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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.51 by root, Tue Apr 24 22:45:38 2012 UTC vs.
Revision 1.66 by root, Fri May 4 14:46:02 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
4 6
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 8#include "xthread.h"
7#include "schmorp.h" 9#include "schmorp.h"
8 10
38 #define PREFER_1_1 1 40 #define PREFER_1_1 1
39#endif 41#endif
40 42
41typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
61 60
61typedef struct mapped * OpenCL__Mapped;
62
62typedef SV *FUTURE; 63typedef SV *FUTURE;
64
65static HV
66 *stash_platform,
67 *stash_device,
68 *stash_subdevice,
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;
63 89
64/*****************************************************************************/ 90/*****************************************************************************/
65 91
66// name must include a leading underscore 92// name must include a leading underscore
67// 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
107} 133}
108 134
109/*****************************************************************************/ 135/*****************************************************************************/
110 136
111/* up to two temporary buffers */ 137/* up to two temporary buffers */
112static void * 138static void * ecb_noinline
113tmpbuf (size_t size) 139tmpbuf (size_t size)
114{ 140{
115 enum { buffers = 3 }; 141 enum { buffers = 4 };
116 static int idx; 142 static int idx;
117 static void *buf [buffers]; 143 static void *buf [buffers];
118 static size_t len [buffers]; 144 static size_t len [buffers];
119 145
120 idx = (idx + 1) % buffers; 146 idx = (idx + 1) % buffers;
193 if (res) \ 219 if (res) \
194 FAIL (name); 220 FAIL (name);
195 221
196/*****************************************************************************/ 222/*****************************************************************************/
197 223
224static SV *
225new_clobj (HV *stash, IV id)
226{
227 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
228}
229
230#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
231#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
232
233/* cl objects are either \$iv, or [$iv, ...] */
234/* they can be upgraded at runtime to the array form */
235static void * ecb_noinline
236SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
237{
238 // sv_derived_from is quite slow :(
239 if (SvROK (sv) && sv_derived_from (sv, pkg))
240 return (void *)SvIV (SvRV (sv));
241
242 croak ("%s: %s is not of type %s", func, svname, pkg);
243}
244
245// the "no-inherit" version of the above
246static void * ecb_noinline
247SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
248{
249 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
250 return (void *)SvIV (SvRV (sv));
251
252 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
253}
254
255/*****************************************************************************/
256
198static cl_context_properties * 257static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 258SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 259{
201 if (!sv || !SvOK (sv)) 260 if (!sv || !SvOK (sv))
202 if (extra) 261 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 262 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
210 int i, len = av_len (av) + 1; 269 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 270 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 271 cl_context_properties *l = p;
213 272
214 if (len & 1) 273 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 274 croak ("%s: %s is not a property list (must contain an even number of elements)", func, svname);
216 275
217 while (extracount--) 276 while (extracount--)
218 *l++ = *extra++; 277 *l++ = *extra++;
219 278
220 for (i = 0; i < len; i += 2) 279 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 282 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 283 cl_context_properties v = SvIV (p_sv); // code below can override
225 284
226 switch (t) 285 switch (t)
227 { 286 {
287 case CL_CONTEXT_PLATFORM:
288 if (SvROK (p_sv))
289 v = (cl_context_properties)SvCLOBJ (func, svname, p_sv, "OpenCL::Platform");
290 break;
291
228 case CL_GLX_DISPLAY_KHR: 292 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 293 if (!SvOK (p_sv))
230 { 294 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 295 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 296 if (func)
259 323
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 324 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
261} 325}
262 326
263/*****************************************************************************/ 327/*****************************************************************************/
264
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr))
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{
274 if (SvROK (sv) && sv_derived_from (sv, pkg))
275 {
276 SV *rv = SvRV (sv);
277
278 if (SvTYPE (rv) == SVt_PVAV)
279 rv = AvARRAY (rv)[0];
280
281 return (void *)SvIV (SvRV (sv));
282 }
283
284 croak ("%s: %s is not of type %s", func, svname, pkg);
285}
286
287static void
288CLOBJ_push (SV *self, SV *data)
289{
290 SV *rv = SvRV (self);
291
292 if (SvTYPE (rv) != SVt_PVAV)
293 {
294 AV *av = newAV ();
295 av_push (av, rv);
296 rv = (SV *)av;
297 SvRV_set (self, rv);
298 }
299
300 av_push ((AV *)rv, data);
301}
302
303static SV *
304sv_struct (STRLEN size)
305{
306 SV *sv = newSV (size);
307 SvPOK_only (sv);
308 return sv;
309}
310
311static void *
312CLOBJ_push_struct (SV *self, STRLEN size)
313{
314 SV *sv = sv_struct (size);
315 CLOBJ_push (self, sv);
316 return SvPVX (sv);
317}
318
319/*****************************************************************************/
320/* callback stuff */ 328/* callback stuff */
321 329
322/* default context callback, log to stderr */ 330/* default context callback, log to stderr */
323static void CL_CALLBACK 331static void CL_CALLBACK
324context_default_notify (const char *msg, const void *info, size_t cb, void *data) 332context_default_notify (const char *msg, const void *info, size_t cb, void *data)
343static void (*eq_signal_func)(void *signal_arg, int value); 351static void (*eq_signal_func)(void *signal_arg, int value);
344static void *eq_signal_arg; 352static void *eq_signal_arg;
345static xmutex_t eq_lock = X_MUTEX_INIT; 353static xmutex_t eq_lock = X_MUTEX_INIT;
346static eq_item *eq_head, *eq_tail; 354static eq_item *eq_head, *eq_tail;
347 355
348static void 356static void ecb_noinline
349eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 357eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
350{ 358{
351 eq_item *item = malloc (sizeof (eq_item)); 359 eq_item *item = malloc (sizeof (eq_item));
352
353 printf ("enq(%p,%p,%p,%p,%p)\n", vtbl, cb, data1, data2, data3);//D
354 360
355 item->next = 0; 361 item->next = 0;
356 item->vtbl = vtbl; 362 item->vtbl = vtbl;
357 item->cb = cb; 363 item->cb = cb;
358 item->data1 = data1; 364 item->data1 = data1;
383 X_UNLOCK (eq_lock); 389 X_UNLOCK (eq_lock);
384 390
385 return res; 391 return res;
386} 392}
387 393
388#if 0
389static void
390mem_free (pTHX_ void *p)
391{
392 free (p);
393}
394//SAVEDESTRUCTOR_X (mem_free, item);
395#endif
396
397static void 394static void
398eq_poll (void) 395eq_poll (void)
399{ 396{
400 eq_item *item; 397 eq_item *item;
401 398
428eq_poll_interrupt (pTHX_ void *c_arg, int value) 425eq_poll_interrupt (pTHX_ void *c_arg, int value)
429{ 426{
430 eq_poll (); 427 eq_poll ();
431} 428}
432 429
430/*****************************************************************************/
433/* context notify */ 431/* context notify */
434 432
435static void 433static void ecb_noinline
436eq_context_push (void *data1, void *data2, void *data3) 434eq_context_push (void *data1, void *data2, void *data3)
437{ 435{
438 dSP; 436 dSP;
439 PUSHs (sv_2mortal (newSVpv (data1, 0))); 437 PUSHs (sv_2mortal (newSVpv (data1, 0)));
440 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 438 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
441 PUTBACK; 439 PUTBACK;
440
441 free (data1);
442 free (data2);
442} 443}
443 444
444static eq_vtbl eq_context_vtbl = { 0, eq_context_push }; 445static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
445 446
447static void CL_CALLBACK
448eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
449{
450 void *pvt_copy = malloc (cb);
451 memcpy (pvt_copy, pvt, cb);
452 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
453}
454
455#define CONTEXT_NOTIFY_CALLBACK \
456 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
457 void *user_data = 0; \
458 \
459 if (SvOK (notify)) \
460 { \
461 pfn_notify = eq_context_notify; \
462 user_data = s_get_cv (notify); \
463 }
464
465static SV * ecb_noinline
466new_clobj_context (cl_context ctx, void *user_data)
467{
468 SV *sv = new_clobj (stash_context, (IV)ctx);
469
470 if (user_data)
471 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
472
473 return sv;
474}
475
476#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
477
478/*****************************************************************************/
446/* build/compile/link notify */ 479/* build/compile/link notify */
447 480
448static void 481static void
449eq_program_push (void *data1, void *data2, void *data3) 482eq_program_push (void *data1, void *data2, void *data3)
450{ 483{
451 dSP; 484 dSP;
452 PUSH_CLOBJ ("OpenCL::Program", data1); 485 PUSH_CLOBJ (stash_program, data1);
453 PUTBACK; 486 PUTBACK;
454} 487}
455 488
456static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 489static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
457 490
480 else 513 else
481 clReleaseProgram (arg->program); 514 clReleaseProgram (arg->program);
482 515
483 free (arg->options); 516 free (arg->options);
484 free (arg); 517 free (arg);
518
519 return 0;
485} 520}
486 521
487static void 522static void
488build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 523build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
489{ 524{
497 532
498 xthread_t id; 533 xthread_t id;
499 thread_create (&id, build_program_thread, arg); 534 thread_create (&id, build_program_thread, arg);
500} 535}
501 536
537/*****************************************************************************/
502/* event objects */ 538/* event objects */
503 539
504static void 540static void
505eq_event_push (void *data1, void *data2, void *data3) 541eq_event_push (void *data1, void *data2, void *data3)
506{ 542{
507 dSP; 543 dSP;
508 PUSH_CLOBJ ("OpenCL::Event", data1); 544 PUSH_CLOBJ (stash_event, data1);
509 PUSHs (sv_2mortal (newSViv ((IV)data2))); 545 PUSHs (sv_2mortal (newSViv ((IV)data2)));
510 PUTBACK; 546 PUTBACK;
511} 547}
512 548
513static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 549static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
514 550
515static void CL_CALLBACK 551static void CL_CALLBACK
516eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data) 552eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
517{ 553{
518 clRetainEvent (event); 554 clRetainEvent (event);
519 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)event_command_exec_status, 0); 555 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
520} 556}
521 557
522/*****************************************************************************/ 558/*****************************************************************************/
559/* utilities for XS code */
523 560
524static size_t 561static size_t
525img_row_pitch (cl_mem img) 562img_row_pitch (cl_mem img)
526{ 563{
527 size_t res; 564 size_t res;
528 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 565 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
529 return res; 566 return res;
530} 567}
531 568
532static cl_event * 569static cl_event * ecb_noinline
533event_list (SV **items, cl_uint *rcount) 570event_list (SV **items, cl_uint *rcount, cl_event extra)
534{ 571{
535 cl_uint count = *rcount; 572 cl_uint count = *rcount;
536 573
537 if (!count) 574 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
575 *rcount = count = 0;
576
577 if (!count && !extra)
538 return 0; 578 return 0;
539 579
540 cl_event *list = tmpbuf (sizeof (cl_event) * count); 580 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
541 int i = 0; 581 int i = 0;
542 582
543 do 583 while (count--)
544 {
545 --count;
546 if (SvOK (items [count])) 584 if (SvOK (items [count]))
547 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 585 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
548 } 586
549 while (count); 587 if (extra)
588 list [i++] = extra;
550 589
551 *rcount = i; 590 *rcount = i;
552 591
553 return i ? list : 0; 592 return i ? list : 0;
554} 593}
555 594
556#define EVENT_LIST(items,count) \ 595#define EVENT_LIST(skip) \
557 cl_uint event_list_count = (count); \ 596 cl_uint event_list_count = items - (skip); \
558 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 597 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
559 598
560#define INFO(class) \ 599#define INFO(class) \
561{ \ 600{ \
562 size_t size; \ 601 size_t size; \
563 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 602 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
566 SvPOK_only (sv); \ 605 SvPOK_only (sv); \
567 SvCUR_set (sv, size); \ 606 SvCUR_set (sv, size); \
568 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 607 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
569 XPUSHs (sv); \ 608 XPUSHs (sv); \
570} 609}
610
611/*****************************************************************************/
612/* mapped_xxx */
613
614static OpenCL__Mapped
615SvMAPPED (SV *sv)
616{
617 // no typechecking atm., keep your fingers crossed
618 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
619}
620
621struct mapped
622{
623 cl_command_queue queue;
624 cl_mem memobj;
625 void *ptr;
626 size_t cb;
627 cl_event event;
628 size_t row_pitch;
629 size_t slice_pitch;
630};
631
632static SV *
633mapped_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)
634{
635 SV *data = newSV (0);
636 SvUPGRADE (data, SVt_PVMG);
637
638 OpenCL__Mapped mapped;
639 New (0, mapped, 1, struct mapped);
640
641 clRetainCommandQueue (queue);
642
643 mapped->queue = queue;
644 mapped->memobj = memobj;
645 mapped->ptr = ptr;
646 mapped->cb = cb;
647 mapped->event = ev;
648 mapped->row_pitch = row_pitch;
649 mapped->slice_pitch = slice_pitch;
650
651 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
652
653 if (SvLEN (data))
654 Safefree (data);
655
656 SvPVX (data) = (char *)ptr;
657 SvCUR_set (data, cb);
658 SvLEN_set (data, 0);
659 SvPOK_only (data);
660
661 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
662
663 if (!(flags & CL_MAP_WRITE))
664 SvREADONLY_on (data);
665
666 return obj;
667}
668
669static void
670mapped_detach (SV *sv, OpenCL__Mapped mapped)
671{
672 SV *data = SvRV (sv);
673
674 // the next check checks both before AND after detach, where SvPVX should STILL be 0
675 if (SvPVX (data) != (char *)mapped->ptr)
676 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
677 else
678 {
679 SvREADONLY_off (data);
680 SvCUR_set (data, 0);
681 SvPVX (data) = 0;
682 SvOK_off (data);
683 }
684
685 mapped->ptr = 0;
686}
687
688static void
689mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
690{
691 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
692 cl_event ev;
693
694 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
695
696 clReleaseEvent (mapped->event);
697 mapped->event = ev;
698
699 mapped_detach (self, mapped);
700}
701
702/*****************************************************************************/
571 703
572MODULE = OpenCL PACKAGE = OpenCL 704MODULE = OpenCL PACKAGE = OpenCL
573 705
574PROTOTYPES: ENABLE 706PROTOTYPES: ENABLE
575 707
603 }; 735 };
604 736
605 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 737 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
606 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 738 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
607 739
740 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
741 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
742 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
743 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
744 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
745 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
746 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
747 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
748 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
749 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
750 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
751 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
752 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
753 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
754 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
755 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
756 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
757 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
758 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
759 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
760 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
761 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
762 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
763
608 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 764 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
609} 765}
610 766
611cl_int 767cl_int
612errno () 768errno ()
614 RETVAL = res; 770 RETVAL = res;
615 OUTPUT: 771 OUTPUT:
616 RETVAL 772 RETVAL
617 773
618const char * 774const char *
619err2str (cl_int err) 775err2str (cl_int err = res)
620 776
621const char * 777const char *
622enum2str (cl_uint value) 778enum2str (cl_uint value)
623 779
624void 780void
632 list = tmpbuf (sizeof (*list) * count); 788 list = tmpbuf (sizeof (*list) * count);
633 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 789 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
634 790
635 EXTEND (SP, count); 791 EXTEND (SP, count);
636 for (i = 0; i < count; ++i) 792 for (i = 0; i < count; ++i)
637 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 793 PUSH_CLOBJ (stash_platform, list [i]);
638 794
639void 795void
640context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 796context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
641 PPCODE: 797 PPCODE:
798 CONTEXT_NOTIFY_CALLBACK;
642 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 799 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
643 XPUSH_CLOBJ ("OpenCL::Context", ctx); 800 XPUSH_CLOBJ_CONTEXT;
644 801
645#if 0
646
647void 802void
648context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 803context (FUTURE properties, FUTURE devices, FUTURE notify)
649 PPCODE: 804 PPCODE:
650 /* der Gipfel der Kunst */ 805 /* der Gipfel der Kunst */
651 806
652#endif
653
654void 807void
655wait_for_events (...) 808wait_for_events (...)
656 CODE: 809 CODE:
657 EVENT_LIST (0, items); 810 EVENT_LIST (0);
658 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 811 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
659 812
660PROTOTYPES: DISABLE 813PROTOTYPES: DISABLE
661 814
662MODULE = OpenCL PACKAGE = OpenCL::Platform 815MODULE = OpenCL PACKAGE = OpenCL::Platform
705 list = tmpbuf (sizeof (*list) * count); 858 list = tmpbuf (sizeof (*list) * count);
706 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 859 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
707 860
708 EXTEND (SP, count); 861 EXTEND (SP, count);
709 for (i = 0; i < count; ++i) 862 for (i = 0; i < count; ++i)
710 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 863 PUSH_CLOBJ (stash_device, list [i]);
711 864
712void 865void
713context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = 0) 866context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
714 PPCODE: 867 PPCODE:
715 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 868 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
716 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 869 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
870
871 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
872 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
717 873
718 AV *av = (AV *)SvRV (devices); 874 AV *av = (AV *)SvRV (devices);
719 cl_uint num_devices = av_len (av) + 1; 875 cl_uint num_devices = av_len (av) + 1;
720 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 876 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
721 877
722 int i; 878 int i;
723 for (i = num_devices; i--; ) 879 for (i = num_devices; i--; )
724 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 880 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
725 881
726 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; 882 CONTEXT_NOTIFY_CALLBACK;
727 void *user_data = 0;
728
729 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res)); 883 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res));
730 XPUSH_CLOBJ ("OpenCL::Context", ctx); 884 XPUSH_CLOBJ_CONTEXT;
731 885
732void 886void
733context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 887context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
734 PPCODE: 888 PPCODE:
735 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 889 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
736 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 890 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
891
892 CONTEXT_NOTIFY_CALLBACK;
737 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 893 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
738 XPUSH_CLOBJ ("OpenCL::Context", ctx); 894 XPUSH_CLOBJ_CONTEXT;
739 895
740MODULE = OpenCL PACKAGE = OpenCL::Device 896MODULE = OpenCL PACKAGE = OpenCL::Device
741 897
742void 898void
743info (OpenCL::Device self, cl_device_info name) 899info (OpenCL::Device self, cl_device_info name)
744 PPCODE: 900 PPCODE:
745 INFO (Device) 901 INFO (Device)
902
903#if CL_VERSION_1_2
904
905void
906sub_devices (OpenCL::Device self, SV *properties)
907 PPCODE:
908 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
909 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
910
911 properties = SvRV (properties);
912
913 cl_uint count = av_len ((AV *)properties) + 1;
914 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
915
916 int i;
917 for (i = 0; i < count; ++i)
918 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
919
920 props [count] = 0;
921
922 cl_uint num_devices;
923 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
924 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
925 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
926
927 EXTEND (SP, num_devices);
928 for (i = 0; i < count; ++i)
929 PUSH_CLOBJ (stash_subdevice, list [i]);
930
931#endif
746 932
747#BEGIN:device 933#BEGIN:device
748 934
749void 935void
750type (OpenCL::Device self) 936type (OpenCL::Device self)
751 PPCODE: 937 PPCODE:
752 cl_device_type value [1]; 938 cl_device_type value [1];
753 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 939 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
754 EXTEND (SP, 1); 940 EXTEND (SP, 1);
755 const int i = 0; 941 const int i = 0;
756 PUSHs (sv_2mortal (newSViv (value [i]))); 942 PUSHs (sv_2mortal (newSVuv (value [i])));
757 943
758void 944void
759vendor_id (OpenCL::Device self) 945vendor_id (OpenCL::Device self)
760 ALIAS: 946 ALIAS:
761 vendor_id = CL_DEVICE_VENDOR_ID 947 vendor_id = CL_DEVICE_VENDOR_ID
906 PPCODE: 1092 PPCODE:
907 cl_command_queue_properties value [1]; 1093 cl_command_queue_properties value [1];
908 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1094 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
909 EXTEND (SP, 1); 1095 EXTEND (SP, 1);
910 const int i = 0; 1096 const int i = 0;
911 PUSHs (sv_2mortal (newSViv (value [i]))); 1097 PUSHs (sv_2mortal (newSVuv (value [i])));
912 1098
913void 1099void
914platform (OpenCL::Device self) 1100platform (OpenCL::Device self)
915 PPCODE: 1101 PPCODE:
916 cl_platform_id value [1]; 1102 cl_platform_id value [1];
917 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1103 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
918 EXTEND (SP, 1); 1104 EXTEND (SP, 1);
919 const int i = 0; 1105 const int i = 0;
920 { 1106 {
921 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1107 PUSH_CLOBJ (stash_platform, value [i]);
922 } 1108 }
923 1109
924void 1110void
925name (OpenCL::Device self) 1111name (OpenCL::Device self)
926 ALIAS: 1112 ALIAS:
945 cl_device_id value [1]; 1131 cl_device_id value [1];
946 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1132 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
947 EXTEND (SP, 1); 1133 EXTEND (SP, 1);
948 const int i = 0; 1134 const int i = 0;
949 { 1135 {
950 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1136 PUSH_CLOBJ (stash_device, value [i]);
951 } 1137 }
952 1138
953void 1139void
954partition_types_ext (OpenCL::Device self) 1140partition_types_ext (OpenCL::Device self)
955 ALIAS: 1141 ALIAS:
966 for (i = 0; i < n; ++i) 1152 for (i = 0; i < n; ++i)
967 PUSHs (sv_2mortal (newSVuv (value [i]))); 1153 PUSHs (sv_2mortal (newSVuv (value [i])));
968 1154
969#END:device 1155#END:device
970 1156
1157MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1158
1159#if CL_VERSION_1_2
1160
1161void
1162DESTROY (OpenCL::SubDevice self)
1163 CODE:
1164 clReleaseDevice (self);
1165
1166#endif
1167
971MODULE = OpenCL PACKAGE = OpenCL::Context 1168MODULE = OpenCL PACKAGE = OpenCL::Context
972 1169
973void 1170void
974DESTROY (OpenCL::Context context) 1171DESTROY (OpenCL::Context self)
975 CODE: 1172 CODE:
976 clReleaseContext (context); 1173 clReleaseContext (self);
977 1174
978void 1175void
979info (OpenCL::Context self, cl_context_info name) 1176info (OpenCL::Context self, cl_context_info name)
980 PPCODE: 1177 PPCODE:
981 INFO (Context) 1178 INFO (Context)
982 1179
983void 1180void
984queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1181queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
985 PPCODE: 1182 PPCODE:
986 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1183 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
987 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1184 XPUSH_CLOBJ (stash_queue, queue);
988 1185
989void 1186void
990user_event (OpenCL::Context self) 1187user_event (OpenCL::Context self)
991 PPCODE: 1188 PPCODE:
992 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1189 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
993 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1190 XPUSH_CLOBJ (stash_userevent, ev);
994 1191
995void 1192void
996buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1193buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
997 PPCODE: 1194 PPCODE:
998 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1195 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
999 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1196 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
1000 1197
1001 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1198 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1002 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1199 XPUSH_CLOBJ (stash_bufferobj, mem);
1003 1200
1004void 1201void
1005buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1202buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
1006 PPCODE: 1203 PPCODE:
1007 STRLEN len; 1204 STRLEN len;
1008 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1205 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1009 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1206 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
1010 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1207 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
1011 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1208 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
1012 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1209 XPUSH_CLOBJ (stash_bufferobj, mem);
1013 1210
1014#if CL_VERSION_1_2 1211#if CL_VERSION_1_2
1015 1212
1016void 1213void
1017image (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) 1214image (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)
1018 PPCODE: 1215 PPCODE:
1019 STRLEN len; 1216 STRLEN len;
1020 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1217 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1021 const cl_image_format format = { channel_order, channel_type }; 1218 const cl_image_format format = { channel_order, channel_type };
1022 const cl_image_desc desc = { 1219 const cl_image_desc desc = {
1025 array_size, row_pitch, slice_pitch, 1222 array_size, row_pitch, slice_pitch,
1026 num_mip_level, num_samples, 1223 num_mip_level, num_samples,
1027 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1224 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1028 }; 1225 };
1029 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1226 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1030 char *klass = "OpenCL::Image"; 1227 HV *stash = stash_image;
1031 switch (type) 1228 switch (type)
1032 { 1229 {
1033 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1230 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1034 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1231 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1035 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1232 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1036 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1233 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1037 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1234 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1038 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1235 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1039 } 1236 }
1040 XPUSH_CLOBJ (klass, mem); 1237 XPUSH_CLOBJ (stash, mem);
1041 1238
1042#endif 1239#endif
1043 1240
1044void 1241void
1045image2d (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) 1242image2d (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)
1051 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1248 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1052#else 1249#else
1053 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1250 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1054 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1251 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1055#endif 1252#endif
1056 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1253 XPUSH_CLOBJ (stash_image2d, mem);
1057 1254
1058void 1255void
1059image3d (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) 1256image3d (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)
1060 PPCODE: 1257 PPCODE:
1061 STRLEN len; 1258 STRLEN len;
1065 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1262 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1066#else 1263#else
1067 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1264 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1068 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1265 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1069#endif 1266#endif
1070 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1267 XPUSH_CLOBJ (stash_image3d, mem);
1071 1268
1072#if cl_apple_gl_sharing || cl_khr_gl_sharing 1269#if cl_apple_gl_sharing || cl_khr_gl_sharing
1073 1270
1074void 1271void
1075gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1272gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1076 PPCODE: 1273 PPCODE:
1077 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1274 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1078 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1275 XPUSH_CLOBJ (stash_bufferobj, mem);
1079 1276
1080void 1277void
1081gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1278gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1082 PPCODE: 1279 PPCODE:
1083 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1280 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1084 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1281 XPUSH_CLOBJ (stash_image2d, mem);
1085 1282
1086#if CL_VERSION_1_2 1283#if CL_VERSION_1_2
1087 1284
1088void 1285void
1089gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1286gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1090 ALIAS: 1287 ALIAS:
1091 PPCODE: 1288 PPCODE:
1092 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1289 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1093 cl_gl_object_type type; 1290 cl_gl_object_type type;
1094 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1291 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1095 char *klass = "OpenCL::Memory"; 1292 HV *stash = stash_memory;
1096 switch (type) 1293 switch (type)
1097 { 1294 {
1098 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1295 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1099 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1296 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1100 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1297 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1101 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1298 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1102 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1299 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1103 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1300 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1104 } 1301 }
1105 XPUSH_CLOBJ (klass, mem); 1302 XPUSH_CLOBJ (stash, mem);
1106 1303
1107#endif 1304#endif
1108 1305
1109void 1306void
1110gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1307gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1112#if PREFER_1_1 1309#if PREFER_1_1
1113 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1310 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1114#else 1311#else
1115 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1312 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1116#endif 1313#endif
1117 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1314 XPUSH_CLOBJ (stash_image2d, mem);
1118 1315
1119void 1316void
1120gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1317gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1121 PPCODE: 1318 PPCODE:
1122#if PREFER_1_1 1319#if PREFER_1_1
1123 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1320 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1124#else 1321#else
1125 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1322 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1126#endif 1323#endif
1127 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1324 XPUSH_CLOBJ (stash_image3d, mem);
1128 1325
1129#endif 1326#endif
1130 1327
1131void 1328void
1132supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1329supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1152 1349
1153void 1350void
1154sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1351sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1155 PPCODE: 1352 PPCODE:
1156 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1353 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1157 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1354 XPUSH_CLOBJ (stash_sampler, sampler);
1158 1355
1159void 1356void
1160program_with_source (OpenCL::Context self, SV *program) 1357program_with_source (OpenCL::Context self, SV *program)
1161 PPCODE: 1358 PPCODE:
1162 STRLEN len; 1359 STRLEN len;
1163 size_t len2; 1360 size_t len2;
1164 const char *ptr = SvPVbyte (program, len); 1361 const char *ptr = SvPVbyte (program, len);
1165 1362
1166 len2 = len; 1363 len2 = len;
1167 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1364 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1168 XPUSH_CLOBJ ("OpenCL::Program", prog); 1365 XPUSH_CLOBJ (stash_program, prog);
1366
1367void
1368program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1369 PPCODE:
1370 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1371 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects");
1372
1373 devices = SvRV (devices);
1374
1375 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1376 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1377
1378 binaries = SvRV (binaries);
1379
1380 if (av_len ((AV *)devices) != av_len ((AV *)binaries))
1381 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1382
1383 int count = av_len ((AV *)devices) + 1;
1384 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1385 size_t *length_list = tmpbuf (sizeof (*length_list) * count);
1386 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count);
1387 cl_int *status_list = tmpbuf (sizeof (*status_list) * count);
1388
1389 int i;
1390 for (i = 0; i < count; ++i)
1391 {
1392 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1393 STRLEN len;
1394 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1395 length_list [i] = len;
1396 }
1397
1398 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res));
1399
1400 EXTEND (SP, 2);
1401 PUSH_CLOBJ (stash_program, prog);
1402
1403 if (GIMME_V == G_ARRAY)
1404 {
1405 AV *av = newAV ();
1406 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1407
1408 for (i = count; i--; )
1409 av_store (av, i, newSViv (status_list [i]));
1410 }
1411
1412#if CL_VERSION_1_2
1413
1414void
1415program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1416 PPCODE:
1417 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1418 croak ("OpenCL::Context::program_with_built_in_kernels: devices must be specified as reference to an array of device objects");
1419
1420 devices = SvRV (devices);
1421
1422 int count = av_len ((AV *)devices) + 1;
1423 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1424
1425 int i;
1426 for (i = 0; i < count; ++i)
1427 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_built_in_kernels", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1428
1429 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, count, device_list, SvPVbyte_nolen (kernel_names), &res));
1430
1431 XPUSH_CLOBJ (stash_program, prog);
1432
1433#endif
1169 1434
1170#BEGIN:context 1435#BEGIN:context
1171 1436
1172void 1437void
1173reference_count (OpenCL::Context self) 1438reference_count (OpenCL::Context self)
1190 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1455 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1191 int i, n = size / sizeof (*value); 1456 int i, n = size / sizeof (*value);
1192 EXTEND (SP, n); 1457 EXTEND (SP, n);
1193 for (i = 0; i < n; ++i) 1458 for (i = 0; i < n; ++i)
1194 { 1459 {
1195 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1460 PUSH_CLOBJ (stash_device, value [i]);
1196 } 1461 }
1197 1462
1198void 1463void
1199properties (OpenCL::Context self) 1464properties (OpenCL::Context self)
1200 PPCODE: 1465 PPCODE:
1215DESTROY (OpenCL::Queue self) 1480DESTROY (OpenCL::Queue self)
1216 CODE: 1481 CODE:
1217 clReleaseCommandQueue (self); 1482 clReleaseCommandQueue (self);
1218 1483
1219void 1484void
1220enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1485read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1486 ALIAS:
1487 enqueue_read_buffer = 0
1221 PPCODE: 1488 PPCODE:
1222 cl_event ev = 0; 1489 cl_event ev = 0;
1223 EVENT_LIST (6, items - 6); 1490 EVENT_LIST (6);
1224 1491
1225 SvUPGRADE (data, SVt_PV); 1492 SvUPGRADE (data, SVt_PV);
1226 SvGROW (data, len); 1493 SvGROW (data, len);
1227 SvPOK_only (data); 1494 SvPOK_only (data);
1228 SvCUR_set (data, len); 1495 SvCUR_set (data, len);
1229 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1496 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1230 1497
1231 if (ev) 1498 if (ev)
1232 XPUSH_CLOBJ ("OpenCL::Event", ev); 1499 XPUSH_CLOBJ (stash_event, ev);
1233 1500
1234void 1501void
1235enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1502write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1503 ALIAS:
1504 enqueue_write_buffer = 0
1236 PPCODE: 1505 PPCODE:
1237 cl_event ev = 0; 1506 cl_event ev = 0;
1238 STRLEN len; 1507 STRLEN len;
1239 char *ptr = SvPVbyte (data, len); 1508 char *ptr = SvPVbyte (data, len);
1240 EVENT_LIST (5, items - 5); 1509 EVENT_LIST (5);
1241 1510
1242 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1511 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1243 1512
1244 if (ev) 1513 if (ev)
1245 XPUSH_CLOBJ ("OpenCL::Event", ev); 1514 XPUSH_CLOBJ (stash_event, ev);
1246 1515
1247#if CL_VERSION_1_2 1516#if CL_VERSION_1_2
1248 1517
1249void 1518void
1250enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1519fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1520 ALIAS:
1521 enqueue_fill_buffer = 0
1251 PPCODE: 1522 PPCODE:
1252 cl_event ev = 0; 1523 cl_event ev = 0;
1253 STRLEN len; 1524 STRLEN len;
1254 char *ptr = SvPVbyte (data, len); 1525 char *ptr = SvPVbyte (data, len);
1255 EVENT_LIST (5, items - 5); 1526 EVENT_LIST (5);
1256 1527
1257 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1528 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1258 1529
1259 if (ev) 1530 if (ev)
1260 XPUSH_CLOBJ ("OpenCL::Event", ev); 1531 XPUSH_CLOBJ (stash_event, ev);
1261 1532
1262void 1533void
1263enqueue_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, ...) 1534fill_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, ...)
1535 ALIAS:
1536 enqueue_fill_image = 0
1264 PPCODE: 1537 PPCODE:
1265 cl_event ev = 0; 1538 cl_event ev = 0;
1266 STRLEN len;
1267 const size_t origin [3] = { x, y, z }; 1539 const size_t origin [3] = { x, y, z };
1268 const size_t region [3] = { width, height, depth }; 1540 const size_t region [3] = { width, height, depth };
1269 EVENT_LIST (12, items - 12); 1541 EVENT_LIST (12);
1270 1542
1271 const cl_float c_f [4] = { r, g, b, a }; 1543 const cl_float c_f [4] = { r, g, b, a };
1272 const cl_uint c_u [4] = { r, g, b, a }; 1544 const cl_uint c_u [4] = { r, g, b, a };
1273 const cl_int c_s [4] = { r, g, b, a }; 1545 const cl_int c_s [4] = { r, g, b, a };
1274 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1546 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1277 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1549 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1278 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1550 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1279 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1551 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1280 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1552 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1281 1553
1282 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1554 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1283 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1555 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1284 1556
1285 if (ev) 1557 if (ev)
1286 XPUSH_CLOBJ ("OpenCL::Event", ev); 1558 XPUSH_CLOBJ (stash_event, ev);
1287 1559
1288#endif 1560#endif
1289 1561
1290void 1562void
1291enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1563copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1564 ALIAS:
1565 enqueue_copy_buffer = 0
1292 PPCODE: 1566 PPCODE:
1293 cl_event ev = 0; 1567 cl_event ev = 0;
1294 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
1295 1569
1296 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1570 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1297 1571
1298 if (ev) 1572 if (ev)
1299 XPUSH_CLOBJ ("OpenCL::Event", ev); 1573 XPUSH_CLOBJ (stash_event, ev);
1300 1574
1301void 1575void
1302enqueue_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, ...) 1576read_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, ...)
1577 ALIAS:
1578 enqueue_read_buffer_rect = 0
1303 PPCODE: 1579 PPCODE:
1304 cl_event ev = 0; 1580 cl_event ev = 0;
1305 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1581 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1306 const size_t host_origin[3] = { host_x, host_y, host_z }; 1582 const size_t host_origin[3] = { host_x, host_y, host_z };
1307 const size_t region[3] = { width, height, depth }; 1583 const size_t region[3] = { width, height, depth };
1308 EVENT_LIST (17, items - 17); 1584 EVENT_LIST (17);
1309 1585
1310 if (!buf_row_pitch) 1586 if (!buf_row_pitch)
1311 buf_row_pitch = region [0]; 1587 buf_row_pitch = region [0];
1312 1588
1313 if (!buf_slice_pitch) 1589 if (!buf_slice_pitch)
1326 SvPOK_only (data); 1602 SvPOK_only (data);
1327 SvCUR_set (data, len); 1603 SvCUR_set (data, len);
1328 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)); 1604 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));
1329 1605
1330 if (ev) 1606 if (ev)
1331 XPUSH_CLOBJ ("OpenCL::Event", ev); 1607 XPUSH_CLOBJ (stash_event, ev);
1332 1608
1333void 1609void
1334enqueue_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, ...) 1610write_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, ...)
1611 ALIAS:
1612 enqueue_write_buffer_rect = 0
1335 PPCODE: 1613 PPCODE:
1336 cl_event ev = 0; 1614 cl_event ev = 0;
1337 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1615 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1338 const size_t host_origin[3] = { host_x, host_y, host_z }; 1616 const size_t host_origin[3] = { host_x, host_y, host_z };
1339 const size_t region[3] = { width, height, depth }; 1617 const size_t region[3] = { width, height, depth };
1340 STRLEN len; 1618 STRLEN len;
1341 char *ptr = SvPVbyte (data, len); 1619 char *ptr = SvPVbyte (data, len);
1342 EVENT_LIST (17, items - 17); 1620 EVENT_LIST (17);
1343 1621
1344 if (!buf_row_pitch) 1622 if (!buf_row_pitch)
1345 buf_row_pitch = region [0]; 1623 buf_row_pitch = region [0];
1346 1624
1347 if (!buf_slice_pitch) 1625 if (!buf_slice_pitch)
1359 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1637 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1360 1638
1361 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)); 1639 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));
1362 1640
1363 if (ev) 1641 if (ev)
1364 XPUSH_CLOBJ ("OpenCL::Event", ev); 1642 XPUSH_CLOBJ (stash_event, ev);
1365 1643
1366void 1644void
1367enqueue_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, ...) 1645copy_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, ...)
1646 ALIAS:
1647 enqueue_copy_buffer_rect = 0
1368 PPCODE: 1648 PPCODE:
1369 cl_event ev = 0; 1649 cl_event ev = 0;
1370 const size_t src_origin[3] = { src_x, src_y, src_z }; 1650 const size_t src_origin[3] = { src_x, src_y, src_z };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1651 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth }; 1652 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16); 1653 EVENT_LIST (16);
1374 1654
1375 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1655 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));
1376 1656
1377 if (ev) 1657 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev); 1658 XPUSH_CLOBJ (stash_event, ev);
1379 1659
1380void 1660void
1381enqueue_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, ...) 1661read_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, ...)
1662 ALIAS:
1663 enqueue_read_image = 0
1382 PPCODE: 1664 PPCODE:
1383 cl_event ev = 0; 1665 cl_event ev = 0;
1384 const size_t src_origin[3] = { src_x, src_y, src_z }; 1666 const size_t src_origin[3] = { src_x, src_y, src_z };
1385 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1386 EVENT_LIST (12, items - 12); 1668 EVENT_LIST (12);
1387 1669
1388 if (!row_pitch) 1670 if (!row_pitch)
1389 row_pitch = img_row_pitch (src); 1671 row_pitch = img_row_pitch (src);
1390 1672
1391 if (depth > 1 && !slice_pitch) 1673 if (depth > 1 && !slice_pitch)
1398 SvPOK_only (data); 1680 SvPOK_only (data);
1399 SvCUR_set (data, len); 1681 SvCUR_set (data, len);
1400 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)); 1682 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));
1401 1683
1402 if (ev) 1684 if (ev)
1403 XPUSH_CLOBJ ("OpenCL::Event", ev); 1685 XPUSH_CLOBJ (stash_event, ev);
1404 1686
1405void 1687void
1406enqueue_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, ...) 1688write_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, ...)
1689 ALIAS:
1690 enqueue_write_image = 0
1407 PPCODE: 1691 PPCODE:
1408 cl_event ev = 0; 1692 cl_event ev = 0;
1409 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1693 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1410 const size_t region[3] = { width, height, depth }; 1694 const size_t region[3] = { width, height, depth };
1411 STRLEN len; 1695 STRLEN len;
1412 char *ptr = SvPVbyte (data, len); 1696 char *ptr = SvPVbyte (data, len);
1413 EVENT_LIST (12, items - 12); 1697 EVENT_LIST (12);
1414 1698
1415 if (!row_pitch) 1699 if (!row_pitch)
1416 row_pitch = img_row_pitch (dst); 1700 row_pitch = img_row_pitch (dst);
1417 1701
1418 if (depth > 1 && !slice_pitch) 1702 if (depth > 1 && !slice_pitch)
1424 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1708 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1425 1709
1426 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)); 1710 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));
1427 1711
1428 if (ev) 1712 if (ev)
1429 XPUSH_CLOBJ ("OpenCL::Event", ev); 1713 XPUSH_CLOBJ (stash_event, ev);
1430 1714
1431void 1715void
1432enqueue_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, ...) 1716copy_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, ...)
1717 ALIAS:
1718 enqueue_copy_image = 0
1433 PPCODE: 1719 PPCODE:
1434 cl_event ev = 0; 1720 cl_event ev = 0;
1435 const size_t src_origin[3] = { src_x, src_y, src_z }; 1721 const size_t src_origin[3] = { src_x, src_y, src_z };
1436 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1722 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1437 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
1438 EVENT_LIST (12, items - 12); 1724 EVENT_LIST (12);
1439 1725
1440 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1726 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1727
1442 if (ev) 1728 if (ev)
1443 XPUSH_CLOBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
1444 1730
1445void 1731void
1446enqueue_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, ...) 1732copy_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, ...)
1733 ALIAS:
1734 enqueue_copy_image_to_buffer = 0
1447 PPCODE: 1735 PPCODE:
1448 cl_event ev = 0; 1736 cl_event ev = 0;
1449 const size_t src_origin[3] = { src_x, src_y, src_z }; 1737 const size_t src_origin[3] = { src_x, src_y, src_z };
1738 const size_t region [3] = { width, height, depth };
1739 EVENT_LIST (10);
1740
1741 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1742
1743 if (ev)
1744 XPUSH_CLOBJ (stash_event, ev);
1745
1746void
1747copy_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, ...)
1748 ALIAS:
1749 enqueue_copy_buffer_to_image = 0
1750 PPCODE:
1751 cl_event ev = 0;
1752 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1753 const size_t region [3] = { width, height, depth };
1754 EVENT_LIST (10);
1755
1756 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1757
1758 if (ev)
1759 XPUSH_CLOBJ (stash_event, ev);
1760
1761void
1762map_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, ...)
1763 ALIAS:
1764 enqueue_map_buffer = 0
1765 PPCODE:
1766 cl_event ev;
1767 EVENT_LIST (6);
1768 size_t cb = SvIV (cb_);
1769
1770 if (!SvOK (cb_))
1771 {
1772 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1773 cb -= offset;
1774 }
1775
1776 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1777 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1778
1779void
1780map_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, ...)
1781 ALIAS:
1782 enqueue_map_image = 0
1783 PPCODE:
1784 size_t width = SvIV (width_);
1785 if (!SvOK (width_))
1786 {
1787 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1788 width -= x;
1789 }
1790
1791 size_t height = SvIV (width_);
1792 if (!SvOK (height_))
1793 {
1794 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1795 height -= y;
1796 }
1797
1798 size_t depth = SvIV (width_);
1799 if (!SvOK (depth_))
1800 {
1801 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1802 depth -= z;
1803
1804 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1805 if (!depth)
1806 depth = 1;
1807 }
1808
1809 const size_t origin[3] = { x, y, z };
1450 const size_t region[3] = { width, height, depth }; 1810 const size_t region[3] = { width, height, depth };
1811 size_t row_pitch, slice_pitch;
1451 EVENT_LIST (10, items - 10); 1812 EVENT_LIST (10);
1452 1813
1814 cl_event ev;
1815 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));
1816
1817 size_t cb = slice_pitch ? slice_pitch * region [2]
1818 : row_pitch ? row_pitch * region [1]
1819 : region [0];
1820
1821 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1822
1823void
1824unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1825 PPCODE:
1826 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1827 if (GIMME_V != G_VOID)
1828 {
1829 clRetainEvent (mapped->event);
1830 XPUSH_CLOBJ (stash_event, mapped->event);
1831 }
1832
1833void
1834task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1835 ALIAS:
1836 enqueue_task = 0
1837 PPCODE:
1838 cl_event ev = 0;
1839 EVENT_LIST (2);
1840
1453 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1841 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1454 1842
1455 if (ev) 1843 if (ev)
1456 XPUSH_CLOBJ ("OpenCL::Event", ev); 1844 XPUSH_CLOBJ (stash_event, ev);
1457 1845
1458void 1846void
1459enqueue_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, ...)
1460 PPCODE:
1461 cl_event ev = 0;
1462 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1463 const size_t region[3] = { width, height, depth };
1464 EVENT_LIST (10, items - 10);
1465
1466 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1467
1468 if (ev)
1469 XPUSH_CLOBJ ("OpenCL::Event", ev);
1470
1471void
1472enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1473 PPCODE:
1474 cl_event ev = 0;
1475 EVENT_LIST (2, items - 2);
1476
1477 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1478
1479 if (ev)
1480 XPUSH_CLOBJ ("OpenCL::Event", ev);
1481
1482void
1483enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1847nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1848 ALIAS:
1849 enqueue_nd_range_kernel = 0
1484 PPCODE: 1850 PPCODE:
1485 cl_event ev = 0; 1851 cl_event ev = 0;
1486 size_t *gwo = 0, *gws, *lws = 0; 1852 size_t *gwo = 0, *gws, *lws = 0;
1487 int gws_len; 1853 int gws_len;
1488 size_t *lists; 1854 size_t *lists;
1489 int i; 1855 int i;
1490 EVENT_LIST (5, items - 5); 1856 EVENT_LIST (5);
1491 1857
1492 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1858 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1493 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1859 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1494 1860
1495 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1861 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1496 1862
1497 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1863 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1498 1864
1499 gws = lists + gws_len * 0; 1865 gws = lists + gws_len * 0;
1500 for (i = 0; i < gws_len; ++i) 1866 for (i = 0; i < gws_len; ++i)
1867 {
1501 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1868 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1869 // at least nvidia crashes for 0-sized work group sizes, work around
1870 if (!gws [i])
1871 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1872 }
1502 1873
1503 if (SvOK (global_work_offset)) 1874 if (SvOK (global_work_offset))
1504 { 1875 {
1505 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1876 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1506 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1877 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1514 } 1885 }
1515 1886
1516 if (SvOK (local_work_size)) 1887 if (SvOK (local_work_size))
1517 { 1888 {
1518 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1889 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1519 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1890 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1520 1891
1521 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1892 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1522 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1893 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1523 1894
1524 lws = lists + gws_len * 2; 1895 lws = lists + gws_len * 2;
1525 for (i = 0; i < gws_len; ++i) 1896 for (i = 0; i < gws_len; ++i)
1897 {
1526 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1898 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1899 // at least nvidia crashes for 0-sized work group sizes, work around
1900 if (!lws [i])
1901 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1902 }
1527 } 1903 }
1528 1904
1529 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1905 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1530 1906
1531 if (ev) 1907 if (ev)
1532 XPUSH_CLOBJ ("OpenCL::Event", ev); 1908 XPUSH_CLOBJ (stash_event, ev);
1909
1910#if CL_VERSION_1_2
1911
1912void
1913migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1914 ALIAS:
1915 enqueue_migrate_mem_objects = 0
1916 PPCODE:
1917 EVENT_LIST (3);
1918
1919 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1920 croak ("OpenCL::Queue::migrate_mem_objects: objects must be an array reference with OpenCL::Memory objects");
1921
1922 objects = SvRV (objects);
1923
1924 cl_uint object_count = av_len ((AV *)objects) + 1;
1925 cl_mem *object_list = tmpbuf (sizeof (*object_list) * object_count);
1926
1927 int i;
1928 for (i = object_count; i--; )
1929 object_list [i] = SvCLOBJ ("OpenCL::Queue::migrate_mem_objects", "objects", *av_fetch ((AV *)objects, i, 0), "OpenCL::Memory");
1930
1931 cl_event ev = 0;
1932 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, object_count, object_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1933
1934 if (ev)
1935 XPUSH_CLOBJ (stash_event, ev);
1936
1937#endif
1533 1938
1534#if cl_apple_gl_sharing || cl_khr_gl_sharing 1939#if cl_apple_gl_sharing || cl_khr_gl_sharing
1535 1940
1536void 1941void
1537enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1942acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1538 ALIAS: 1943 ALIAS:
1944 release_gl_objects = 1
1945 enqueue_acquire_gl_objects = 0
1539 enqueue_release_gl_objects = 1 1946 enqueue_release_gl_objects = 1
1540 PPCODE: 1947 PPCODE:
1541 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1948 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1542 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1949 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1543 1950
1544 cl_event ev = 0; 1951 cl_event ev = 0;
1545 EVENT_LIST (2, items - 2); 1952 EVENT_LIST (2);
1546 AV *av = (AV *)SvRV (objects); 1953 AV *av = (AV *)SvRV (objects);
1547 cl_uint num_objects = av_len (av) + 1; 1954 cl_uint num_objects = av_len (av) + 1;
1548 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1955 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1956
1549 int i; 1957 int i;
1550
1551 for (i = num_objects; i--; ) 1958 for (i = num_objects; i--; )
1552 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1959 object_list [i] = SvCLOBJ ("OpenCL::Queue::acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1553 1960
1554 if (ix) 1961 if (ix)
1555 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1962 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1556 else 1963 else
1557 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1964 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1558 1965
1559 if (ev) 1966 if (ev)
1560 XPUSH_CLOBJ ("OpenCL::Event", ev); 1967 XPUSH_CLOBJ (stash_event, ev);
1561 1968
1562#endif 1969#endif
1563 1970
1564void 1971void
1565enqueue_wait_for_events (OpenCL::Queue self, ...) 1972wait_for_events (OpenCL::Queue self, ...)
1973 ALIAS:
1974 enqueue_wait_for_events = 0
1566 CODE: 1975 CODE:
1567 EVENT_LIST (1, items - 1); 1976 EVENT_LIST (1);
1568#if PREFER_1_1 1977#if PREFER_1_1
1569 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1978 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1570#else 1979#else
1571 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 1980 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1572#endif 1981#endif
1573 1982
1574void 1983void
1575enqueue_marker (OpenCL::Queue self, ...) 1984marker (OpenCL::Queue self, ...)
1985 ALIAS:
1986 enqueue_marker = 0
1576 PPCODE: 1987 PPCODE:
1577 cl_event ev = 0; 1988 cl_event ev = 0;
1578 EVENT_LIST (1, items - 1); 1989 EVENT_LIST (1);
1579#if PREFER_1_1 1990#if PREFER_1_1
1580 if (!event_list_count) 1991 if (!event_list_count)
1581 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 1992 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1582 else 1993 else
1583#if CL_VERSION_1_2 1994#if CL_VERSION_1_2
1590#endif 2001#endif
1591#else 2002#else
1592 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2003 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1593#endif 2004#endif
1594 if (ev) 2005 if (ev)
1595 XPUSH_CLOBJ ("OpenCL::Event", ev); 2006 XPUSH_CLOBJ (stash_event, ev);
1596 2007
1597void 2008void
1598enqueue_barrier (OpenCL::Queue self, ...) 2009barrier (OpenCL::Queue self, ...)
2010 ALIAS:
2011 enqueue_barrier = 0
1599 PPCODE: 2012 PPCODE:
1600 cl_event ev = 0; 2013 cl_event ev = 0;
1601 EVENT_LIST (1, items - 1); 2014 EVENT_LIST (1);
1602#if PREFER_1_1 2015#if PREFER_1_1
1603 if (!event_list_count && GIMME_V == G_VOID) 2016 if (!event_list_count && GIMME_V == G_VOID)
1604 NEED_SUCCESS (EnqueueBarrier, (self)); 2017 NEED_SUCCESS (EnqueueBarrier, (self));
1605 else 2018 else
1606#if CL_VERSION_1_2 2019#if CL_VERSION_1_2
1616#endif 2029#endif
1617#else 2030#else
1618 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2031 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1619#endif 2032#endif
1620 if (ev) 2033 if (ev)
1621 XPUSH_CLOBJ ("OpenCL::Event", ev); 2034 XPUSH_CLOBJ (stash_event, ev);
1622 2035
1623void 2036void
1624flush (OpenCL::Queue self) 2037flush (OpenCL::Queue self)
1625 CODE: 2038 CODE:
1626 NEED_SUCCESS (Flush, (self)); 2039 NEED_SUCCESS (Flush, (self));
1644 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2057 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1645 EXTEND (SP, 1); 2058 EXTEND (SP, 1);
1646 const int i = 0; 2059 const int i = 0;
1647 { 2060 {
1648 NEED_SUCCESS (RetainContext, (value [i])); 2061 NEED_SUCCESS (RetainContext, (value [i]));
1649 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2062 PUSH_CLOBJ (stash_context, value [i]);
1650 } 2063 }
1651 2064
1652void 2065void
1653device (OpenCL::Queue self) 2066device (OpenCL::Queue self)
1654 PPCODE: 2067 PPCODE:
1655 cl_device_id value [1]; 2068 cl_device_id value [1];
1656 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2069 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1657 EXTEND (SP, 1); 2070 EXTEND (SP, 1);
1658 const int i = 0; 2071 const int i = 0;
1659 { 2072 {
1660 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2073 PUSH_CLOBJ (stash_device, value [i]);
1661 } 2074 }
1662 2075
1663void 2076void
1664reference_count (OpenCL::Queue self) 2077reference_count (OpenCL::Queue self)
1665 PPCODE: 2078 PPCODE:
1674 PPCODE: 2087 PPCODE:
1675 cl_command_queue_properties value [1]; 2088 cl_command_queue_properties value [1];
1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2089 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1677 EXTEND (SP, 1); 2090 EXTEND (SP, 1);
1678 const int i = 0; 2091 const int i = 0;
1679 PUSHs (sv_2mortal (newSViv (value [i]))); 2092 PUSHs (sv_2mortal (newSVuv (value [i])));
1680 2093
1681#END:command_queue 2094#END:command_queue
1682 2095
1683MODULE = OpenCL PACKAGE = OpenCL::Memory 2096MODULE = OpenCL PACKAGE = OpenCL::Memory
1684 2097
1699 PPCODE: 2112 PPCODE:
1700 cl_mem_object_type value [1]; 2113 cl_mem_object_type value [1];
1701 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2114 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1702 EXTEND (SP, 1); 2115 EXTEND (SP, 1);
1703 const int i = 0; 2116 const int i = 0;
1704 PUSHs (sv_2mortal (newSViv (value [i]))); 2117 PUSHs (sv_2mortal (newSVuv (value [i])));
1705 2118
1706void 2119void
1707flags (OpenCL::Memory self) 2120flags (OpenCL::Memory self)
1708 PPCODE: 2121 PPCODE:
1709 cl_mem_flags value [1]; 2122 cl_mem_flags value [1];
1710 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2123 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1711 EXTEND (SP, 1); 2124 EXTEND (SP, 1);
1712 const int i = 0; 2125 const int i = 0;
1713 PUSHs (sv_2mortal (newSViv (value [i]))); 2126 PUSHs (sv_2mortal (newSVuv (value [i])));
1714 2127
1715void 2128void
1716size (OpenCL::Memory self) 2129size (OpenCL::Memory self)
1717 ALIAS: 2130 ALIAS:
1718 size = CL_MEM_SIZE 2131 size = CL_MEM_SIZE
1752 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2165 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1753 EXTEND (SP, 1); 2166 EXTEND (SP, 1);
1754 const int i = 0; 2167 const int i = 0;
1755 { 2168 {
1756 NEED_SUCCESS (RetainContext, (value [i])); 2169 NEED_SUCCESS (RetainContext, (value [i]));
1757 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2170 PUSH_CLOBJ (stash_context, value [i]);
1758 } 2171 }
1759 2172
1760void 2173void
1761associated_memobject (OpenCL::Memory self) 2174associated_memobject (OpenCL::Memory self)
1762 PPCODE: 2175 PPCODE:
1764 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2177 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1765 EXTEND (SP, 1); 2178 EXTEND (SP, 1);
1766 const int i = 0; 2179 const int i = 0;
1767 { 2180 {
1768 NEED_SUCCESS (RetainMemObject, (value [i])); 2181 NEED_SUCCESS (RetainMemObject, (value [i]));
1769 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2182 PUSH_CLOBJ (stash_memory, value [i]);
1770 } 2183 }
1771 2184
1772#END:mem 2185#END:mem
1773 2186
1774#if cl_apple_gl_sharing || cl_khr_gl_sharing 2187#if cl_apple_gl_sharing || cl_khr_gl_sharing
1794 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2207 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1795 2208
1796 cl_buffer_region crdata = { origin, size }; 2209 cl_buffer_region crdata = { origin, size };
1797 2210
1798 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2211 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1799 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2212 XPUSH_CLOBJ (stash_buffer, mem);
1800 2213
1801MODULE = OpenCL PACKAGE = OpenCL::Image 2214MODULE = OpenCL PACKAGE = OpenCL::Image
1802 2215
1803void 2216void
1804image_info (OpenCL::Image self, cl_image_info name) 2217image_info (OpenCL::Image self, cl_image_info name)
1890 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2303 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1891 EXTEND (SP, 1); 2304 EXTEND (SP, 1);
1892 const int i = 0; 2305 const int i = 0;
1893 { 2306 {
1894 NEED_SUCCESS (RetainContext, (value [i])); 2307 NEED_SUCCESS (RetainContext, (value [i]));
1895 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2308 PUSH_CLOBJ (stash_context, value [i]);
1896 } 2309 }
1897 2310
1898void 2311void
1899normalized_coords (OpenCL::Sampler self) 2312normalized_coords (OpenCL::Sampler self)
1900 PPCODE: 2313 PPCODE:
1901 cl_addressing_mode value [1]; 2314 cl_addressing_mode value [1];
1902 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2315 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1903 EXTEND (SP, 1); 2316 EXTEND (SP, 1);
1904 const int i = 0; 2317 const int i = 0;
1905 PUSHs (sv_2mortal (newSViv (value [i]))); 2318 PUSHs (sv_2mortal (newSVuv (value [i])));
1906 2319
1907void 2320void
1908addressing_mode (OpenCL::Sampler self) 2321addressing_mode (OpenCL::Sampler self)
1909 PPCODE: 2322 PPCODE:
1910 cl_filter_mode value [1]; 2323 cl_filter_mode value [1];
1911 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2324 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1912 EXTEND (SP, 1); 2325 EXTEND (SP, 1);
1913 const int i = 0; 2326 const int i = 0;
1914 PUSHs (sv_2mortal (newSViv (value [i]))); 2327 PUSHs (sv_2mortal (newSVuv (value [i])));
1915 2328
1916void 2329void
1917filter_mode (OpenCL::Sampler self) 2330filter_mode (OpenCL::Sampler self)
1918 PPCODE: 2331 PPCODE:
1919 cl_bool value [1]; 2332 cl_bool value [1];
2011 2424
2012void 2425void
2013kernel (OpenCL::Program program, SV *function) 2426kernel (OpenCL::Program program, SV *function)
2014 PPCODE: 2427 PPCODE:
2015 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2428 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2016 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2429 XPUSH_CLOBJ (stash_kernel, kernel);
2017 2430
2018void 2431void
2019kernels_in_program (OpenCL::Program program) 2432kernels_in_program (OpenCL::Program program)
2020 PPCODE: 2433 PPCODE:
2021 cl_uint num_kernels; 2434 cl_uint num_kernels;
2024 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2437 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2025 2438
2026 int i; 2439 int i;
2027 EXTEND (SP, num_kernels); 2440 EXTEND (SP, num_kernels);
2028 for (i = 0; i < num_kernels; ++i) 2441 for (i = 0; i < num_kernels; ++i)
2029 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2442 PUSH_CLOBJ (stash_kernel, kernels [i]);
2030 2443
2031void 2444void
2032info (OpenCL::Program self, cl_program_info name) 2445info (OpenCL::Program self, cl_program_info name)
2033 PPCODE: 2446 PPCODE:
2034 INFO (Program) 2447 INFO (Program)
2082 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2495 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2083 EXTEND (SP, 1); 2496 EXTEND (SP, 1);
2084 const int i = 0; 2497 const int i = 0;
2085 { 2498 {
2086 NEED_SUCCESS (RetainContext, (value [i])); 2499 NEED_SUCCESS (RetainContext, (value [i]));
2087 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2500 PUSH_CLOBJ (stash_context, value [i]);
2088 } 2501 }
2089 2502
2090void 2503void
2091devices (OpenCL::Program self) 2504devices (OpenCL::Program self)
2092 PPCODE: 2505 PPCODE:
2096 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2509 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2097 int i, n = size / sizeof (*value); 2510 int i, n = size / sizeof (*value);
2098 EXTEND (SP, n); 2511 EXTEND (SP, n);
2099 for (i = 0; i < n; ++i) 2512 for (i = 0; i < n; ++i)
2100 { 2513 {
2101 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2514 PUSH_CLOBJ (stash_device, value [i]);
2102 } 2515 }
2103 2516
2104void 2517void
2105source (OpenCL::Program self) 2518source (OpenCL::Program self)
2106 PPCODE: 2519 PPCODE:
2132DESTROY (OpenCL::Kernel self) 2545DESTROY (OpenCL::Kernel self)
2133 CODE: 2546 CODE:
2134 clReleaseKernel (self); 2547 clReleaseKernel (self);
2135 2548
2136void 2549void
2550setf (OpenCL::Kernel self, const char *format, ...)
2551 CODE:
2552 int i;
2553 for (i = 2; ; ++i)
2554 {
2555 while (*format == ' ')
2556 ++format;
2557
2558 char type = *format++;
2559
2560 if (!type)
2561 break;
2562
2563 if (i >= items)
2564 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2565
2566 SV *sv = ST (i);
2567
2568 union
2569 {
2570 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2571 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2572 cl_half ch; cl_float cf; cl_double cd;
2573 cl_mem cm;
2574 cl_sampler ca;
2575 size_t cz;
2576 cl_event ce;
2577 } arg;
2578 size_t size;
2579 int nullarg = 0;
2580
2581 switch (type)
2582 {
2583 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2584 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2585 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2586 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2587 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2588 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2589 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2590 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2591
2592 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2593 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2594 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2595
2596 case 'z': nullarg = 1; size = SvIV (sv); break;
2597
2598 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2599 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2600 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2601
2602 default:
2603 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2604 }
2605
2606 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2607 if (res)
2608 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2609 }
2610
2611 if (i != items)
2612 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2613
2614void
2137set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2615set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2138 CODE: 2616 CODE:
2139 clSetKernelArg (self, idx, sizeof (value), &value); 2617 clSetKernelArg (self, idx, sizeof (value), &value);
2140 2618
2141void 2619void
2189 clSetKernelArg (self, idx, sizeof (value), &value); 2667 clSetKernelArg (self, idx, sizeof (value), &value);
2190 2668
2191void 2669void
2192set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2670set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2193 CODE: 2671 CODE:
2194 clSetKernelArg (self, idx, sizeof (value), &value); 2672 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2195 2673
2196void 2674void
2197set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2675set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2198 CODE: 2676 CODE:
2199 clSetKernelArg (self, idx, sizeof (value), &value); 2677 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2200 2678
2201void 2679void
2202set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2680set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2681 ALIAS:
2682 set_image2d = 0
2683 set_image3d = 0
2203 CODE: 2684 CODE:
2204 clSetKernelArg (self, idx, sizeof (value), &value); 2685 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2205
2206void
2207set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
2208 CODE:
2209 clSetKernelArg (self, idx, sizeof (value), &value);
2210 2686
2211void 2687void
2212set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2688set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2213 CODE: 2689 CODE:
2214 clSetKernelArg (self, idx, sizeof (value), &value); 2690 clSetKernelArg (self, idx, sizeof (value), &value);
2260 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2736 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2261 EXTEND (SP, 1); 2737 EXTEND (SP, 1);
2262 const int i = 0; 2738 const int i = 0;
2263 { 2739 {
2264 NEED_SUCCESS (RetainContext, (value [i])); 2740 NEED_SUCCESS (RetainContext, (value [i]));
2265 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2741 PUSH_CLOBJ (stash_context, value [i]);
2266 } 2742 }
2267 2743
2268void 2744void
2269program (OpenCL::Kernel self) 2745program (OpenCL::Kernel self)
2270 PPCODE: 2746 PPCODE:
2272 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2748 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2273 EXTEND (SP, 1); 2749 EXTEND (SP, 1);
2274 const int i = 0; 2750 const int i = 0;
2275 { 2751 {
2276 NEED_SUCCESS (RetainProgram, (value [i])); 2752 NEED_SUCCESS (RetainProgram, (value [i]));
2277 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2753 PUSH_CLOBJ (stash_program, value [i]);
2278 } 2754 }
2279 2755
2280#END:kernel 2756#END:kernel
2281 2757
2282void 2758void
2329 const int i = 0; 2805 const int i = 0;
2330 PUSHs (sv_2mortal (newSVuv (value [i]))); 2806 PUSHs (sv_2mortal (newSVuv (value [i])));
2331 2807
2332#END:kernel_work_group 2808#END:kernel_work_group
2333 2809
2810#if CL_VERSION_1_2
2811
2812void
2813arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2814 PPCODE:
2815 size_t size;
2816 NEED_SUCCESS (GetKernelArgInfo, (self, name, 0, 0, &size));
2817 SV *sv = sv_2mortal (newSV (size));
2818 SvUPGRADE (sv, SVt_PV);
2819 SvPOK_only (sv);
2820 SvCUR_set (sv, size);
2821 NEED_SUCCESS (GetKernelArgInfo, (self, name, size, SvPVX (sv), 0));
2822 XPUSHs (sv);
2823
2824#BEGIN:kernel_arg
2825
2826#END:kernel_arg
2827
2828#endif
2829
2334MODULE = OpenCL PACKAGE = OpenCL::Event 2830MODULE = OpenCL PACKAGE = OpenCL::Event
2335 2831
2336void 2832void
2337DESTROY (OpenCL::Event self) 2833DESTROY (OpenCL::Event self)
2338 CODE: 2834 CODE:
2362 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2858 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2363 EXTEND (SP, 1); 2859 EXTEND (SP, 1);
2364 const int i = 0; 2860 const int i = 0;
2365 { 2861 {
2366 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2862 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2367 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2863 PUSH_CLOBJ (stash_queue, value [i]);
2368 } 2864 }
2369 2865
2370void 2866void
2371command_type (OpenCL::Event self) 2867command_type (OpenCL::Event self)
2372 PPCODE: 2868 PPCODE:
2395 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2891 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2396 EXTEND (SP, 1); 2892 EXTEND (SP, 1);
2397 const int i = 0; 2893 const int i = 0;
2398 { 2894 {
2399 NEED_SUCCESS (RetainContext, (value [i])); 2895 NEED_SUCCESS (RetainContext, (value [i]));
2400 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2896 PUSH_CLOBJ (stash_context, value [i]);
2401 } 2897 }
2402 2898
2403#END:event 2899#END:event
2404 2900
2405void 2901void
2430void 2926void
2431set_status (OpenCL::UserEvent self, cl_int execution_status) 2927set_status (OpenCL::UserEvent self, cl_int execution_status)
2432 CODE: 2928 CODE:
2433 clSetUserEventStatus (self, execution_status); 2929 clSetUserEventStatus (self, execution_status);
2434 2930
2931MODULE = OpenCL PACKAGE = OpenCL::Mapped
2932
2933void
2934DESTROY (SV *self)
2935 CODE:
2936 OpenCL__Mapped mapped = SvMAPPED (self);
2937
2938 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2939 mapped_detach (self, mapped);
2940
2941 clReleaseCommandQueue (mapped->queue);
2942 clReleaseEvent (mapped->event);
2943 Safefree (mapped);
2944
2945void
2946unmap (OpenCL::Mapped self, ...)
2947 CODE:
2948 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2949
2950bool
2951mapped (OpenCL::Mapped self)
2952 CODE:
2953 RETVAL = !!self->ptr;
2954 OUTPUT:
2955 RETVAL
2956
2957void
2958wait (OpenCL::Mapped self)
2959 PPCODE:
2960 if (self->event)
2961 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2962
2963void
2964event (OpenCL::Mapped self)
2965 PPCODE:
2966 if (!self->event)
2967 XSRETURN_UNDEF;
2968
2969 clRetainEvent (self->event);
2970 XPUSH_CLOBJ (stash_event, self->event);
2971
2972size_t
2973size (OpenCL::Mapped self)
2974 CODE:
2975 RETVAL = self->cb;
2976 OUTPUT:
2977 RETVAL
2978
2979IV
2980ptr (OpenCL::Mapped self)
2981 CODE:
2982 RETVAL = PTR2IV (self->ptr);
2983 OUTPUT:
2984 RETVAL
2985
2986void
2987set (OpenCL::Mapped self, size_t offset, SV *data)
2988 CODE:
2989 STRLEN len;
2990 const char *ptr = SvPVbyte (data, len);
2991
2992 if (offset + len > self->cb)
2993 croak ("OpenCL::Mapped::set out of bound condition detected");
2994
2995 memcpy (offset + (char *)self->ptr, ptr, len);
2996
2997MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2998
2999MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3000
3001IV
3002row_pitch (OpenCL::Mapped self)
3003 ALIAS:
3004 slice_pitch = 1
3005 CODE:
3006 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3007 OUTPUT:
3008 RETVAL
3009
3010
3011

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines