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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.48 by root, Tue Apr 24 14:24:42 2012 UTC vs.
Revision 1.64 by root, Thu May 3 23:30:08 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#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
4 10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
7#endif 13#endif
8 14
40typedef cl_command_queue OpenCL__Queue; 46typedef cl_command_queue OpenCL__Queue;
41typedef cl_mem OpenCL__Memory; 47typedef cl_mem OpenCL__Memory;
42typedef cl_mem OpenCL__Buffer; 48typedef cl_mem OpenCL__Buffer;
43typedef cl_mem OpenCL__BufferObj; 49typedef cl_mem OpenCL__BufferObj;
44typedef cl_mem OpenCL__Image; 50typedef cl_mem OpenCL__Image;
45typedef cl_mem OpenCL__Image2D;
46typedef cl_mem OpenCL__Image3D;
47typedef cl_mem OpenCL__Memory_ornull; 51typedef cl_mem OpenCL__Memory_ornull;
48typedef cl_mem OpenCL__Buffer_ornull; 52typedef cl_mem OpenCL__Buffer_ornull;
49typedef cl_mem OpenCL__Image_ornull; 53typedef cl_mem OpenCL__Image_ornull;
50typedef cl_mem OpenCL__Image2D_ornull;
51typedef cl_mem OpenCL__Image3D_ornull;
52typedef cl_sampler OpenCL__Sampler; 54typedef cl_sampler OpenCL__Sampler;
53typedef cl_program OpenCL__Program; 55typedef cl_program OpenCL__Program;
54typedef cl_kernel OpenCL__Kernel; 56typedef cl_kernel OpenCL__Kernel;
55typedef cl_event OpenCL__Event; 57typedef cl_event OpenCL__Event;
56typedef cl_event OpenCL__UserEvent; 58typedef cl_event OpenCL__UserEvent;
57 59
60typedef struct mapped * OpenCL__Mapped;
61
58typedef SV *FUTURE; 62typedef SV *FUTURE;
63
64static HV
65 *stash_platform,
66 *stash_device,
67 *stash_context,
68 *stash_queue,
69 *stash_program,
70 *stash_kernel,
71 *stash_sampler,
72 *stash_event,
73 *stash_userevent,
74 *stash_memory,
75 *stash_buffer,
76 *stash_bufferobj,
77 *stash_image,
78 *stash_image1d,
79 *stash_image1darray,
80 *stash_image1dbuffer,
81 *stash_image2d,
82 *stash_image2darray,
83 *stash_image3d,
84 *stash_mapped,
85 *stash_mappedbuffer,
86 *stash_mappedimage;
59 87
60/*****************************************************************************/ 88/*****************************************************************************/
61 89
62// name must include a leading underscore 90// name must include a leading underscore
63// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 91// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
103} 131}
104 132
105/*****************************************************************************/ 133/*****************************************************************************/
106 134
107/* up to two temporary buffers */ 135/* up to two temporary buffers */
108static void * 136static void * ecb_noinline
109tmpbuf (size_t size) 137tmpbuf (size_t size)
110{ 138{
111 enum { buffers = 3 }; 139 enum { buffers = 4 };
112 static int idx; 140 static int idx;
113 static void *buf [buffers]; 141 static void *buf [buffers];
114 static size_t len [buffers]; 142 static size_t len [buffers];
115 143
116 idx = (idx + 1) % buffers; 144 idx = (idx + 1) % buffers;
189 if (res) \ 217 if (res) \
190 FAIL (name); 218 FAIL (name);
191 219
192/*****************************************************************************/ 220/*****************************************************************************/
193 221
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 * 222static cl_context_properties * ecb_noinline
209SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 223SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
210{ 224{
211 if (!sv || !SvOK (sv)) 225 if (!sv || !SvOK (sv))
212 if (extra) 226 if (extra)
213 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 227 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
270 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 284 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
271} 285}
272 286
273/*****************************************************************************/ 287/*****************************************************************************/
274 288
289static SV *
290new_clobj (HV *stash, IV id)
291{
292 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
293}
294
295#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
296#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
297
298/* cl objects are either \$iv, or [$iv, ...] */
299/* they can be upgraded at runtime to the array form */
300static void * ecb_noinline
301SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
302{
303 // sv_derived_from is quite slow :(
304 if (SvROK (sv) && sv_derived_from (sv, pkg))
305 return (void *)SvIV (SvRV (sv));
306
307 croak ("%s: %s is not of type %s", func, svname, pkg);
308}
309
310// the "no-inherit" version of the above
311static void * ecb_noinline
312SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
313{
314 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
315 return (void *)SvIV (SvRV (sv));
316
317 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
318}
319
320/*****************************************************************************/
321/* callback stuff */
322
323/* default context callback, log to stderr */
324static void CL_CALLBACK
325context_default_notify (const char *msg, const void *info, size_t cb, void *data)
326{
327 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
328}
329
330typedef struct
331{
332 int free_cb;
333 void (*push)(void *data1, void *data2, void *data3);
334} eq_vtbl;
335
336typedef struct eq_item
337{
338 struct eq_item *next;
339 eq_vtbl *vtbl;
340 SV *cb;
341 void *data1, *data2, *data3;
342} eq_item;
343
344static void (*eq_signal_func)(void *signal_arg, int value);
345static void *eq_signal_arg;
346static xmutex_t eq_lock = X_MUTEX_INIT;
347static eq_item *eq_head, *eq_tail;
348
349static void ecb_noinline
350eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
351{
352 eq_item *item = malloc (sizeof (eq_item));
353
354 item->next = 0;
355 item->vtbl = vtbl;
356 item->cb = cb;
357 item->data1 = data1;
358 item->data2 = data2;
359 item->data3 = data3;
360
361 X_LOCK (eq_lock);
362
363 *(eq_head ? &eq_tail->next : &eq_head) = item;
364 eq_tail = item;
365
366 X_UNLOCK (eq_lock);
367
368 eq_signal_func (eq_signal_arg, 0);
369}
370
371static eq_item *
372eq_dec (void)
373{
374 eq_item *res;
375
376 X_LOCK (eq_lock);
377
378 res = eq_head;
379 if (res)
380 eq_head = res->next;
381
382 X_UNLOCK (eq_lock);
383
384 return res;
385}
386
387static void
388eq_poll (void)
389{
390 eq_item *item;
391
392 while ((item = eq_dec ()))
393 {
394 ENTER;
395 SAVETMPS;
396
397 dSP;
398 PUSHMARK (SP);
399 EXTEND (SP, 2);
400
401 if (item->vtbl->free_cb)
402 sv_2mortal (item->cb);
403
404 PUTBACK;
405 item->vtbl->push (item->data1, item->data2, item->data3);
406
407 SV *cb = item->cb;
408 free (item);
409
410 call_sv (cb, G_DISCARD | G_VOID);
411
412 FREETMPS;
413 LEAVE;
414 }
415}
416
417static void
418eq_poll_interrupt (pTHX_ void *c_arg, int value)
419{
420 eq_poll ();
421}
422
423/*****************************************************************************/
424/* context notify */
425
426static void ecb_noinline
427eq_context_push (void *data1, void *data2, void *data3)
428{
429 dSP;
430 PUSHs (sv_2mortal (newSVpv (data1, 0)));
431 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
432 PUTBACK;
433
434 free (data1);
435 free (data2);
436}
437
438static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
439
440static void CL_CALLBACK
441eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
442{
443 void *pvt_copy = malloc (cb);
444 memcpy (pvt_copy, pvt, cb);
445 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
446}
447
448#define CONTEXT_NOTIFY_CALLBACK \
449 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
450 void *user_data = 0; \
451 \
452 if (SvOK (notify)) \
453 { \
454 pfn_notify = eq_context_notify; \
455 user_data = s_get_cv (notify); \
456 }
457
458static SV * ecb_noinline
459new_clobj_context (cl_context ctx, void *user_data)
460{
461 SV *sv = new_clobj (stash_context, (IV)ctx);
462
463 if (user_data)
464 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
465
466 return sv;
467}
468
469#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
470
471/*****************************************************************************/
472/* build/compile/link notify */
473
474static void
475eq_program_push (void *data1, void *data2, void *data3)
476{
477 dSP;
478 PUSH_CLOBJ (stash_program, data1);
479 PUTBACK;
480}
481
482static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
483
484static void CL_CALLBACK
485eq_program_notify (cl_program program, void *user_data)
486{
487 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
488}
489
490struct build_args
491{
492 cl_program program;
493 char *options;
494 void *user_data;
495 cl_uint num_devices;
496};
497
498X_THREAD_PROC (build_program_thread)
499{
500 struct build_args *arg = thr_arg;
501
502 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
503
504 if (arg->user_data)
505 eq_program_notify (arg->program, arg->user_data);
506 else
507 clReleaseProgram (arg->program);
508
509 free (arg->options);
510 free (arg);
511
512 return 0;
513}
514
515static void
516build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
517{
518 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
519
520 arg->program = program;
521 arg->options = strdup (options);
522 arg->user_data = user_data;
523 arg->num_devices = num_devices;
524 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
525
526 xthread_t id;
527 thread_create (&id, build_program_thread, arg);
528}
529
530/*****************************************************************************/
531/* event objects */
532
533static void
534eq_event_push (void *data1, void *data2, void *data3)
535{
536 dSP;
537 PUSH_CLOBJ (stash_event, data1);
538 PUSHs (sv_2mortal (newSViv ((IV)data2)));
539 PUTBACK;
540}
541
542static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
543
544static void CL_CALLBACK
545eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
546{
547 clRetainEvent (event);
548 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
549}
550
551/*****************************************************************************/
552/* utilities for XS code */
553
275static size_t 554static size_t
276img_row_pitch (cl_mem img) 555img_row_pitch (cl_mem img)
277{ 556{
278 size_t res; 557 size_t res;
279 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 558 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
280 return res; 559 return res;
281} 560}
282 561
283static cl_event * 562static cl_event * ecb_noinline
284event_list (SV **items, cl_uint *rcount) 563event_list (SV **items, cl_uint *rcount, cl_event extra)
285{ 564{
286 cl_uint count = *rcount; 565 cl_uint count = *rcount;
287 566
288 if (!count) 567 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
568 *rcount = count = 0;
569
570 if (!count && !extra)
289 return 0; 571 return 0;
290 572
291 cl_event *list = tmpbuf (sizeof (cl_event) * count); 573 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
292 int i = 0; 574 int i = 0;
293 575
294 do 576 while (count--)
295 {
296 --count;
297 if (SvOK (items [count])) 577 if (SvOK (items [count]))
298 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 578 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
299 } 579
300 while (count); 580 if (extra)
581 list [i++] = extra;
301 582
302 *rcount = i; 583 *rcount = i;
303 584
304 return i ? list : 0; 585 return i ? list : 0;
305} 586}
306 587
307#define EVENT_LIST(items,count) \ 588#define EVENT_LIST(skip) \
308 cl_uint event_list_count = (count); \ 589 cl_uint event_list_count = items - (skip); \
309 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 590 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
310 591
311#define INFO(class) \ 592#define INFO(class) \
312{ \ 593{ \
313 size_t size; \ 594 size_t size; \
314 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 595 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
318 SvCUR_set (sv, size); \ 599 SvCUR_set (sv, size); \
319 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 600 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
320 XPUSHs (sv); \ 601 XPUSHs (sv); \
321} 602}
322 603
604/*****************************************************************************/
605/* mapped_xxx */
606
607static OpenCL__Mapped
608SvMAPPED (SV *sv)
609{
610 // no typechecking atm., keep your fingers crossed
611 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
612}
613
614struct mapped
615{
616 cl_command_queue queue;
617 cl_mem memobj;
618 void *ptr;
619 size_t cb;
620 cl_event event;
621 size_t row_pitch;
622 size_t slice_pitch;
623};
624
625static SV *
626mapped_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)
627{
628 SV *data = newSV (0);
629 SvUPGRADE (data, SVt_PVMG);
630
631 OpenCL__Mapped mapped;
632 New (0, mapped, 1, struct mapped);
633
634 clRetainCommandQueue (queue);
635
636 mapped->queue = queue;
637 mapped->memobj = memobj;
638 mapped->ptr = ptr;
639 mapped->cb = cb;
640 mapped->event = ev;
641 mapped->row_pitch = row_pitch;
642 mapped->slice_pitch = slice_pitch;
643
644 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
645
646 if (SvLEN (data))
647 Safefree (data);
648
649 SvPVX (data) = (char *)ptr;
650 SvCUR_set (data, cb);
651 SvLEN_set (data, 0);
652 SvPOK_only (data);
653
654 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
655
656 if (!(flags & CL_MAP_WRITE))
657 SvREADONLY_on (data);
658
659 return obj;
660}
661
662static void
663mapped_detach (SV *sv, OpenCL__Mapped mapped)
664{
665 SV *data = SvRV (sv);
666
667 // the next check checks both before AND after detach, where SvPVX should STILL be 0
668 if (SvPVX (data) != (char *)mapped->ptr)
669 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
670 else
671 {
672 SvREADONLY_off (data);
673 SvCUR_set (data, 0);
674 SvPVX (data) = 0;
675 SvOK_off (data);
676 }
677
678 mapped->ptr = 0;
679}
680
681static void
682mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
683{
684 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
685 cl_event ev;
686
687 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
688
689 clReleaseEvent (mapped->event);
690 mapped->event = ev;
691
692 mapped_detach (self, mapped);
693}
694
695/*****************************************************************************/
696
323MODULE = OpenCL PACKAGE = OpenCL 697MODULE = OpenCL PACKAGE = OpenCL
324 698
325PROTOTYPES: ENABLE 699PROTOTYPES: ENABLE
700
701void
702poll ()
703 CODE:
704 eq_poll ();
705
706void
707_eq_initialise (IV func, IV arg)
708 CODE:
709 eq_signal_func = (void (*)(void *, int))func;
710 eq_signal_arg = (void*)arg;
326 711
327BOOT: 712BOOT:
328{ 713{
329 HV *stash = gv_stashpv ("OpenCL", 1); 714 HV *stash = gv_stashpv ("OpenCL", 1);
330 static const ivstr *civ, const_iv[] = { 715 static const ivstr *civ, const_iv[] = {
339 { sizeof (cl_half ), "SIZEOF_HALF" }, 724 { sizeof (cl_half ), "SIZEOF_HALF" },
340 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 725 { sizeof (cl_float ), "SIZEOF_FLOAT" },
341 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 726 { sizeof (cl_double), "SIZEOF_DOUBLE" },
342#include "constiv.h" 727#include "constiv.h"
343 }; 728 };
729
344 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 730 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)); 731 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
732
733 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
734 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
735 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
736 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
737 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
738 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
739 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
740 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
741 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
742 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
743 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
744 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
745 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
746 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
747 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
748 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
749 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
750 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
751 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
752 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
753 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
754 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
755
756 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
346} 757}
347 758
348cl_int 759cl_int
349errno () 760errno ()
350 CODE: 761 CODE:
351 RETVAL = res; 762 RETVAL = res;
352 OUTPUT: 763 OUTPUT:
353 RETVAL 764 RETVAL
354 765
355const char * 766const char *
356err2str (cl_int err) 767err2str (cl_int err = res)
357 768
358const char * 769const char *
359enum2str (cl_uint value) 770enum2str (cl_uint value)
360 771
361void 772void
369 list = tmpbuf (sizeof (*list) * count); 780 list = tmpbuf (sizeof (*list) * count);
370 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 781 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
371 782
372 EXTEND (SP, count); 783 EXTEND (SP, count);
373 for (i = 0; i < count; ++i) 784 for (i = 0; i < count; ++i)
374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 785 PUSH_CLOBJ (stash_platform, (IV)list [i]);
375 786
376void 787void
377context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 788context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
378 PPCODE: 789 PPCODE:
790 CONTEXT_NOTIFY_CALLBACK;
379 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 791 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
380 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 792 XPUSH_CLOBJ_CONTEXT;
381 793
382#if 0
383
384void 794void
385context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 795context (FUTURE properties, FUTURE devices, FUTURE notify)
386 PPCODE: 796 PPCODE:
387 /* der Gipfel der Kunst */ 797 /* der Gipfel der Kunst */
388 798
389#endif
390
391void 799void
392wait_for_events (...) 800wait_for_events (...)
393 CODE: 801 CODE:
394 EVENT_LIST (0, items); 802 EVENT_LIST (0);
395 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 803 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
396 804
397PROTOTYPES: DISABLE 805PROTOTYPES: DISABLE
398 806
399MODULE = OpenCL PACKAGE = OpenCL::Platform 807MODULE = OpenCL PACKAGE = OpenCL::Platform
442 list = tmpbuf (sizeof (*list) * count); 850 list = tmpbuf (sizeof (*list) * count);
443 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 851 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
444 852
445 EXTEND (SP, count); 853 EXTEND (SP, count);
446 for (i = 0; i < count; ++i) 854 for (i = 0; i < count; ++i)
447 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 855 PUSH_CLOBJ (stash_device, list [i]);
448 856
449void 857void
450context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 858context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
451 PPCODE: 859 PPCODE:
452 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 860 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
453 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 861 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
862
863 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
864 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
454 865
455 AV *av = (AV *)SvRV (devices); 866 AV *av = (AV *)SvRV (devices);
456 cl_uint num_devices = av_len (av) + 1; 867 cl_uint num_devices = av_len (av) + 1;
457 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 868 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
869
458 int i; 870 int i;
459
460 for (i = num_devices; i--; ) 871 for (i = num_devices; i--; )
461 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 872 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
462 873
874 CONTEXT_NOTIFY_CALLBACK;
463 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 875 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res));
464 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 876 XPUSH_CLOBJ_CONTEXT;
465 877
466void 878void
467context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 879context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
468 PPCODE: 880 PPCODE:
469 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 881 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); 882 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
883
884 CONTEXT_NOTIFY_CALLBACK;
471 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 885 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
472 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 886 XPUSH_CLOBJ_CONTEXT;
473 887
474MODULE = OpenCL PACKAGE = OpenCL::Device 888MODULE = OpenCL PACKAGE = OpenCL::Device
475 889
476void 890void
477info (OpenCL::Device self, cl_device_info name) 891info (OpenCL::Device self, cl_device_info name)
485 PPCODE: 899 PPCODE:
486 cl_device_type value [1]; 900 cl_device_type value [1];
487 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
488 EXTEND (SP, 1); 902 EXTEND (SP, 1);
489 const int i = 0; 903 const int i = 0;
490 PUSHs (sv_2mortal (newSViv (value [i]))); 904 PUSHs (sv_2mortal (newSVuv (value [i])));
491 905
492void 906void
493vendor_id (OpenCL::Device self) 907vendor_id (OpenCL::Device self)
494 ALIAS: 908 ALIAS:
495 vendor_id = CL_DEVICE_VENDOR_ID 909 vendor_id = CL_DEVICE_VENDOR_ID
640 PPCODE: 1054 PPCODE:
641 cl_command_queue_properties value [1]; 1055 cl_command_queue_properties value [1];
642 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1056 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
643 EXTEND (SP, 1); 1057 EXTEND (SP, 1);
644 const int i = 0; 1058 const int i = 0;
645 PUSHs (sv_2mortal (newSViv (value [i]))); 1059 PUSHs (sv_2mortal (newSVuv (value [i])));
646 1060
647void 1061void
648platform (OpenCL::Device self) 1062platform (OpenCL::Device self)
649 PPCODE: 1063 PPCODE:
650 cl_platform_id value [1]; 1064 cl_platform_id value [1];
651 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1065 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
652 EXTEND (SP, 1); 1066 EXTEND (SP, 1);
653 const int i = 0; 1067 const int i = 0;
654 { 1068 {
655 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1069 PUSH_CLOBJ (stash_platform, value [i]);
656 } 1070 }
657 1071
658void 1072void
659name (OpenCL::Device self) 1073name (OpenCL::Device self)
660 ALIAS: 1074 ALIAS:
679 cl_device_id value [1]; 1093 cl_device_id value [1];
680 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1094 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
681 EXTEND (SP, 1); 1095 EXTEND (SP, 1);
682 const int i = 0; 1096 const int i = 0;
683 { 1097 {
684 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1098 PUSH_CLOBJ (stash_device, value [i]);
685 } 1099 }
686 1100
687void 1101void
688partition_types_ext (OpenCL::Device self) 1102partition_types_ext (OpenCL::Device self)
689 ALIAS: 1103 ALIAS:
716 1130
717void 1131void
718queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1132queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
719 PPCODE: 1133 PPCODE:
720 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1134 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
721 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1135 XPUSH_CLOBJ (stash_queue, queue);
722 1136
723void 1137void
724user_event (OpenCL::Context self) 1138user_event (OpenCL::Context self)
725 PPCODE: 1139 PPCODE:
726 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1140 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
727 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1141 XPUSH_CLOBJ (stash_userevent, ev);
728 1142
729void 1143void
730buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1144buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
731 PPCODE: 1145 PPCODE:
732 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1146 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?"); 1147 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
734 1148
735 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1149 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
736 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1150 XPUSH_CLOBJ (stash_bufferobj, mem);
737 1151
738void 1152void
739buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1153buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
740 PPCODE: 1154 PPCODE:
741 STRLEN len; 1155 STRLEN len;
742 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1156 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
743 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1157 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?"); 1158 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)); 1159 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
746 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1160 XPUSH_CLOBJ (stash_bufferobj, mem);
747 1161
748#if CL_VERSION_1_2 1162#if CL_VERSION_1_2
749 1163
750void 1164void
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) 1165image (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: 1166 PPCODE:
753 STRLEN len; 1167 STRLEN len;
754 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1168 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
755 const cl_image_format format = { channel_order, channel_type }; 1169 const cl_image_format format = { channel_order, channel_type };
756 const cl_image_desc desc = { 1170 const cl_image_desc desc = {
757 type, 1171 type,
758 width, height, depth, 1172 width, height, depth,
759 array_size, row_pitch, slice_pitch, 1173 array_size, row_pitch, slice_pitch,
760 num_mip_level, num_samples, 1174 num_mip_level, num_samples,
761 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvPTROBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1175 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
762 }; 1176 };
763 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1177 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
764 char *klass = "OpenCL::Image"; 1178 HV *stash = stash_image;
765 switch (type) 1179 switch (type)
766 { 1180 {
767 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1181 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
768 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1182 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
769 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1183 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
770 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1184 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
771 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1185 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
772 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1186 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
773 } 1187 }
774 XPUSH_NEW_OBJ (klass, mem); 1188 XPUSH_CLOBJ (stash, mem);
775 1189
776#endif 1190#endif
777 1191
778void 1192void
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) 1193image2d (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)); 1199 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
786#else 1200#else
787 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1201 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)); 1202 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
789#endif 1203#endif
790 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1204 XPUSH_CLOBJ (stash_image2d, mem);
791 1205
792void 1206void
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) 1207image3d (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: 1208 PPCODE:
795 STRLEN len; 1209 STRLEN len;
799 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1213 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
800#else 1214#else
801 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1215 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)); 1216 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
803#endif 1217#endif
804 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1218 XPUSH_CLOBJ (stash_image3d, mem);
805 1219
806#if cl_apple_gl_sharing || cl_khr_gl_sharing 1220#if cl_apple_gl_sharing || cl_khr_gl_sharing
807 1221
808void 1222void
809gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1223gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
810 PPCODE: 1224 PPCODE:
811 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1225 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
812 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1226 XPUSH_CLOBJ (stash_bufferobj, mem);
813 1227
814void 1228void
815gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1229gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
816 PPCODE: 1230 PPCODE:
817 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1231 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
818 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1232 XPUSH_CLOBJ (stash_image2d, mem);
819 1233
820#if CL_VERSION_1_2 1234#if CL_VERSION_1_2
821 1235
822void 1236void
823gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1237gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
824 ALIAS: 1238 ALIAS:
825 PPCODE: 1239 PPCODE:
826 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1240 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
827 cl_gl_object_type type; 1241 cl_gl_object_type type;
828 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1242 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
829 char *klass = "OpenCL::Memory"; 1243 HV *stash = stash_memory;
830 switch (type) 1244 switch (type)
831 { 1245 {
832 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1246 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
833 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1247 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
834 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1248 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
835 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1249 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
836 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1250 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
837 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1251 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
838 } 1252 }
839 XPUSH_NEW_OBJ (klass, mem); 1253 XPUSH_CLOBJ (stash, mem);
840 1254
841#endif 1255#endif
842 1256
843void 1257void
844gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1258gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
846#if PREFER_1_1 1260#if PREFER_1_1
847 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1261 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
848#else 1262#else
849 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1263 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
850#endif 1264#endif
851 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1265 XPUSH_CLOBJ (stash_image2d, mem);
852 1266
853void 1267void
854gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1268gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
855 PPCODE: 1269 PPCODE:
856#if PREFER_1_1 1270#if PREFER_1_1
857 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
858#else 1272#else
859 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1273 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
860#endif 1274#endif
861 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1275 XPUSH_CLOBJ (stash_image3d, mem);
862 1276
863#endif 1277#endif
864 1278
865void 1279void
866supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1280supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
886 1300
887void 1301void
888sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1302sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
889 PPCODE: 1303 PPCODE:
890 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1304 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
891 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1305 XPUSH_CLOBJ (stash_sampler, sampler);
892 1306
893void 1307void
894program_with_source (OpenCL::Context self, SV *program) 1308program_with_source (OpenCL::Context self, SV *program)
895 PPCODE: 1309 PPCODE:
896 STRLEN len; 1310 STRLEN len;
897 size_t len2; 1311 size_t len2;
898 const char *ptr = SvPVbyte (program, len); 1312 const char *ptr = SvPVbyte (program, len);
899 1313
900 len2 = len; 1314 len2 = len;
901 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1315 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
902 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1316 XPUSH_CLOBJ (stash_program, prog);
1317
1318void
1319program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1320 PPCODE:
1321 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1322 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects");
1323
1324 devices = SvRV (devices);
1325
1326 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1327 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1328
1329 binaries = SvRV (binaries);
1330
1331 if (av_len ((AV *)devices) != av_len ((AV *)binaries))
1332 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1333
1334 int count = av_len ((AV *)devices) + 1;
1335 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1336 size_t *length_list = tmpbuf (sizeof (*length_list) * count);
1337 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count);
1338 cl_int *status_list = tmpbuf (sizeof (*status_list) * count);
1339
1340 int i;
1341 for (i = 0; i < count; ++i)
1342 {
1343 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1344 STRLEN len;
1345 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1346 length_list [i] = len;
1347 }
1348
1349 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res));
1350
1351 EXTEND (SP, 2);
1352 PUSH_CLOBJ (stash_program, prog);
1353
1354 if (GIMME_V == G_ARRAY)
1355 {
1356 AV *av = newAV ();
1357 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1358
1359 for (i = count; i--; )
1360 av_store (av, i, newSViv (status_list [i]));
1361 }
903 1362
904#BEGIN:context 1363#BEGIN:context
905 1364
906void 1365void
907reference_count (OpenCL::Context self) 1366reference_count (OpenCL::Context self)
924 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1383 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
925 int i, n = size / sizeof (*value); 1384 int i, n = size / sizeof (*value);
926 EXTEND (SP, n); 1385 EXTEND (SP, n);
927 for (i = 0; i < n; ++i) 1386 for (i = 0; i < n; ++i)
928 { 1387 {
929 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1388 PUSH_CLOBJ (stash_device, value [i]);
930 } 1389 }
931 1390
932void 1391void
933properties (OpenCL::Context self) 1392properties (OpenCL::Context self)
934 PPCODE: 1393 PPCODE:
949DESTROY (OpenCL::Queue self) 1408DESTROY (OpenCL::Queue self)
950 CODE: 1409 CODE:
951 clReleaseCommandQueue (self); 1410 clReleaseCommandQueue (self);
952 1411
953void 1412void
954enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1413read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1414 ALIAS:
1415 enqueue_read_buffer = 0
955 PPCODE: 1416 PPCODE:
956 cl_event ev = 0; 1417 cl_event ev = 0;
957 EVENT_LIST (6, items - 6); 1418 EVENT_LIST (6);
958 1419
959 SvUPGRADE (data, SVt_PV); 1420 SvUPGRADE (data, SVt_PV);
960 SvGROW (data, len); 1421 SvGROW (data, len);
961 SvPOK_only (data); 1422 SvPOK_only (data);
962 SvCUR_set (data, len); 1423 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)); 1424 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
964 1425
965 if (ev) 1426 if (ev)
966 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1427 XPUSH_CLOBJ (stash_event, ev);
967 1428
968void 1429void
969enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1430write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1431 ALIAS:
1432 enqueue_write_buffer = 0
970 PPCODE: 1433 PPCODE:
971 cl_event ev = 0; 1434 cl_event ev = 0;
972 STRLEN len; 1435 STRLEN len;
973 char *ptr = SvPVbyte (data, len); 1436 char *ptr = SvPVbyte (data, len);
974 EVENT_LIST (5, items - 5); 1437 EVENT_LIST (5);
975 1438
976 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1439 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
977 1440
978 if (ev) 1441 if (ev)
979 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1442 XPUSH_CLOBJ (stash_event, ev);
980 1443
981#if CL_VERSION_1_2 1444#if CL_VERSION_1_2
982 1445
983void 1446void
984enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1447fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1448 ALIAS:
1449 enqueue_fill_buffer = 0
985 PPCODE: 1450 PPCODE:
986 cl_event ev = 0; 1451 cl_event ev = 0;
987 STRLEN len; 1452 STRLEN len;
988 char *ptr = SvPVbyte (data, len); 1453 char *ptr = SvPVbyte (data, len);
989 EVENT_LIST (5, items - 5); 1454 EVENT_LIST (5);
990 1455
991 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1456 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
992 1457
993 if (ev) 1458 if (ev)
994 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1459 XPUSH_CLOBJ (stash_event, ev);
995 1460
996void 1461void
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, ...) 1462fill_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, ...)
1463 ALIAS:
1464 enqueue_fill_image = 0
998 PPCODE: 1465 PPCODE:
999 cl_event ev = 0; 1466 cl_event ev = 0;
1000 STRLEN len;
1001 const size_t origin [3] = { x, y, z }; 1467 const size_t origin [3] = { x, y, z };
1002 const size_t region [3] = { width, height, depth }; 1468 const size_t region [3] = { width, height, depth };
1003 EVENT_LIST (12, items - 12); 1469 EVENT_LIST (12);
1004 1470
1005 const cl_float c_f [4] = { r, g, b, a }; 1471 const cl_float c_f [4] = { r, g, b, a };
1006 const cl_uint c_u [4] = { r, g, b, a }; 1472 const cl_uint c_u [4] = { r, g, b, a };
1007 const cl_int c_s [4] = { r, g, b, a }; 1473 const cl_int c_s [4] = { r, g, b, a };
1008 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1474 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1009 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 }; 1475 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1010 cl_image_format format; 1476 cl_image_format format;
1011 NEED_SUCCESS (clGetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1477 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1012 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1478 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) 1479 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."); 1480 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1015 1481
1016 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1482 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)); 1483 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1018 1484
1019 if (ev) 1485 if (ev)
1020 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1486 XPUSH_CLOBJ (stash_event, ev);
1021 1487
1022#endif 1488#endif
1023 1489
1024void 1490void
1025enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1491copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1492 ALIAS:
1493 enqueue_copy_buffer = 0
1026 PPCODE: 1494 PPCODE:
1027 cl_event ev = 0; 1495 cl_event ev = 0;
1028 EVENT_LIST (6, items - 6); 1496 EVENT_LIST (6);
1029 1497
1030 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1498 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1031 1499
1032 if (ev) 1500 if (ev)
1033 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1501 XPUSH_CLOBJ (stash_event, ev);
1034 1502
1035void 1503void
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, ...) 1504read_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, ...)
1505 ALIAS:
1506 enqueue_read_buffer_rect = 0
1037 PPCODE: 1507 PPCODE:
1038 cl_event ev = 0; 1508 cl_event ev = 0;
1039 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1509 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 }; 1510 const size_t host_origin[3] = { host_x, host_y, host_z };
1041 const size_t region[3] = { width, height, depth }; 1511 const size_t region[3] = { width, height, depth };
1042 EVENT_LIST (17, items - 17); 1512 EVENT_LIST (17);
1043 1513
1044 if (!buf_row_pitch) 1514 if (!buf_row_pitch)
1045 buf_row_pitch = region [0]; 1515 buf_row_pitch = region [0];
1046 1516
1047 if (!buf_slice_pitch) 1517 if (!buf_slice_pitch)
1060 SvPOK_only (data); 1530 SvPOK_only (data);
1061 SvCUR_set (data, len); 1531 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)); 1532 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 1533
1064 if (ev) 1534 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1535 XPUSH_CLOBJ (stash_event, ev);
1066 1536
1067void 1537void
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, ...) 1538write_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, ...)
1539 ALIAS:
1540 enqueue_write_buffer_rect = 0
1069 PPCODE: 1541 PPCODE:
1070 cl_event ev = 0; 1542 cl_event ev = 0;
1071 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1543 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 }; 1544 const size_t host_origin[3] = { host_x, host_y, host_z };
1073 const size_t region[3] = { width, height, depth }; 1545 const size_t region[3] = { width, height, depth };
1074 STRLEN len; 1546 STRLEN len;
1075 char *ptr = SvPVbyte (data, len); 1547 char *ptr = SvPVbyte (data, len);
1076 EVENT_LIST (17, items - 17); 1548 EVENT_LIST (17);
1077 1549
1078 if (!buf_row_pitch) 1550 if (!buf_row_pitch)
1079 buf_row_pitch = region [0]; 1551 buf_row_pitch = region [0];
1080 1552
1081 if (!buf_slice_pitch) 1553 if (!buf_slice_pitch)
1093 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1565 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1094 1566
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)); 1567 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 1568
1097 if (ev) 1569 if (ev)
1098 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1570 XPUSH_CLOBJ (stash_event, ev);
1099 1571
1100void 1572void
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, ...) 1573copy_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, ...)
1574 ALIAS:
1575 enqueue_copy_buffer_rect = 0
1102 PPCODE: 1576 PPCODE:
1103 cl_event ev = 0; 1577 cl_event ev = 0;
1104 const size_t src_origin[3] = { src_x, src_y, src_z }; 1578 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 }; 1579 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1106 const size_t region[3] = { width, height, depth }; 1580 const size_t region[3] = { width, height, depth };
1107 EVENT_LIST (16, items - 16); 1581 EVENT_LIST (16);
1108 1582
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)); 1583 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 1584
1111 if (ev) 1585 if (ev)
1112 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1586 XPUSH_CLOBJ (stash_event, ev);
1113 1587
1114void 1588void
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, ...) 1589read_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, ...)
1590 ALIAS:
1591 enqueue_read_image = 0
1116 PPCODE: 1592 PPCODE:
1117 cl_event ev = 0; 1593 cl_event ev = 0;
1118 const size_t src_origin[3] = { src_x, src_y, src_z }; 1594 const size_t src_origin[3] = { src_x, src_y, src_z };
1119 const size_t region[3] = { width, height, depth }; 1595 const size_t region[3] = { width, height, depth };
1120 EVENT_LIST (12, items - 12); 1596 EVENT_LIST (12);
1121 1597
1122 if (!row_pitch) 1598 if (!row_pitch)
1123 row_pitch = img_row_pitch (src); 1599 row_pitch = img_row_pitch (src);
1124 1600
1125 if (depth > 1 && !slice_pitch) 1601 if (depth > 1 && !slice_pitch)
1132 SvPOK_only (data); 1608 SvPOK_only (data);
1133 SvCUR_set (data, len); 1609 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)); 1610 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 1611
1136 if (ev) 1612 if (ev)
1137 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1613 XPUSH_CLOBJ (stash_event, ev);
1138 1614
1139void 1615void
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, ...) 1616write_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, ...)
1617 ALIAS:
1618 enqueue_write_image = 0
1141 PPCODE: 1619 PPCODE:
1142 cl_event ev = 0; 1620 cl_event ev = 0;
1143 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1621 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1144 const size_t region[3] = { width, height, depth }; 1622 const size_t region[3] = { width, height, depth };
1145 STRLEN len; 1623 STRLEN len;
1146 char *ptr = SvPVbyte (data, len); 1624 char *ptr = SvPVbyte (data, len);
1147 EVENT_LIST (12, items - 12); 1625 EVENT_LIST (12);
1148 1626
1149 if (!row_pitch) 1627 if (!row_pitch)
1150 row_pitch = img_row_pitch (dst); 1628 row_pitch = img_row_pitch (dst);
1151 1629
1152 if (depth > 1 && !slice_pitch) 1630 if (depth > 1 && !slice_pitch)
1158 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1636 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1159 1637
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)); 1638 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 1639
1162 if (ev) 1640 if (ev)
1163 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1641 XPUSH_CLOBJ (stash_event, ev);
1164 1642
1165void 1643void
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, ...) 1644copy_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, ...)
1645 ALIAS:
1646 enqueue_copy_image = 0
1167 PPCODE: 1647 PPCODE:
1168 cl_event ev = 0; 1648 cl_event ev = 0;
1169 const size_t src_origin[3] = { src_x, src_y, src_z }; 1649 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 }; 1650 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1171 const size_t region[3] = { width, height, depth }; 1651 const size_t region[3] = { width, height, depth };
1172 EVENT_LIST (12, items - 12); 1652 EVENT_LIST (12);
1173 1653
1174 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1654 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1175 1655
1176 if (ev) 1656 if (ev)
1177 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1657 XPUSH_CLOBJ (stash_event, ev);
1178 1658
1179void 1659void
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, ...) 1660copy_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, ...)
1661 ALIAS:
1662 enqueue_copy_image_to_buffer = 0
1181 PPCODE: 1663 PPCODE:
1182 cl_event ev = 0; 1664 cl_event ev = 0;
1183 const size_t src_origin[3] = { src_x, src_y, src_z }; 1665 const size_t src_origin[3] = { src_x, src_y, src_z };
1666 const size_t region [3] = { width, height, depth };
1667 EVENT_LIST (10);
1668
1669 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1670
1671 if (ev)
1672 XPUSH_CLOBJ (stash_event, ev);
1673
1674void
1675copy_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, ...)
1676 ALIAS:
1677 enqueue_copy_buffer_to_image = 0
1678 PPCODE:
1679 cl_event ev = 0;
1680 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1681 const size_t region [3] = { width, height, depth };
1682 EVENT_LIST (10);
1683
1684 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1685
1686 if (ev)
1687 XPUSH_CLOBJ (stash_event, ev);
1688
1689void
1690map_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, SV *cb_ = &PL_sv_undef, ...)
1691 ALIAS:
1692 enqueue_map_buffer = 0
1693 PPCODE:
1694 cl_event ev;
1695 EVENT_LIST (6);
1696 size_t cb = SvIV (cb_);
1697
1698 if (!SvOK (cb_))
1699 {
1700 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1701 cb -= offset;
1702 }
1703
1704 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1705 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1706
1707void
1708map_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, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1709 ALIAS:
1710 enqueue_map_image = 0
1711 PPCODE:
1712 size_t width = SvIV (width_);
1713 if (!SvOK (width_))
1714 {
1715 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1716 width -= x;
1717 }
1718
1719 size_t height = SvIV (width_);
1720 if (!SvOK (height_))
1721 {
1722 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1723 height -= y;
1724 }
1725
1726 size_t depth = SvIV (width_);
1727 if (!SvOK (depth_))
1728 {
1729 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1730 depth -= z;
1731
1732 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1733 if (!depth)
1734 depth = 1;
1735 }
1736
1737 const size_t origin[3] = { x, y, z };
1184 const size_t region[3] = { width, height, depth }; 1738 const size_t region[3] = { width, height, depth };
1739 size_t row_pitch, slice_pitch;
1185 EVENT_LIST (10, items - 10); 1740 EVENT_LIST (10);
1186 1741
1742 cl_event ev;
1743 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));
1744
1745 size_t cb = slice_pitch ? slice_pitch * region [2]
1746 : row_pitch ? row_pitch * region [1]
1747 : region [0];
1748
1749 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1750
1751void
1752unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1753 PPCODE:
1754 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1755 if (GIMME_V != G_VOID)
1756 {
1757 clRetainEvent (mapped->event);
1758 XPUSH_CLOBJ (stash_event, mapped->event);
1759 }
1760
1761void
1762task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1763 ALIAS:
1764 enqueue_task = 0
1765 PPCODE:
1766 cl_event ev = 0;
1767 EVENT_LIST (2);
1768
1187 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1769 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1188 1770
1189 if (ev) 1771 if (ev)
1190 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1772 XPUSH_CLOBJ (stash_event, ev);
1191 1773
1192void 1774void
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, ...)
1194 PPCODE:
1195 cl_event ev = 0;
1196 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1197 const size_t region[3] = { width, height, depth };
1198 EVENT_LIST (10, items - 10);
1199
1200 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1201
1202 if (ev)
1203 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1204
1205void
1206enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1207 PPCODE:
1208 cl_event ev = 0;
1209 EVENT_LIST (2, items - 2);
1210
1211 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1212
1213 if (ev)
1214 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1215
1216void
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, ...) 1775nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1776 ALIAS:
1777 enqueue_nd_range_kernel = 0
1218 PPCODE: 1778 PPCODE:
1219 cl_event ev = 0; 1779 cl_event ev = 0;
1220 size_t *gwo = 0, *gws, *lws = 0; 1780 size_t *gwo = 0, *gws, *lws = 0;
1221 int gws_len; 1781 int gws_len;
1222 size_t *lists; 1782 size_t *lists;
1223 int i; 1783 int i;
1224 EVENT_LIST (5, items - 5); 1784 EVENT_LIST (5);
1225 1785
1226 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1786 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1227 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1787 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1228 1788
1229 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1789 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1230 1790
1231 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1791 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1232 1792
1233 gws = lists + gws_len * 0; 1793 gws = lists + gws_len * 0;
1234 for (i = 0; i < gws_len; ++i) 1794 for (i = 0; i < gws_len; ++i)
1795 {
1235 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1796 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1797 // at least nvidia crashes for 0-sized work group sizes, work around
1798 if (!gws [i])
1799 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1800 }
1236 1801
1237 if (SvOK (global_work_offset)) 1802 if (SvOK (global_work_offset))
1238 { 1803 {
1239 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1804 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"); 1805 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1248 } 1813 }
1249 1814
1250 if (SvOK (local_work_size)) 1815 if (SvOK (local_work_size))
1251 { 1816 {
1252 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1817 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"); 1818 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1254 1819
1255 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1820 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"); 1821 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1257 1822
1258 lws = lists + gws_len * 2; 1823 lws = lists + gws_len * 2;
1259 for (i = 0; i < gws_len; ++i) 1824 for (i = 0; i < gws_len; ++i)
1825 {
1260 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1826 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1827 // at least nvidia crashes for 0-sized work group sizes, work around
1828 if (!lws [i])
1829 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1830 }
1261 } 1831 }
1262 1832
1263 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1833 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1264 1834
1265 if (ev) 1835 if (ev)
1266 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1836 XPUSH_CLOBJ (stash_event, ev);
1267 1837
1268#if cl_apple_gl_sharing || cl_khr_gl_sharing 1838#if cl_apple_gl_sharing || cl_khr_gl_sharing
1269 1839
1270void 1840void
1271enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1841acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1272 ALIAS: 1842 ALIAS:
1843 release_gl_objects = 1
1844 enqueue_acquire_gl_objects = 0
1273 enqueue_release_gl_objects = 1 1845 enqueue_release_gl_objects = 1
1274 PPCODE: 1846 PPCODE:
1275 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1847 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"); 1848 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1277 1849
1278 cl_event ev = 0; 1850 cl_event ev = 0;
1279 EVENT_LIST (2, items - 2); 1851 EVENT_LIST (2);
1280 AV *av = (AV *)SvRV (objects); 1852 AV *av = (AV *)SvRV (objects);
1281 cl_uint num_objects = av_len (av) + 1; 1853 cl_uint num_objects = av_len (av) + 1;
1282 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1854 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1283 int i; 1855 int i;
1284 1856
1285 for (i = num_objects; i--; ) 1857 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"); 1858 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1287 1859
1288 if (ix) 1860 if (ix)
1289 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1861 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1290 else 1862 else
1291 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1863 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1292 1864
1293 if (ev) 1865 if (ev)
1294 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1866 XPUSH_CLOBJ (stash_event, ev);
1295 1867
1296#endif 1868#endif
1297 1869
1298void 1870void
1299enqueue_wait_for_events (OpenCL::Queue self, ...) 1871wait_for_events (OpenCL::Queue self, ...)
1872 ALIAS:
1873 enqueue_wait_for_events = 0
1300 CODE: 1874 CODE:
1301 EVENT_LIST (1, items - 1); 1875 EVENT_LIST (1);
1302#if PREFER_1_1 1876#if PREFER_1_1
1303 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1877 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1304#else 1878#else
1305 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 1879 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1306#endif 1880#endif
1307 1881
1308void 1882void
1309enqueue_marker (OpenCL::Queue self, ...) 1883marker (OpenCL::Queue self, ...)
1884 ALIAS:
1885 enqueue_marker = 0
1310 PPCODE: 1886 PPCODE:
1311 cl_event ev = 0; 1887 cl_event ev = 0;
1312 EVENT_LIST (1, items - 1); 1888 EVENT_LIST (1);
1313#if PREFER_1_1 1889#if PREFER_1_1
1314 if (!event_list_count) 1890 if (!event_list_count)
1315 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 1891 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1316 else 1892 else
1317#if CL_VERSION_1_2 1893#if CL_VERSION_1_2
1318 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1894 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1319#else 1895#else
1896 {
1320 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); // also a barrier 1897 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1898 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1899 }
1321#endif 1900#endif
1322#else 1901#else
1323 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1902 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1324#endif 1903#endif
1325 if (ev) 1904 if (ev)
1326 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1905 XPUSH_CLOBJ (stash_event, ev);
1327 1906
1328void 1907void
1329enqueue_barrier (OpenCL::Queue self, ...) 1908barrier (OpenCL::Queue self, ...)
1909 ALIAS:
1910 enqueue_barrier = 0
1330 PPCODE: 1911 PPCODE:
1331 cl_event ev = 0; 1912 cl_event ev = 0;
1332 EVENT_LIST (1, items - 1); 1913 EVENT_LIST (1);
1333#if PREFER_1_1 1914#if PREFER_1_1
1334 if (!event_list_count && GIMME_V == G_VOID) 1915 if (!event_list_count && GIMME_V == G_VOID)
1335 NEED_SUCCESS (EnqueueBarrier, (self)); 1916 NEED_SUCCESS (EnqueueBarrier, (self));
1336 else 1917 else
1337#if CL_VERSION_1_2 1918#if CL_VERSION_1_2
1338 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1919 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1339#else 1920#else
1340 { 1921 {
1341 if (event_list_count) 1922 if (event_list_count)
1342 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1923 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1343 1924
1344 if (GIMME_V != G_VOID) 1925 if (GIMME_V != G_VOID)
1345 NEED_SUCCESS (EnqueueMarker, (self, &ev)); 1926 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1346 } 1927 }
1347#endif 1928#endif
1348#else 1929#else
1349 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1930 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1350#endif 1931#endif
1351 if (ev) 1932 if (ev)
1352 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1933 XPUSH_CLOBJ (stash_event, ev);
1353 1934
1354void 1935void
1355flush (OpenCL::Queue self) 1936flush (OpenCL::Queue self)
1356 CODE: 1937 CODE:
1357 NEED_SUCCESS (Flush, (self)); 1938 NEED_SUCCESS (Flush, (self));
1375 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1956 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1376 EXTEND (SP, 1); 1957 EXTEND (SP, 1);
1377 const int i = 0; 1958 const int i = 0;
1378 { 1959 {
1379 NEED_SUCCESS (RetainContext, (value [i])); 1960 NEED_SUCCESS (RetainContext, (value [i]));
1380 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1961 PUSH_CLOBJ (stash_context, value [i]);
1381 } 1962 }
1382 1963
1383void 1964void
1384device (OpenCL::Queue self) 1965device (OpenCL::Queue self)
1385 PPCODE: 1966 PPCODE:
1386 cl_device_id value [1]; 1967 cl_device_id value [1];
1387 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1968 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1388 EXTEND (SP, 1); 1969 EXTEND (SP, 1);
1389 const int i = 0; 1970 const int i = 0;
1390 { 1971 {
1391 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1972 PUSH_CLOBJ (stash_device, value [i]);
1392 } 1973 }
1393 1974
1394void 1975void
1395reference_count (OpenCL::Queue self) 1976reference_count (OpenCL::Queue self)
1396 PPCODE: 1977 PPCODE:
1405 PPCODE: 1986 PPCODE:
1406 cl_command_queue_properties value [1]; 1987 cl_command_queue_properties value [1];
1407 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1988 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1408 EXTEND (SP, 1); 1989 EXTEND (SP, 1);
1409 const int i = 0; 1990 const int i = 0;
1410 PUSHs (sv_2mortal (newSViv (value [i]))); 1991 PUSHs (sv_2mortal (newSVuv (value [i])));
1411 1992
1412#END:command_queue 1993#END:command_queue
1413 1994
1414MODULE = OpenCL PACKAGE = OpenCL::Memory 1995MODULE = OpenCL PACKAGE = OpenCL::Memory
1415 1996
1430 PPCODE: 2011 PPCODE:
1431 cl_mem_object_type value [1]; 2012 cl_mem_object_type value [1];
1432 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2013 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1433 EXTEND (SP, 1); 2014 EXTEND (SP, 1);
1434 const int i = 0; 2015 const int i = 0;
1435 PUSHs (sv_2mortal (newSViv (value [i]))); 2016 PUSHs (sv_2mortal (newSVuv (value [i])));
1436 2017
1437void 2018void
1438flags (OpenCL::Memory self) 2019flags (OpenCL::Memory self)
1439 PPCODE: 2020 PPCODE:
1440 cl_mem_flags value [1]; 2021 cl_mem_flags value [1];
1441 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2022 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1442 EXTEND (SP, 1); 2023 EXTEND (SP, 1);
1443 const int i = 0; 2024 const int i = 0;
1444 PUSHs (sv_2mortal (newSViv (value [i]))); 2025 PUSHs (sv_2mortal (newSVuv (value [i])));
1445 2026
1446void 2027void
1447size (OpenCL::Memory self) 2028size (OpenCL::Memory self)
1448 ALIAS: 2029 ALIAS:
1449 size = CL_MEM_SIZE 2030 size = CL_MEM_SIZE
1483 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2064 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1484 EXTEND (SP, 1); 2065 EXTEND (SP, 1);
1485 const int i = 0; 2066 const int i = 0;
1486 { 2067 {
1487 NEED_SUCCESS (RetainContext, (value [i])); 2068 NEED_SUCCESS (RetainContext, (value [i]));
1488 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2069 PUSH_CLOBJ (stash_context, value [i]);
1489 } 2070 }
1490 2071
1491void 2072void
1492associated_memobject (OpenCL::Memory self) 2073associated_memobject (OpenCL::Memory self)
1493 PPCODE: 2074 PPCODE:
1495 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2076 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1496 EXTEND (SP, 1); 2077 EXTEND (SP, 1);
1497 const int i = 0; 2078 const int i = 0;
1498 { 2079 {
1499 NEED_SUCCESS (RetainMemObject, (value [i])); 2080 NEED_SUCCESS (RetainMemObject, (value [i]));
1500 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2081 PUSH_CLOBJ (stash_memory, value [i]);
1501 } 2082 }
1502 2083
1503#END:mem 2084#END:mem
1504 2085
1505#if cl_apple_gl_sharing || cl_khr_gl_sharing 2086#if cl_apple_gl_sharing || cl_khr_gl_sharing
1525 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2106 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1526 2107
1527 cl_buffer_region crdata = { origin, size }; 2108 cl_buffer_region crdata = { origin, size };
1528 2109
1529 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2110 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1530 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2111 XPUSH_CLOBJ (stash_buffer, mem);
1531 2112
1532MODULE = OpenCL PACKAGE = OpenCL::Image 2113MODULE = OpenCL PACKAGE = OpenCL::Image
1533 2114
1534void 2115void
1535image_info (OpenCL::Image self, cl_image_info name) 2116image_info (OpenCL::Image self, cl_image_info name)
1536 PPCODE: 2117 PPCODE:
1537 INFO (Image) 2118 INFO (Image)
2119
2120void
2121format (OpenCL::Image self)
2122 PPCODE:
2123 cl_image_format format;
2124 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2125 EXTEND (SP, 2);
2126 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2127 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1538 2128
1539#BEGIN:image 2129#BEGIN:image
1540 2130
1541void 2131void
1542element_size (OpenCL::Image self) 2132element_size (OpenCL::Image self)
1612 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2202 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1613 EXTEND (SP, 1); 2203 EXTEND (SP, 1);
1614 const int i = 0; 2204 const int i = 0;
1615 { 2205 {
1616 NEED_SUCCESS (RetainContext, (value [i])); 2206 NEED_SUCCESS (RetainContext, (value [i]));
1617 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2207 PUSH_CLOBJ (stash_context, value [i]);
1618 } 2208 }
1619 2209
1620void 2210void
1621normalized_coords (OpenCL::Sampler self) 2211normalized_coords (OpenCL::Sampler self)
1622 PPCODE: 2212 PPCODE:
1623 cl_addressing_mode value [1]; 2213 cl_addressing_mode value [1];
1624 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2214 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1625 EXTEND (SP, 1); 2215 EXTEND (SP, 1);
1626 const int i = 0; 2216 const int i = 0;
1627 PUSHs (sv_2mortal (newSViv (value [i]))); 2217 PUSHs (sv_2mortal (newSVuv (value [i])));
1628 2218
1629void 2219void
1630addressing_mode (OpenCL::Sampler self) 2220addressing_mode (OpenCL::Sampler self)
1631 PPCODE: 2221 PPCODE:
1632 cl_filter_mode value [1]; 2222 cl_filter_mode value [1];
1633 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2223 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1634 EXTEND (SP, 1); 2224 EXTEND (SP, 1);
1635 const int i = 0; 2225 const int i = 0;
1636 PUSHs (sv_2mortal (newSViv (value [i]))); 2226 PUSHs (sv_2mortal (newSVuv (value [i])));
1637 2227
1638void 2228void
1639filter_mode (OpenCL::Sampler self) 2229filter_mode (OpenCL::Sampler self)
1640 PPCODE: 2230 PPCODE:
1641 cl_bool value [1]; 2231 cl_bool value [1];
1652DESTROY (OpenCL::Program self) 2242DESTROY (OpenCL::Program self)
1653 CODE: 2243 CODE:
1654 clReleaseProgram (self); 2244 clReleaseProgram (self);
1655 2245
1656void 2246void
1657build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2247build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2248 ALIAS:
2249 build_async = 1
1658 CODE: 2250 CODE:
2251 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2252 void *user_data = 0;
2253 cl_uint num_devices = 0;
2254 cl_device_id *device_list = 0;
2255
2256 if (SvOK (devices))
2257 {
2258 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2259 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2260
2261 AV *av = (AV *)SvRV (devices);
2262 num_devices = av_len (av) + 1;
2263
2264 if (num_devices)
2265 {
2266 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2267 int count;
2268 for (count = 0; count < num_devices; ++count)
2269 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2270 }
2271 }
2272
2273 if (SvOK (notify))
2274 {
2275 NEED_SUCCESS (RetainProgram, (self));
2276 pfn_notify = eq_program_notify;
2277 user_data = SvREFCNT_inc (s_get_cv (notify));
2278 }
2279
2280 if (ix)
2281 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2282 else
1659 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2283 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1660 2284
1661void 2285void
1662build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2286build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1663 PPCODE: 2287 PPCODE:
1664 size_t size; 2288 size_t size;
1699 2323
1700void 2324void
1701kernel (OpenCL::Program program, SV *function) 2325kernel (OpenCL::Program program, SV *function)
1702 PPCODE: 2326 PPCODE:
1703 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2327 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1704 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2328 XPUSH_CLOBJ (stash_kernel, kernel);
1705 2329
1706void 2330void
1707kernels_in_program (OpenCL::Program program) 2331kernels_in_program (OpenCL::Program program)
1708 PPCODE: 2332 PPCODE:
1709 cl_uint num_kernels; 2333 cl_uint num_kernels;
1712 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2336 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
1713 2337
1714 int i; 2338 int i;
1715 EXTEND (SP, num_kernels); 2339 EXTEND (SP, num_kernels);
1716 for (i = 0; i < num_kernels; ++i) 2340 for (i = 0; i < num_kernels; ++i)
1717 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Kernel", kernels [i])); 2341 PUSH_CLOBJ (stash_kernel, kernels [i]);
1718 2342
1719void 2343void
1720info (OpenCL::Program self, cl_program_info name) 2344info (OpenCL::Program self, cl_program_info name)
1721 PPCODE: 2345 PPCODE:
1722 INFO (Program) 2346 INFO (Program)
1770 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2394 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1771 EXTEND (SP, 1); 2395 EXTEND (SP, 1);
1772 const int i = 0; 2396 const int i = 0;
1773 { 2397 {
1774 NEED_SUCCESS (RetainContext, (value [i])); 2398 NEED_SUCCESS (RetainContext, (value [i]));
1775 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2399 PUSH_CLOBJ (stash_context, value [i]);
1776 } 2400 }
1777 2401
1778void 2402void
1779devices (OpenCL::Program self) 2403devices (OpenCL::Program self)
1780 PPCODE: 2404 PPCODE:
1784 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2408 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1785 int i, n = size / sizeof (*value); 2409 int i, n = size / sizeof (*value);
1786 EXTEND (SP, n); 2410 EXTEND (SP, n);
1787 for (i = 0; i < n; ++i) 2411 for (i = 0; i < n; ++i)
1788 { 2412 {
1789 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2413 PUSH_CLOBJ (stash_device, value [i]);
1790 } 2414 }
1791 2415
1792void 2416void
1793source (OpenCL::Program self) 2417source (OpenCL::Program self)
1794 PPCODE: 2418 PPCODE:
1820DESTROY (OpenCL::Kernel self) 2444DESTROY (OpenCL::Kernel self)
1821 CODE: 2445 CODE:
1822 clReleaseKernel (self); 2446 clReleaseKernel (self);
1823 2447
1824void 2448void
2449setf (OpenCL::Kernel self, const char *format, ...)
2450 CODE:
2451 int i;
2452 for (i = 2; ; ++i)
2453 {
2454 while (*format == ' ')
2455 ++format;
2456
2457 char type = *format++;
2458
2459 if (!type)
2460 break;
2461
2462 if (i >= items)
2463 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2464
2465 SV *sv = ST (i);
2466
2467 union
2468 {
2469 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2470 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2471 cl_half ch; cl_float cf; cl_double cd;
2472 cl_mem cm;
2473 cl_sampler ca;
2474 size_t cz;
2475 cl_event ce;
2476 } arg;
2477 size_t size;
2478 int nullarg = 0;
2479
2480 switch (type)
2481 {
2482 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2483 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2484 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2485 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2486 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2487 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2488 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2489 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2490
2491 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2492 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2493 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2494
2495 case 'z': nullarg = 1; size = SvIV (sv); break;
2496
2497 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2498 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2499 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2500
2501 default:
2502 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2503 }
2504
2505 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2506 if (res)
2507 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2508 }
2509
2510 if (i != items)
2511 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2512
2513void
1825set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2514set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1826 CODE: 2515 CODE:
1827 clSetKernelArg (self, idx, sizeof (value), &value); 2516 clSetKernelArg (self, idx, sizeof (value), &value);
1828 2517
1829void 2518void
1877 clSetKernelArg (self, idx, sizeof (value), &value); 2566 clSetKernelArg (self, idx, sizeof (value), &value);
1878 2567
1879void 2568void
1880set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2569set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1881 CODE: 2570 CODE:
1882 clSetKernelArg (self, idx, sizeof (value), &value); 2571 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1883 2572
1884void 2573void
1885set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2574set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1886 CODE: 2575 CODE:
1887 clSetKernelArg (self, idx, sizeof (value), &value); 2576 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1888 2577
1889void 2578void
1890set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2579set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2580 ALIAS:
2581 set_image2d = 0
2582 set_image3d = 0
1891 CODE: 2583 CODE:
1892 clSetKernelArg (self, idx, sizeof (value), &value); 2584 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1893
1894void
1895set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1896 CODE:
1897 clSetKernelArg (self, idx, sizeof (value), &value);
1898 2585
1899void 2586void
1900set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2587set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1901 CODE: 2588 CODE:
1902 clSetKernelArg (self, idx, sizeof (value), &value); 2589 clSetKernelArg (self, idx, sizeof (value), &value);
1948 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2635 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1949 EXTEND (SP, 1); 2636 EXTEND (SP, 1);
1950 const int i = 0; 2637 const int i = 0;
1951 { 2638 {
1952 NEED_SUCCESS (RetainContext, (value [i])); 2639 NEED_SUCCESS (RetainContext, (value [i]));
1953 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2640 PUSH_CLOBJ (stash_context, value [i]);
1954 } 2641 }
1955 2642
1956void 2643void
1957program (OpenCL::Kernel self) 2644program (OpenCL::Kernel self)
1958 PPCODE: 2645 PPCODE:
1960 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2647 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1961 EXTEND (SP, 1); 2648 EXTEND (SP, 1);
1962 const int i = 0; 2649 const int i = 0;
1963 { 2650 {
1964 NEED_SUCCESS (RetainProgram, (value [i])); 2651 NEED_SUCCESS (RetainProgram, (value [i]));
1965 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2652 PUSH_CLOBJ (stash_program, value [i]);
1966 } 2653 }
1967 2654
1968#END:kernel 2655#END:kernel
1969 2656
1970void 2657void
2030wait (OpenCL::Event self) 2717wait (OpenCL::Event self)
2031 CODE: 2718 CODE:
2032 clWaitForEvents (1, &self); 2719 clWaitForEvents (1, &self);
2033 2720
2034void 2721void
2722cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2723 CODE:
2724 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2725
2726void
2035info (OpenCL::Event self, cl_event_info name) 2727info (OpenCL::Event self, cl_event_info name)
2036 PPCODE: 2728 PPCODE:
2037 INFO (Event) 2729 INFO (Event)
2038 2730
2039#BEGIN:event 2731#BEGIN:event
2045 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2737 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2046 EXTEND (SP, 1); 2738 EXTEND (SP, 1);
2047 const int i = 0; 2739 const int i = 0;
2048 { 2740 {
2049 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2741 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2050 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2742 PUSH_CLOBJ (stash_queue, value [i]);
2051 } 2743 }
2052 2744
2053void 2745void
2054command_type (OpenCL::Event self) 2746command_type (OpenCL::Event self)
2055 PPCODE: 2747 PPCODE:
2078 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2770 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2079 EXTEND (SP, 1); 2771 EXTEND (SP, 1);
2080 const int i = 0; 2772 const int i = 0;
2081 { 2773 {
2082 NEED_SUCCESS (RetainContext, (value [i])); 2774 NEED_SUCCESS (RetainContext, (value [i]));
2083 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2775 PUSH_CLOBJ (stash_context, value [i]);
2084 } 2776 }
2085 2777
2086#END:event 2778#END:event
2087 2779
2088void 2780void
2113void 2805void
2114set_status (OpenCL::UserEvent self, cl_int execution_status) 2806set_status (OpenCL::UserEvent self, cl_int execution_status)
2115 CODE: 2807 CODE:
2116 clSetUserEventStatus (self, execution_status); 2808 clSetUserEventStatus (self, execution_status);
2117 2809
2810MODULE = OpenCL PACKAGE = OpenCL::Mapped
2811
2812void
2813DESTROY (SV *self)
2814 CODE:
2815 OpenCL__Mapped mapped = SvMAPPED (self);
2816
2817 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2818 mapped_detach (self, mapped);
2819
2820 clReleaseCommandQueue (mapped->queue);
2821 clReleaseEvent (mapped->event);
2822 Safefree (mapped);
2823
2824void
2825unmap (OpenCL::Mapped self, ...)
2826 CODE:
2827 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2828
2829bool
2830mapped (OpenCL::Mapped self)
2831 CODE:
2832 RETVAL = !!self->ptr;
2833 OUTPUT:
2834 RETVAL
2835
2836void
2837wait (OpenCL::Mapped self)
2838 PPCODE:
2839 if (self->event)
2840 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2841
2842void
2843event (OpenCL::Mapped self)
2844 PPCODE:
2845 if (!self->event)
2846 XSRETURN_UNDEF;
2847
2848 clRetainEvent (self->event);
2849 XPUSH_CLOBJ (stash_event, self->event);
2850
2851size_t
2852size (OpenCL::Mapped self)
2853 CODE:
2854 RETVAL = self->cb;
2855 OUTPUT:
2856 RETVAL
2857
2858IV
2859ptr (OpenCL::Mapped self)
2860 CODE:
2861 RETVAL = PTR2IV (self->ptr);
2862 OUTPUT:
2863 RETVAL
2864
2865void
2866set (OpenCL::Mapped self, size_t offset, SV *data)
2867 CODE:
2868 STRLEN len;
2869 const char *ptr = SvPVbyte (data, len);
2870
2871 if (offset + len > self->cb)
2872 croak ("OpenCL::Mapped::set out of bound condition detected");
2873
2874 memcpy (offset + (char *)self->ptr, ptr, len);
2875
2876MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2877
2878MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2879
2880IV
2881row_pitch (OpenCL::Mapped self)
2882 ALIAS:
2883 slice_pitch = 1
2884 CODE:
2885 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2886 OUTPUT:
2887 RETVAL
2888
2889
2890

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines