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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.39 by root, Sat Apr 21 19:48:58 2012 UTC vs.
Revision 1.61 by root, Tue May 1 16:37:23 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h"
7#include "schmorp.h"
4 8
5#ifdef I_DLFCN 9#ifdef I_DLFCN
6 #include <dlfcn.h> 10 #include <dlfcn.h>
7#endif 11#endif
8 12
11// this. This breaks 100% of the opencl 1.1 apps, for what reason? 15// this. This breaks 100% of the opencl 1.1 apps, for what reason?
12// after all, the functions are deprecated, not removed. 16// after all, the functions are deprecated, not removed.
13// in addition, you cannot test for this in any future-proof way. 17// in addition, you cannot test for this in any future-proof way.
14// each time a new opencl version comes out, you need to make a new 18// each time a new opencl version comes out, you need to make a new
15// release. 19// release.
16#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
17#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */ 20#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
21
22#ifndef PREFER_1_1
23 #define PREFER_1_1 1
24#endif
25
26#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif
18 29
19#ifdef __APPLE__ 30#ifdef __APPLE__
20 #include <OpenCL/opencl.h> 31 #include <OpenCL/opencl.h>
21#else 32#else
22 #include <CL/opencl.h> 33 #include <CL/opencl.h>
23#endif 34#endif
24 35
25#if 0 // testing
26#undef CL_USE_DEPRECATED_OPENCL_1_1_APIS
27#undef CL_VERSION_1_2 36#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1
38 #define PREFER_1_1 1
28#endif 39#endif
29 40
30typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
31typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
32typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
46typedef cl_program OpenCL__Program; 57typedef cl_program OpenCL__Program;
47typedef cl_kernel OpenCL__Kernel; 58typedef cl_kernel OpenCL__Kernel;
48typedef cl_event OpenCL__Event; 59typedef cl_event OpenCL__Event;
49typedef cl_event OpenCL__UserEvent; 60typedef cl_event OpenCL__UserEvent;
50 61
62typedef struct mapped * OpenCL__Mapped;
63
51typedef SV *FUTURE; 64typedef SV *FUTURE;
65
66static HV
67 *stash_platform,
68 *stash_device,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
52 89
53/*****************************************************************************/ 90/*****************************************************************************/
54 91
55// name must include a leading underscore 92// name must include a leading underscore
56// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
182 if (res) \ 219 if (res) \
183 FAIL (name); 220 FAIL (name);
184 221
185/*****************************************************************************/ 222/*****************************************************************************/
186 223
187#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
188#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
189
190static void *
191SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
192{
193 if (SvROK (sv) && sv_derived_from (sv, pkg))
194 return (void *)SvIV (SvRV (sv));
195
196 croak ("%s: %s is not of type %s", func, svname, pkg);
197}
198
199/*****************************************************************************/
200
201static cl_context_properties * 224static cl_context_properties *
202SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
203{ 226{
204 if (!sv || !SvOK (sv)) 227 if (!sv || !SvOK (sv))
205 if (extra) 228 if (extra)
263 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
264} 287}
265 288
266/*****************************************************************************/ 289/*****************************************************************************/
267 290
291static SV *
292new_clobj (HV *stash, IV id)
293{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
295}
296
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
299
300/* cl objects are either \$iv, or [$iv, ...] */
301/* they can be upgraded at runtime to the array form */
302static void *
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
304{
305 if (SvROK (sv) && sv_derived_from (sv, pkg))
306 return (void *)SvIV (SvRV (sv));
307
308 croak ("%s: %s is not of type %s", func, svname, pkg);
309}
310
311/*****************************************************************************/
312/* callback stuff */
313
314/* default context callback, log to stderr */
315static void CL_CALLBACK
316context_default_notify (const char *msg, const void *info, size_t cb, void *data)
317{
318 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
319}
320
321typedef struct
322{
323 int free_cb;
324 void (*push)(void *data1, void *data2, void *data3);
325} eq_vtbl;
326
327typedef struct eq_item
328{
329 struct eq_item *next;
330 eq_vtbl *vtbl;
331 SV *cb;
332 void *data1, *data2, *data3;
333} eq_item;
334
335static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail;
339
340static void
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{
343 eq_item *item = malloc (sizeof (eq_item));
344
345 item->next = 0;
346 item->vtbl = vtbl;
347 item->cb = cb;
348 item->data1 = data1;
349 item->data2 = data2;
350 item->data3 = data3;
351
352 X_LOCK (eq_lock);
353
354 *(eq_head ? &eq_tail->next : &eq_head) = item;
355 eq_tail = item;
356
357 X_UNLOCK (eq_lock);
358
359 eq_signal_func (eq_signal_arg, 0);
360}
361
362static eq_item *
363eq_dec (void)
364{
365 eq_item *res;
366
367 X_LOCK (eq_lock);
368
369 res = eq_head;
370 if (res)
371 eq_head = res->next;
372
373 X_UNLOCK (eq_lock);
374
375 return res;
376}
377
378static void
379eq_poll (void)
380{
381 eq_item *item;
382
383 while ((item = eq_dec ()))
384 {
385 ENTER;
386 SAVETMPS;
387
388 dSP;
389 PUSHMARK (SP);
390 EXTEND (SP, 2);
391
392 if (item->vtbl->free_cb)
393 sv_2mortal (item->cb);
394
395 PUTBACK;
396 item->vtbl->push (item->data1, item->data2, item->data3);
397
398 SV *cb = item->cb;
399 free (item);
400
401 call_sv (cb, G_DISCARD | G_VOID);
402
403 FREETMPS;
404 LEAVE;
405 }
406}
407
408static void
409eq_poll_interrupt (pTHX_ void *c_arg, int value)
410{
411 eq_poll ();
412}
413
414/*****************************************************************************/
415/* context notify */
416
417static void
418eq_context_push (void *data1, void *data2, void *data3)
419{
420 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
423 PUTBACK;
424
425 free (data1);
426 free (data2);
427}
428
429static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
430
431static void CL_CALLBACK
432eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
433{
434 void *pvt_copy = malloc (cb);
435 memcpy (pvt_copy, pvt, cb);
436 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
437}
438
439#define CONTEXT_NOTIFY_CALLBACK \
440 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
441 void *user_data = 0; \
442 \
443 if (SvOK (notify)) \
444 { \
445 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \
447 }
448
449static SV *
450new_clobj_context (cl_context ctx, void *user_data)
451{
452 SV *sv = new_clobj (stash_context, (IV)ctx);
453
454 if (user_data)
455 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
456
457 return sv;
458}
459
460#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
461
462/*****************************************************************************/
463/* build/compile/link notify */
464
465static void
466eq_program_push (void *data1, void *data2, void *data3)
467{
468 dSP;
469 PUSH_CLOBJ (stash_program, data1);
470 PUTBACK;
471}
472
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474
475static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data)
477{
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
479}
480
481struct build_args
482{
483 cl_program program;
484 char *options;
485 void *user_data;
486 cl_uint num_devices;
487};
488
489X_THREAD_PROC (build_program_thread)
490{
491 struct build_args *arg = thr_arg;
492
493 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
494
495 if (arg->user_data)
496 eq_program_notify (arg->program, arg->user_data);
497 else
498 clReleaseProgram (arg->program);
499
500 free (arg->options);
501 free (arg);
502}
503
504static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{
507 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
508
509 arg->program = program;
510 arg->options = strdup (options);
511 arg->user_data = user_data;
512 arg->num_devices = num_devices;
513 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
514
515 xthread_t id;
516 thread_create (&id, build_program_thread, arg);
517}
518
519/*****************************************************************************/
520/* event objects */
521
522static void
523eq_event_push (void *data1, void *data2, void *data3)
524{
525 dSP;
526 PUSH_CLOBJ (stash_event, data1);
527 PUSHs (sv_2mortal (newSViv ((IV)data2)));
528 PUTBACK;
529}
530
531static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
532
533static void CL_CALLBACK
534eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
535{
536 clRetainEvent (event);
537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
538}
539
540/*****************************************************************************/
541/* mapped_xxx */
542
543static OpenCL__Mapped
544SvMAPPED (SV *sv)
545{
546 // no typechecking atm., keep your fingers crossed
547 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
548}
549
550struct mapped
551{
552 cl_command_queue queue;
553 cl_mem memobj;
554 void *ptr;
555 size_t cb;
556 cl_event event;
557 size_t row_pitch;
558 size_t slice_pitch;
559};
560
561static SV *
562mapped_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)
563{
564 SV *data = newSV (0);
565 SvUPGRADE (data, SVt_PVMG);
566
567 OpenCL__Mapped mapped;
568 New (0, mapped, 1, struct mapped);
569
570 clRetainCommandQueue (queue);
571
572 mapped->queue = queue;
573 mapped->memobj = memobj;
574 mapped->ptr = ptr;
575 mapped->cb = cb;
576 mapped->event = ev;
577 mapped->row_pitch = row_pitch;
578 mapped->slice_pitch = slice_pitch;
579
580 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
581
582 if (!(flags & CL_MAP_WRITE))
583 SvREADONLY_on (data);
584
585 if (SvLEN (data))
586 Safefree (data);
587
588 SvPVX (data) = (char *)ptr;
589 SvCUR_set (data, cb);
590 SvLEN_set (data, 0);
591 SvPOK_only (data);
592
593 return sv_2mortal (sv_bless (newRV_noinc (data), stash));
594}
595
596static void
597mapped_detach (SV *sv, OpenCL__Mapped mapped)
598{
599 SV *data = SvRV (sv);
600
601 if (SvPVX (data) != (char *)mapped->ptr)
602 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
603 else
604 {
605 SvREADONLY_off (data);
606 SvCUR_set (data, 0);
607 SvPVX (data) = 0;
608 SvOK_off (data);
609 }
610
611 mapped->ptr = 0;
612}
613
614/*****************************************************************************/
615
268static size_t 616static size_t
269img_row_pitch (cl_mem img) 617img_row_pitch (cl_mem img)
270{ 618{
271 size_t res; 619 size_t res;
272 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 620 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
273 return res; 621 return res;
274} 622}
275 623
276static cl_event * 624static cl_event *
277event_list (SV **items, cl_uint *rcount) 625event_list (SV **items, cl_uint *rcount, cl_event extra)
278{ 626{
279 cl_uint count = *rcount; 627 cl_uint count = *rcount;
280 628
281 if (!count) 629 if (!count && !extra)
282 return 0; 630 return 0;
283 631
284 cl_event *list = tmpbuf (sizeof (cl_event) * count); 632 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
285 int i = 0; 633 int i = 0;
286 634
287 do 635 while (count)
288 { 636 {
289 --count; 637 --count;
290 if (SvOK (items [count])) 638 if (SvOK (items [count]))
291 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 639 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
292 } 640 }
293 while (count); 641
642 if (extra)
643 list [i++] = extra;
294 644
295 *rcount = i; 645 *rcount = i;
296 646
297 return i ? list : 0; 647 return i ? list : 0;
298} 648}
299 649
300#define EVENT_LIST(items,count) \ 650#define EVENT_LIST(skip) \
301 cl_uint event_list_count = (count); \ 651 cl_uint event_list_count = items - skip; \
302 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 652 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
303 653
304#define INFO(class) \ 654#define INFO(class) \
305{ \ 655{ \
306 size_t size; \ 656 size_t size; \
307 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 657 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
314} 664}
315 665
316MODULE = OpenCL PACKAGE = OpenCL 666MODULE = OpenCL PACKAGE = OpenCL
317 667
318PROTOTYPES: ENABLE 668PROTOTYPES: ENABLE
669
670void
671poll ()
672 CODE:
673 eq_poll ();
674
675void
676_eq_initialise (IV func, IV arg)
677 CODE:
678 eq_signal_func = (void (*)(void *, int))func;
679 eq_signal_arg = (void*)arg;
319 680
320BOOT: 681BOOT:
321{ 682{
322 HV *stash = gv_stashpv ("OpenCL", 1); 683 HV *stash = gv_stashpv ("OpenCL", 1);
323 static const ivstr *civ, const_iv[] = { 684 static const ivstr *civ, const_iv[] = {
332 { sizeof (cl_half ), "SIZEOF_HALF" }, 693 { sizeof (cl_half ), "SIZEOF_HALF" },
333 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 694 { sizeof (cl_float ), "SIZEOF_FLOAT" },
334 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 695 { sizeof (cl_double), "SIZEOF_DOUBLE" },
335#include "constiv.h" 696#include "constiv.h"
336 }; 697 };
698
337 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 699 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
338 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 700 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
701
702 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
703 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
704 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
705 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
706 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
707 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
708 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
709 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
710 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
711 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
712 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
713 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
714 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
715 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
716 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
717 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
718 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
719 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
720 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
721 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
722 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
723 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
724
725 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
339} 726}
340 727
341cl_int 728cl_int
342errno () 729errno ()
343 CODE: 730 CODE:
344 RETVAL = res; 731 RETVAL = res;
345 OUTPUT: 732 OUTPUT:
346 RETVAL 733 RETVAL
347 734
348const char * 735const char *
349err2str (cl_int err) 736err2str (cl_int err = res)
350 737
351const char * 738const char *
352enum2str (cl_uint value) 739enum2str (cl_uint value)
353 740
354void 741void
362 list = tmpbuf (sizeof (*list) * count); 749 list = tmpbuf (sizeof (*list) * count);
363 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 750 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
364 751
365 EXTEND (SP, count); 752 EXTEND (SP, count);
366 for (i = 0; i < count; ++i) 753 for (i = 0; i < count; ++i)
367 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 754 PUSH_CLOBJ (stash_platform, (IV)list [i]);
368 755
369void 756void
370context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 757context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
371 PPCODE: 758 PPCODE:
759 CONTEXT_NOTIFY_CALLBACK;
372 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 760 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
373 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 761 XPUSH_CLOBJ_CONTEXT;
374 762
375#if 0
376
377void 763void
378context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 764context (FUTURE properties, FUTURE devices, FUTURE notify)
379 PPCODE: 765 PPCODE:
380 /* der Gipfel der Kunst */ 766 /* der Gipfel der Kunst */
381 767
768void
769wait_for_events (...)
770 CODE:
771 EVENT_LIST (0);
772 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
773
774PROTOTYPES: DISABLE
775
776MODULE = OpenCL PACKAGE = OpenCL::Platform
777
778void
779info (OpenCL::Platform self, cl_platform_info name)
780 PPCODE:
781 INFO (Platform)
782
783void
784unload_compiler (OpenCL::Platform self)
785 CODE:
786#if CL_VERSION_1_2
787 clUnloadPlatformCompiler (self);
382#endif 788#endif
383
384void
385wait_for_events (...)
386 CODE:
387 EVENT_LIST (0, items);
388 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
389
390PROTOTYPES: DISABLE
391
392MODULE = OpenCL PACKAGE = OpenCL::Platform
393
394void
395info (OpenCL::Platform self, cl_platform_info name)
396 PPCODE:
397 INFO (Platform)
398 789
399#BEGIN:platform 790#BEGIN:platform
400 791
401void 792void
402profile (OpenCL::Platform self) 793profile (OpenCL::Platform self)
428 list = tmpbuf (sizeof (*list) * count); 819 list = tmpbuf (sizeof (*list) * count);
429 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 820 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
430 821
431 EXTEND (SP, count); 822 EXTEND (SP, count);
432 for (i = 0; i < count; ++i) 823 for (i = 0; i < count; ++i)
433 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 824 PUSH_CLOBJ (stash_device, list [i]);
434 825
435void 826void
436context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 827context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
437 PPCODE: 828 PPCODE:
438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 829 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
439 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 830 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
440 831
441 AV *av = (AV *)SvRV (devices); 832 AV *av = (AV *)SvRV (devices);
442 cl_uint num_devices = av_len (av) + 1; 833 cl_uint num_devices = av_len (av) + 1;
443 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 834 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
835
444 int i; 836 int i;
445
446 for (i = num_devices; i--; ) 837 for (i = num_devices; i--; )
447 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 838 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
448 839
840 CONTEXT_NOTIFY_CALLBACK;
449 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 841 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
450 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 842 XPUSH_CLOBJ_CONTEXT;
451 843
452void 844void
453context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 845context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
454 PPCODE: 846 PPCODE:
455 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 847 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
456 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 848 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
849
850 CONTEXT_NOTIFY_CALLBACK;
457 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 851 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
458 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 852 XPUSH_CLOBJ_CONTEXT;
459 853
460MODULE = OpenCL PACKAGE = OpenCL::Device 854MODULE = OpenCL PACKAGE = OpenCL::Device
461 855
462void 856void
463info (OpenCL::Device self, cl_device_info name) 857info (OpenCL::Device self, cl_device_info name)
471 PPCODE: 865 PPCODE:
472 cl_device_type value [1]; 866 cl_device_type value [1];
473 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 867 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
474 EXTEND (SP, 1); 868 EXTEND (SP, 1);
475 const int i = 0; 869 const int i = 0;
476 PUSHs (sv_2mortal (newSViv (value [i]))); 870 PUSHs (sv_2mortal (newSVuv (value [i])));
477 871
478void 872void
479vendor_id (OpenCL::Device self) 873vendor_id (OpenCL::Device self)
480 ALIAS: 874 ALIAS:
481 vendor_id = CL_DEVICE_VENDOR_ID 875 vendor_id = CL_DEVICE_VENDOR_ID
502 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 896 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
503 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 897 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
504 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 898 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
505 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 899 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
506 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 900 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
507 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 901 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
508 PPCODE: 902 PPCODE:
509 cl_uint value [1]; 903 cl_uint value [1];
510 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 904 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
511 EXTEND (SP, 1); 905 EXTEND (SP, 1);
512 const int i = 0; 906 const int i = 0;
626 PPCODE: 1020 PPCODE:
627 cl_command_queue_properties value [1]; 1021 cl_command_queue_properties value [1];
628 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1022 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
629 EXTEND (SP, 1); 1023 EXTEND (SP, 1);
630 const int i = 0; 1024 const int i = 0;
631 PUSHs (sv_2mortal (newSViv (value [i]))); 1025 PUSHs (sv_2mortal (newSVuv (value [i])));
632 1026
633void 1027void
634platform (OpenCL::Device self) 1028platform (OpenCL::Device self)
635 PPCODE: 1029 PPCODE:
636 cl_platform_id value [1]; 1030 cl_platform_id value [1];
637 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1031 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
638 EXTEND (SP, 1); 1032 EXTEND (SP, 1);
639 const int i = 0; 1033 const int i = 0;
640 { 1034 {
641 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1035 PUSH_CLOBJ (stash_platform, value [i]);
642 } 1036 }
643 1037
644void 1038void
645name (OpenCL::Device self) 1039name (OpenCL::Device self)
646 ALIAS: 1040 ALIAS:
665 cl_device_id value [1]; 1059 cl_device_id value [1];
666 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1060 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
667 EXTEND (SP, 1); 1061 EXTEND (SP, 1);
668 const int i = 0; 1062 const int i = 0;
669 { 1063 {
670 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1064 PUSH_CLOBJ (stash_device, value [i]);
671 } 1065 }
672 1066
673void 1067void
674partition_types_ext (OpenCL::Device self) 1068partition_types_ext (OpenCL::Device self)
675 ALIAS: 1069 ALIAS:
702 1096
703void 1097void
704queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1098queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
705 PPCODE: 1099 PPCODE:
706 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1100 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
707 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1101 XPUSH_CLOBJ (stash_queue, queue);
708 1102
709void 1103void
710user_event (OpenCL::Context self) 1104user_event (OpenCL::Context self)
711 PPCODE: 1105 PPCODE:
712 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1106 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
713 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1107 XPUSH_CLOBJ (stash_userevent, ev);
714 1108
715void 1109void
716buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1110buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
717 PPCODE: 1111 PPCODE:
718 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1112 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
719 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1113 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
720 1114
721 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1115 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
722 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1116 XPUSH_CLOBJ (stash_bufferobj, mem);
723 1117
724void 1118void
725buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1119buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
726 PPCODE: 1120 PPCODE:
727 STRLEN len; 1121 STRLEN len;
728 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1122 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
729 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1123 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
730 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1124 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
731 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1125 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
732 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1126 XPUSH_CLOBJ (stash_bufferobj, mem);
733 1127
734#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS 1128#if CL_VERSION_1_2
735 1129
736void 1130void
737image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1131image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
738 PPCODE: 1132 PPCODE:
739 STRLEN len; 1133 STRLEN len;
740 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1134 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
741 const cl_image_format format = { channel_order, channel_type }; 1135 const cl_image_format format = { channel_order, channel_type };
1136 const cl_image_desc desc = {
1137 type,
1138 width, height, depth,
1139 array_size, row_pitch, slice_pitch,
1140 num_mip_level, num_samples,
1141 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1142 };
742 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1143 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
743 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1144 HV *stash = stash_image;
1145 switch (type)
1146 {
1147 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1148 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1149 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1150 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1151 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1152 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1153 }
1154 XPUSH_CLOBJ (stash, mem);
744 1155
1156#endif
1157
745void 1158void
746image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1159image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
747 PPCODE: 1160 PPCODE:
748 STRLEN len; 1161 STRLEN len;
749 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1162 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
750 const cl_image_format format = { channel_order, channel_type }; 1163 const cl_image_format format = { channel_order, channel_type };
1164#if PREFER_1_1
1165 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1166#else
1167 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1168 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1169#endif
1170 XPUSH_CLOBJ (stash_image2d, mem);
1171
1172void
1173image3d (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)
1174 PPCODE:
1175 STRLEN len;
1176 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1177 const cl_image_format format = { channel_order, channel_type };
1178#if PREFER_1_1
751 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1179 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
752 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1180#else
753 1181 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1182 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
754#endif 1183#endif
1184 XPUSH_CLOBJ (stash_image3d, mem);
755 1185
756#if cl_apple_gl_sharing || cl_khr_gl_sharing 1186#if cl_apple_gl_sharing || cl_khr_gl_sharing
757 1187
758void 1188void
759gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1189gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
760 PPCODE: 1190 PPCODE:
761 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1191 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
762 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1192 XPUSH_CLOBJ (stash_bufferobj, mem);
763 1193
764#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS 1194void
1195gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1196 PPCODE:
1197 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1198 XPUSH_CLOBJ (stash_image2d, mem);
1199
1200#if CL_VERSION_1_2
1201
1202void
1203gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1204 ALIAS:
1205 PPCODE:
1206 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1207 cl_gl_object_type type;
1208 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1209 HV *stash = stash_memory;
1210 switch (type)
1211 {
1212 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1213 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1214 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1215 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1216 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1217 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1218 }
1219 XPUSH_CLOBJ (stash, mem);
1220
1221#endif
765 1222
766void 1223void
767gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1224gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
768 PPCODE: 1225 PPCODE:
1226#if PREFER_1_1
769 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1227 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
770 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1228#else
1229 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1230#endif
1231 XPUSH_CLOBJ (stash_image2d, mem);
771 1232
772void 1233void
773gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1234gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
774 PPCODE: 1235 PPCODE:
1236#if PREFER_1_1
775 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1237 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
776 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1238#else
777 1239 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
778#endif 1240#endif
779 1241 XPUSH_CLOBJ (stash_image3d, mem);
780#if CL_VERSION_1_2
781
782void
783gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
784 PPCODE:
785 char *klass = "OpenCL::Memory";
786 cl_gl_object_type t;
787 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
788 NEED_SUCCESS (GetGLObjectInfo, (mem, &t, 0));
789 switch (t)
790 {
791 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break;
792 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break;
793 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break;
794 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break;
795 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break;
796 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break;
797 }
798 XPUSH_NEW_OBJ (klass, mem);
799
800#endif
801
802void
803gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
804 PPCODE:
805 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
806 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
807 1242
808#endif 1243#endif
809 1244
810void 1245void
811supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1246supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
831 1266
832void 1267void
833sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1268sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
834 PPCODE: 1269 PPCODE:
835 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1270 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
836 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1271 XPUSH_CLOBJ (stash_sampler, sampler);
837 1272
838void 1273void
839program_with_source (OpenCL::Context self, SV *program) 1274program_with_source (OpenCL::Context self, SV *program)
840 PPCODE: 1275 PPCODE:
841 STRLEN len; 1276 STRLEN len;
842 size_t len2; 1277 size_t len2;
843 const char *ptr = SvPVbyte (program, len); 1278 const char *ptr = SvPVbyte (program, len);
844 1279
845 len2 = len; 1280 len2 = len;
846 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1281 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
847 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1282 XPUSH_CLOBJ (stash_program, prog);
848 1283
849#BEGIN:context 1284#BEGIN:context
850 1285
851void 1286void
852reference_count (OpenCL::Context self) 1287reference_count (OpenCL::Context self)
869 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1304 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
870 int i, n = size / sizeof (*value); 1305 int i, n = size / sizeof (*value);
871 EXTEND (SP, n); 1306 EXTEND (SP, n);
872 for (i = 0; i < n; ++i) 1307 for (i = 0; i < n; ++i)
873 { 1308 {
874 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1309 PUSH_CLOBJ (stash_device, value [i]);
875 } 1310 }
876 1311
877void 1312void
878properties (OpenCL::Context self) 1313properties (OpenCL::Context self)
879 PPCODE: 1314 PPCODE:
894DESTROY (OpenCL::Queue self) 1329DESTROY (OpenCL::Queue self)
895 CODE: 1330 CODE:
896 clReleaseCommandQueue (self); 1331 clReleaseCommandQueue (self);
897 1332
898void 1333void
899enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1334read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1335 ALIAS:
1336 enqueue_read_buffer = 0
900 PPCODE: 1337 PPCODE:
901 cl_event ev = 0; 1338 cl_event ev = 0;
902 EVENT_LIST (6, items - 6); 1339 EVENT_LIST (6);
903 1340
904 SvUPGRADE (data, SVt_PV); 1341 SvUPGRADE (data, SVt_PV);
905 SvGROW (data, len); 1342 SvGROW (data, len);
906 SvPOK_only (data); 1343 SvPOK_only (data);
907 SvCUR_set (data, len); 1344 SvCUR_set (data, len);
908 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1345 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
909 1346
910 if (ev) 1347 if (ev)
911 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1348 XPUSH_CLOBJ (stash_event, ev);
912 1349
913void 1350void
914enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1351write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1352 ALIAS:
1353 enqueue_write_buffer = 0
915 PPCODE: 1354 PPCODE:
916 cl_event ev = 0; 1355 cl_event ev = 0;
917 STRLEN len; 1356 STRLEN len;
918 char *ptr = SvPVbyte (data, len); 1357 char *ptr = SvPVbyte (data, len);
919 EVENT_LIST (5, items - 5); 1358 EVENT_LIST (5);
920 1359
921 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1360 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
922 1361
923 if (ev) 1362 if (ev)
924 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1363 XPUSH_CLOBJ (stash_event, ev);
925 1364
1365#if CL_VERSION_1_2
1366
926void 1367void
927enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1368fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1369 ALIAS:
1370 enqueue_fill_buffer = 0
928 PPCODE: 1371 PPCODE:
929 cl_event ev = 0; 1372 cl_event ev = 0;
1373 STRLEN len;
1374 char *ptr = SvPVbyte (data, len);
1375 EVENT_LIST (5);
1376
1377 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1378
1379 if (ev)
1380 XPUSH_CLOBJ (stash_event, ev);
1381
1382void
1383fill_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, ...)
1384 ALIAS:
1385 enqueue_fill_image = 0
1386 PPCODE:
1387 cl_event ev = 0;
1388 STRLEN len;
1389 const size_t origin [3] = { x, y, z };
1390 const size_t region [3] = { width, height, depth };
1391 EVENT_LIST (12);
1392
1393 const cl_float c_f [4] = { r, g, b, a };
1394 const cl_uint c_u [4] = { r, g, b, a };
1395 const cl_int c_s [4] = { r, g, b, a };
1396 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1397 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1398 cl_image_format format;
1399 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1400 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1401 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1402 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1403
1404 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1405 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1406
1407 if (ev)
1408 XPUSH_CLOBJ (stash_event, ev);
1409
1410#endif
1411
1412void
1413copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1414 ALIAS:
1415 enqueue_copy_buffer = 0
1416 PPCODE:
1417 cl_event ev = 0;
930 EVENT_LIST (6, items - 6); 1418 EVENT_LIST (6);
931 1419
932 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1420 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
933 1421
934 if (ev) 1422 if (ev)
935 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1423 XPUSH_CLOBJ (stash_event, ev);
936 1424
937void 1425void
938enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1426read_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, ...)
1427 ALIAS:
1428 enqueue_read_buffer_rect = 0
939 PPCODE: 1429 PPCODE:
940 cl_event ev = 0; 1430 cl_event ev = 0;
941 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1431 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
942 const size_t host_origin[3] = { host_x, host_y, host_z }; 1432 const size_t host_origin[3] = { host_x, host_y, host_z };
943 const size_t region[3] = { width, height, depth }; 1433 const size_t region[3] = { width, height, depth };
944 EVENT_LIST (17, items - 17); 1434 EVENT_LIST (17);
945 1435
946 if (!buf_row_pitch) 1436 if (!buf_row_pitch)
947 buf_row_pitch = region [0]; 1437 buf_row_pitch = region [0];
948 1438
949 if (!buf_slice_pitch) 1439 if (!buf_slice_pitch)
962 SvPOK_only (data); 1452 SvPOK_only (data);
963 SvCUR_set (data, len); 1453 SvCUR_set (data, len);
964 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1454 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
965 1455
966 if (ev) 1456 if (ev)
967 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1457 XPUSH_CLOBJ (stash_event, ev);
968 1458
969void 1459void
970enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1460write_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, ...)
1461 ALIAS:
1462 enqueue_write_buffer_rect = 0
971 PPCODE: 1463 PPCODE:
972 cl_event ev = 0; 1464 cl_event ev = 0;
973 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1465 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
974 const size_t host_origin[3] = { host_x, host_y, host_z }; 1466 const size_t host_origin[3] = { host_x, host_y, host_z };
975 const size_t region[3] = { width, height, depth }; 1467 const size_t region[3] = { width, height, depth };
976 STRLEN len; 1468 STRLEN len;
977 char *ptr = SvPVbyte (data, len); 1469 char *ptr = SvPVbyte (data, len);
978 EVENT_LIST (17, items - 17); 1470 EVENT_LIST (17);
979 1471
980 if (!buf_row_pitch) 1472 if (!buf_row_pitch)
981 buf_row_pitch = region [0]; 1473 buf_row_pitch = region [0];
982 1474
983 if (!buf_slice_pitch) 1475 if (!buf_slice_pitch)
995 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1487 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
996 1488
997 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1489 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
998 1490
999 if (ev) 1491 if (ev)
1000 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1492 XPUSH_CLOBJ (stash_event, ev);
1001 1493
1002void 1494void
1003enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1495copy_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, ...)
1496 ALIAS:
1497 enqueue_copy_buffer_rect = 0
1004 PPCODE: 1498 PPCODE:
1005 cl_event ev = 0; 1499 cl_event ev = 0;
1006 const size_t src_origin[3] = { src_x, src_y, src_z }; 1500 const size_t src_origin[3] = { src_x, src_y, src_z };
1007 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1501 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1008 const size_t region[3] = { width, height, depth }; 1502 const size_t region[3] = { width, height, depth };
1009 EVENT_LIST (16, items - 16); 1503 EVENT_LIST (16);
1010 1504
1011 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1505 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1012 1506
1013 if (ev) 1507 if (ev)
1014 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1508 XPUSH_CLOBJ (stash_event, ev);
1015 1509
1016void 1510void
1017enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1511read_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, ...)
1512 ALIAS:
1513 enqueue_read_image = 0
1018 PPCODE: 1514 PPCODE:
1019 cl_event ev = 0; 1515 cl_event ev = 0;
1020 const size_t src_origin[3] = { src_x, src_y, src_z }; 1516 const size_t src_origin[3] = { src_x, src_y, src_z };
1021 const size_t region[3] = { width, height, depth }; 1517 const size_t region[3] = { width, height, depth };
1022 EVENT_LIST (12, items - 12); 1518 EVENT_LIST (12);
1023 1519
1024 if (!row_pitch) 1520 if (!row_pitch)
1025 row_pitch = img_row_pitch (src); 1521 row_pitch = img_row_pitch (src);
1026 1522
1027 if (depth > 1 && !slice_pitch) 1523 if (depth > 1 && !slice_pitch)
1034 SvPOK_only (data); 1530 SvPOK_only (data);
1035 SvCUR_set (data, len); 1531 SvCUR_set (data, len);
1036 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1532 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1037 1533
1038 if (ev) 1534 if (ev)
1039 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1535 XPUSH_CLOBJ (stash_event, ev);
1040 1536
1041void 1537void
1042enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1538write_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, ...)
1539 ALIAS:
1540 enqueue_write_image = 0
1043 PPCODE: 1541 PPCODE:
1044 cl_event ev = 0; 1542 cl_event ev = 0;
1045 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1543 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1046 const size_t region[3] = { width, height, depth }; 1544 const size_t region[3] = { width, height, depth };
1047 STRLEN len; 1545 STRLEN len;
1048 char *ptr = SvPVbyte (data, len); 1546 char *ptr = SvPVbyte (data, len);
1049 EVENT_LIST (12, items - 12); 1547 EVENT_LIST (12);
1050 1548
1051 if (!row_pitch) 1549 if (!row_pitch)
1052 row_pitch = img_row_pitch (dst); 1550 row_pitch = img_row_pitch (dst);
1053 1551
1054 if (depth > 1 && !slice_pitch) 1552 if (depth > 1 && !slice_pitch)
1060 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1558 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1061 1559
1062 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1560 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1063 1561
1064 if (ev) 1562 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1563 XPUSH_CLOBJ (stash_event, ev);
1066 1564
1067void 1565void
1068enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1566copy_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, ...)
1567 ALIAS:
1568 enqueue_copy_image = 0
1069 PPCODE: 1569 PPCODE:
1070 cl_event ev = 0; 1570 cl_event ev = 0;
1071 const size_t src_origin[3] = { src_x, src_y, src_z }; 1571 const size_t src_origin[3] = { src_x, src_y, src_z };
1072 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1572 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1073 const size_t region[3] = { width, height, depth }; 1573 const size_t region[3] = { width, height, depth };
1074 EVENT_LIST (12, items - 12); 1574 EVENT_LIST (12);
1075 1575
1076 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1077 1577
1078 if (ev) 1578 if (ev)
1079 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
1080 1580
1081void 1581void
1082enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1582copy_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, ...)
1583 ALIAS:
1584 enqueue_copy_image_to_buffer = 0
1083 PPCODE: 1585 PPCODE:
1084 cl_event ev = 0; 1586 cl_event ev = 0;
1085 const size_t src_origin[3] = { src_x, src_y, src_z }; 1587 const size_t src_origin[3] = { src_x, src_y, src_z };
1588 const size_t region [3] = { width, height, depth };
1589 EVENT_LIST (10);
1590
1591 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1592
1593 if (ev)
1594 XPUSH_CLOBJ (stash_event, ev);
1595
1596void
1597copy_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, ...)
1598 ALIAS:
1599 enqueue_copy_buffer_to_image = 0
1600 PPCODE:
1601 cl_event ev = 0;
1602 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1603 const size_t region [3] = { width, height, depth };
1604 EVENT_LIST (10);
1605
1606 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1607
1608 if (ev)
1609 XPUSH_CLOBJ (stash_event, ev);
1610
1611void
1612map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, size_t cb = 0, ...)
1613 ALIAS:
1614 enqueue_map_buffer = 0
1615 PPCODE:
1616 cl_event ev;
1617 EVENT_LIST (7);
1618
1619 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1620 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1621
1622void
1623map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, size_t width = 0, size_t height = 0, size_t depth = 0, ...)
1624 ALIAS:
1625 enqueue_map_image = 0
1626 PPCODE:
1627 cl_event ev;
1628 const size_t origin[3] = { x, y, z };
1086 const size_t region[3] = { width, height, depth }; 1629 const size_t region[3] = { width, height, depth };
1630 size_t row_pitch, slice_pitch;
1087 EVENT_LIST (10, items - 10); 1631 EVENT_LIST (11);
1088 1632
1633 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));
1634
1635 size_t cb = slice_pitch ? slice_pitch * region [2]
1636 : row_pitch ? row_pitch * region [1]
1637 : region [0];
1638
1639 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1640
1641void
1642unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1643 PPCODE:
1644 cl_uint event_list_count = items - 2;
1645 cl_event *event_list_ptr = event_list (&ST (2), &event_list_count, mapped->event);
1646
1647 NEED_SUCCESS (EnqueueUnmapMemObject, (self, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &mapped->event));
1648 mapped_detach (ST (1), mapped);
1649
1650 if (GIMME_V != G_VOID)
1651 XPUSH_CLOBJ (stash_event, mapped->event);
1652
1653void
1654task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1655 ALIAS:
1656 enqueue_task = 0
1657 PPCODE:
1658 cl_event ev = 0;
1659 EVENT_LIST (2);
1660
1089 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1661 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1090 1662
1091 if (ev) 1663 if (ev)
1092 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1664 XPUSH_CLOBJ (stash_event, ev);
1093 1665
1094void 1666void
1095enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1096 PPCODE:
1097 cl_event ev = 0;
1098 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1099 const size_t region[3] = { width, height, depth };
1100 EVENT_LIST (10, items - 10);
1101
1102 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1103
1104 if (ev)
1105 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1106
1107void
1108enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1109 PPCODE:
1110 cl_event ev = 0;
1111 EVENT_LIST (2, items - 2);
1112
1113 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1114
1115 if (ev)
1116 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1117
1118void
1119enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1667nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1668 ALIAS:
1669 enqueue_nd_range_kernel = 0
1120 PPCODE: 1670 PPCODE:
1121 cl_event ev = 0; 1671 cl_event ev = 0;
1122 size_t *gwo = 0, *gws, *lws = 0; 1672 size_t *gwo = 0, *gws, *lws = 0;
1123 int gws_len; 1673 int gws_len;
1124 size_t *lists; 1674 size_t *lists;
1125 int i; 1675 int i;
1126 EVENT_LIST (5, items - 5); 1676 EVENT_LIST (5);
1127 1677
1128 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1678 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1129 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1679 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1130 1680
1131 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1681 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1132 1682
1133 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1683 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1134 1684
1135 gws = lists + gws_len * 0; 1685 gws = lists + gws_len * 0;
1136 for (i = 0; i < gws_len; ++i) 1686 for (i = 0; i < gws_len; ++i)
1687 {
1137 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1688 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1689 // at least nvidia crashes for 0-sized work group sizes, work around
1690 if (!gws [i])
1691 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1692 }
1138 1693
1139 if (SvOK (global_work_offset)) 1694 if (SvOK (global_work_offset))
1140 { 1695 {
1141 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1696 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1142 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1697 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1150 } 1705 }
1151 1706
1152 if (SvOK (local_work_size)) 1707 if (SvOK (local_work_size))
1153 { 1708 {
1154 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1709 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1155 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1710 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1156 1711
1157 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1712 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1158 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1713 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1159 1714
1160 lws = lists + gws_len * 2; 1715 lws = lists + gws_len * 2;
1161 for (i = 0; i < gws_len; ++i) 1716 for (i = 0; i < gws_len; ++i)
1717 {
1162 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1718 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1719 // at least nvidia crashes for 0-sized work group sizes, work around
1720 if (!lws [i])
1721 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1722 }
1163 } 1723 }
1164 1724
1165 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1725 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1166 1726
1167 if (ev) 1727 if (ev)
1168 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1728 XPUSH_CLOBJ (stash_event, ev);
1169 1729
1170#if cl_apple_gl_sharing || cl_khr_gl_sharing 1730#if cl_apple_gl_sharing || cl_khr_gl_sharing
1171 1731
1172void 1732void
1173enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1733acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1174 ALIAS: 1734 ALIAS:
1735 release_gl_objects = 1
1736 enqueue_acquire_gl_objects = 0
1175 enqueue_release_gl_objects = 1 1737 enqueue_release_gl_objects = 1
1176 PPCODE: 1738 PPCODE:
1177 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1739 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1178 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1740 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1179 1741
1180 cl_event ev = 0; 1742 cl_event ev = 0;
1181 EVENT_LIST (2, items - 2); 1743 EVENT_LIST (2);
1182 AV *av = (AV *)SvRV (objects); 1744 AV *av = (AV *)SvRV (objects);
1183 cl_uint num_objects = av_len (av) + 1; 1745 cl_uint num_objects = av_len (av) + 1;
1184 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1746 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1185 int i; 1747 int i;
1186 1748
1187 for (i = num_objects; i--; ) 1749 for (i = num_objects; i--; )
1188 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1750 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1189 1751
1190 if (ix) 1752 if (ix)
1191 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1753 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1192 else 1754 else
1193 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1755 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1194 1756
1195 if (ev) 1757 if (ev)
1196 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1758 XPUSH_CLOBJ (stash_event, ev);
1197 1759
1198#endif 1760#endif
1199 1761
1200#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS
1201
1202void 1762void
1203enqueue_wait_for_events (OpenCL::Queue self, ...) 1763wait_for_events (OpenCL::Queue self, ...)
1764 ALIAS:
1765 enqueue_wait_for_events = 0
1204 CODE: 1766 CODE:
1205 EVENT_LIST (1, items - 1); 1767 EVENT_LIST (1);
1768#if PREFER_1_1
1206 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1769 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1207 1770#else
1771 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1208#endif 1772#endif
1209 1773
1210void 1774void
1211enqueue_marker (OpenCL::Queue self, ...) 1775marker (OpenCL::Queue self, ...)
1776 ALIAS:
1777 enqueue_marker = 0
1212 PPCODE: 1778 PPCODE:
1213 cl_event ev = 0; 1779 cl_event ev = 0;
1214 EVENT_LIST (1, items - 1); 1780 EVENT_LIST (1);
1781#if PREFER_1_1
1782 if (!event_list_count)
1783 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1784 else
1215#if CL_VERSION_1_2 1785#if CL_VERSION_1_2
1786 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1787#else
1788 {
1789 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1790 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1791 }
1792#endif
1793#else
1216 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1794 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1217#else
1218 if (event_list_count)
1219 croak ("OpenCL::Queue->enqueue_marker does not support a wait list in OpenCL 1.1 - upgrade to 1.2");
1220 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1221#endif 1795#endif
1222 if (ev) 1796 if (ev)
1223 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1797 XPUSH_CLOBJ (stash_event, ev);
1224 1798
1225void 1799void
1226enqueue_barrier (OpenCL::Queue self, ...) 1800barrier (OpenCL::Queue self, ...)
1801 ALIAS:
1802 enqueue_barrier = 0
1227 PPCODE: 1803 PPCODE:
1228 cl_event ev = 0; 1804 cl_event ev = 0;
1229 EVENT_LIST (1, items - 1); 1805 EVENT_LIST (1);
1806#if PREFER_1_1
1807 if (!event_list_count && GIMME_V == G_VOID)
1808 NEED_SUCCESS (EnqueueBarrier, (self));
1809 else
1230#if CL_VERSION_1_2 1810#if CL_VERSION_1_2
1231 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, &ev)); 1811 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1232#else 1812#else
1813 {
1233 if (event_list_count) 1814 if (event_list_count)
1234 croak ("OpenCL::Queue->enqueue_barrier does not support a wait list in OpenCL 1.1 - upgrade to 1.2"); 1815 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1816
1235 if (GIMME_V != G_VOID) 1817 if (GIMME_V != G_VOID)
1236 croak ("OpenCL::Queue->enqueue_barrier does not return an event object in OpenCL 1.1 - upgrade to 1.2");
1237 NEED_SUCCESS (EnqueueBarrier, (self)); 1818 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1819 }
1820#endif
1821#else
1822 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1238#endif 1823#endif
1239 if (ev) 1824 if (ev)
1240 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1241 1826
1242void 1827void
1243flush (OpenCL::Queue self) 1828flush (OpenCL::Queue self)
1244 CODE: 1829 CODE:
1245 NEED_SUCCESS (Flush, (self)); 1830 NEED_SUCCESS (Flush, (self));
1263 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1848 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1264 EXTEND (SP, 1); 1849 EXTEND (SP, 1);
1265 const int i = 0; 1850 const int i = 0;
1266 { 1851 {
1267 NEED_SUCCESS (RetainContext, (value [i])); 1852 NEED_SUCCESS (RetainContext, (value [i]));
1268 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1853 PUSH_CLOBJ (stash_context, value [i]);
1269 } 1854 }
1270 1855
1271void 1856void
1272device (OpenCL::Queue self) 1857device (OpenCL::Queue self)
1273 PPCODE: 1858 PPCODE:
1274 cl_device_id value [1]; 1859 cl_device_id value [1];
1275 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1860 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1276 EXTEND (SP, 1); 1861 EXTEND (SP, 1);
1277 const int i = 0; 1862 const int i = 0;
1278 { 1863 {
1279 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1864 PUSH_CLOBJ (stash_device, value [i]);
1280 } 1865 }
1281 1866
1282void 1867void
1283reference_count (OpenCL::Queue self) 1868reference_count (OpenCL::Queue self)
1284 PPCODE: 1869 PPCODE:
1293 PPCODE: 1878 PPCODE:
1294 cl_command_queue_properties value [1]; 1879 cl_command_queue_properties value [1];
1295 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1880 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1296 EXTEND (SP, 1); 1881 EXTEND (SP, 1);
1297 const int i = 0; 1882 const int i = 0;
1298 PUSHs (sv_2mortal (newSViv (value [i]))); 1883 PUSHs (sv_2mortal (newSVuv (value [i])));
1299 1884
1300#END:command_queue 1885#END:command_queue
1301 1886
1302MODULE = OpenCL PACKAGE = OpenCL::Memory 1887MODULE = OpenCL PACKAGE = OpenCL::Memory
1303 1888
1318 PPCODE: 1903 PPCODE:
1319 cl_mem_object_type value [1]; 1904 cl_mem_object_type value [1];
1320 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 1905 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1321 EXTEND (SP, 1); 1906 EXTEND (SP, 1);
1322 const int i = 0; 1907 const int i = 0;
1323 PUSHs (sv_2mortal (newSViv (value [i]))); 1908 PUSHs (sv_2mortal (newSVuv (value [i])));
1324 1909
1325void 1910void
1326flags (OpenCL::Memory self) 1911flags (OpenCL::Memory self)
1327 PPCODE: 1912 PPCODE:
1328 cl_mem_flags value [1]; 1913 cl_mem_flags value [1];
1329 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 1914 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1330 EXTEND (SP, 1); 1915 EXTEND (SP, 1);
1331 const int i = 0; 1916 const int i = 0;
1332 PUSHs (sv_2mortal (newSViv (value [i]))); 1917 PUSHs (sv_2mortal (newSVuv (value [i])));
1333 1918
1334void 1919void
1335size (OpenCL::Memory self) 1920size (OpenCL::Memory self)
1336 ALIAS: 1921 ALIAS:
1337 size = CL_MEM_SIZE 1922 size = CL_MEM_SIZE
1371 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1956 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1372 EXTEND (SP, 1); 1957 EXTEND (SP, 1);
1373 const int i = 0; 1958 const int i = 0;
1374 { 1959 {
1375 NEED_SUCCESS (RetainContext, (value [i])); 1960 NEED_SUCCESS (RetainContext, (value [i]));
1376 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1961 PUSH_CLOBJ (stash_context, value [i]);
1377 } 1962 }
1378 1963
1379void 1964void
1380associated_memobject (OpenCL::Memory self) 1965associated_memobject (OpenCL::Memory self)
1381 PPCODE: 1966 PPCODE:
1383 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1968 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1384 EXTEND (SP, 1); 1969 EXTEND (SP, 1);
1385 const int i = 0; 1970 const int i = 0;
1386 { 1971 {
1387 NEED_SUCCESS (RetainMemObject, (value [i])); 1972 NEED_SUCCESS (RetainMemObject, (value [i]));
1388 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1973 PUSH_CLOBJ (stash_memory, value [i]);
1389 } 1974 }
1390 1975
1391#END:mem 1976#END:mem
1392 1977
1393#if cl_apple_gl_sharing || cl_khr_gl_sharing 1978#if cl_apple_gl_sharing || cl_khr_gl_sharing
1413 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 1998 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1414 1999
1415 cl_buffer_region crdata = { origin, size }; 2000 cl_buffer_region crdata = { origin, size };
1416 2001
1417 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2002 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1418 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2003 XPUSH_CLOBJ (stash_buffer, mem);
1419 2004
1420MODULE = OpenCL PACKAGE = OpenCL::Image 2005MODULE = OpenCL PACKAGE = OpenCL::Image
1421 2006
1422void 2007void
1423image_info (OpenCL::Image self, cl_image_info name) 2008image_info (OpenCL::Image self, cl_image_info name)
1424 PPCODE: 2009 PPCODE:
1425 INFO (Image) 2010 INFO (Image)
2011
2012void
2013format (OpenCL::Image self)
2014 PPCODE:
2015 cl_image_format format;
2016 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2017 EXTEND (SP, 2);
2018 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2019 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1426 2020
1427#BEGIN:image 2021#BEGIN:image
1428 2022
1429void 2023void
1430element_size (OpenCL::Image self) 2024element_size (OpenCL::Image self)
1500 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2094 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1501 EXTEND (SP, 1); 2095 EXTEND (SP, 1);
1502 const int i = 0; 2096 const int i = 0;
1503 { 2097 {
1504 NEED_SUCCESS (RetainContext, (value [i])); 2098 NEED_SUCCESS (RetainContext, (value [i]));
1505 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2099 PUSH_CLOBJ (stash_context, value [i]);
1506 } 2100 }
1507 2101
1508void 2102void
1509normalized_coords (OpenCL::Sampler self) 2103normalized_coords (OpenCL::Sampler self)
1510 PPCODE: 2104 PPCODE:
1511 cl_addressing_mode value [1]; 2105 cl_addressing_mode value [1];
1512 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2106 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1513 EXTEND (SP, 1); 2107 EXTEND (SP, 1);
1514 const int i = 0; 2108 const int i = 0;
1515 PUSHs (sv_2mortal (newSViv (value [i]))); 2109 PUSHs (sv_2mortal (newSVuv (value [i])));
1516 2110
1517void 2111void
1518addressing_mode (OpenCL::Sampler self) 2112addressing_mode (OpenCL::Sampler self)
1519 PPCODE: 2113 PPCODE:
1520 cl_filter_mode value [1]; 2114 cl_filter_mode value [1];
1521 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2115 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1522 EXTEND (SP, 1); 2116 EXTEND (SP, 1);
1523 const int i = 0; 2117 const int i = 0;
1524 PUSHs (sv_2mortal (newSViv (value [i]))); 2118 PUSHs (sv_2mortal (newSVuv (value [i])));
1525 2119
1526void 2120void
1527filter_mode (OpenCL::Sampler self) 2121filter_mode (OpenCL::Sampler self)
1528 PPCODE: 2122 PPCODE:
1529 cl_bool value [1]; 2123 cl_bool value [1];
1540DESTROY (OpenCL::Program self) 2134DESTROY (OpenCL::Program self)
1541 CODE: 2135 CODE:
1542 clReleaseProgram (self); 2136 clReleaseProgram (self);
1543 2137
1544void 2138void
1545build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2139build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2140 ALIAS:
2141 build_async = 1
1546 CODE: 2142 CODE:
2143 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2144 void *user_data = 0;
2145 cl_uint num_devices = 0;
2146 cl_device_id *device_list = 0;
2147
2148 if (SvOK (devices))
2149 {
2150 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2151 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2152
2153 AV *av = (AV *)SvRV (devices);
2154 num_devices = av_len (av) + 1;
2155
2156 if (num_devices)
2157 {
2158 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2159 int count;
2160 for (count = 0; count < num_devices; ++count)
2161 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2162 }
2163 }
2164
2165 if (SvOK (notify))
2166 {
2167 NEED_SUCCESS (RetainProgram, (self));
2168 pfn_notify = eq_program_notify;
2169 user_data = SvREFCNT_inc (s_get_cv (notify));
2170 }
2171
2172 if (ix)
2173 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2174 else
1547 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2175 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1548 2176
1549void 2177void
1550build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2178build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1551 PPCODE: 2179 PPCODE:
1552 size_t size; 2180 size_t size;
1587 2215
1588void 2216void
1589kernel (OpenCL::Program program, SV *function) 2217kernel (OpenCL::Program program, SV *function)
1590 PPCODE: 2218 PPCODE:
1591 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2219 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1592 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2220 XPUSH_CLOBJ (stash_kernel, kernel);
2221
2222void
2223kernels_in_program (OpenCL::Program program)
2224 PPCODE:
2225 cl_uint num_kernels;
2226 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2227 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2228 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2229
2230 int i;
2231 EXTEND (SP, num_kernels);
2232 for (i = 0; i < num_kernels; ++i)
2233 PUSH_CLOBJ (stash_kernel, kernels [i]);
1593 2234
1594void 2235void
1595info (OpenCL::Program self, cl_program_info name) 2236info (OpenCL::Program self, cl_program_info name)
1596 PPCODE: 2237 PPCODE:
1597 INFO (Program) 2238 INFO (Program)
1645 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2286 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1646 EXTEND (SP, 1); 2287 EXTEND (SP, 1);
1647 const int i = 0; 2288 const int i = 0;
1648 { 2289 {
1649 NEED_SUCCESS (RetainContext, (value [i])); 2290 NEED_SUCCESS (RetainContext, (value [i]));
1650 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2291 PUSH_CLOBJ (stash_context, value [i]);
1651 } 2292 }
1652 2293
1653void 2294void
1654devices (OpenCL::Program self) 2295devices (OpenCL::Program self)
1655 PPCODE: 2296 PPCODE:
1659 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2300 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1660 int i, n = size / sizeof (*value); 2301 int i, n = size / sizeof (*value);
1661 EXTEND (SP, n); 2302 EXTEND (SP, n);
1662 for (i = 0; i < n; ++i) 2303 for (i = 0; i < n; ++i)
1663 { 2304 {
1664 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2305 PUSH_CLOBJ (stash_device, value [i]);
1665 } 2306 }
1666 2307
1667void 2308void
1668source (OpenCL::Program self) 2309source (OpenCL::Program self)
1669 PPCODE: 2310 PPCODE:
1695DESTROY (OpenCL::Kernel self) 2336DESTROY (OpenCL::Kernel self)
1696 CODE: 2337 CODE:
1697 clReleaseKernel (self); 2338 clReleaseKernel (self);
1698 2339
1699void 2340void
2341setf (OpenCL::Kernel self, const char *format, ...)
2342 CODE:
2343 int i;
2344 for (i = 2; ; ++i)
2345 {
2346 while (*format == ' ')
2347 ++format;
2348
2349 char type = *format++;
2350
2351 if (!type)
2352 break;
2353
2354 if (i >= items)
2355 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2356
2357 SV *sv = ST (i);
2358
2359 union
2360 {
2361 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2362 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2363 cl_half ch; cl_float cf; cl_double cd;
2364 cl_mem cm;
2365 cl_sampler ca;
2366 size_t cz;
2367 cl_event ce;
2368 } arg;
2369 size_t size;
2370 int nullarg = 0;
2371
2372 switch (type)
2373 {
2374 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2375 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2376 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2377 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2378 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2379 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2380 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2381 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2382
2383 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2384 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2385 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2386
2387 case 'z': nullarg = 1; size = SvIV (sv); break;
2388
2389 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2390 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2391 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2392
2393 default:
2394 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2395 }
2396
2397 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2398 if (res)
2399 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2400 }
2401
2402 if (i != items)
2403 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2404
2405void
1700set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2406set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1701 CODE: 2407 CODE:
1702 clSetKernelArg (self, idx, sizeof (value), &value); 2408 clSetKernelArg (self, idx, sizeof (value), &value);
1703 2409
1704void 2410void
1752 clSetKernelArg (self, idx, sizeof (value), &value); 2458 clSetKernelArg (self, idx, sizeof (value), &value);
1753 2459
1754void 2460void
1755set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2461set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1756 CODE: 2462 CODE:
1757 clSetKernelArg (self, idx, sizeof (value), &value); 2463 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1758 2464
1759void 2465void
1760set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2466set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1761 CODE: 2467 CODE:
1762 clSetKernelArg (self, idx, sizeof (value), &value); 2468 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1763 2469
1764void 2470void
1765set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2471set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2472 ALIAS:
2473 set_image2d = 0
2474 set_image3d = 0
1766 CODE: 2475 CODE:
1767 clSetKernelArg (self, idx, sizeof (value), &value); 2476 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1768
1769void
1770set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1771 CODE:
1772 clSetKernelArg (self, idx, sizeof (value), &value);
1773 2477
1774void 2478void
1775set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2479set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1776 CODE: 2480 CODE:
1777 clSetKernelArg (self, idx, sizeof (value), &value); 2481 clSetKernelArg (self, idx, sizeof (value), &value);
1823 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2527 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1824 EXTEND (SP, 1); 2528 EXTEND (SP, 1);
1825 const int i = 0; 2529 const int i = 0;
1826 { 2530 {
1827 NEED_SUCCESS (RetainContext, (value [i])); 2531 NEED_SUCCESS (RetainContext, (value [i]));
1828 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2532 PUSH_CLOBJ (stash_context, value [i]);
1829 } 2533 }
1830 2534
1831void 2535void
1832program (OpenCL::Kernel self) 2536program (OpenCL::Kernel self)
1833 PPCODE: 2537 PPCODE:
1835 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2539 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1836 EXTEND (SP, 1); 2540 EXTEND (SP, 1);
1837 const int i = 0; 2541 const int i = 0;
1838 { 2542 {
1839 NEED_SUCCESS (RetainProgram, (value [i])); 2543 NEED_SUCCESS (RetainProgram, (value [i]));
1840 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2544 PUSH_CLOBJ (stash_program, value [i]);
1841 } 2545 }
1842 2546
1843#END:kernel 2547#END:kernel
1844 2548
1845void 2549void
1905wait (OpenCL::Event self) 2609wait (OpenCL::Event self)
1906 CODE: 2610 CODE:
1907 clWaitForEvents (1, &self); 2611 clWaitForEvents (1, &self);
1908 2612
1909void 2613void
2614cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2615 CODE:
2616 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2617
2618void
1910info (OpenCL::Event self, cl_event_info name) 2619info (OpenCL::Event self, cl_event_info name)
1911 PPCODE: 2620 PPCODE:
1912 INFO (Event) 2621 INFO (Event)
1913 2622
1914#BEGIN:event 2623#BEGIN:event
1920 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2629 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1921 EXTEND (SP, 1); 2630 EXTEND (SP, 1);
1922 const int i = 0; 2631 const int i = 0;
1923 { 2632 {
1924 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2633 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1925 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2634 PUSH_CLOBJ (stash_queue, value [i]);
1926 } 2635 }
1927 2636
1928void 2637void
1929command_type (OpenCL::Event self) 2638command_type (OpenCL::Event self)
1930 PPCODE: 2639 PPCODE:
1953 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2662 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1954 EXTEND (SP, 1); 2663 EXTEND (SP, 1);
1955 const int i = 0; 2664 const int i = 0;
1956 { 2665 {
1957 NEED_SUCCESS (RetainContext, (value [i])); 2666 NEED_SUCCESS (RetainContext, (value [i]));
1958 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2667 PUSH_CLOBJ (stash_context, value [i]);
1959 } 2668 }
1960 2669
1961#END:event 2670#END:event
1962 2671
1963void 2672void
1988void 2697void
1989set_status (OpenCL::UserEvent self, cl_int execution_status) 2698set_status (OpenCL::UserEvent self, cl_int execution_status)
1990 CODE: 2699 CODE:
1991 clSetUserEventStatus (self, execution_status); 2700 clSetUserEventStatus (self, execution_status);
1992 2701
2702MODULE = OpenCL PACKAGE = OpenCL::Mapped
2703
2704void
2705DESTROY (SV *self)
2706 CODE:
2707 OpenCL__Mapped mapped = SvMAPPED (self);
2708
2709 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2710 mapped_detach (self, mapped);
2711
2712 clReleaseCommandQueue (mapped->queue);
2713
2714 if (mapped->event)
2715 clReleaseEvent (mapped->event);
2716
2717 Safefree (mapped);
2718
2719bool
2720mapped (OpenCL::Mapped self)
2721 CODE:
2722 RETVAL = !!self->ptr;
2723 OUTPUT:
2724 RETVAL
2725
2726void
2727wait (OpenCL::Mapped self)
2728 PPCODE:
2729 if (self->event)
2730 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2731
2732void
2733event (OpenCL::Mapped self)
2734 PPCODE:
2735 if (!self->event)
2736 XSRETURN_UNDEF;
2737
2738 clRetainEvent (self->event);
2739 XPUSH_CLOBJ (stash_event, self->event);
2740
2741size_t
2742size (OpenCL::Mapped self)
2743 CODE:
2744 RETVAL = self->cb;
2745 OUTPUT:
2746 RETVAL
2747
2748IV
2749ptr (OpenCL::Mapped self)
2750 CODE:
2751 RETVAL = PTR2IV (self->ptr);
2752 OUTPUT:
2753 RETVAL
2754
2755MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2756
2757MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2758
2759IV
2760row_pitch (OpenCL::Mapped self)
2761 ALIAS:
2762 slice_pitch = 1
2763 CODE:
2764 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2765 OUTPUT:
2766 RETVAL
2767
2768
2769

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines