj_context (cl_context ctx, void *user_data)
420{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx);
422
423 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425
426 return sv;
427}
428
429#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
430
431/*****************************************************************************/
432/* build/compile/link notify */
433
434static void
435eq_program_push (void *data1, void *data2, void *data3)
436{
437 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1);
439 PUTBACK;
440}
441
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443
444static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data)
446{
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
448}
449
450struct build_args
451{
452 cl_program program;
453 char *options;
454 void *user_data;
455 cl_uint num_devices;
456};
457
458X_THREAD_PROC (build_program_thread)
459{
460 struct build_args *arg = thr_arg;
461
462 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
463
464 if (arg->user_data)
465 eq_program_notify (arg->program, arg->user_data);
466 else
467 clReleaseProgram (arg->program);
468
469 free (arg->options);
470 free (arg);
471}
472
473static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{
476 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
477
478 arg->program = program;
479 arg->options = strdup (options);
480 arg->user_data = user_data;
481 arg->num_devices = num_devices;
482 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
483
484 xthread_t id;
485 thread_create (&id, build_program_thread, arg);
486}
487
488/*****************************************************************************/
489/* event objects */
490
491static void
492eq_event_push (void *data1, void *data2, void *data3)
493{
494 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK;
498}
499
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
501
502static void CL_CALLBACK
503eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
504{
505 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507}
508
509/*****************************************************************************/
510
275static size_t 511static size_t
276img_row_pitch (cl_mem img) 512img_row_pitch (cl_mem img)
277{ 513{
278 size_t res; 514 size_t res;
279 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
293 529
294 do 530 do
295 { 531 {
296 --count; 532 --count;
297 if (SvOK (items [count])) 533 if (SvOK (items [count]))
298 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
299 } 535 }
300 while (count); 536 while (count);
301 537
302 *rcount = i; 538 *rcount = i;
303 539
321} 557}
322 558
323MODULE = OpenCL PACKAGE = OpenCL 559MODULE = OpenCL PACKAGE = OpenCL
324 560
325PROTOTYPES: ENABLE 561PROTOTYPES: ENABLE
562
563void
564poll ()
565 CODE:
566 eq_poll ();
567
568void
569_eq_initialise (IV func, IV arg)
570 CODE:
571 eq_signal_func = (void (*)(void *, int))func;
572 eq_signal_arg = (void*)arg;
326 573
327BOOT: 574BOOT:
328{ 575{
329 HV *stash = gv_stashpv ("OpenCL", 1); 576 HV *stash = gv_stashpv ("OpenCL", 1);
330 static const ivstr *civ, const_iv[] = { 577 static const ivstr *civ, const_iv[] = {
339 { sizeof (cl_half ), "SIZEOF_HALF" }, 586 { sizeof (cl_half ), "SIZEOF_HALF" },
340 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 587 { sizeof (cl_float ), "SIZEOF_FLOAT" },
341 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 588 { sizeof (cl_double), "SIZEOF_DOUBLE" },
342#include "constiv.h" 589#include "constiv.h"
343 }; 590 };
591
344 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
345 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 593 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
594
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
346} 596}
347 597
348cl_int 598cl_int
349errno () 599errno ()
350 CODE: 600 CODE:
351 RETVAL = res; 601 RETVAL = res;
352 OUTPUT: 602 OUTPUT:
353 RETVAL 603 RETVAL
354 604
355const char * 605const char *
356err2str (cl_int err) 606err2str (cl_int err = res)
357 607
358const char * 608const char *
359enum2str (cl_uint value) 609enum2str (cl_uint value)
360 610
361void 611void
369 list = tmpbuf (sizeof (*list) * count); 619 list = tmpbuf (sizeof (*list) * count);
370 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
371 621
372 EXTEND (SP, count); 622 EXTEND (SP, count);
373 for (i = 0; i < count; ++i) 623 for (i = 0; i < count; ++i)
374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 624 PUSH_CLOBJ ("OpenCL::Platform", list [i]);
375 625
376void 626void
377context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
378 PPCODE: 628 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK;
379 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
380 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 631 XPUSH_CLOBJ_CONTEXT;
381 632
382#if 0
383
384void 633void
385context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 634context (FUTURE properties, FUTURE devices, FUTURE notify)
386 PPCODE: 635 PPCODE:
387 /* der Gipfel der Kunst */ 636 /* der Gipfel der Kunst */
388
389#endif
390 637
391void 638void
392wait_for_events (...) 639wait_for_events (...)
393 CODE: 640 CODE:
394 EVENT_LIST (0, items); 641 EVENT_LIST (0, items);
445 EXTEND (SP, count); 692 EXTEND (SP, count);
446 for (i = 0; i < count; ++i) 693 for (i = 0; i < count; ++i)
447 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i]));
448 695
449void 696void
450context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
451 PPCODE: 698 PPCODE:
452 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
453 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
454 701
455 AV *av = (AV *)SvRV (devices); 702 AV *av = (AV *)SvRV (devices);
456 cl_uint num_devices = av_len (av) + 1; 703 cl_uint num_devices = av_len (av) + 1;
457 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
705
458 int i; 706 int i;
459
460 for (i = num_devices; i--; ) 707 for (i = num_devices; i--; )
461 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
462 709
710 CONTEXT_NOTIFY_CALLBACK;
463 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
464 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 712 XPUSH_CLOBJ_CONTEXT;
465 713
466void 714void
467context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
468 PPCODE: 716 PPCODE:
469 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
470 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719
720 CONTEXT_NOTIFY_CALLBACK;
471 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
472 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 722 XPUSH_CLOBJ_CONTEXT;
473 723
474MODULE = OpenCL PACKAGE = OpenCL::Device 724MODULE = OpenCL PACKAGE = OpenCL::Device
475 725
476void 726void
477info (OpenCL::Device self, cl_device_info name) 727info (OpenCL::Device self, cl_device_info name)
650 cl_platform_id value [1]; 900 cl_platform_id value [1];
651 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
652 EXTEND (SP, 1); 902 EXTEND (SP, 1);
653 const int i = 0; 903 const int i = 0;
654 { 904 {
655 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 905 PUSH_CLOBJ ("OpenCL::Platform", value [i]);
656 } 906 }
657 907
658void 908void
659name (OpenCL::Device self) 909name (OpenCL::Device self)
660 ALIAS: 910 ALIAS:
679 cl_device_id value [1]; 929 cl_device_id value [1];
680 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
681 EXTEND (SP, 1); 931 EXTEND (SP, 1);
682 const int i = 0; 932 const int i = 0;
683 { 933 {
684 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 934 PUSH_CLOBJ ("OpenCL::Device", value [i]);
685 } 935 }
686 936
687void 937void
688partition_types_ext (OpenCL::Device self) 938partition_types_ext (OpenCL::Device self)
689 ALIAS: 939 ALIAS:
716 966
717void 967void
718queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
719 PPCODE: 969 PPCODE:
720 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
721 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 971 XPUSH_CLOBJ ("OpenCL::Queue", queue);
722 972
723void 973void
724user_event (OpenCL::Context self) 974user_event (OpenCL::Context self)
725 PPCODE: 975 PPCODE:
726 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
727 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev);
728 978
729void 979void
730buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
731 PPCODE: 981 PPCODE:
732 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
733 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 983 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
734 984
735 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
736 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
737 987
738void 988void
739buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
740 PPCODE: 990 PPCODE:
741 STRLEN len; 991 STRLEN len;
742 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
743 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
744 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 994 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
745 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 995 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
746 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
747 997
748#if CL_VERSION_1_2 998#if CL_VERSION_1_2
749 999
750void 1000void
751image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1001image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
752 PPCODE: 1002 PPCODE:
753 STRLEN len; 1003 STRLEN len;
754 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
755 const cl_image_format format = { channel_order, channel_type }; 1005 const cl_image_format format = { channel_order, channel_type };
756 const cl_image_desc desc = { 1006 const cl_image_desc desc = {
757 type, 1007 type,
758 width, height, depth, 1008 width, height, depth,
759 array_size, row_pitch, slice_pitch, 1009 array_size, row_pitch, slice_pitch,
760 num_mip_level, num_samples, 1010 num_mip_level, num_samples,
761 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvPTROBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
762 }; 1012 };
763 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
764 char *klass = "OpenCL::Image"; 1014 char *klass = "OpenCL::Image";
765 switch (type) 1015 switch (type)
766 { 1016 {
769 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break;
770 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break;
771 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break;
772 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break;
773 } 1023 }
774 XPUSH_NEW_OBJ (klass, mem); 1024 XPUSH_CLOBJ (klass, mem);
775 1025
776#endif 1026#endif
777 1027
778void 1028void
779image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1029image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
785 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1035 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
786#else 1036#else
787 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
788 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1038 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
789#endif 1039#endif
790 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
791 1041
792void 1042void
793image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1043image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
794 PPCODE: 1044 PPCODE:
795 STRLEN len; 1045 STRLEN len;
799 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
800#else 1050#else
801 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
802 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1052 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
803#endif 1053#endif
804 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
805 1055
806#if cl_apple_gl_sharing || cl_khr_gl_sharing 1056#if cl_apple_gl_sharing || cl_khr_gl_sharing
807 1057
808void 1058void
809gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
810 PPCODE: 1060 PPCODE:
811 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
812 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
813 1063
814void 1064void
815gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
816 PPCODE: 1066 PPCODE:
817 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
818 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
819 1069
820#if CL_VERSION_1_2 1070#if CL_VERSION_1_2
821 1071
822void 1072void
823gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
834 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break;
835 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break;
836 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break;
837 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break;
838 } 1088 }
839 XPUSH_NEW_OBJ (klass, mem); 1089 XPUSH_CLOBJ (klass, mem);
840 1090
841#endif 1091#endif
842 1092
843void 1093void
844gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
846#if PREFER_1_1 1096#if PREFER_1_1
847 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
848#else 1098#else
849 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
850#endif 1100#endif
851 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
852 1102
853void 1103void
854gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
855 PPCODE: 1105 PPCODE:
856#if PREFER_1_1 1106#if PREFER_1_1
857 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
858#else 1108#else
859 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
860#endif 1110#endif
861 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
862 1112
863#endif 1113#endif
864 1114
865void 1115void
866supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
886 1136
887void 1137void
888sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
889 PPCODE: 1139 PPCODE:
890 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
891 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler);
892 1142
893void 1143void
894program_with_source (OpenCL::Context self, SV *program) 1144program_with_source (OpenCL::Context self, SV *program)
895 PPCODE: 1145 PPCODE:
896 STRLEN len; 1146 STRLEN len;
897 size_t len2; 1147 size_t len2;
898 const char *ptr = SvPVbyte (program, len); 1148 const char *ptr = SvPVbyte (program, len);
899 1149
900 len2 = len; 1150 len2 = len;
901 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
902 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1152 XPUSH_CLOBJ ("OpenCL::Program", prog);
903 1153
904#BEGIN:context 1154#BEGIN:context
905 1155
906void 1156void
907reference_count (OpenCL::Context self) 1157reference_count (OpenCL::Context self)
924 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
925 int i, n = size / sizeof (*value); 1175 int i, n = size / sizeof (*value);
926 EXTEND (SP, n); 1176 EXTEND (SP, n);
927 for (i = 0; i < n; ++i) 1177 for (i = 0; i < n; ++i)
928 { 1178 {
929 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1179 PUSH_CLOBJ ("OpenCL::Device", value [i]);
930 } 1180 }
931 1181
932void 1182void
933properties (OpenCL::Context self) 1183properties (OpenCL::Context self)
934 PPCODE: 1184 PPCODE:
949DESTROY (OpenCL::Queue self) 1199DESTROY (OpenCL::Queue self)
950 CODE: 1200 CODE:
951 clReleaseCommandQueue (self); 1201 clReleaseCommandQueue (self);
952 1202
953void 1203void
954enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1204read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1205 ALIAS:
1206 enqueue_read_buffer = 0
955 PPCODE: 1207 PPCODE:
956 cl_event ev = 0; 1208 cl_event ev = 0;
957 EVENT_LIST (6, items - 6); 1209 EVENT_LIST (6, items - 6);
958 1210
959 SvUPGRADE (data, SVt_PV); 1211 SvUPGRADE (data, SVt_PV);
961 SvPOK_only (data); 1213 SvPOK_only (data);
962 SvCUR_set (data, len); 1214 SvCUR_set (data, len);
963 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1215 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
964 1216
965 if (ev) 1217 if (ev)
966 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1218 XPUSH_CLOBJ ("OpenCL::Event", ev);
967 1219
968void 1220void
969enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1221write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1222 ALIAS:
1223 enqueue_write_buffer = 0
970 PPCODE: 1224 PPCODE:
971 cl_event ev = 0; 1225 cl_event ev = 0;
972 STRLEN len; 1226 STRLEN len;
973 char *ptr = SvPVbyte (data, len); 1227 char *ptr = SvPVbyte (data, len);
974 EVENT_LIST (5, items - 5); 1228 EVENT_LIST (5, items - 5);
975 1229
976 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1230 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
977 1231
978 if (ev) 1232 if (ev)
979 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1233 XPUSH_CLOBJ ("OpenCL::Event", ev);
980 1234
981#if CL_VERSION_1_2 1235#if CL_VERSION_1_2
982 1236
983void 1237void
984enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1238fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1239 ALIAS:
1240 enqueue_fill_buffer = 0
985 PPCODE: 1241 PPCODE:
986 cl_event ev = 0; 1242 cl_event ev = 0;
987 STRLEN len; 1243 STRLEN len;
988 char *ptr = SvPVbyte (data, len); 1244 char *ptr = SvPVbyte (data, len);
989 EVENT_LIST (5, items - 5); 1245 EVENT_LIST (5, items - 5);
990 1246
991 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1247 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
992 1248
993 if (ev) 1249 if (ev)
994 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1250 XPUSH_CLOBJ ("OpenCL::Event", ev);
995 1251
996void 1252void
997enqueue_fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 1253fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1254 ALIAS:
1255 enqueue_fill_image = 0
998 PPCODE: 1256 PPCODE:
999 cl_event ev = 0; 1257 cl_event ev = 0;
1000 STRLEN len; 1258 STRLEN len;
1001 const size_t origin [3] = { x, y, z }; 1259 const size_t origin [3] = { x, y, z };
1002 const size_t region [3] = { width, height, depth }; 1260 const size_t region [3] = { width, height, depth };
1011 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1269 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1012 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1270 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1013 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1271 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1014 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1272 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1015 1273
1016 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1274 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1017 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1275 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1018 1276
1019 if (ev) 1277 if (ev)
1020 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1278 XPUSH_CLOBJ ("OpenCL::Event", ev);
1021 1279
1022#endif 1280#endif
1023 1281
1024void 1282void
1025enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1283copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1284 ALIAS:
1285 enqueue_copy_buffer = 0
1026 PPCODE: 1286 PPCODE:
1027 cl_event ev = 0; 1287 cl_event ev = 0;
1028 EVENT_LIST (6, items - 6); 1288 EVENT_LIST (6, items - 6);
1029 1289
1030 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1290 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1031 1291
1032 if (ev) 1292 if (ev)
1033 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1293 XPUSH_CLOBJ ("OpenCL::Event", ev);
1034 1294
1035void 1295void
1036enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1296read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1297 ALIAS:
1298 enqueue_read_buffer_rect = 0
1037 PPCODE: 1299 PPCODE:
1038 cl_event ev = 0; 1300 cl_event ev = 0;
1039 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1301 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1040 const size_t host_origin[3] = { host_x, host_y, host_z }; 1302 const size_t host_origin[3] = { host_x, host_y, host_z };
1041 const size_t region[3] = { width, height, depth }; 1303 const size_t region[3] = { width, height, depth };
1060 SvPOK_only (data); 1322 SvPOK_only (data);
1061 SvCUR_set (data, len); 1323 SvCUR_set (data, len);
1062 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1324 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1063 1325
1064 if (ev) 1326 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1327 XPUSH_CLOBJ ("OpenCL::Event", ev);
1066 1328
1067void 1329void
1068enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1330write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1331 ALIAS:
1332 enqueue_write_buffer_rect = 0
1069 PPCODE: 1333 PPCODE:
1070 cl_event ev = 0; 1334 cl_event ev = 0;
1071 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1335 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1072 const size_t host_origin[3] = { host_x, host_y, host_z }; 1336 const size_t host_origin[3] = { host_x, host_y, host_z };
1073 const size_t region[3] = { width, height, depth }; 1337 const size_t region[3] = { width, height, depth };
1093 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1357 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1094 1358
1095 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1359 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1096 1360
1097 if (ev) 1361 if (ev)
1098 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1362 XPUSH_CLOBJ ("OpenCL::Event", ev);
1099 1363
1100void 1364void
1101enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1365copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1366 ALIAS:
1367 enqueue_copy_buffer_rect = 0
1102 PPCODE: 1368 PPCODE:
1103 cl_event ev = 0; 1369 cl_event ev = 0;
1104 const size_t src_origin[3] = { src_x, src_y, src_z }; 1370 const size_t src_origin[3] = { src_x, src_y, src_z };
1105 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1106 const size_t region[3] = { width, height, depth }; 1372 const size_t region[3] = { width, height, depth };
1107 EVENT_LIST (16, items - 16); 1373 EVENT_LIST (16, items - 16);
1108 1374
1109 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1375 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1110 1376
1111 if (ev) 1377 if (ev)
1112 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1378 XPUSH_CLOBJ ("OpenCL::Event", ev);
1113 1379
1114void 1380void
1115enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1381read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1382 ALIAS:
1383 enqueue_read_image = 0
1116 PPCODE: 1384 PPCODE:
1117 cl_event ev = 0; 1385 cl_event ev = 0;
1118 const size_t src_origin[3] = { src_x, src_y, src_z }; 1386 const size_t src_origin[3] = { src_x, src_y, src_z };
1119 const size_t region[3] = { width, height, depth }; 1387 const size_t region[3] = { width, height, depth };
1120 EVENT_LIST (12, items - 12); 1388 EVENT_LIST (12, items - 12);
1132 SvPOK_only (data); 1400 SvPOK_only (data);
1133 SvCUR_set (data, len); 1401 SvCUR_set (data, len);
1134 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1402 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1135 1403
1136 if (ev) 1404 if (ev)
1137 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1405 XPUSH_CLOBJ ("OpenCL::Event", ev);
1138 1406
1139void 1407void
1140enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1408write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1409 ALIAS:
1410 enqueue_write_image = 0
1141 PPCODE: 1411 PPCODE:
1142 cl_event ev = 0; 1412 cl_event ev = 0;
1143 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1413 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1144 const size_t region[3] = { width, height, depth }; 1414 const size_t region[3] = { width, height, depth };
1145 STRLEN len; 1415 STRLEN len;
1158 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1428 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1159 1429
1160 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1430 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1161 1431
1162 if (ev) 1432 if (ev)
1163 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1433 XPUSH_CLOBJ ("OpenCL::Event", ev);
1164 1434
1165void 1435void
1166enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1436copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1437 ALIAS:
1438 enqueue_copy_image = 0
1167 PPCODE: 1439 PPCODE:
1168 cl_event ev = 0; 1440 cl_event ev = 0;
1169 const size_t src_origin[3] = { src_x, src_y, src_z }; 1441 const size_t src_origin[3] = { src_x, src_y, src_z };
1170 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1442 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1171 const size_t region[3] = { width, height, depth }; 1443 const size_t region[3] = { width, height, depth };
1172 EVENT_LIST (12, items - 12); 1444 EVENT_LIST (12, items - 12);
1173 1445
1174 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1446 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1175 1447
1176 if (ev) 1448 if (ev)
1177 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1449 XPUSH_CLOBJ ("OpenCL::Event", ev);
1178 1450
1179void 1451void
1180enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1452copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1453 ALIAS:
1454 enqueue_copy_image_to_buffer = 0
1181 PPCODE: 1455 PPCODE:
1182 cl_event ev = 0; 1456 cl_event ev = 0;
1183 const size_t src_origin[3] = { src_x, src_y, src_z }; 1457 const size_t src_origin[3] = { src_x, src_y, src_z };
1184 const size_t region[3] = { width, height, depth }; 1458 const size_t region[3] = { width, height, depth };
1185 EVENT_LIST (10, items - 10); 1459 EVENT_LIST (10, items - 10);
1186 1460
1187 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1461 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1188 1462
1189 if (ev) 1463 if (ev)
1190 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
1191 1465
1192void 1466void
1193enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1467copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1468 ALIAS:
1469 enqueue_copy_buffer_to_image = 0
1194 PPCODE: 1470 PPCODE:
1195 cl_event ev = 0; 1471 cl_event ev = 0;
1196 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1472 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1197 const size_t region[3] = { width, height, depth }; 1473 const size_t region[3] = { width, height, depth };
1198 EVENT_LIST (10, items - 10); 1474 EVENT_LIST (10, items - 10);
1199 1475
1200 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1476 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1201 1477
1202 if (ev) 1478 if (ev)
1203 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1479 XPUSH_CLOBJ ("OpenCL::Event", ev);
1204 1480
1205void 1481void
1206enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1482task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1483 ALIAS:
1484 enqueue_task = 0
1207 PPCODE: 1485 PPCODE:
1208 cl_event ev = 0; 1486 cl_event ev = 0;
1209 EVENT_LIST (2, items - 2); 1487 EVENT_LIST (2, items - 2);
1210 1488
1211 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1489 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1212 1490
1213 if (ev) 1491 if (ev)
1214 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1492 XPUSH_CLOBJ ("OpenCL::Event", ev);
1215 1493
1216void 1494void
1217enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1495nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1496 ALIAS:
1497 enqueue_nd_range_kernel = 0
1218 PPCODE: 1498 PPCODE:
1219 cl_event ev = 0; 1499 cl_event ev = 0;
1220 size_t *gwo = 0, *gws, *lws = 0; 1500 size_t *gwo = 0, *gws, *lws = 0;
1221 int gws_len; 1501 int gws_len;
1222 size_t *lists; 1502 size_t *lists;
1230 1510
1231 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1511 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1232 1512
1233 gws = lists + gws_len * 0; 1513 gws = lists + gws_len * 0;
1234 for (i = 0; i < gws_len; ++i) 1514 for (i = 0; i < gws_len; ++i)
1515 {
1235 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1516 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1517 // at least nvidia crashes for 0-sized work group sizes, work around
1518 if (!gws [i])
1519 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1520 }
1236 1521
1237 if (SvOK (global_work_offset)) 1522 if (SvOK (global_work_offset))
1238 { 1523 {
1239 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1524 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1240 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1525 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1248 } 1533 }
1249 1534
1250 if (SvOK (local_work_size)) 1535 if (SvOK (local_work_size))
1251 { 1536 {
1252 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1537 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1253 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1538 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1254 1539
1255 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1540 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1256 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1541 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1257 1542
1258 lws = lists + gws_len * 2; 1543 lws = lists + gws_len * 2;
1259 for (i = 0; i < gws_len; ++i) 1544 for (i = 0; i < gws_len; ++i)
1545 {
1260 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1546 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1547 // at least nvidia crashes for 0-sized work group sizes, work around
1548 if (!lws [i])
1549 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1550 }
1261 } 1551 }
1262 1552
1263 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1553 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1264 1554
1265 if (ev) 1555 if (ev)
1266 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1556 XPUSH_CLOBJ ("OpenCL::Event", ev);
1267 1557
1268#if cl_apple_gl_sharing || cl_khr_gl_sharing 1558#if cl_apple_gl_sharing || cl_khr_gl_sharing
1269 1559
1270void 1560void
1271enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1561acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1562 ALIAS:
1563 enqueue_acquire_gl_objects = 0
1272 ALIAS: 1564 ALIAS:
1273 enqueue_release_gl_objects = 1 1565 enqueue_release_gl_objects = 1
1274 PPCODE: 1566 PPCODE:
1275 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1567 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1276 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1568 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1281 cl_uint num_objects = av_len (av) + 1; 1573 cl_uint num_objects = av_len (av) + 1;
1282 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1574 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1283 int i; 1575 int i;
1284 1576
1285 for (i = num_objects; i--; ) 1577 for (i = num_objects; i--; )
1286 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1578 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1287 1579
1288 if (ix) 1580 if (ix)
1289 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1581 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1290 else 1582 else
1291 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1583 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1292 1584
1293 if (ev) 1585 if (ev)
1294 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1586 XPUSH_CLOBJ ("OpenCL::Event", ev);
1295 1587
1296#endif 1588#endif
1297 1589
1298void 1590void
1299enqueue_wait_for_events (OpenCL::Queue self, ...) 1591wait_for_events (OpenCL::Queue self, ...)
1592 ALIAS:
1593 enqueue_wait_for_events = 0
1300 CODE: 1594 CODE:
1301 EVENT_LIST (1, items - 1); 1595 EVENT_LIST (1, items - 1);
1302#if PREFER_1_1 1596#if PREFER_1_1
1303 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1597 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1304#else 1598#else
1305 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 1599 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1306#endif 1600#endif
1307 1601
1308void 1602void
1309enqueue_marker (OpenCL::Queue self, ...) 1603marker (OpenCL::Queue self, ...)
1604 ALIAS:
1605 enqueue_marker = 0
1310 PPCODE: 1606 PPCODE:
1311 cl_event ev = 0; 1607 cl_event ev = 0;
1312 EVENT_LIST (1, items - 1); 1608 EVENT_LIST (1, items - 1);
1313#if PREFER_1_1 1609#if PREFER_1_1
1314 if (!event_list_count) 1610 if (!event_list_count)
1324#endif 1620#endif
1325#else 1621#else
1326 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1622 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1327#endif 1623#endif
1328 if (ev) 1624 if (ev)
1329 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1625 XPUSH_CLOBJ ("OpenCL::Event", ev);
1330 1626
1331void 1627void
1332enqueue_barrier (OpenCL::Queue self, ...) 1628barrier (OpenCL::Queue self, ...)
1629 ALIAS:
1630 enqueue_barrier = 0
1333 PPCODE: 1631 PPCODE:
1334 cl_event ev = 0; 1632 cl_event ev = 0;
1335 EVENT_LIST (1, items - 1); 1633 EVENT_LIST (1, items - 1);
1336#if PREFER_1_1 1634#if PREFER_1_1
1337 if (!event_list_count && GIMME_V == G_VOID) 1635 if (!event_list_count && GIMME_V == G_VOID)
1350#endif 1648#endif
1351#else 1649#else
1352 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1650 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1353#endif 1651#endif
1354 if (ev) 1652 if (ev)
1355 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1653 XPUSH_CLOBJ ("OpenCL::Event", ev);
1356 1654
1357void 1655void
1358flush (OpenCL::Queue self) 1656flush (OpenCL::Queue self)
1359 CODE: 1657 CODE:
1360 NEED_SUCCESS (Flush, (self)); 1658 NEED_SUCCESS (Flush, (self));
1378 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1379 EXTEND (SP, 1); 1677 EXTEND (SP, 1);
1380 const int i = 0; 1678 const int i = 0;
1381 { 1679 {
1382 NEED_SUCCESS (RetainContext, (value [i])); 1680 NEED_SUCCESS (RetainContext, (value [i]));
1383 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1681 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1384 } 1682 }
1385 1683
1386void 1684void
1387device (OpenCL::Queue self) 1685device (OpenCL::Queue self)
1388 PPCODE: 1686 PPCODE:
1389 cl_device_id value [1]; 1687 cl_device_id value [1];
1390 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1688 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1391 EXTEND (SP, 1); 1689 EXTEND (SP, 1);
1392 const int i = 0; 1690 const int i = 0;
1393 { 1691 {
1394 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1692 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1395 } 1693 }
1396 1694
1397void 1695void
1398reference_count (OpenCL::Queue self) 1696reference_count (OpenCL::Queue self)
1399 PPCODE: 1697 PPCODE:
1486 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1784 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1487 EXTEND (SP, 1); 1785 EXTEND (SP, 1);
1488 const int i = 0; 1786 const int i = 0;
1489 { 1787 {
1490 NEED_SUCCESS (RetainContext, (value [i])); 1788 NEED_SUCCESS (RetainContext, (value [i]));
1491 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1789 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1492 } 1790 }
1493 1791
1494void 1792void
1495associated_memobject (OpenCL::Memory self) 1793associated_memobject (OpenCL::Memory self)
1496 PPCODE: 1794 PPCODE:
1498 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1796 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1499 EXTEND (SP, 1); 1797 EXTEND (SP, 1);
1500 const int i = 0; 1798 const int i = 0;
1501 { 1799 {
1502 NEED_SUCCESS (RetainMemObject, (value [i])); 1800 NEED_SUCCESS (RetainMemObject, (value [i]));
1503 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1801 PUSH_CLOBJ ("OpenCL::Memory", value [i]);
1504 } 1802 }
1505 1803
1506#END:mem 1804#END:mem
1507 1805
1508#if cl_apple_gl_sharing || cl_khr_gl_sharing 1806#if cl_apple_gl_sharing || cl_khr_gl_sharing
1528 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 1826 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1529 1827
1530 cl_buffer_region crdata = { origin, size }; 1828 cl_buffer_region crdata = { origin, size };
1531 1829
1532 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 1830 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1533 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1831 XPUSH_CLOBJ ("OpenCL::Buffer", mem);
1534 1832
1535MODULE = OpenCL PACKAGE = OpenCL::Image 1833MODULE = OpenCL PACKAGE = OpenCL::Image
1536 1834
1537void 1835void
1538image_info (OpenCL::Image self, cl_image_info name) 1836image_info (OpenCL::Image self, cl_image_info name)
1624 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 1922 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1625 EXTEND (SP, 1); 1923 EXTEND (SP, 1);
1626 const int i = 0; 1924 const int i = 0;
1627 { 1925 {
1628 NEED_SUCCESS (RetainContext, (value [i])); 1926 NEED_SUCCESS (RetainContext, (value [i]));
1629 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1927 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1630 } 1928 }
1631 1929
1632void 1930void
1633normalized_coords (OpenCL::Sampler self) 1931normalized_coords (OpenCL::Sampler self)
1634 PPCODE: 1932 PPCODE:
1664DESTROY (OpenCL::Program self) 1962DESTROY (OpenCL::Program self)
1665 CODE: 1963 CODE:
1666 clReleaseProgram (self); 1964 clReleaseProgram (self);
1667 1965
1668void 1966void
1669build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 1967build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1968 ALIAS:
1969 build_async = 1
1670 CODE: 1970 CODE:
1971 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1972 void *user_data = 0;
1973 cl_uint num_devices = 0;
1974 cl_device_id *device_list = 0;
1975
1976 if (SvOK (devices))
1977 {
1978 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1979 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
1980
1981 AV *av = (AV *)SvRV (devices);
1982 num_devices = av_len (av) + 1;
1983
1984 if (num_devices)
1985 {
1986 device_list = tmpbuf (sizeof (*device_list) * num_devices);
1987 int count;
1988 for (count = 0; count < num_devices; ++count)
1989 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1990 }
1991 }
1992
1993 if (SvOK (notify))
1994 {
1995 NEED_SUCCESS (RetainProgram, (self));
1996 pfn_notify = eq_program_notify;
1997 user_data = SvREFCNT_inc (s_get_cv (notify));
1998 }
1999
2000 if (ix)
2001 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2002 else
1671 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2003 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1672 2004
1673void 2005void
1674build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2006build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1675 PPCODE: 2007 PPCODE:
1676 size_t size; 2008 size_t size;
1711 2043
1712void 2044void
1713kernel (OpenCL::Program program, SV *function) 2045kernel (OpenCL::Program program, SV *function)
1714 PPCODE: 2046 PPCODE:
1715 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2047 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1716 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2048 XPUSH_CLOBJ ("OpenCL::Kernel", kernel);
1717 2049
1718void 2050void
1719kernels_in_program (OpenCL::Program program) 2051kernels_in_program (OpenCL::Program program)
1720 PPCODE: 2052 PPCODE:
1721 cl_uint num_kernels; 2053 cl_uint num_kernels;
1724 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2056 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
1725 2057
1726 int i; 2058 int i;
1727 EXTEND (SP, num_kernels); 2059 EXTEND (SP, num_kernels);
1728 for (i = 0; i < num_kernels; ++i) 2060 for (i = 0; i < num_kernels; ++i)
1729 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Kernel", kernels [i])); 2061 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]);
1730 2062
1731void 2063void
1732info (OpenCL::Program self, cl_program_info name) 2064info (OpenCL::Program self, cl_program_info name)
1733 PPCODE: 2065 PPCODE:
1734 INFO (Program) 2066 INFO (Program)
1782 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2114 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1783 EXTEND (SP, 1); 2115 EXTEND (SP, 1);
1784 const int i = 0; 2116 const int i = 0;
1785 { 2117 {
1786 NEED_SUCCESS (RetainContext, (value [i])); 2118 NEED_SUCCESS (RetainContext, (value [i]));
1787 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2119 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1788 } 2120 }
1789 2121
1790void 2122void
1791devices (OpenCL::Program self) 2123devices (OpenCL::Program self)
1792 PPCODE: 2124 PPCODE:
1796 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2128 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1797 int i, n = size / sizeof (*value); 2129 int i, n = size / sizeof (*value);
1798 EXTEND (SP, n); 2130 EXTEND (SP, n);
1799 for (i = 0; i < n; ++i) 2131 for (i = 0; i < n; ++i)
1800 { 2132 {
1801 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2133 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1802 } 2134 }
1803 2135
1804void 2136void
1805source (OpenCL::Program self) 2137source (OpenCL::Program self)
1806 PPCODE: 2138 PPCODE:
1832DESTROY (OpenCL::Kernel self) 2164DESTROY (OpenCL::Kernel self)
1833 CODE: 2165 CODE:
1834 clReleaseKernel (self); 2166 clReleaseKernel (self);
1835 2167
1836void 2168void
2169setf (OpenCL::Kernel self, const char *format, ...)
2170 CODE:
2171 int i;
2172 for (i = 2; ; ++i)
2173 {
2174 while (*format == ' ')
2175 ++format;
2176
2177 char type = *format++;
2178
2179 if (!type)
2180 break;
2181
2182 if (i >= items)
2183 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2184
2185 SV *sv = ST (i);
2186
2187 union
2188 {
2189 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2190 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2191 cl_half ch; cl_float cf; cl_double cd;
2192 cl_mem cm;
2193 cl_sampler ca;
2194 size_t cz;
2195 cl_event ce;
2196 } arg;
2197 size_t size;
2198 int nullarg = 0;
2199
2200 switch (type)
2201 {
2202 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2203 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2204 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2205 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2206 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2207 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2208 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2209 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2210
2211 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2212 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2213 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2214
2215 case 'z': nullarg = 1; size = SvIV (sv); break;
2216
2217 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2218 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2219 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2220
2221 default:
2222 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2223 }
2224
2225 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2226 if (res)
2227 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2228 }
2229
2230 if (i != items)
2231 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2232
2233void
1837set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2234set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1838 CODE: 2235 CODE:
1839 clSetKernelArg (self, idx, sizeof (value), &value); 2236 clSetKernelArg (self, idx, sizeof (value), &value);
1840 2237
1841void 2238void
1889 clSetKernelArg (self, idx, sizeof (value), &value); 2286 clSetKernelArg (self, idx, sizeof (value), &value);
1890 2287
1891void 2288void
1892set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2289set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1893 CODE: 2290 CODE:
1894 clSetKernelArg (self, idx, sizeof (value), &value); 2291 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1895 2292
1896void 2293void
1897set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2294set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1898 CODE: 2295 CODE:
1899 clSetKernelArg (self, idx, sizeof (value), &value); 2296 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1900 2297
1901void 2298void
1902set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2299set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2300 ALIAS:
2301 set_image2d = 0
2302 set_image3d = 0
1903 CODE: 2303 CODE:
1904 clSetKernelArg (self, idx, sizeof (value), &value); 2304 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1905
1906void
1907set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1908 CODE:
1909 clSetKernelArg (self, idx, sizeof (value), &value);
1910 2305
1911void 2306void
1912set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2307set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1913 CODE: 2308 CODE:
1914 clSetKernelArg (self, idx, sizeof (value), &value); 2309 clSetKernelArg (self, idx, sizeof (value), &value);
1960 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2355 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1961 EXTEND (SP, 1); 2356 EXTEND (SP, 1);
1962 const int i = 0; 2357 const int i = 0;
1963 { 2358 {
1964 NEED_SUCCESS (RetainContext, (value [i])); 2359 NEED_SUCCESS (RetainContext, (value [i]));
1965 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2360 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1966 } 2361 }
1967 2362
1968void 2363void
1969program (OpenCL::Kernel self) 2364program (OpenCL::Kernel self)
1970 PPCODE: 2365 PPCODE:
1972 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2367 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1973 EXTEND (SP, 1); 2368 EXTEND (SP, 1);
1974 const int i = 0; 2369 const int i = 0;
1975 { 2370 {
1976 NEED_SUCCESS (RetainProgram, (value [i])); 2371 NEED_SUCCESS (RetainProgram, (value [i]));
1977 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2372 PUSH_CLOBJ ("OpenCL::Program", value [i]);
1978 } 2373 }
1979 2374
1980#END:kernel 2375#END:kernel
1981 2376
1982void 2377void
2042wait (OpenCL::Event self) 2437wait (OpenCL::Event self)
2043 CODE: 2438 CODE:
2044 clWaitForEvents (1, &self); 2439 clWaitForEvents (1, &self);
2045 2440
2046void 2441void
2442cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2443 CODE:
2444 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2445
2446void
2047info (OpenCL::Event self, cl_event_info name) 2447info (OpenCL::Event self, cl_event_info name)
2048 PPCODE: 2448 PPCODE:
2049 INFO (Event) 2449 INFO (Event)
2050 2450
2051#BEGIN:event 2451#BEGIN:event
2057 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2457 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2058 EXTEND (SP, 1); 2458 EXTEND (SP, 1);
2059 const int i = 0; 2459 const int i = 0;
2060 { 2460 {
2061 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2461 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2062 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2462 PUSH_CLOBJ ("OpenCL::Queue", value [i]);
2063 } 2463 }
2064 2464
2065void 2465void
2066command_type (OpenCL::Event self) 2466command_type (OpenCL::Event self)
2067 PPCODE: 2467 PPCODE:
2090 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2490 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2091 EXTEND (SP, 1); 2491 EXTEND (SP, 1);
2092 const int i = 0; 2492 const int i = 0;
2093 { 2493 {
2094 NEED_SUCCESS (RetainContext, (value [i])); 2494 NEED_SUCCESS (RetainContext, (value [i]));
2095 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2495 PUSH_CLOBJ ("OpenCL::Context", value [i]);
2096 } 2496 }
2097 2497
2098#END:event 2498#END:event
2099 2499
2100void 2500void

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines