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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.41 by root, Sat Apr 21 19:53:29 2012 UTC vs.
Revision 1.63 by root, Tue May 1 22:25:13 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
11// this. This breaks 100% of the opencl 1.1 apps, for what reason? 15// this. This breaks 100% of the opencl 1.1 apps, for what reason?
12// after all, the functions are deprecated, not removed. 16// after all, the functions are deprecated, not removed.
13// in addition, you cannot test for this in any future-proof way. 17// in addition, you cannot test for this in any future-proof way.
14// each time a new opencl version comes out, you need to make a new 18// each time a new opencl version comes out, you need to make a new
15// release. 19// release.
16#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
17#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */ 20#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
21
22#ifndef PREFER_1_1
23 #define PREFER_1_1 1
24#endif
25
26#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif
18 29
19#ifdef __APPLE__ 30#ifdef __APPLE__
20 #include <OpenCL/opencl.h> 31 #include <OpenCL/opencl.h>
21#else 32#else
22 #include <CL/opencl.h> 33 #include <CL/opencl.h>
23#endif 34#endif
24 35
25#if 0 // testing
26#undef CL_USE_DEPRECATED_OPENCL_1_1_APIS
27#undef CL_VERSION_1_2 36#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1
38 #define PREFER_1_1 1
28#endif 39#endif
29 40
30typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
31typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
32typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
46typedef cl_program OpenCL__Program; 57typedef cl_program OpenCL__Program;
47typedef cl_kernel OpenCL__Kernel; 58typedef cl_kernel OpenCL__Kernel;
48typedef cl_event OpenCL__Event; 59typedef cl_event OpenCL__Event;
49typedef cl_event OpenCL__UserEvent; 60typedef cl_event OpenCL__UserEvent;
50 61
62typedef struct mapped * OpenCL__Mapped;
63
51typedef SV *FUTURE; 64typedef SV *FUTURE;
65
66static HV
67 *stash_platform,
68 *stash_device,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
52 89
53/*****************************************************************************/ 90/*****************************************************************************/
54 91
55// name must include a leading underscore 92// name must include a leading underscore
56// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
182 if (res) \ 219 if (res) \
183 FAIL (name); 220 FAIL (name);
184 221
185/*****************************************************************************/ 222/*****************************************************************************/
186 223
187#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
188#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
189
190static void *
191SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
192{
193 if (SvROK (sv) && sv_derived_from (sv, pkg))
194 return (void *)SvIV (SvRV (sv));
195
196 croak ("%s: %s is not of type %s", func, svname, pkg);
197}
198
199/*****************************************************************************/
200
201static cl_context_properties * 224static cl_context_properties *
202SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
203{ 226{
204 if (!sv || !SvOK (sv)) 227 if (!sv || !SvOK (sv))
205 if (extra) 228 if (extra)
263 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
264} 287}
265 288
266/*****************************************************************************/ 289/*****************************************************************************/
267 290
291static SV *
292new_clobj (HV *stash, IV id)
293{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
295}
296
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
299
300/* cl objects are either \$iv, or [$iv, ...] */
301/* they can be upgraded at runtime to the array form */
302static void *
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
304{
305 if (SvROK (sv) && sv_derived_from (sv, pkg))
306 return (void *)SvIV (SvRV (sv));
307
308 croak ("%s: %s is not of type %s", func, svname, pkg);
309}
310
311/*****************************************************************************/
312/* callback stuff */
313
314/* default context callback, log to stderr */
315static void CL_CALLBACK
316context_default_notify (const char *msg, const void *info, size_t cb, void *data)
317{
318 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
319}
320
321typedef struct
322{
323 int free_cb;
324 void (*push)(void *data1, void *data2, void *data3);
325} eq_vtbl;
326
327typedef struct eq_item
328{
329 struct eq_item *next;
330 eq_vtbl *vtbl;
331 SV *cb;
332 void *data1, *data2, *data3;
333} eq_item;
334
335static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail;
339
340static void
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{
343 eq_item *item = malloc (sizeof (eq_item));
344
345 item->next = 0;
346 item->vtbl = vtbl;
347 item->cb = cb;
348 item->data1 = data1;
349 item->data2 = data2;
350 item->data3 = data3;
351
352 X_LOCK (eq_lock);
353
354 *(eq_head ? &eq_tail->next : &eq_head) = item;
355 eq_tail = item;
356
357 X_UNLOCK (eq_lock);
358
359 eq_signal_func (eq_signal_arg, 0);
360}
361
362static eq_item *
363eq_dec (void)
364{
365 eq_item *res;
366
367 X_LOCK (eq_lock);
368
369 res = eq_head;
370 if (res)
371 eq_head = res->next;
372
373 X_UNLOCK (eq_lock);
374
375 return res;
376}
377
378static void
379eq_poll (void)
380{
381 eq_item *item;
382
383 while ((item = eq_dec ()))
384 {
385 ENTER;
386 SAVETMPS;
387
388 dSP;
389 PUSHMARK (SP);
390 EXTEND (SP, 2);
391
392 if (item->vtbl->free_cb)
393 sv_2mortal (item->cb);
394
395 PUTBACK;
396 item->vtbl->push (item->data1, item->data2, item->data3);
397
398 SV *cb = item->cb;
399 free (item);
400
401 call_sv (cb, G_DISCARD | G_VOID);
402
403 FREETMPS;
404 LEAVE;
405 }
406}
407
408static void
409eq_poll_interrupt (pTHX_ void *c_arg, int value)
410{
411 eq_poll ();
412}
413
414/*****************************************************************************/
415/* context notify */
416
417static void
418eq_context_push (void *data1, void *data2, void *data3)
419{
420 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
423 PUTBACK;
424
425 free (data1);
426 free (data2);
427}
428
429static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
430
431static void CL_CALLBACK
432eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
433{
434 void *pvt_copy = malloc (cb);
435 memcpy (pvt_copy, pvt, cb);
436 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
437}
438
439#define CONTEXT_NOTIFY_CALLBACK \
440 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
441 void *user_data = 0; \
442 \
443 if (SvOK (notify)) \
444 { \
445 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \
447 }
448
449static SV *
450new_clobj_context (cl_context ctx, void *user_data)
451{
452 SV *sv = new_clobj (stash_context, (IV)ctx);
453
454 if (user_data)
455 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
456
457 return sv;
458}
459
460#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
461
462/*****************************************************************************/
463/* build/compile/link notify */
464
465static void
466eq_program_push (void *data1, void *data2, void *data3)
467{
468 dSP;
469 PUSH_CLOBJ (stash_program, data1);
470 PUTBACK;
471}
472
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474
475static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data)
477{
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
479}
480
481struct build_args
482{
483 cl_program program;
484 char *options;
485 void *user_data;
486 cl_uint num_devices;
487};
488
489X_THREAD_PROC (build_program_thread)
490{
491 struct build_args *arg = thr_arg;
492
493 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
494
495 if (arg->user_data)
496 eq_program_notify (arg->program, arg->user_data);
497 else
498 clReleaseProgram (arg->program);
499
500 free (arg->options);
501 free (arg);
502}
503
504static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{
507 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
508
509 arg->program = program;
510 arg->options = strdup (options);
511 arg->user_data = user_data;
512 arg->num_devices = num_devices;
513 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
514
515 xthread_t id;
516 thread_create (&id, build_program_thread, arg);
517}
518
519/*****************************************************************************/
520/* event objects */
521
522static void
523eq_event_push (void *data1, void *data2, void *data3)
524{
525 dSP;
526 PUSH_CLOBJ (stash_event, data1);
527 PUSHs (sv_2mortal (newSViv ((IV)data2)));
528 PUTBACK;
529}
530
531static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
532
533static void CL_CALLBACK
534eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
535{
536 clRetainEvent (event);
537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
538}
539
540/*****************************************************************************/
541/* utilities for XS code */
542
268static size_t 543static size_t
269img_row_pitch (cl_mem img) 544img_row_pitch (cl_mem img)
270{ 545{
271 size_t res; 546 size_t res;
272 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
273 return res; 548 return res;
274} 549}
275 550
276static cl_event * 551static cl_event *
277event_list (SV **items, cl_uint *rcount) 552event_list (SV **items, cl_uint *rcount, cl_event extra)
278{ 553{
279 cl_uint count = *rcount; 554 cl_uint count = *rcount;
280 555
281 if (!count) 556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed
557 count = 0;
558
559 if (!count && !extra)
282 return 0; 560 return 0;
283 561
284 cl_event *list = tmpbuf (sizeof (cl_event) * count); 562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
285 int i = 0; 563 int i = 0;
286 564
287 do 565 while (count--)
288 {
289 --count;
290 if (SvOK (items [count])) 566 if (SvOK (items [count]))
291 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
292 } 568
293 while (count); 569 if (extra)
570 list [i++] = extra;
294 571
295 *rcount = i; 572 *rcount = i;
296 573
297 return i ? list : 0; 574 return i ? list : 0;
298} 575}
299 576
300#define EVENT_LIST(items,count) \ 577#define EVENT_LIST(skip) \
301 cl_uint event_list_count = (count); \ 578 cl_uint event_list_count = items - (skip); \
302 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
303 580
304#define INFO(class) \ 581#define INFO(class) \
305{ \ 582{ \
306 size_t size; \ 583 size_t size; \
307 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
311 SvCUR_set (sv, size); \ 588 SvCUR_set (sv, size); \
312 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 589 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
313 XPUSHs (sv); \ 590 XPUSHs (sv); \
314} 591}
315 592
593/*****************************************************************************/
594/* mapped_xxx */
595
596static OpenCL__Mapped
597SvMAPPED (SV *sv)
598{
599 // no typechecking atm., keep your fingers crossed
600 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
601}
602
603struct mapped
604{
605 cl_command_queue queue;
606 cl_mem memobj;
607 void *ptr;
608 size_t cb;
609 cl_event event;
610 size_t row_pitch;
611 size_t slice_pitch;
612};
613
614static SV *
615mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
616{
617 SV *data = newSV (0);
618 SvUPGRADE (data, SVt_PVMG);
619
620 OpenCL__Mapped mapped;
621 New (0, mapped, 1, struct mapped);
622
623 clRetainCommandQueue (queue);
624
625 mapped->queue = queue;
626 mapped->memobj = memobj;
627 mapped->ptr = ptr;
628 mapped->cb = cb;
629 mapped->event = ev;
630 mapped->row_pitch = row_pitch;
631 mapped->slice_pitch = slice_pitch;
632
633 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
634
635 if (SvLEN (data))
636 Safefree (data);
637
638 SvPVX (data) = (char *)ptr;
639 SvCUR_set (data, cb);
640 SvLEN_set (data, 0);
641 SvPOK_only (data);
642
643 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
644
645 if (!(flags & CL_MAP_WRITE))
646 SvREADONLY_on (data);
647
648 return obj;
649}
650
651static void
652mapped_detach (SV *sv, OpenCL__Mapped mapped)
653{
654 SV *data = SvRV (sv);
655
656 // the next check checks both before AND after detach, where SvPVX should STILL be 0
657 if (SvPVX (data) != (char *)mapped->ptr)
658 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
659 else
660 {
661 SvREADONLY_off (data);
662 SvCUR_set (data, 0);
663 SvPVX (data) = 0;
664 SvOK_off (data);
665 }
666
667 mapped->ptr = 0;
668}
669
670static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
674 cl_event ev;
675
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677
678 clReleaseEvent (mapped->event);
679 mapped->event = ev;
680
681 mapped_detach (self, mapped);
682}
683
684/*****************************************************************************/
685
316MODULE = OpenCL PACKAGE = OpenCL 686MODULE = OpenCL PACKAGE = OpenCL
317 687
318PROTOTYPES: ENABLE 688PROTOTYPES: ENABLE
689
690void
691poll ()
692 CODE:
693 eq_poll ();
694
695void
696_eq_initialise (IV func, IV arg)
697 CODE:
698 eq_signal_func = (void (*)(void *, int))func;
699 eq_signal_arg = (void*)arg;
319 700
320BOOT: 701BOOT:
321{ 702{
322 HV *stash = gv_stashpv ("OpenCL", 1); 703 HV *stash = gv_stashpv ("OpenCL", 1);
323 static const ivstr *civ, const_iv[] = { 704 static const ivstr *civ, const_iv[] = {
332 { sizeof (cl_half ), "SIZEOF_HALF" }, 713 { sizeof (cl_half ), "SIZEOF_HALF" },
333 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 714 { sizeof (cl_float ), "SIZEOF_FLOAT" },
334 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 715 { sizeof (cl_double), "SIZEOF_DOUBLE" },
335#include "constiv.h" 716#include "constiv.h"
336 }; 717 };
718
337 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
338 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
729 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
730 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
731 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
732 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
733 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
734 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
735 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
736 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
737 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
738 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
739 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
740 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
741 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
742 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
743 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
744
745 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
339} 746}
340 747
341cl_int 748cl_int
342errno () 749errno ()
343 CODE: 750 CODE:
344 RETVAL = res; 751 RETVAL = res;
345 OUTPUT: 752 OUTPUT:
346 RETVAL 753 RETVAL
347 754
348const char * 755const char *
349err2str (cl_int err) 756err2str (cl_int err = res)
350 757
351const char * 758const char *
352enum2str (cl_uint value) 759enum2str (cl_uint value)
353 760
354void 761void
362 list = tmpbuf (sizeof (*list) * count); 769 list = tmpbuf (sizeof (*list) * count);
363 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
364 771
365 EXTEND (SP, count); 772 EXTEND (SP, count);
366 for (i = 0; i < count; ++i) 773 for (i = 0; i < count; ++i)
367 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 774 PUSH_CLOBJ (stash_platform, (IV)list [i]);
368 775
369void 776void
370context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
371 PPCODE: 778 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK;
372 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 780 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
373 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 781 XPUSH_CLOBJ_CONTEXT;
374 782
375#if 0
376
377void 783void
378context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 784context (FUTURE properties, FUTURE devices, FUTURE notify)
379 PPCODE: 785 PPCODE:
380 /* der Gipfel der Kunst */ 786 /* der Gipfel der Kunst */
381 787
788void
789wait_for_events (...)
790 CODE:
791 EVENT_LIST (0);
792 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
793
794PROTOTYPES: DISABLE
795
796MODULE = OpenCL PACKAGE = OpenCL::Platform
797
798void
799info (OpenCL::Platform self, cl_platform_info name)
800 PPCODE:
801 INFO (Platform)
802
803void
804unload_compiler (OpenCL::Platform self)
805 CODE:
806#if CL_VERSION_1_2
807 clUnloadPlatformCompiler (self);
382#endif 808#endif
383
384void
385wait_for_events (...)
386 CODE:
387 EVENT_LIST (0, items);
388 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
389
390PROTOTYPES: DISABLE
391
392MODULE = OpenCL PACKAGE = OpenCL::Platform
393
394void
395info (OpenCL::Platform self, cl_platform_info name)
396 PPCODE:
397 INFO (Platform)
398 809
399#BEGIN:platform 810#BEGIN:platform
400 811
401void 812void
402profile (OpenCL::Platform self) 813profile (OpenCL::Platform self)
428 list = tmpbuf (sizeof (*list) * count); 839 list = tmpbuf (sizeof (*list) * count);
429 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 840 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
430 841
431 EXTEND (SP, count); 842 EXTEND (SP, count);
432 for (i = 0; i < count; ++i) 843 for (i = 0; i < count; ++i)
433 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 844 PUSH_CLOBJ (stash_device, list [i]);
434 845
435void 846void
436context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
437 PPCODE: 848 PPCODE:
438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
439 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 850 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
440 851
441 AV *av = (AV *)SvRV (devices); 852 AV *av = (AV *)SvRV (devices);
442 cl_uint num_devices = av_len (av) + 1; 853 cl_uint num_devices = av_len (av) + 1;
443 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
855
444 int i; 856 int i;
445
446 for (i = num_devices; i--; ) 857 for (i = num_devices; i--; )
447 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
448 859
860 CONTEXT_NOTIFY_CALLBACK;
449 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 861 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
450 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 862 XPUSH_CLOBJ_CONTEXT;
451 863
452void 864void
453context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 865context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
454 PPCODE: 866 PPCODE:
455 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
456 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 868 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
869
870 CONTEXT_NOTIFY_CALLBACK;
457 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 871 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
458 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 872 XPUSH_CLOBJ_CONTEXT;
459 873
460MODULE = OpenCL PACKAGE = OpenCL::Device 874MODULE = OpenCL PACKAGE = OpenCL::Device
461 875
462void 876void
463info (OpenCL::Device self, cl_device_info name) 877info (OpenCL::Device self, cl_device_info name)
471 PPCODE: 885 PPCODE:
472 cl_device_type value [1]; 886 cl_device_type value [1];
473 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 887 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
474 EXTEND (SP, 1); 888 EXTEND (SP, 1);
475 const int i = 0; 889 const int i = 0;
476 PUSHs (sv_2mortal (newSViv (value [i]))); 890 PUSHs (sv_2mortal (newSVuv (value [i])));
477 891
478void 892void
479vendor_id (OpenCL::Device self) 893vendor_id (OpenCL::Device self)
480 ALIAS: 894 ALIAS:
481 vendor_id = CL_DEVICE_VENDOR_ID 895 vendor_id = CL_DEVICE_VENDOR_ID
626 PPCODE: 1040 PPCODE:
627 cl_command_queue_properties value [1]; 1041 cl_command_queue_properties value [1];
628 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1042 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
629 EXTEND (SP, 1); 1043 EXTEND (SP, 1);
630 const int i = 0; 1044 const int i = 0;
631 PUSHs (sv_2mortal (newSViv (value [i]))); 1045 PUSHs (sv_2mortal (newSVuv (value [i])));
632 1046
633void 1047void
634platform (OpenCL::Device self) 1048platform (OpenCL::Device self)
635 PPCODE: 1049 PPCODE:
636 cl_platform_id value [1]; 1050 cl_platform_id value [1];
637 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1051 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
638 EXTEND (SP, 1); 1052 EXTEND (SP, 1);
639 const int i = 0; 1053 const int i = 0;
640 { 1054 {
641 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1055 PUSH_CLOBJ (stash_platform, value [i]);
642 } 1056 }
643 1057
644void 1058void
645name (OpenCL::Device self) 1059name (OpenCL::Device self)
646 ALIAS: 1060 ALIAS:
665 cl_device_id value [1]; 1079 cl_device_id value [1];
666 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1080 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
667 EXTEND (SP, 1); 1081 EXTEND (SP, 1);
668 const int i = 0; 1082 const int i = 0;
669 { 1083 {
670 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1084 PUSH_CLOBJ (stash_device, value [i]);
671 } 1085 }
672 1086
673void 1087void
674partition_types_ext (OpenCL::Device self) 1088partition_types_ext (OpenCL::Device self)
675 ALIAS: 1089 ALIAS:
702 1116
703void 1117void
704queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1118queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
705 PPCODE: 1119 PPCODE:
706 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1120 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
707 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1121 XPUSH_CLOBJ (stash_queue, queue);
708 1122
709void 1123void
710user_event (OpenCL::Context self) 1124user_event (OpenCL::Context self)
711 PPCODE: 1125 PPCODE:
712 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1126 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
713 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1127 XPUSH_CLOBJ (stash_userevent, ev);
714 1128
715void 1129void
716buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1130buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
717 PPCODE: 1131 PPCODE:
718 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1132 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
719 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1133 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
720 1134
721 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1135 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
722 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1136 XPUSH_CLOBJ (stash_bufferobj, mem);
723 1137
724void 1138void
725buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1139buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
726 PPCODE: 1140 PPCODE:
727 STRLEN len; 1141 STRLEN len;
728 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1142 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
729 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1143 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
730 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1144 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
731 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1145 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
732 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1146 XPUSH_CLOBJ (stash_bufferobj, mem);
733 1147
734#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS 1148#if CL_VERSION_1_2
735 1149
736void 1150void
737image2d (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) 1151image (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)
738 PPCODE: 1152 PPCODE:
739 STRLEN len; 1153 STRLEN len;
740 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1154 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
741 const cl_image_format format = { channel_order, channel_type }; 1155 const cl_image_format format = { channel_order, channel_type };
1156 const cl_image_desc desc = {
1157 type,
1158 width, height, depth,
1159 array_size, row_pitch, slice_pitch,
1160 num_mip_level, num_samples,
1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1162 };
742 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
743 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1164 HV *stash = stash_image;
1165 switch (type)
1166 {
1167 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1168 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1169 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1170 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1171 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1172 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1173 }
1174 XPUSH_CLOBJ (stash, mem);
744 1175
1176#endif
1177
745void 1178void
746image3d (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) 1179image2d (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)
747 PPCODE: 1180 PPCODE:
748 STRLEN len; 1181 STRLEN len;
749 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1182 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
750 const cl_image_format format = { channel_order, channel_type }; 1183 const cl_image_format format = { channel_order, channel_type };
1184#if PREFER_1_1
1185 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1186#else
1187 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1188 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1189#endif
1190 XPUSH_CLOBJ (stash_image2d, mem);
1191
1192void
1193image3d (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)
1194 PPCODE:
1195 STRLEN len;
1196 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1197 const cl_image_format format = { channel_order, channel_type };
1198#if PREFER_1_1
751 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1199 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
752 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1200#else
753 1201 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1202 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
754#endif 1203#endif
1204 XPUSH_CLOBJ (stash_image3d, mem);
755 1205
756#if cl_apple_gl_sharing || cl_khr_gl_sharing 1206#if cl_apple_gl_sharing || cl_khr_gl_sharing
757 1207
758void 1208void
759gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1209gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
760 PPCODE: 1210 PPCODE:
761 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1211 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
762 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1212 XPUSH_CLOBJ (stash_bufferobj, mem);
763 1213
764void 1214void
765gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1215gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
766 PPCODE: 1216 PPCODE:
767 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1217 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
768 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1218 XPUSH_CLOBJ (stash_image2d, mem);
769 1219
770#if CL_VERSION_1_2 1220#if CL_VERSION_1_2
771 1221
772void 1222void
773gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1223gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1224 ALIAS:
774 PPCODE: 1225 PPCODE:
775 char *klass = "OpenCL::Memory"; 1226 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
776 cl_gl_object_type t; 1227 cl_gl_object_type type;
1228 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1229 HV *stash = stash_memory;
1230 switch (type)
1231 {
1232 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1233 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1234 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1235 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1236 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1237 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1238 }
1239 XPUSH_CLOBJ (stash, mem);
1240
1241#endif
1242
1243void
1244gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1245 PPCODE:
1246#if PREFER_1_1
777 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1247 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
778 NEED_SUCCESS (GetGLObjectInfo, (mem, &t, 0)); 1248#else
779 switch (t) 1249 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
780 {
781 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break;
782 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break;
783 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break;
784 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break;
785 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break;
786 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break;
787 }
788 XPUSH_NEW_OBJ (klass, mem);
789
790#endif 1250#endif
791 1251 XPUSH_CLOBJ (stash_image2d, mem);
792#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS
793
794void
795gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
796 PPCODE:
797 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
798 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
799 1252
800void 1253void
801gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1254gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
802 PPCODE: 1255 PPCODE:
1256#if PREFER_1_1
803 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1257 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
804 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1258#else
805 1259 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
806#endif 1260#endif
1261 XPUSH_CLOBJ (stash_image3d, mem);
807 1262
808#endif 1263#endif
809 1264
810void 1265void
811supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1266supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
831 1286
832void 1287void
833sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1288sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
834 PPCODE: 1289 PPCODE:
835 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1290 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
836 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1291 XPUSH_CLOBJ (stash_sampler, sampler);
837 1292
838void 1293void
839program_with_source (OpenCL::Context self, SV *program) 1294program_with_source (OpenCL::Context self, SV *program)
840 PPCODE: 1295 PPCODE:
841 STRLEN len; 1296 STRLEN len;
842 size_t len2; 1297 size_t len2;
843 const char *ptr = SvPVbyte (program, len); 1298 const char *ptr = SvPVbyte (program, len);
844 1299
845 len2 = len; 1300 len2 = len;
846 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
847 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1302 XPUSH_CLOBJ (stash_program, prog);
848 1303
849#BEGIN:context 1304#BEGIN:context
850 1305
851void 1306void
852reference_count (OpenCL::Context self) 1307reference_count (OpenCL::Context self)
869 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1324 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
870 int i, n = size / sizeof (*value); 1325 int i, n = size / sizeof (*value);
871 EXTEND (SP, n); 1326 EXTEND (SP, n);
872 for (i = 0; i < n; ++i) 1327 for (i = 0; i < n; ++i)
873 { 1328 {
874 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1329 PUSH_CLOBJ (stash_device, value [i]);
875 } 1330 }
876 1331
877void 1332void
878properties (OpenCL::Context self) 1333properties (OpenCL::Context self)
879 PPCODE: 1334 PPCODE:
894DESTROY (OpenCL::Queue self) 1349DESTROY (OpenCL::Queue self)
895 CODE: 1350 CODE:
896 clReleaseCommandQueue (self); 1351 clReleaseCommandQueue (self);
897 1352
898void 1353void
899enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS:
1356 enqueue_read_buffer = 0
900 PPCODE: 1357 PPCODE:
901 cl_event ev = 0; 1358 cl_event ev = 0;
902 EVENT_LIST (6, items - 6); 1359 EVENT_LIST (6);
903 1360
904 SvUPGRADE (data, SVt_PV); 1361 SvUPGRADE (data, SVt_PV);
905 SvGROW (data, len); 1362 SvGROW (data, len);
906 SvPOK_only (data); 1363 SvPOK_only (data);
907 SvCUR_set (data, len); 1364 SvCUR_set (data, len);
908 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1365 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
909 1366
910 if (ev) 1367 if (ev)
911 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1368 XPUSH_CLOBJ (stash_event, ev);
912 1369
913void 1370void
914enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS:
1373 enqueue_write_buffer = 0
915 PPCODE: 1374 PPCODE:
916 cl_event ev = 0; 1375 cl_event ev = 0;
917 STRLEN len; 1376 STRLEN len;
918 char *ptr = SvPVbyte (data, len); 1377 char *ptr = SvPVbyte (data, len);
919 EVENT_LIST (5, items - 5); 1378 EVENT_LIST (5);
920 1379
921 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1380 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
922 1381
923 if (ev) 1382 if (ev)
924 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1383 XPUSH_CLOBJ (stash_event, ev);
925 1384
1385#if CL_VERSION_1_2
1386
926void 1387void
927enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS:
1390 enqueue_fill_buffer = 0
928 PPCODE: 1391 PPCODE:
929 cl_event ev = 0; 1392 cl_event ev = 0;
1393 STRLEN len;
1394 char *ptr = SvPVbyte (data, len);
1395 EVENT_LIST (5);
1396
1397 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1398
1399 if (ev)
1400 XPUSH_CLOBJ (stash_event, ev);
1401
1402void
1403fill_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, ...)
1404 ALIAS:
1405 enqueue_fill_image = 0
1406 PPCODE:
1407 cl_event ev = 0;
1408 STRLEN len;
1409 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412
1413 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1417 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1418 cl_image_format format;
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1421 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1422 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1425 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426
1427 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev);
1429
1430#endif
1431
1432void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS:
1435 enqueue_copy_buffer = 0
1436 PPCODE:
1437 cl_event ev = 0;
930 EVENT_LIST (6, items - 6); 1438 EVENT_LIST (6);
931 1439
932 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1440 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
933 1441
934 if (ev) 1442 if (ev)
935 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1443 XPUSH_CLOBJ (stash_event, ev);
936 1444
937void 1445void
938enqueue_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, ...) 1446read_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, ...)
1447 ALIAS:
1448 enqueue_read_buffer_rect = 0
939 PPCODE: 1449 PPCODE:
940 cl_event ev = 0; 1450 cl_event ev = 0;
941 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
942 const size_t host_origin[3] = { host_x, host_y, host_z }; 1452 const size_t host_origin[3] = { host_x, host_y, host_z };
943 const size_t region[3] = { width, height, depth }; 1453 const size_t region[3] = { width, height, depth };
944 EVENT_LIST (17, items - 17); 1454 EVENT_LIST (17);
945 1455
946 if (!buf_row_pitch) 1456 if (!buf_row_pitch)
947 buf_row_pitch = region [0]; 1457 buf_row_pitch = region [0];
948 1458
949 if (!buf_slice_pitch) 1459 if (!buf_slice_pitch)
962 SvPOK_only (data); 1472 SvPOK_only (data);
963 SvCUR_set (data, len); 1473 SvCUR_set (data, len);
964 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)); 1474 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));
965 1475
966 if (ev) 1476 if (ev)
967 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1477 XPUSH_CLOBJ (stash_event, ev);
968 1478
969void 1479void
970enqueue_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, ...) 1480write_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, ...)
1481 ALIAS:
1482 enqueue_write_buffer_rect = 0
971 PPCODE: 1483 PPCODE:
972 cl_event ev = 0; 1484 cl_event ev = 0;
973 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
974 const size_t host_origin[3] = { host_x, host_y, host_z }; 1486 const size_t host_origin[3] = { host_x, host_y, host_z };
975 const size_t region[3] = { width, height, depth }; 1487 const size_t region[3] = { width, height, depth };
976 STRLEN len; 1488 STRLEN len;
977 char *ptr = SvPVbyte (data, len); 1489 char *ptr = SvPVbyte (data, len);
978 EVENT_LIST (17, items - 17); 1490 EVENT_LIST (17);
979 1491
980 if (!buf_row_pitch) 1492 if (!buf_row_pitch)
981 buf_row_pitch = region [0]; 1493 buf_row_pitch = region [0];
982 1494
983 if (!buf_slice_pitch) 1495 if (!buf_slice_pitch)
995 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
996 1508
997 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)); 1509 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));
998 1510
999 if (ev) 1511 if (ev)
1000 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1512 XPUSH_CLOBJ (stash_event, ev);
1001 1513
1002void 1514void
1003enqueue_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, ...) 1515copy_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, ...)
1516 ALIAS:
1517 enqueue_copy_buffer_rect = 0
1004 PPCODE: 1518 PPCODE:
1005 cl_event ev = 0; 1519 cl_event ev = 0;
1006 const size_t src_origin[3] = { src_x, src_y, src_z }; 1520 const size_t src_origin[3] = { src_x, src_y, src_z };
1007 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1521 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1008 const size_t region[3] = { width, height, depth }; 1522 const size_t region[3] = { width, height, depth };
1009 EVENT_LIST (16, items - 16); 1523 EVENT_LIST (16);
1010 1524
1011 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)); 1525 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));
1012 1526
1013 if (ev) 1527 if (ev)
1014 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1528 XPUSH_CLOBJ (stash_event, ev);
1015 1529
1016void 1530void
1017enqueue_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, ...) 1531read_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, ...)
1532 ALIAS:
1533 enqueue_read_image = 0
1018 PPCODE: 1534 PPCODE:
1019 cl_event ev = 0; 1535 cl_event ev = 0;
1020 const size_t src_origin[3] = { src_x, src_y, src_z }; 1536 const size_t src_origin[3] = { src_x, src_y, src_z };
1021 const size_t region[3] = { width, height, depth }; 1537 const size_t region[3] = { width, height, depth };
1022 EVENT_LIST (12, items - 12); 1538 EVENT_LIST (12);
1023 1539
1024 if (!row_pitch) 1540 if (!row_pitch)
1025 row_pitch = img_row_pitch (src); 1541 row_pitch = img_row_pitch (src);
1026 1542
1027 if (depth > 1 && !slice_pitch) 1543 if (depth > 1 && !slice_pitch)
1034 SvPOK_only (data); 1550 SvPOK_only (data);
1035 SvCUR_set (data, len); 1551 SvCUR_set (data, len);
1036 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)); 1552 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));
1037 1553
1038 if (ev) 1554 if (ev)
1039 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1555 XPUSH_CLOBJ (stash_event, ev);
1040 1556
1041void 1557void
1042enqueue_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, ...) 1558write_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, ...)
1559 ALIAS:
1560 enqueue_write_image = 0
1043 PPCODE: 1561 PPCODE:
1044 cl_event ev = 0; 1562 cl_event ev = 0;
1045 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1046 const size_t region[3] = { width, height, depth }; 1564 const size_t region[3] = { width, height, depth };
1047 STRLEN len; 1565 STRLEN len;
1048 char *ptr = SvPVbyte (data, len); 1566 char *ptr = SvPVbyte (data, len);
1049 EVENT_LIST (12, items - 12); 1567 EVENT_LIST (12);
1050 1568
1051 if (!row_pitch) 1569 if (!row_pitch)
1052 row_pitch = img_row_pitch (dst); 1570 row_pitch = img_row_pitch (dst);
1053 1571
1054 if (depth > 1 && !slice_pitch) 1572 if (depth > 1 && !slice_pitch)
1060 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1061 1579
1062 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)); 1580 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));
1063 1581
1064 if (ev) 1582 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1583 XPUSH_CLOBJ (stash_event, ev);
1066 1584
1067void 1585void
1068enqueue_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, ...) 1586copy_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, ...)
1587 ALIAS:
1588 enqueue_copy_image = 0
1069 PPCODE: 1589 PPCODE:
1070 cl_event ev = 0; 1590 cl_event ev = 0;
1071 const size_t src_origin[3] = { src_x, src_y, src_z }; 1591 const size_t src_origin[3] = { src_x, src_y, src_z };
1072 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1592 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1073 const size_t region[3] = { width, height, depth }; 1593 const size_t region[3] = { width, height, depth };
1074 EVENT_LIST (12, items - 12); 1594 EVENT_LIST (12);
1075 1595
1076 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1596 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1077 1597
1078 if (ev) 1598 if (ev)
1079 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1599 XPUSH_CLOBJ (stash_event, ev);
1080 1600
1081void 1601void
1082enqueue_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, ...) 1602copy_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, ...)
1603 ALIAS:
1604 enqueue_copy_image_to_buffer = 0
1083 PPCODE: 1605 PPCODE:
1084 cl_event ev = 0; 1606 cl_event ev = 0;
1085 const size_t src_origin[3] = { src_x, src_y, src_z }; 1607 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610
1611 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612
1613 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev);
1615
1616void
1617copy_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, ...)
1618 ALIAS:
1619 enqueue_copy_buffer_to_image = 0
1620 PPCODE:
1621 cl_event ev = 0;
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625
1626 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627
1628 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev);
1630
1631void
1632map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, size_t cb = 0, ...)
1633 ALIAS:
1634 enqueue_map_buffer = 0
1635 PPCODE:
1636 cl_event ev;
1637 EVENT_LIST (6);
1638
1639 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1640 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1641
1642void
1643map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, size_t width = 0, size_t height = 0, size_t depth = 0, ...)
1644 ALIAS:
1645 enqueue_map_image = 0
1646 PPCODE:
1647 cl_event ev;
1648 const size_t origin[3] = { x, y, z };
1086 const size_t region[3] = { width, height, depth }; 1649 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch;
1087 EVENT_LIST (10, items - 10); 1651 EVENT_LIST (10);
1088 1652
1653 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1654
1655 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1]
1657 : region [0];
1658
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1660
1661void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID)
1666 {
1667 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event);
1669 }
1670
1671void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS:
1674 enqueue_task = 0
1675 PPCODE:
1676 cl_event ev = 0;
1677 EVENT_LIST (2);
1678
1089 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1679 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1090 1680
1091 if (ev) 1681 if (ev)
1092 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1682 XPUSH_CLOBJ (stash_event, ev);
1093 1683
1094void 1684void
1095enqueue_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, ...)
1096 PPCODE:
1097 cl_event ev = 0;
1098 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1099 const size_t region[3] = { width, height, depth };
1100 EVENT_LIST (10, items - 10);
1101
1102 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1103
1104 if (ev)
1105 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1106
1107void
1108enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1109 PPCODE:
1110 cl_event ev = 0;
1111 EVENT_LIST (2, items - 2);
1112
1113 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1114
1115 if (ev)
1116 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1117
1118void
1119enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1686 ALIAS:
1687 enqueue_nd_range_kernel = 0
1120 PPCODE: 1688 PPCODE:
1121 cl_event ev = 0; 1689 cl_event ev = 0;
1122 size_t *gwo = 0, *gws, *lws = 0; 1690 size_t *gwo = 0, *gws, *lws = 0;
1123 int gws_len; 1691 int gws_len;
1124 size_t *lists; 1692 size_t *lists;
1125 int i; 1693 int i;
1126 EVENT_LIST (5, items - 5); 1694 EVENT_LIST (5);
1127 1695
1128 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1129 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1130 1698
1131 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1132 1700
1133 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1701 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1134 1702
1135 gws = lists + gws_len * 0; 1703 gws = lists + gws_len * 0;
1136 for (i = 0; i < gws_len; ++i) 1704 for (i = 0; i < gws_len; ++i)
1705 {
1137 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1706 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1707 // at least nvidia crashes for 0-sized work group sizes, work around
1708 if (!gws [i])
1709 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1710 }
1138 1711
1139 if (SvOK (global_work_offset)) 1712 if (SvOK (global_work_offset))
1140 { 1713 {
1141 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1714 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1142 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1715 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1150 } 1723 }
1151 1724
1152 if (SvOK (local_work_size)) 1725 if (SvOK (local_work_size))
1153 { 1726 {
1154 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1727 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1155 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1728 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1156 1729
1157 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1730 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1158 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1731 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1159 1732
1160 lws = lists + gws_len * 2; 1733 lws = lists + gws_len * 2;
1161 for (i = 0; i < gws_len; ++i) 1734 for (i = 0; i < gws_len; ++i)
1735 {
1162 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1736 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1737 // at least nvidia crashes for 0-sized work group sizes, work around
1738 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 }
1163 } 1741 }
1164 1742
1165 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1743 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1166 1744
1167 if (ev) 1745 if (ev)
1168 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1746 XPUSH_CLOBJ (stash_event, ev);
1169 1747
1170#if cl_apple_gl_sharing || cl_khr_gl_sharing 1748#if cl_apple_gl_sharing || cl_khr_gl_sharing
1171 1749
1172void 1750void
1173enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1174 ALIAS: 1752 ALIAS:
1753 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0
1175 enqueue_release_gl_objects = 1 1755 enqueue_release_gl_objects = 1
1176 PPCODE: 1756 PPCODE:
1177 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1178 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1179 1759
1180 cl_event ev = 0; 1760 cl_event ev = 0;
1181 EVENT_LIST (2, items - 2); 1761 EVENT_LIST (2);
1182 AV *av = (AV *)SvRV (objects); 1762 AV *av = (AV *)SvRV (objects);
1183 cl_uint num_objects = av_len (av) + 1; 1763 cl_uint num_objects = av_len (av) + 1;
1184 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1764 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1185 int i; 1765 int i;
1186 1766
1187 for (i = num_objects; i--; ) 1767 for (i = num_objects; i--; )
1188 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1768 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1189 1769
1190 if (ix) 1770 if (ix)
1191 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1192 else 1772 else
1193 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1194 1774
1195 if (ev) 1775 if (ev)
1196 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1776 XPUSH_CLOBJ (stash_event, ev);
1197 1777
1198#endif 1778#endif
1199 1779
1200#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS
1201
1202void 1780void
1203enqueue_wait_for_events (OpenCL::Queue self, ...) 1781wait_for_events (OpenCL::Queue self, ...)
1782 ALIAS:
1783 enqueue_wait_for_events = 0
1204 CODE: 1784 CODE:
1205 EVENT_LIST (1, items - 1); 1785 EVENT_LIST (1);
1786#if PREFER_1_1
1206 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1787 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1207 1788#else
1789 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1208#endif 1790#endif
1209 1791
1210void 1792void
1211enqueue_marker (OpenCL::Queue self, ...) 1793marker (OpenCL::Queue self, ...)
1794 ALIAS:
1795 enqueue_marker = 0
1212 PPCODE: 1796 PPCODE:
1213 cl_event ev = 0; 1797 cl_event ev = 0;
1214 EVENT_LIST (1, items - 1); 1798 EVENT_LIST (1);
1799#if PREFER_1_1
1800 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else
1215#if CL_VERSION_1_2 1803#if CL_VERSION_1_2
1804 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1805#else
1806 {
1807 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1808 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1809 }
1810#endif
1811#else
1216 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1812 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1217#else
1218 if (event_list_count)
1219 croak ("OpenCL::Queue->enqueue_marker does not support a wait list in OpenCL 1.1 - upgrade to 1.2");
1220 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1221#endif 1813#endif
1222 if (ev) 1814 if (ev)
1223 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1815 XPUSH_CLOBJ (stash_event, ev);
1224 1816
1225void 1817void
1226enqueue_barrier (OpenCL::Queue self, ...) 1818barrier (OpenCL::Queue self, ...)
1819 ALIAS:
1820 enqueue_barrier = 0
1227 PPCODE: 1821 PPCODE:
1228 cl_event ev = 0; 1822 cl_event ev = 0;
1229 EVENT_LIST (1, items - 1); 1823 EVENT_LIST (1);
1824#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID)
1826 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else
1230#if CL_VERSION_1_2 1828#if CL_VERSION_1_2
1231 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, &ev)); 1829 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1232#else 1830#else
1831 {
1233 if (event_list_count) 1832 if (event_list_count)
1234 croak ("OpenCL::Queue->enqueue_barrier does not support a wait list in OpenCL 1.1 - upgrade to 1.2"); 1833 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1834
1235 if (GIMME_V != G_VOID) 1835 if (GIMME_V != G_VOID)
1236 croak ("OpenCL::Queue->enqueue_barrier does not return an event object in OpenCL 1.1 - upgrade to 1.2");
1237 NEED_SUCCESS (EnqueueBarrier, (self)); 1836 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1837 }
1838#endif
1839#else
1840 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1238#endif 1841#endif
1239 if (ev) 1842 if (ev)
1240 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1843 XPUSH_CLOBJ (stash_event, ev);
1241 1844
1242void 1845void
1243flush (OpenCL::Queue self) 1846flush (OpenCL::Queue self)
1244 CODE: 1847 CODE:
1245 NEED_SUCCESS (Flush, (self)); 1848 NEED_SUCCESS (Flush, (self));
1263 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1866 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1264 EXTEND (SP, 1); 1867 EXTEND (SP, 1);
1265 const int i = 0; 1868 const int i = 0;
1266 { 1869 {
1267 NEED_SUCCESS (RetainContext, (value [i])); 1870 NEED_SUCCESS (RetainContext, (value [i]));
1268 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1871 PUSH_CLOBJ (stash_context, value [i]);
1269 } 1872 }
1270 1873
1271void 1874void
1272device (OpenCL::Queue self) 1875device (OpenCL::Queue self)
1273 PPCODE: 1876 PPCODE:
1274 cl_device_id value [1]; 1877 cl_device_id value [1];
1275 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1878 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1276 EXTEND (SP, 1); 1879 EXTEND (SP, 1);
1277 const int i = 0; 1880 const int i = 0;
1278 { 1881 {
1279 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1882 PUSH_CLOBJ (stash_device, value [i]);
1280 } 1883 }
1281 1884
1282void 1885void
1283reference_count (OpenCL::Queue self) 1886reference_count (OpenCL::Queue self)
1284 PPCODE: 1887 PPCODE:
1293 PPCODE: 1896 PPCODE:
1294 cl_command_queue_properties value [1]; 1897 cl_command_queue_properties value [1];
1295 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1898 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1296 EXTEND (SP, 1); 1899 EXTEND (SP, 1);
1297 const int i = 0; 1900 const int i = 0;
1298 PUSHs (sv_2mortal (newSViv (value [i]))); 1901 PUSHs (sv_2mortal (newSVuv (value [i])));
1299 1902
1300#END:command_queue 1903#END:command_queue
1301 1904
1302MODULE = OpenCL PACKAGE = OpenCL::Memory 1905MODULE = OpenCL PACKAGE = OpenCL::Memory
1303 1906
1318 PPCODE: 1921 PPCODE:
1319 cl_mem_object_type value [1]; 1922 cl_mem_object_type value [1];
1320 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 1923 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1321 EXTEND (SP, 1); 1924 EXTEND (SP, 1);
1322 const int i = 0; 1925 const int i = 0;
1323 PUSHs (sv_2mortal (newSViv (value [i]))); 1926 PUSHs (sv_2mortal (newSVuv (value [i])));
1324 1927
1325void 1928void
1326flags (OpenCL::Memory self) 1929flags (OpenCL::Memory self)
1327 PPCODE: 1930 PPCODE:
1328 cl_mem_flags value [1]; 1931 cl_mem_flags value [1];
1329 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 1932 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1330 EXTEND (SP, 1); 1933 EXTEND (SP, 1);
1331 const int i = 0; 1934 const int i = 0;
1332 PUSHs (sv_2mortal (newSViv (value [i]))); 1935 PUSHs (sv_2mortal (newSVuv (value [i])));
1333 1936
1334void 1937void
1335size (OpenCL::Memory self) 1938size (OpenCL::Memory self)
1336 ALIAS: 1939 ALIAS:
1337 size = CL_MEM_SIZE 1940 size = CL_MEM_SIZE
1371 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1974 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1372 EXTEND (SP, 1); 1975 EXTEND (SP, 1);
1373 const int i = 0; 1976 const int i = 0;
1374 { 1977 {
1375 NEED_SUCCESS (RetainContext, (value [i])); 1978 NEED_SUCCESS (RetainContext, (value [i]));
1376 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1979 PUSH_CLOBJ (stash_context, value [i]);
1377 } 1980 }
1378 1981
1379void 1982void
1380associated_memobject (OpenCL::Memory self) 1983associated_memobject (OpenCL::Memory self)
1381 PPCODE: 1984 PPCODE:
1383 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1986 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1384 EXTEND (SP, 1); 1987 EXTEND (SP, 1);
1385 const int i = 0; 1988 const int i = 0;
1386 { 1989 {
1387 NEED_SUCCESS (RetainMemObject, (value [i])); 1990 NEED_SUCCESS (RetainMemObject, (value [i]));
1388 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1991 PUSH_CLOBJ (stash_memory, value [i]);
1389 } 1992 }
1390 1993
1391#END:mem 1994#END:mem
1392 1995
1393#if cl_apple_gl_sharing || cl_khr_gl_sharing 1996#if cl_apple_gl_sharing || cl_khr_gl_sharing
1413 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2016 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1414 2017
1415 cl_buffer_region crdata = { origin, size }; 2018 cl_buffer_region crdata = { origin, size };
1416 2019
1417 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2020 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1418 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2021 XPUSH_CLOBJ (stash_buffer, mem);
1419 2022
1420MODULE = OpenCL PACKAGE = OpenCL::Image 2023MODULE = OpenCL PACKAGE = OpenCL::Image
1421 2024
1422void 2025void
1423image_info (OpenCL::Image self, cl_image_info name) 2026image_info (OpenCL::Image self, cl_image_info name)
1424 PPCODE: 2027 PPCODE:
1425 INFO (Image) 2028 INFO (Image)
2029
2030void
2031format (OpenCL::Image self)
2032 PPCODE:
2033 cl_image_format format;
2034 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2035 EXTEND (SP, 2);
2036 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2037 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1426 2038
1427#BEGIN:image 2039#BEGIN:image
1428 2040
1429void 2041void
1430element_size (OpenCL::Image self) 2042element_size (OpenCL::Image self)
1500 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2112 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1501 EXTEND (SP, 1); 2113 EXTEND (SP, 1);
1502 const int i = 0; 2114 const int i = 0;
1503 { 2115 {
1504 NEED_SUCCESS (RetainContext, (value [i])); 2116 NEED_SUCCESS (RetainContext, (value [i]));
1505 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2117 PUSH_CLOBJ (stash_context, value [i]);
1506 } 2118 }
1507 2119
1508void 2120void
1509normalized_coords (OpenCL::Sampler self) 2121normalized_coords (OpenCL::Sampler self)
1510 PPCODE: 2122 PPCODE:
1511 cl_addressing_mode value [1]; 2123 cl_addressing_mode value [1];
1512 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2124 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1513 EXTEND (SP, 1); 2125 EXTEND (SP, 1);
1514 const int i = 0; 2126 const int i = 0;
1515 PUSHs (sv_2mortal (newSViv (value [i]))); 2127 PUSHs (sv_2mortal (newSVuv (value [i])));
1516 2128
1517void 2129void
1518addressing_mode (OpenCL::Sampler self) 2130addressing_mode (OpenCL::Sampler self)
1519 PPCODE: 2131 PPCODE:
1520 cl_filter_mode value [1]; 2132 cl_filter_mode value [1];
1521 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2133 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1522 EXTEND (SP, 1); 2134 EXTEND (SP, 1);
1523 const int i = 0; 2135 const int i = 0;
1524 PUSHs (sv_2mortal (newSViv (value [i]))); 2136 PUSHs (sv_2mortal (newSVuv (value [i])));
1525 2137
1526void 2138void
1527filter_mode (OpenCL::Sampler self) 2139filter_mode (OpenCL::Sampler self)
1528 PPCODE: 2140 PPCODE:
1529 cl_bool value [1]; 2141 cl_bool value [1];
1540DESTROY (OpenCL::Program self) 2152DESTROY (OpenCL::Program self)
1541 CODE: 2153 CODE:
1542 clReleaseProgram (self); 2154 clReleaseProgram (self);
1543 2155
1544void 2156void
1545build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS:
2159 build_async = 1
1546 CODE: 2160 CODE:
2161 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2162 void *user_data = 0;
2163 cl_uint num_devices = 0;
2164 cl_device_id *device_list = 0;
2165
2166 if (SvOK (devices))
2167 {
2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2170
2171 AV *av = (AV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173
2174 if (num_devices)
2175 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2177 int count;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 }
2181 }
2182
2183 if (SvOK (notify))
2184 {
2185 NEED_SUCCESS (RetainProgram, (self));
2186 pfn_notify = eq_program_notify;
2187 user_data = SvREFCNT_inc (s_get_cv (notify));
2188 }
2189
2190 if (ix)
2191 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2192 else
1547 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2193 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1548 2194
1549void 2195void
1550build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1551 PPCODE: 2197 PPCODE:
1552 size_t size; 2198 size_t size;
1587 2233
1588void 2234void
1589kernel (OpenCL::Program program, SV *function) 2235kernel (OpenCL::Program program, SV *function)
1590 PPCODE: 2236 PPCODE:
1591 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2237 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1592 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2238 XPUSH_CLOBJ (stash_kernel, kernel);
2239
2240void
2241kernels_in_program (OpenCL::Program program)
2242 PPCODE:
2243 cl_uint num_kernels;
2244 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2245 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2246 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2247
2248 int i;
2249 EXTEND (SP, num_kernels);
2250 for (i = 0; i < num_kernels; ++i)
2251 PUSH_CLOBJ (stash_kernel, kernels [i]);
1593 2252
1594void 2253void
1595info (OpenCL::Program self, cl_program_info name) 2254info (OpenCL::Program self, cl_program_info name)
1596 PPCODE: 2255 PPCODE:
1597 INFO (Program) 2256 INFO (Program)
1645 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2304 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1646 EXTEND (SP, 1); 2305 EXTEND (SP, 1);
1647 const int i = 0; 2306 const int i = 0;
1648 { 2307 {
1649 NEED_SUCCESS (RetainContext, (value [i])); 2308 NEED_SUCCESS (RetainContext, (value [i]));
1650 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2309 PUSH_CLOBJ (stash_context, value [i]);
1651 } 2310 }
1652 2311
1653void 2312void
1654devices (OpenCL::Program self) 2313devices (OpenCL::Program self)
1655 PPCODE: 2314 PPCODE:
1659 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2318 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1660 int i, n = size / sizeof (*value); 2319 int i, n = size / sizeof (*value);
1661 EXTEND (SP, n); 2320 EXTEND (SP, n);
1662 for (i = 0; i < n; ++i) 2321 for (i = 0; i < n; ++i)
1663 { 2322 {
1664 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2323 PUSH_CLOBJ (stash_device, value [i]);
1665 } 2324 }
1666 2325
1667void 2326void
1668source (OpenCL::Program self) 2327source (OpenCL::Program self)
1669 PPCODE: 2328 PPCODE:
1695DESTROY (OpenCL::Kernel self) 2354DESTROY (OpenCL::Kernel self)
1696 CODE: 2355 CODE:
1697 clReleaseKernel (self); 2356 clReleaseKernel (self);
1698 2357
1699void 2358void
2359setf (OpenCL::Kernel self, const char *format, ...)
2360 CODE:
2361 int i;
2362 for (i = 2; ; ++i)
2363 {
2364 while (*format == ' ')
2365 ++format;
2366
2367 char type = *format++;
2368
2369 if (!type)
2370 break;
2371
2372 if (i >= items)
2373 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2374
2375 SV *sv = ST (i);
2376
2377 union
2378 {
2379 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2380 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2381 cl_half ch; cl_float cf; cl_double cd;
2382 cl_mem cm;
2383 cl_sampler ca;
2384 size_t cz;
2385 cl_event ce;
2386 } arg;
2387 size_t size;
2388 int nullarg = 0;
2389
2390 switch (type)
2391 {
2392 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2393 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2394 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2395 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2396 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2397 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2398 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2399 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2400
2401 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404
2405 case 'z': nullarg = 1; size = SvIV (sv); break;
2406
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2408 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2409 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410
2411 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 }
2414
2415 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2416 if (res)
2417 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2418 }
2419
2420 if (i != items)
2421 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2422
2423void
1700set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2424set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1701 CODE: 2425 CODE:
1702 clSetKernelArg (self, idx, sizeof (value), &value); 2426 clSetKernelArg (self, idx, sizeof (value), &value);
1703 2427
1704void 2428void
1752 clSetKernelArg (self, idx, sizeof (value), &value); 2476 clSetKernelArg (self, idx, sizeof (value), &value);
1753 2477
1754void 2478void
1755set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2479set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1756 CODE: 2480 CODE:
1757 clSetKernelArg (self, idx, sizeof (value), &value); 2481 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1758 2482
1759void 2483void
1760set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2484set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1761 CODE: 2485 CODE:
1762 clSetKernelArg (self, idx, sizeof (value), &value); 2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1763 2487
1764void 2488void
1765set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
1766 CODE: 2493 CODE:
1767 clSetKernelArg (self, idx, sizeof (value), &value); 2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1768
1769void
1770set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1771 CODE:
1772 clSetKernelArg (self, idx, sizeof (value), &value);
1773 2495
1774void 2496void
1775set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1776 CODE: 2498 CODE:
1777 clSetKernelArg (self, idx, sizeof (value), &value); 2499 clSetKernelArg (self, idx, sizeof (value), &value);
1823 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2545 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1824 EXTEND (SP, 1); 2546 EXTEND (SP, 1);
1825 const int i = 0; 2547 const int i = 0;
1826 { 2548 {
1827 NEED_SUCCESS (RetainContext, (value [i])); 2549 NEED_SUCCESS (RetainContext, (value [i]));
1828 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2550 PUSH_CLOBJ (stash_context, value [i]);
1829 } 2551 }
1830 2552
1831void 2553void
1832program (OpenCL::Kernel self) 2554program (OpenCL::Kernel self)
1833 PPCODE: 2555 PPCODE:
1835 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2557 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1836 EXTEND (SP, 1); 2558 EXTEND (SP, 1);
1837 const int i = 0; 2559 const int i = 0;
1838 { 2560 {
1839 NEED_SUCCESS (RetainProgram, (value [i])); 2561 NEED_SUCCESS (RetainProgram, (value [i]));
1840 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2562 PUSH_CLOBJ (stash_program, value [i]);
1841 } 2563 }
1842 2564
1843#END:kernel 2565#END:kernel
1844 2566
1845void 2567void
1905wait (OpenCL::Event self) 2627wait (OpenCL::Event self)
1906 CODE: 2628 CODE:
1907 clWaitForEvents (1, &self); 2629 clWaitForEvents (1, &self);
1908 2630
1909void 2631void
2632cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2633 CODE:
2634 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2635
2636void
1910info (OpenCL::Event self, cl_event_info name) 2637info (OpenCL::Event self, cl_event_info name)
1911 PPCODE: 2638 PPCODE:
1912 INFO (Event) 2639 INFO (Event)
1913 2640
1914#BEGIN:event 2641#BEGIN:event
1920 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1921 EXTEND (SP, 1); 2648 EXTEND (SP, 1);
1922 const int i = 0; 2649 const int i = 0;
1923 { 2650 {
1924 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2651 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1925 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2652 PUSH_CLOBJ (stash_queue, value [i]);
1926 } 2653 }
1927 2654
1928void 2655void
1929command_type (OpenCL::Event self) 2656command_type (OpenCL::Event self)
1930 PPCODE: 2657 PPCODE:
1953 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2680 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1954 EXTEND (SP, 1); 2681 EXTEND (SP, 1);
1955 const int i = 0; 2682 const int i = 0;
1956 { 2683 {
1957 NEED_SUCCESS (RetainContext, (value [i])); 2684 NEED_SUCCESS (RetainContext, (value [i]));
1958 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2685 PUSH_CLOBJ (stash_context, value [i]);
1959 } 2686 }
1960 2687
1961#END:event 2688#END:event
1962 2689
1963void 2690void
1988void 2715void
1989set_status (OpenCL::UserEvent self, cl_int execution_status) 2716set_status (OpenCL::UserEvent self, cl_int execution_status)
1990 CODE: 2717 CODE:
1991 clSetUserEventStatus (self, execution_status); 2718 clSetUserEventStatus (self, execution_status);
1992 2719
2720MODULE = OpenCL PACKAGE = OpenCL::Mapped
2721
2722void
2723DESTROY (SV *self)
2724 CODE:
2725 OpenCL__Mapped mapped = SvMAPPED (self);
2726
2727 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2728 mapped_detach (self, mapped);
2729
2730 clReleaseCommandQueue (mapped->queue);
2731 clReleaseEvent (mapped->event);
2732 Safefree (mapped);
2733
2734void
2735unmap (OpenCL::Mapped self, ...)
2736 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2738
2739bool
2740mapped (OpenCL::Mapped self)
2741 CODE:
2742 RETVAL = !!self->ptr;
2743 OUTPUT:
2744 RETVAL
2745
2746void
2747wait (OpenCL::Mapped self)
2748 PPCODE:
2749 if (self->event)
2750 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2751
2752void
2753event (OpenCL::Mapped self)
2754 PPCODE:
2755 if (!self->event)
2756 XSRETURN_UNDEF;
2757
2758 clRetainEvent (self->event);
2759 XPUSH_CLOBJ (stash_event, self->event);
2760
2761size_t
2762size (OpenCL::Mapped self)
2763 CODE:
2764 RETVAL = self->cb;
2765 OUTPUT:
2766 RETVAL
2767
2768IV
2769ptr (OpenCL::Mapped self)
2770 CODE:
2771 RETVAL = PTR2IV (self->ptr);
2772 OUTPUT:
2773 RETVAL
2774
2775void
2776set (OpenCL::Mapped self, size_t offset, SV *data)
2777 CODE:
2778 STRLEN len;
2779 const char *ptr = SvPVbyte (data, len);
2780
2781 if (offset < 0 || offset + len > self->cb)
2782 croak ("OpenCL::Mapped::set out of bound condition detected");
2783
2784 memcpy (offset + (char *)self->ptr, ptr, len);
2785
2786MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2787
2788MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2789
2790IV
2791row_pitch (OpenCL::Mapped self)
2792 ALIAS:
2793 slice_pitch = 1
2794 CODE:
2795 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2796 OUTPUT:
2797 RETVAL
2798
2799
2800

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines