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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.43 by root, Sat Apr 21 22:16:09 2012 UTC vs.
Revision 1.68 by root, Fri May 4 14:56:50 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
4 10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
7#endif 13#endif
8 14
11// this. This breaks 100% of the opencl 1.1 apps, for what reason? 17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
12// after all, the functions are deprecated, not removed. 18// after all, the functions are deprecated, not removed.
13// in addition, you cannot test for this in any future-proof way. 19// in addition, you cannot test for this in any future-proof way.
14// each time a new opencl version comes out, you need to make a new 20// each time a new opencl version comes out, you need to make a new
15// release. 21// release.
16#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
17#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */ 22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
30#endif
18 31
19#ifdef __APPLE__ 32#ifdef __APPLE__
20 #include <OpenCL/opencl.h> 33 #include <OpenCL/opencl.h>
21#else 34#else
22 #include <CL/opencl.h> 35 #include <CL/opencl.h>
23#endif 36#endif
24 37
25#if 0 // testing
26#undef CL_USE_DEPRECATED_OPENCL_1_1_APIS
27#undef CL_VERSION_1_2 38#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1
40 #define PREFER_1_1 1
28#endif 41#endif
29 42
30typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
31typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
32typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
33typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
34typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
35typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
36typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
37typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
38typedef cl_mem OpenCL__Image2D;
39typedef cl_mem OpenCL__Image3D;
40typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
41typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
42typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
43typedef cl_mem OpenCL__Image2D_ornull;
44typedef cl_mem OpenCL__Image3D_ornull;
45typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
46typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
47typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
48typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
49typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
50 60
61typedef struct mapped * OpenCL__Mapped;
62
51typedef 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;
52 89
53/*****************************************************************************/ 90/*****************************************************************************/
54 91
55// name must include a leading underscore 92// name must include a leading underscore
56// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
96} 133}
97 134
98/*****************************************************************************/ 135/*****************************************************************************/
99 136
100/* up to two temporary buffers */ 137/* up to two temporary buffers */
101static void * 138static void * ecb_noinline
102tmpbuf (size_t size) 139tmpbuf (size_t size)
103{ 140{
104 enum { buffers = 3 }; 141 enum { buffers = 4 };
105 static int idx; 142 static int idx;
106 static void *buf [buffers]; 143 static void *buf [buffers];
107 static size_t len [buffers]; 144 static size_t len [buffers];
108 145
109 idx = (idx + 1) % buffers; 146 idx = (idx + 1) % buffers;
182 if (res) \ 219 if (res) \
183 FAIL (name); 220 FAIL (name);
184 221
185/*****************************************************************************/ 222/*****************************************************************************/
186 223
187#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
188#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
189
190static void * 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
191SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 236SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
192{ 237{
238 // sv_derived_from is quite slow :(
193 if (SvROK (sv) && sv_derived_from (sv, pkg)) 239 if (SvROK (sv) && sv_derived_from (sv, pkg))
194 return (void *)SvIV (SvRV (sv)); 240 return (void *)SvIV (SvRV (sv));
195 241
196 croak ("%s: %s is not of type %s", func, svname, pkg); 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));
197} 253}
198 254
199/*****************************************************************************/ 255/*****************************************************************************/
200 256
201static cl_context_properties * 257static cl_context_properties * ecb_noinline
202SvCONTEXTPROPERTIES (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)
203{ 259{
204 if (!sv || !SvOK (sv)) 260 if (!sv || !SvOK (sv))
205 if (extra) 261 if (extra)
206 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
213 int i, len = av_len (av) + 1; 269 int i, len = av_len (av) + 1;
214 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));
215 cl_context_properties *l = p; 271 cl_context_properties *l = p;
216 272
217 if (len & 1) 273 if (len & 1)
218 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);
219 275
220 while (extracount--) 276 while (extracount--)
221 *l++ = *extra++; 277 *l++ = *extra++;
222 278
223 for (i = 0; i < len; i += 2) 279 for (i = 0; i < len; i += 2)
226 SV *p_sv = *av_fetch (av, i + 1, 0); 282 SV *p_sv = *av_fetch (av, i + 1, 0);
227 cl_context_properties v = SvIV (p_sv); // code below can override 283 cl_context_properties v = SvIV (p_sv); // code below can override
228 284
229 switch (t) 285 switch (t)
230 { 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
231 case CL_GLX_DISPLAY_KHR: 292 case CL_GLX_DISPLAY_KHR:
232 if (!SvOK (p_sv)) 293 if (!SvOK (p_sv))
233 { 294 {
234 void *func = glsym ("_glXGetCurrentDisplay"); 295 void *func = glsym ("_glXGetCurrentDisplay");
235 if (func) 296 if (func)
262 323
263 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);
264} 325}
265 326
266/*****************************************************************************/ 327/*****************************************************************************/
328/* callback stuff */
329
330/* default context callback, log to stderr */
331static void CL_CALLBACK
332context_default_notify (const char *msg, const void *info, size_t cb, void *data)
333{
334 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
335}
336
337typedef struct
338{
339 int free_cb;
340 void (*push)(void *data1, void *data2, void *data3);
341} eq_vtbl;
342
343typedef struct eq_item
344{
345 struct eq_item *next;
346 eq_vtbl *vtbl;
347 SV *cb;
348 void *data1, *data2, *data3;
349} eq_item;
350
351static void (*eq_signal_func)(void *signal_arg, int value);
352static void *eq_signal_arg;
353static xmutex_t eq_lock = X_MUTEX_INIT;
354static eq_item *eq_head, *eq_tail;
355
356static void ecb_noinline
357eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
358{
359 eq_item *item = malloc (sizeof (eq_item));
360
361 item->next = 0;
362 item->vtbl = vtbl;
363 item->cb = cb;
364 item->data1 = data1;
365 item->data2 = data2;
366 item->data3 = data3;
367
368 X_LOCK (eq_lock);
369
370 *(eq_head ? &eq_tail->next : &eq_head) = item;
371 eq_tail = item;
372
373 X_UNLOCK (eq_lock);
374
375 eq_signal_func (eq_signal_arg, 0);
376}
377
378static eq_item *
379eq_dec (void)
380{
381 eq_item *res;
382
383 X_LOCK (eq_lock);
384
385 res = eq_head;
386 if (res)
387 eq_head = res->next;
388
389 X_UNLOCK (eq_lock);
390
391 return res;
392}
393
394static void
395eq_poll (void)
396{
397 eq_item *item;
398
399 while ((item = eq_dec ()))
400 {
401 ENTER;
402 SAVETMPS;
403
404 dSP;
405 PUSHMARK (SP);
406 EXTEND (SP, 2);
407
408 if (item->vtbl->free_cb)
409 sv_2mortal (item->cb);
410
411 PUTBACK;
412 item->vtbl->push (item->data1, item->data2, item->data3);
413
414 SV *cb = item->cb;
415 free (item);
416
417 call_sv (cb, G_DISCARD | G_VOID);
418
419 FREETMPS;
420 LEAVE;
421 }
422}
423
424static void
425eq_poll_interrupt (pTHX_ void *c_arg, int value)
426{
427 eq_poll ();
428}
429
430/*****************************************************************************/
431/* context notify */
432
433static void ecb_noinline
434eq_context_push (void *data1, void *data2, void *data3)
435{
436 dSP;
437 PUSHs (sv_2mortal (newSVpv (data1, 0)));
438 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
439 PUTBACK;
440
441 free (data1);
442 free (data2);
443}
444
445static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
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/*****************************************************************************/
479/* build/compile/link notify */
480
481static void
482eq_program_push (void *data1, void *data2, void *data3)
483{
484 dSP;
485 PUSH_CLOBJ (stash_program, data1);
486 PUTBACK;
487}
488
489static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
490
491static void CL_CALLBACK
492eq_program_notify (cl_program program, void *user_data)
493{
494 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
495}
496
497struct build_args
498{
499 cl_program program;
500 char *options;
501 void *user_data;
502 cl_uint num_devices;
503};
504
505X_THREAD_PROC (build_program_thread)
506{
507 struct build_args *arg = thr_arg;
508
509 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
510
511 if (arg->user_data)
512 eq_program_notify (arg->program, arg->user_data);
513 else
514 clReleaseProgram (arg->program);
515
516 free (arg->options);
517 free (arg);
518
519 return 0;
520}
521
522static void
523build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
524{
525 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
526
527 arg->program = program;
528 arg->options = strdup (options);
529 arg->user_data = user_data;
530 arg->num_devices = num_devices;
531 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
532
533 xthread_t id;
534 thread_create (&id, build_program_thread, arg);
535}
536
537/*****************************************************************************/
538/* event objects */
539
540static void
541eq_event_push (void *data1, void *data2, void *data3)
542{
543 dSP;
544 PUSH_CLOBJ (stash_event, data1);
545 PUSHs (sv_2mortal (newSViv ((IV)data2)));
546 PUTBACK;
547}
548
549static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
550
551static void CL_CALLBACK
552eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
553{
554 clRetainEvent (event);
555 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
556}
557
558/*****************************************************************************/
559/* utilities for XS code */
267 560
268static size_t 561static size_t
269img_row_pitch (cl_mem img) 562img_row_pitch (cl_mem img)
270{ 563{
271 size_t res; 564 size_t res;
272 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 565 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
273 return res; 566 return res;
274} 567}
275 568
276static cl_event * 569static cl_event * ecb_noinline
277event_list (SV **items, cl_uint *rcount) 570event_list (SV **items, cl_uint *rcount, cl_event extra)
278{ 571{
279 cl_uint count = *rcount; 572 cl_uint count = *rcount;
280 573
281 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)
282 return 0; 578 return 0;
283 579
284 cl_event *list = tmpbuf (sizeof (cl_event) * count); 580 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
285 int i = 0; 581 int i = 0;
286 582
287 do 583 while (count--)
288 {
289 --count;
290 if (SvOK (items [count])) 584 if (SvOK (items [count]))
291 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 585 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
292 } 586
293 while (count); 587 if (extra)
588 list [i++] = extra;
294 589
295 *rcount = i; 590 *rcount = i;
296 591
297 return i ? list : 0; 592 return i ? list : 0;
298} 593}
299 594
300#define EVENT_LIST(items,count) \ 595#define EVENT_LIST(skip) \
301 cl_uint event_list_count = (count); \ 596 cl_uint event_list_count = items - (skip); \
302 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)
303 598
304#define INFO(class) \ 599#define INFO(class) \
305{ \ 600{ \
306 size_t size; \ 601 size_t size; \
307 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 602 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
311 SvCUR_set (sv, size); \ 606 SvCUR_set (sv, size); \
312 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 607 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
313 XPUSHs (sv); \ 608 XPUSHs (sv); \
314} 609}
315 610
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/*****************************************************************************/
703
316MODULE = OpenCL PACKAGE = OpenCL 704MODULE = OpenCL PACKAGE = OpenCL
317 705
318PROTOTYPES: ENABLE 706PROTOTYPES: ENABLE
707
708void
709poll ()
710 CODE:
711 eq_poll ();
712
713void
714_eq_initialise (IV func, IV arg)
715 CODE:
716 eq_signal_func = (void (*)(void *, int))func;
717 eq_signal_arg = (void*)arg;
319 718
320BOOT: 719BOOT:
321{ 720{
322 HV *stash = gv_stashpv ("OpenCL", 1); 721 HV *stash = gv_stashpv ("OpenCL", 1);
323 static const ivstr *civ, const_iv[] = { 722 static const ivstr *civ, const_iv[] = {
332 { sizeof (cl_half ), "SIZEOF_HALF" }, 731 { sizeof (cl_half ), "SIZEOF_HALF" },
333 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 732 { sizeof (cl_float ), "SIZEOF_FLOAT" },
334 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 733 { sizeof (cl_double), "SIZEOF_DOUBLE" },
335#include "constiv.h" 734#include "constiv.h"
336 }; 735 };
736
337 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--)
338 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 738 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
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
764 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
339} 765}
340 766
341cl_int 767cl_int
342errno () 768errno ()
343 CODE: 769 CODE:
344 RETVAL = res; 770 RETVAL = res;
345 OUTPUT: 771 OUTPUT:
346 RETVAL 772 RETVAL
347 773
348const char * 774const char *
349err2str (cl_int err) 775err2str (cl_int err = res)
350 776
351const char * 777const char *
352enum2str (cl_uint value) 778enum2str (cl_uint value)
353 779
354void 780void
362 list = tmpbuf (sizeof (*list) * count); 788 list = tmpbuf (sizeof (*list) * count);
363 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 789 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
364 790
365 EXTEND (SP, count); 791 EXTEND (SP, count);
366 for (i = 0; i < count; ++i) 792 for (i = 0; i < count; ++i)
367 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 793 PUSH_CLOBJ (stash_platform, list [i]);
368 794
369void 795void
370context_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)
371 PPCODE: 797 PPCODE:
798 CONTEXT_NOTIFY_CALLBACK;
372 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));
373 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 800 XPUSH_CLOBJ_CONTEXT;
374 801
375#if 0
376
377void 802void
378context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 803context (FUTURE properties, FUTURE devices, FUTURE notify)
379 PPCODE: 804 PPCODE:
380 /* der Gipfel der Kunst */ 805 /* der Gipfel der Kunst */
381 806
807void
808wait_for_events (...)
809 CODE:
810 EVENT_LIST (0);
811 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
812
813PROTOTYPES: DISABLE
814
815MODULE = OpenCL PACKAGE = OpenCL::Platform
816
817void
818info (OpenCL::Platform self, cl_platform_info name)
819 PPCODE:
820 INFO (Platform)
821
822void
823unload_compiler (OpenCL::Platform self)
824 CODE:
825#if CL_VERSION_1_2
826 clUnloadPlatformCompiler (self);
382#endif 827#endif
383
384void
385wait_for_events (...)
386 CODE:
387 EVENT_LIST (0, items);
388 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
389
390PROTOTYPES: DISABLE
391
392MODULE = OpenCL PACKAGE = OpenCL::Platform
393
394void
395info (OpenCL::Platform self, cl_platform_info name)
396 PPCODE:
397 INFO (Platform)
398 828
399#BEGIN:platform 829#BEGIN:platform
400 830
401void 831void
402profile (OpenCL::Platform self) 832profile (OpenCL::Platform self)
428 list = tmpbuf (sizeof (*list) * count); 858 list = tmpbuf (sizeof (*list) * count);
429 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 859 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
430 860
431 EXTEND (SP, count); 861 EXTEND (SP, count);
432 for (i = 0; i < count; ++i) 862 for (i = 0; i < count; ++i)
433 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 863 PUSH_CLOBJ (stash_device, list [i]);
434 864
435void 865void
436context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 866context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
437 PPCODE: 867 PPCODE:
438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 868 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
439 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);
440 873
441 AV *av = (AV *)SvRV (devices); 874 AV *av = (AV *)SvRV (devices);
442 cl_uint num_devices = av_len (av) + 1; 875 cl_uint num_devices = av_len (av) + 1;
443 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);
877
444 int i; 878 int i;
445
446 for (i = num_devices; i--; ) 879 for (i = num_devices; i--; )
447 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 880 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
448 881
882 CONTEXT_NOTIFY_CALLBACK;
449 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 883 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res));
450 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 884 XPUSH_CLOBJ_CONTEXT;
451 885
452void 886void
453context_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)
454 PPCODE: 888 PPCODE:
455 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 889 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
456 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 890 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
891
892 CONTEXT_NOTIFY_CALLBACK;
457 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));
458 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 894 XPUSH_CLOBJ_CONTEXT;
459 895
460MODULE = OpenCL PACKAGE = OpenCL::Device 896MODULE = OpenCL PACKAGE = OpenCL::Device
461 897
462void 898void
463info (OpenCL::Device self, cl_device_info name) 899info (OpenCL::Device self, cl_device_info name)
464 PPCODE: 900 PPCODE:
465 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
466 932
467#BEGIN:device 933#BEGIN:device
468 934
469void 935void
470type (OpenCL::Device self) 936type (OpenCL::Device self)
471 PPCODE: 937 PPCODE:
472 cl_device_type value [1]; 938 cl_device_type value [1];
473 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 939 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
474 EXTEND (SP, 1); 940 EXTEND (SP, 1);
475 const int i = 0; 941 const int i = 0;
476 PUSHs (sv_2mortal (newSViv (value [i]))); 942 PUSHs (sv_2mortal (newSVuv (value [i])));
477 943
478void 944void
479vendor_id (OpenCL::Device self) 945vendor_id (OpenCL::Device self)
480 ALIAS: 946 ALIAS:
481 vendor_id = CL_DEVICE_VENDOR_ID 947 vendor_id = CL_DEVICE_VENDOR_ID
626 PPCODE: 1092 PPCODE:
627 cl_command_queue_properties value [1]; 1093 cl_command_queue_properties value [1];
628 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));
629 EXTEND (SP, 1); 1095 EXTEND (SP, 1);
630 const int i = 0; 1096 const int i = 0;
631 PUSHs (sv_2mortal (newSViv (value [i]))); 1097 PUSHs (sv_2mortal (newSVuv (value [i])));
632 1098
633void 1099void
634platform (OpenCL::Device self) 1100platform (OpenCL::Device self)
635 PPCODE: 1101 PPCODE:
636 cl_platform_id value [1]; 1102 cl_platform_id value [1];
637 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1103 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
638 EXTEND (SP, 1); 1104 EXTEND (SP, 1);
639 const int i = 0; 1105 const int i = 0;
640 { 1106 {
641 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1107 PUSH_CLOBJ (stash_platform, value [i]);
642 } 1108 }
643 1109
644void 1110void
645name (OpenCL::Device self) 1111name (OpenCL::Device self)
646 ALIAS: 1112 ALIAS:
665 cl_device_id value [1]; 1131 cl_device_id value [1];
666 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));
667 EXTEND (SP, 1); 1133 EXTEND (SP, 1);
668 const int i = 0; 1134 const int i = 0;
669 { 1135 {
670 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1136 PUSH_CLOBJ (stash_device, value [i]);
671 } 1137 }
672 1138
673void 1139void
674partition_types_ext (OpenCL::Device self) 1140partition_types_ext (OpenCL::Device self)
675 ALIAS: 1141 ALIAS:
686 for (i = 0; i < n; ++i) 1152 for (i = 0; i < n; ++i)
687 PUSHs (sv_2mortal (newSVuv (value [i]))); 1153 PUSHs (sv_2mortal (newSVuv (value [i])));
688 1154
689#END:device 1155#END:device
690 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
691MODULE = OpenCL PACKAGE = OpenCL::Context 1168MODULE = OpenCL PACKAGE = OpenCL::Context
692 1169
693void 1170void
694DESTROY (OpenCL::Context context) 1171DESTROY (OpenCL::Context self)
695 CODE: 1172 CODE:
696 clReleaseContext (context); 1173 clReleaseContext (self);
697 1174
698void 1175void
699info (OpenCL::Context self, cl_context_info name) 1176info (OpenCL::Context self, cl_context_info name)
700 PPCODE: 1177 PPCODE:
701 INFO (Context) 1178 INFO (Context)
702 1179
703void 1180void
704queue (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)
705 PPCODE: 1182 PPCODE:
706 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));
707 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1184 XPUSH_CLOBJ (stash_queue, queue);
708 1185
709void 1186void
710user_event (OpenCL::Context self) 1187user_event (OpenCL::Context self)
711 PPCODE: 1188 PPCODE:
712 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1189 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
713 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1190 XPUSH_CLOBJ (stash_userevent, ev);
714 1191
715void 1192void
716buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1193buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
717 PPCODE: 1194 PPCODE:
718 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))
719 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?");
720 1197
721 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));
722 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1199 XPUSH_CLOBJ (stash_bufferobj, mem);
723 1200
724void 1201void
725buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1202buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
726 PPCODE: 1203 PPCODE:
727 STRLEN len; 1204 STRLEN len;
728 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1205 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
729 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)))
730 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?");
731 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));
732 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1209 XPUSH_CLOBJ (stash_bufferobj, mem);
733 1210
734#if CL_VERSION_1_2 1211#if CL_VERSION_1_2
735 1212
736void 1213void
737image (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)
738 PPCODE: 1215 PPCODE:
739 STRLEN len; 1216 STRLEN len;
740 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1217 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
741 const cl_image_format format = { channel_order, channel_type }; 1218 const cl_image_format format = { channel_order, channel_type };
742 const cl_image_desc desc = { 1219 const cl_image_desc desc = {
743 type, 1220 type,
744 width, height, depth, 1221 width, height, depth,
745 array_size, row_pitch, slice_pitch, 1222 array_size, row_pitch, slice_pitch,
746 num_mip_level, num_samples, 1223 num_mip_level, num_samples,
747 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvPTROBJ ("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
748 }; 1225 };
749 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));
750 char *klass = "OpenCL::Image"; 1227 HV *stash = stash_image;
751 switch (type) 1228 switch (type)
752 { 1229 {
753 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1230 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
754 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1231 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
755 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1232 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
756 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1233 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
757 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1234 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
758 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1235 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
759 } 1236 }
760 XPUSH_NEW_OBJ (klass, mem); 1237 XPUSH_CLOBJ (stash, mem);
761 1238
762#endif 1239#endif
763 1240
764void 1241void
765image2d (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)
766 PPCODE: 1243 PPCODE:
767 STRLEN len; 1244 STRLEN len;
768 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1245 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
769 const cl_image_format format = { channel_order, channel_type }; 1246 const cl_image_format format = { channel_order, channel_type };
770#if CL_VERSION_1_2 1247#if PREFER_1_1
1248 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1249#else
771 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 };
772 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));
773#else
774 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
775#endif 1252#endif
776 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1253 XPUSH_CLOBJ (stash_image2d, mem);
777 1254
778void 1255void
779image3d (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)
780 PPCODE: 1257 PPCODE:
781 STRLEN len; 1258 STRLEN len;
782 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1259 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
783 const cl_image_format format = { channel_order, channel_type }; 1260 const cl_image_format format = { channel_order, channel_type };
784#if CL_VERSION_1_2 1261#if PREFER_1_1
1262 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1263#else
785 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 };
786 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));
787#else
788 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
789#endif 1266#endif
790 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1267 XPUSH_CLOBJ (stash_image3d, mem);
791 1268
792#if cl_apple_gl_sharing || cl_khr_gl_sharing 1269#if cl_apple_gl_sharing || cl_khr_gl_sharing
793 1270
794void 1271void
795gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1272gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
796 PPCODE: 1273 PPCODE:
797 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1274 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
798 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1275 XPUSH_CLOBJ (stash_bufferobj, mem);
799 1276
800void 1277void
801gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1278gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
802 PPCODE: 1279 PPCODE:
803 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1280 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
804 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1281 XPUSH_CLOBJ (stash_image2d, mem);
805 1282
806#if CL_VERSION_1_2 1283#if CL_VERSION_1_2
807 1284
808void 1285void
809gl_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)
810 ALIAS: 1287 ALIAS:
811 gl_texture2d = 0
812 gl_texture3d = 0
813 PPCODE: 1288 PPCODE:
814 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));
815 cl_gl_object_type type; 1290 cl_gl_object_type type;
816 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1291 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
817 char *klass = "OpenCL::Memory"; 1292 HV *stash = stash_memory;
818 switch (type) 1293 switch (type)
819 { 1294 {
820 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1295 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
821 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1296 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
822 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1297 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
823 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1298 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
824 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1299 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
825 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1300 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
826 } 1301 }
827 XPUSH_NEW_OBJ (klass, mem); 1302 XPUSH_CLOBJ (stash, mem);
828 1303
1304#endif
1305
1306void
1307gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1308 PPCODE:
1309#if PREFER_1_1
1310 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
829#else 1311#else
830
831void
832gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
833 PPCODE:
834 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1312 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
835 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1313#endif
1314 XPUSH_CLOBJ (stash_image2d, mem);
836 1315
837void 1316void
838gl_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)
839 PPCODE: 1318 PPCODE:
1319#if PREFER_1_1
840 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));
841 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1321#else
842 1322 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
843#endif 1323#endif
1324 XPUSH_CLOBJ (stash_image3d, mem);
844 1325
845#endif 1326#endif
846 1327
847void 1328void
848supported_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)
868 1349
869void 1350void
870sampler (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)
871 PPCODE: 1352 PPCODE:
872 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));
873 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1354 XPUSH_CLOBJ (stash_sampler, sampler);
874 1355
875void 1356void
876program_with_source (OpenCL::Context self, SV *program) 1357program_with_source (OpenCL::Context self, SV *program)
877 PPCODE: 1358 PPCODE:
878 STRLEN len; 1359 STRLEN len;
879 size_t len2; 1360 size_t len2;
880 const char *ptr = SvPVbyte (program, len); 1361 const char *ptr = SvPVbyte (program, len);
881 1362
882 len2 = len; 1363 len2 = len;
883 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));
884 XPUSH_NEW_OBJ ("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
885 1434
886#BEGIN:context 1435#BEGIN:context
887 1436
888void 1437void
889reference_count (OpenCL::Context self) 1438reference_count (OpenCL::Context self)
906 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1455 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
907 int i, n = size / sizeof (*value); 1456 int i, n = size / sizeof (*value);
908 EXTEND (SP, n); 1457 EXTEND (SP, n);
909 for (i = 0; i < n; ++i) 1458 for (i = 0; i < n; ++i)
910 { 1459 {
911 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1460 PUSH_CLOBJ (stash_device, value [i]);
912 } 1461 }
913 1462
914void 1463void
915properties (OpenCL::Context self) 1464properties (OpenCL::Context self)
916 PPCODE: 1465 PPCODE:
931DESTROY (OpenCL::Queue self) 1480DESTROY (OpenCL::Queue self)
932 CODE: 1481 CODE:
933 clReleaseCommandQueue (self); 1482 clReleaseCommandQueue (self);
934 1483
935void 1484void
936enqueue_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
937 PPCODE: 1488 PPCODE:
938 cl_event ev = 0; 1489 cl_event ev = 0;
939 EVENT_LIST (6, items - 6); 1490 EVENT_LIST (6);
940 1491
941 SvUPGRADE (data, SVt_PV); 1492 SvUPGRADE (data, SVt_PV);
942 SvGROW (data, len); 1493 SvGROW (data, len);
943 SvPOK_only (data); 1494 SvPOK_only (data);
944 SvCUR_set (data, len); 1495 SvCUR_set (data, len);
945 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));
946 1497
947 if (ev) 1498 if (ev)
948 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1499 XPUSH_CLOBJ (stash_event, ev);
949 1500
950void 1501void
951enqueue_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
952 PPCODE: 1505 PPCODE:
953 cl_event ev = 0; 1506 cl_event ev = 0;
954 STRLEN len; 1507 STRLEN len;
955 char *ptr = SvPVbyte (data, len); 1508 char *ptr = SvPVbyte (data, len);
956 EVENT_LIST (5, items - 5); 1509 EVENT_LIST (5);
957 1510
958 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));
959 1512
960 if (ev) 1513 if (ev)
961 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1514 XPUSH_CLOBJ (stash_event, ev);
962 1515
1516#if CL_VERSION_1_2
1517
963void 1518void
964enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1519fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1520 ALIAS:
1521 enqueue_fill_buffer = 0
965 PPCODE: 1522 PPCODE:
966 cl_event ev = 0; 1523 cl_event ev = 0;
1524 STRLEN len;
1525 char *ptr = SvPVbyte (data, len);
1526 EVENT_LIST (5);
1527
1528 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1529
1530 if (ev)
1531 XPUSH_CLOBJ (stash_event, ev);
1532
1533void
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
1537 PPCODE:
1538 cl_event ev = 0;
1539 const size_t origin [3] = { x, y, z };
1540 const size_t region [3] = { width, height, depth };
1541 EVENT_LIST (12);
1542
1543 const cl_float c_f [4] = { r, g, b, a };
1544 const cl_uint c_u [4] = { r, g, b, a };
1545 const cl_int c_s [4] = { r, g, b, a };
1546 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1547 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1548 cl_image_format format;
1549 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1550 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1551 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1552 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1553
1554 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1555 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1556
1557 if (ev)
1558 XPUSH_CLOBJ (stash_event, ev);
1559
1560#endif
1561
1562void
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
1566 PPCODE:
1567 cl_event ev = 0;
967 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
968 1569
969 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));
970 1571
971 if (ev) 1572 if (ev)
972 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1573 XPUSH_CLOBJ (stash_event, ev);
973 1574
974void 1575void
975enqueue_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
976 PPCODE: 1579 PPCODE:
977 cl_event ev = 0; 1580 cl_event ev = 0;
978 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 };
979 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 };
980 const size_t region[3] = { width, height, depth }; 1583 const size_t region[3] = { width, height, depth };
981 EVENT_LIST (17, items - 17); 1584 EVENT_LIST (17);
982 1585
983 if (!buf_row_pitch) 1586 if (!buf_row_pitch)
984 buf_row_pitch = region [0]; 1587 buf_row_pitch = region [0];
985 1588
986 if (!buf_slice_pitch) 1589 if (!buf_slice_pitch)
999 SvPOK_only (data); 1602 SvPOK_only (data);
1000 SvCUR_set (data, len); 1603 SvCUR_set (data, len);
1001 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));
1002 1605
1003 if (ev) 1606 if (ev)
1004 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1607 XPUSH_CLOBJ (stash_event, ev);
1005 1608
1006void 1609void
1007enqueue_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
1008 PPCODE: 1613 PPCODE:
1009 cl_event ev = 0; 1614 cl_event ev = 0;
1010 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 };
1011 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 };
1012 const size_t region[3] = { width, height, depth }; 1617 const size_t region[3] = { width, height, depth };
1013 STRLEN len; 1618 STRLEN len;
1014 char *ptr = SvPVbyte (data, len); 1619 char *ptr = SvPVbyte (data, len);
1015 EVENT_LIST (17, items - 17); 1620 EVENT_LIST (17);
1016 1621
1017 if (!buf_row_pitch) 1622 if (!buf_row_pitch)
1018 buf_row_pitch = region [0]; 1623 buf_row_pitch = region [0];
1019 1624
1020 if (!buf_slice_pitch) 1625 if (!buf_slice_pitch)
1032 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1637 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1033 1638
1034 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));
1035 1640
1036 if (ev) 1641 if (ev)
1037 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1642 XPUSH_CLOBJ (stash_event, ev);
1038 1643
1039void 1644void
1040enqueue_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
1041 PPCODE: 1648 PPCODE:
1042 cl_event ev = 0; 1649 cl_event ev = 0;
1043 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 };
1044 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 };
1045 const size_t region[3] = { width, height, depth }; 1652 const size_t region[3] = { width, height, depth };
1046 EVENT_LIST (16, items - 16); 1653 EVENT_LIST (16);
1047 1654
1048 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));
1049 1656
1050 if (ev) 1657 if (ev)
1051 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1658 XPUSH_CLOBJ (stash_event, ev);
1052 1659
1053void 1660void
1054enqueue_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
1055 PPCODE: 1664 PPCODE:
1056 cl_event ev = 0; 1665 cl_event ev = 0;
1057 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 };
1058 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1059 EVENT_LIST (12, items - 12); 1668 EVENT_LIST (12);
1060 1669
1061 if (!row_pitch) 1670 if (!row_pitch)
1062 row_pitch = img_row_pitch (src); 1671 row_pitch = img_row_pitch (src);
1063 1672
1064 if (depth > 1 && !slice_pitch) 1673 if (depth > 1 && !slice_pitch)
1071 SvPOK_only (data); 1680 SvPOK_only (data);
1072 SvCUR_set (data, len); 1681 SvCUR_set (data, len);
1073 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));
1074 1683
1075 if (ev) 1684 if (ev)
1076 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1685 XPUSH_CLOBJ (stash_event, ev);
1077 1686
1078void 1687void
1079enqueue_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
1080 PPCODE: 1691 PPCODE:
1081 cl_event ev = 0; 1692 cl_event ev = 0;
1082 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 };
1083 const size_t region[3] = { width, height, depth }; 1694 const size_t region[3] = { width, height, depth };
1084 STRLEN len; 1695 STRLEN len;
1085 char *ptr = SvPVbyte (data, len); 1696 char *ptr = SvPVbyte (data, len);
1086 EVENT_LIST (12, items - 12); 1697 EVENT_LIST (12);
1087 1698
1088 if (!row_pitch) 1699 if (!row_pitch)
1089 row_pitch = img_row_pitch (dst); 1700 row_pitch = img_row_pitch (dst);
1090 1701
1091 if (depth > 1 && !slice_pitch) 1702 if (depth > 1 && !slice_pitch)
1097 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1708 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1098 1709
1099 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));
1100 1711
1101 if (ev) 1712 if (ev)
1102 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1713 XPUSH_CLOBJ (stash_event, ev);
1103 1714
1104void 1715void
1105enqueue_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
1106 PPCODE: 1719 PPCODE:
1107 cl_event ev = 0; 1720 cl_event ev = 0;
1108 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 };
1109 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 };
1110 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
1111 EVENT_LIST (12, items - 12); 1724 EVENT_LIST (12);
1112 1725
1113 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));
1114 1727
1115 if (ev) 1728 if (ev)
1116 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
1117 1730
1118void 1731void
1119enqueue_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
1120 PPCODE: 1735 PPCODE:
1121 cl_event ev = 0; 1736 cl_event ev = 0;
1122 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 };
1123 const size_t region[3] = { width, height, depth }; 1810 const size_t region[3] = { width, height, depth };
1811 size_t row_pitch, slice_pitch;
1124 EVENT_LIST (10, items - 10); 1812 EVENT_LIST (10);
1125 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
1126 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));
1127 1842
1128 if (ev) 1843 if (ev)
1129 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1844 XPUSH_CLOBJ (stash_event, ev);
1130 1845
1131void 1846void
1132enqueue_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, ...)
1133 PPCODE:
1134 cl_event ev = 0;
1135 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1136 const size_t region[3] = { width, height, depth };
1137 EVENT_LIST (10, items - 10);
1138
1139 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1140
1141 if (ev)
1142 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1143
1144void
1145enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1146 PPCODE:
1147 cl_event ev = 0;
1148 EVENT_LIST (2, items - 2);
1149
1150 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1151
1152 if (ev)
1153 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1154
1155void
1156enqueue_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
1157 PPCODE: 1850 PPCODE:
1158 cl_event ev = 0; 1851 cl_event ev = 0;
1159 size_t *gwo = 0, *gws, *lws = 0; 1852 size_t *gwo = 0, *gws, *lws = 0;
1160 int gws_len; 1853 int gws_len;
1161 size_t *lists; 1854 size_t *lists;
1162 int i; 1855 int i;
1163 EVENT_LIST (5, items - 5); 1856 EVENT_LIST (5);
1164 1857
1165 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)
1166 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1859 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1167 1860
1168 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1861 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1169 1862
1170 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1863 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1171 1864
1172 gws = lists + gws_len * 0; 1865 gws = lists + gws_len * 0;
1173 for (i = 0; i < gws_len; ++i) 1866 for (i = 0; i < gws_len; ++i)
1867 {
1174 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 }
1175 1873
1176 if (SvOK (global_work_offset)) 1874 if (SvOK (global_work_offset))
1177 { 1875 {
1178 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)
1179 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");
1187 } 1885 }
1188 1886
1189 if (SvOK (local_work_size)) 1887 if (SvOK (local_work_size))
1190 { 1888 {
1191 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)
1192 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");
1193 1891
1194 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1892 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1195 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");
1196 1894
1197 lws = lists + gws_len * 2; 1895 lws = lists + gws_len * 2;
1198 for (i = 0; i < gws_len; ++i) 1896 for (i = 0; i < gws_len; ++i)
1897 {
1199 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 }
1200 } 1903 }
1201 1904
1202 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));
1203 1906
1204 if (ev) 1907 if (ev)
1205 XPUSH_NEW_OBJ ("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
1206 1938
1207#if cl_apple_gl_sharing || cl_khr_gl_sharing 1939#if cl_apple_gl_sharing || cl_khr_gl_sharing
1208 1940
1209void 1941void
1210enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1942acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1211 ALIAS: 1943 ALIAS:
1944 release_gl_objects = 1
1945 enqueue_acquire_gl_objects = 0
1212 enqueue_release_gl_objects = 1 1946 enqueue_release_gl_objects = 1
1213 PPCODE: 1947 PPCODE:
1214 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1948 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1215 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");
1216 1950
1217 cl_event ev = 0; 1951 cl_event ev = 0;
1218 EVENT_LIST (2, items - 2); 1952 EVENT_LIST (2);
1219 AV *av = (AV *)SvRV (objects); 1953 AV *av = (AV *)SvRV (objects);
1220 cl_uint num_objects = av_len (av) + 1; 1954 cl_uint num_objects = av_len (av) + 1;
1221 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1955 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1956
1222 int i; 1957 int i;
1223
1224 for (i = num_objects; i--; ) 1958 for (i = num_objects; i--; )
1225 object_list [i] = SvPTROBJ ("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");
1226 1960
1227 if (ix) 1961 if (ix)
1228 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));
1229 else 1963 else
1230 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));
1231 1965
1232 if (ev) 1966 if (ev)
1233 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1967 XPUSH_CLOBJ (stash_event, ev);
1234 1968
1235#endif 1969#endif
1236 1970
1237#if !CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS
1238
1239void 1971void
1240enqueue_wait_for_events (OpenCL::Queue self, ...) 1972wait_for_events (OpenCL::Queue self, ...)
1973 ALIAS:
1974 enqueue_wait_for_events = 0
1241 CODE: 1975 CODE:
1242 EVENT_LIST (1, items - 1); 1976 EVENT_LIST (1);
1977#if PREFER_1_1
1243 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1978 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1244 1979#else
1980 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1245#endif 1981#endif
1246 1982
1247void 1983void
1248enqueue_marker (OpenCL::Queue self, ...) 1984marker (OpenCL::Queue self, ...)
1985 ALIAS:
1986 enqueue_marker = 0
1249 PPCODE: 1987 PPCODE:
1250 cl_event ev = 0; 1988 cl_event ev = 0;
1251 EVENT_LIST (1, items - 1); 1989 EVENT_LIST (1);
1990#if PREFER_1_1
1991 if (!event_list_count)
1992 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1993 else
1252#if CL_VERSION_1_2 1994#if CL_VERSION_1_2
1995 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1996#else
1997 {
1998 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1999 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2000 }
2001#endif
2002#else
1253 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));
1254#else
1255 if (event_list_count)
1256 croak ("OpenCL::Queue->enqueue_marker does not support a wait list in OpenCL 1.1 - upgrade to 1.2");
1257 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1258#endif 2004#endif
1259 if (ev) 2005 if (ev)
1260 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2006 XPUSH_CLOBJ (stash_event, ev);
1261 2007
1262void 2008void
1263enqueue_barrier (OpenCL::Queue self, ...) 2009barrier (OpenCL::Queue self, ...)
2010 ALIAS:
2011 enqueue_barrier = 0
1264 PPCODE: 2012 PPCODE:
1265 cl_event ev = 0; 2013 cl_event ev = 0;
1266 EVENT_LIST (1, items - 1); 2014 EVENT_LIST (1);
2015#if PREFER_1_1
2016 if (!event_list_count && GIMME_V == G_VOID)
2017 NEED_SUCCESS (EnqueueBarrier, (self));
2018 else
1267#if CL_VERSION_1_2 2019#if CL_VERSION_1_2
1268 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, &ev)); 2020 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1269#else 2021#else
2022 {
1270 if (event_list_count) 2023 if (event_list_count)
1271 croak ("OpenCL::Queue->enqueue_barrier does not support a wait list in OpenCL 1.1 - upgrade to 1.2"); 2024 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2025
1272 if (GIMME_V != G_VOID) 2026 if (GIMME_V != G_VOID)
1273 croak ("OpenCL::Queue->enqueue_barrier does not return an event object in OpenCL 1.1 - upgrade to 1.2");
1274 NEED_SUCCESS (EnqueueBarrier, (self)); 2027 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2028 }
2029#endif
2030#else
2031 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1275#endif 2032#endif
1276 if (ev) 2033 if (ev)
1277 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2034 XPUSH_CLOBJ (stash_event, ev);
1278 2035
1279void 2036void
1280flush (OpenCL::Queue self) 2037flush (OpenCL::Queue self)
1281 CODE: 2038 CODE:
1282 NEED_SUCCESS (Flush, (self)); 2039 NEED_SUCCESS (Flush, (self));
1300 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2057 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1301 EXTEND (SP, 1); 2058 EXTEND (SP, 1);
1302 const int i = 0; 2059 const int i = 0;
1303 { 2060 {
1304 NEED_SUCCESS (RetainContext, (value [i])); 2061 NEED_SUCCESS (RetainContext, (value [i]));
1305 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2062 PUSH_CLOBJ (stash_context, value [i]);
1306 } 2063 }
1307 2064
1308void 2065void
1309device (OpenCL::Queue self) 2066device (OpenCL::Queue self)
1310 PPCODE: 2067 PPCODE:
1311 cl_device_id value [1]; 2068 cl_device_id value [1];
1312 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2069 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1313 EXTEND (SP, 1); 2070 EXTEND (SP, 1);
1314 const int i = 0; 2071 const int i = 0;
1315 { 2072 {
1316 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2073 PUSH_CLOBJ (stash_device, value [i]);
1317 } 2074 }
1318 2075
1319void 2076void
1320reference_count (OpenCL::Queue self) 2077reference_count (OpenCL::Queue self)
1321 PPCODE: 2078 PPCODE:
1330 PPCODE: 2087 PPCODE:
1331 cl_command_queue_properties value [1]; 2088 cl_command_queue_properties value [1];
1332 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2089 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1333 EXTEND (SP, 1); 2090 EXTEND (SP, 1);
1334 const int i = 0; 2091 const int i = 0;
1335 PUSHs (sv_2mortal (newSViv (value [i]))); 2092 PUSHs (sv_2mortal (newSVuv (value [i])));
1336 2093
1337#END:command_queue 2094#END:command_queue
1338 2095
1339MODULE = OpenCL PACKAGE = OpenCL::Memory 2096MODULE = OpenCL PACKAGE = OpenCL::Memory
1340 2097
1355 PPCODE: 2112 PPCODE:
1356 cl_mem_object_type value [1]; 2113 cl_mem_object_type value [1];
1357 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2114 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1358 EXTEND (SP, 1); 2115 EXTEND (SP, 1);
1359 const int i = 0; 2116 const int i = 0;
1360 PUSHs (sv_2mortal (newSViv (value [i]))); 2117 PUSHs (sv_2mortal (newSVuv (value [i])));
1361 2118
1362void 2119void
1363flags (OpenCL::Memory self) 2120flags (OpenCL::Memory self)
1364 PPCODE: 2121 PPCODE:
1365 cl_mem_flags value [1]; 2122 cl_mem_flags value [1];
1366 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2123 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1367 EXTEND (SP, 1); 2124 EXTEND (SP, 1);
1368 const int i = 0; 2125 const int i = 0;
1369 PUSHs (sv_2mortal (newSViv (value [i]))); 2126 PUSHs (sv_2mortal (newSVuv (value [i])));
1370 2127
1371void 2128void
1372size (OpenCL::Memory self) 2129size (OpenCL::Memory self)
1373 ALIAS: 2130 ALIAS:
1374 size = CL_MEM_SIZE 2131 size = CL_MEM_SIZE
1408 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2165 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1409 EXTEND (SP, 1); 2166 EXTEND (SP, 1);
1410 const int i = 0; 2167 const int i = 0;
1411 { 2168 {
1412 NEED_SUCCESS (RetainContext, (value [i])); 2169 NEED_SUCCESS (RetainContext, (value [i]));
1413 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2170 PUSH_CLOBJ (stash_context, value [i]);
1414 } 2171 }
1415 2172
1416void 2173void
1417associated_memobject (OpenCL::Memory self) 2174associated_memobject (OpenCL::Memory self)
1418 PPCODE: 2175 PPCODE:
1420 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));
1421 EXTEND (SP, 1); 2178 EXTEND (SP, 1);
1422 const int i = 0; 2179 const int i = 0;
1423 { 2180 {
1424 NEED_SUCCESS (RetainMemObject, (value [i])); 2181 NEED_SUCCESS (RetainMemObject, (value [i]));
1425 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2182 PUSH_CLOBJ (stash_memory, value [i]);
1426 } 2183 }
1427 2184
1428#END:mem 2185#END:mem
1429 2186
1430#if cl_apple_gl_sharing || cl_khr_gl_sharing 2187#if cl_apple_gl_sharing || cl_khr_gl_sharing
1450 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!");
1451 2208
1452 cl_buffer_region crdata = { origin, size }; 2209 cl_buffer_region crdata = { origin, size };
1453 2210
1454 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));
1455 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2212 XPUSH_CLOBJ (stash_buffer, mem);
1456 2213
1457MODULE = OpenCL PACKAGE = OpenCL::Image 2214MODULE = OpenCL PACKAGE = OpenCL::Image
1458 2215
1459void 2216void
1460image_info (OpenCL::Image self, cl_image_info name) 2217image_info (OpenCL::Image self, cl_image_info name)
1461 PPCODE: 2218 PPCODE:
1462 INFO (Image) 2219 INFO (Image)
2220
2221void
2222format (OpenCL::Image self)
2223 PPCODE:
2224 cl_image_format format;
2225 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2226 EXTEND (SP, 2);
2227 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2228 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1463 2229
1464#BEGIN:image 2230#BEGIN:image
1465 2231
1466void 2232void
1467element_size (OpenCL::Image self) 2233element_size (OpenCL::Image self)
1537 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2303 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1538 EXTEND (SP, 1); 2304 EXTEND (SP, 1);
1539 const int i = 0; 2305 const int i = 0;
1540 { 2306 {
1541 NEED_SUCCESS (RetainContext, (value [i])); 2307 NEED_SUCCESS (RetainContext, (value [i]));
1542 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2308 PUSH_CLOBJ (stash_context, value [i]);
1543 } 2309 }
1544 2310
1545void 2311void
1546normalized_coords (OpenCL::Sampler self) 2312normalized_coords (OpenCL::Sampler self)
1547 PPCODE: 2313 PPCODE:
1548 cl_addressing_mode value [1]; 2314 cl_addressing_mode value [1];
1549 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));
1550 EXTEND (SP, 1); 2316 EXTEND (SP, 1);
1551 const int i = 0; 2317 const int i = 0;
1552 PUSHs (sv_2mortal (newSViv (value [i]))); 2318 PUSHs (sv_2mortal (newSVuv (value [i])));
1553 2319
1554void 2320void
1555addressing_mode (OpenCL::Sampler self) 2321addressing_mode (OpenCL::Sampler self)
1556 PPCODE: 2322 PPCODE:
1557 cl_filter_mode value [1]; 2323 cl_filter_mode value [1];
1558 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));
1559 EXTEND (SP, 1); 2325 EXTEND (SP, 1);
1560 const int i = 0; 2326 const int i = 0;
1561 PUSHs (sv_2mortal (newSViv (value [i]))); 2327 PUSHs (sv_2mortal (newSVuv (value [i])));
1562 2328
1563void 2329void
1564filter_mode (OpenCL::Sampler self) 2330filter_mode (OpenCL::Sampler self)
1565 PPCODE: 2331 PPCODE:
1566 cl_bool value [1]; 2332 cl_bool value [1];
1577DESTROY (OpenCL::Program self) 2343DESTROY (OpenCL::Program self)
1578 CODE: 2344 CODE:
1579 clReleaseProgram (self); 2345 clReleaseProgram (self);
1580 2346
1581void 2347void
1582build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2348build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2349 ALIAS:
2350 build_async = 1
1583 CODE: 2351 CODE:
2352 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2353 void *user_data = 0;
2354 cl_uint num_devices = 0;
2355 cl_device_id *device_list = 0;
2356
2357 if (SvOK (devices))
2358 {
2359 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2360 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2361
2362 AV *av = (AV *)SvRV (devices);
2363 num_devices = av_len (av) + 1;
2364
2365 if (num_devices)
2366 {
2367 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2368 int count;
2369 for (count = 0; count < num_devices; ++count)
2370 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2371 }
2372 }
2373
2374 if (SvOK (notify))
2375 {
2376 NEED_SUCCESS (RetainProgram, (self));
2377 pfn_notify = eq_program_notify;
2378 user_data = SvREFCNT_inc (s_get_cv (notify));
2379 }
2380
2381 if (ix)
2382 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2383 else
1584 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2384 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1585 2385
1586void 2386void
1587build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2387build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1588 PPCODE: 2388 PPCODE:
1589 size_t size; 2389 size_t size;
1624 2424
1625void 2425void
1626kernel (OpenCL::Program program, SV *function) 2426kernel (OpenCL::Program program, SV *function)
1627 PPCODE: 2427 PPCODE:
1628 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));
1629 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2429 XPUSH_CLOBJ (stash_kernel, kernel);
2430
2431void
2432kernels_in_program (OpenCL::Program program)
2433 PPCODE:
2434 cl_uint num_kernels;
2435 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2436 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2437 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2438
2439 int i;
2440 EXTEND (SP, num_kernels);
2441 for (i = 0; i < num_kernels; ++i)
2442 PUSH_CLOBJ (stash_kernel, kernels [i]);
1630 2443
1631void 2444void
1632info (OpenCL::Program self, cl_program_info name) 2445info (OpenCL::Program self, cl_program_info name)
1633 PPCODE: 2446 PPCODE:
1634 INFO (Program) 2447 INFO (Program)
1682 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2495 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1683 EXTEND (SP, 1); 2496 EXTEND (SP, 1);
1684 const int i = 0; 2497 const int i = 0;
1685 { 2498 {
1686 NEED_SUCCESS (RetainContext, (value [i])); 2499 NEED_SUCCESS (RetainContext, (value [i]));
1687 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2500 PUSH_CLOBJ (stash_context, value [i]);
1688 } 2501 }
1689 2502
1690void 2503void
1691devices (OpenCL::Program self) 2504devices (OpenCL::Program self)
1692 PPCODE: 2505 PPCODE:
1696 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2509 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1697 int i, n = size / sizeof (*value); 2510 int i, n = size / sizeof (*value);
1698 EXTEND (SP, n); 2511 EXTEND (SP, n);
1699 for (i = 0; i < n; ++i) 2512 for (i = 0; i < n; ++i)
1700 { 2513 {
1701 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2514 PUSH_CLOBJ (stash_device, value [i]);
1702 } 2515 }
1703 2516
1704void 2517void
1705source (OpenCL::Program self) 2518source (OpenCL::Program self)
1706 PPCODE: 2519 PPCODE:
1732DESTROY (OpenCL::Kernel self) 2545DESTROY (OpenCL::Kernel self)
1733 CODE: 2546 CODE:
1734 clReleaseKernel (self); 2547 clReleaseKernel (self);
1735 2548
1736void 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
1737set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2615set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1738 CODE: 2616 CODE:
1739 clSetKernelArg (self, idx, sizeof (value), &value); 2617 clSetKernelArg (self, idx, sizeof (value), &value);
1740 2618
1741void 2619void
1789 clSetKernelArg (self, idx, sizeof (value), &value); 2667 clSetKernelArg (self, idx, sizeof (value), &value);
1790 2668
1791void 2669void
1792set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2670set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1793 CODE: 2671 CODE:
1794 clSetKernelArg (self, idx, sizeof (value), &value); 2672 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1795 2673
1796void 2674void
1797set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2675set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1798 CODE: 2676 CODE:
1799 clSetKernelArg (self, idx, sizeof (value), &value); 2677 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1800 2678
1801void 2679void
1802set_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
1803 CODE: 2684 CODE:
1804 clSetKernelArg (self, idx, sizeof (value), &value); 2685 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1805
1806void
1807set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1808 CODE:
1809 clSetKernelArg (self, idx, sizeof (value), &value);
1810 2686
1811void 2687void
1812set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2688set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1813 CODE: 2689 CODE:
1814 clSetKernelArg (self, idx, sizeof (value), &value); 2690 clSetKernelArg (self, idx, sizeof (value), &value);
1860 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2736 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1861 EXTEND (SP, 1); 2737 EXTEND (SP, 1);
1862 const int i = 0; 2738 const int i = 0;
1863 { 2739 {
1864 NEED_SUCCESS (RetainContext, (value [i])); 2740 NEED_SUCCESS (RetainContext, (value [i]));
1865 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2741 PUSH_CLOBJ (stash_context, value [i]);
1866 } 2742 }
1867 2743
1868void 2744void
1869program (OpenCL::Kernel self) 2745program (OpenCL::Kernel self)
1870 PPCODE: 2746 PPCODE:
1872 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2748 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1873 EXTEND (SP, 1); 2749 EXTEND (SP, 1);
1874 const int i = 0; 2750 const int i = 0;
1875 { 2751 {
1876 NEED_SUCCESS (RetainProgram, (value [i])); 2752 NEED_SUCCESS (RetainProgram, (value [i]));
1877 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2753 PUSH_CLOBJ (stash_program, value [i]);
1878 } 2754 }
1879 2755
1880#END:kernel 2756#END:kernel
1881 2757
1882void 2758void
1929 const int i = 0; 2805 const int i = 0;
1930 PUSHs (sv_2mortal (newSVuv (value [i]))); 2806 PUSHs (sv_2mortal (newSVuv (value [i])));
1931 2807
1932#END:kernel_work_group 2808#END:kernel_work_group
1933 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, idx, 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, idx, name, size, SvPVX (sv), 0));
2822 XPUSHs (sv);
2823
2824#BEGIN:kernel_arg
2825
2826void
2827arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2828 PPCODE:
2829 cl_kernel_arg_address_qualifier value [1];
2830 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2831 EXTEND (SP, 1);
2832 const int i = 0;
2833 PUSHs (sv_2mortal (newSVuv (value [i])));
2834
2835void
2836arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2837 PPCODE:
2838 cl_kernel_arg_access_qualifier value [1];
2839 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2840 EXTEND (SP, 1);
2841 const int i = 0;
2842 PUSHs (sv_2mortal (newSVuv (value [i])));
2843
2844void
2845arg_type_name (OpenCL::Kernel self, cl_uint idx)
2846 ALIAS:
2847 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2848 arg_name = CL_KERNEL_ARG_NAME
2849 PPCODE:
2850 size_t size;
2851 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2852 char *value = tmpbuf (size);
2853 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2854 EXTEND (SP, 1);
2855 const int i = 0;
2856 PUSHs (sv_2mortal (newSVpv (value, 0)));
2857
2858void
2859arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2860 PPCODE:
2861 cl_kernel_arg_type_qualifier value [1];
2862 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2863 EXTEND (SP, 1);
2864 const int i = 0;
2865 PUSHs (sv_2mortal (newSVuv (value [i])));
2866
2867#END:kernel_arg
2868
2869#endif
2870
1934MODULE = OpenCL PACKAGE = OpenCL::Event 2871MODULE = OpenCL PACKAGE = OpenCL::Event
1935 2872
1936void 2873void
1937DESTROY (OpenCL::Event self) 2874DESTROY (OpenCL::Event self)
1938 CODE: 2875 CODE:
1940 2877
1941void 2878void
1942wait (OpenCL::Event self) 2879wait (OpenCL::Event self)
1943 CODE: 2880 CODE:
1944 clWaitForEvents (1, &self); 2881 clWaitForEvents (1, &self);
2882
2883void
2884cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2885 CODE:
2886 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1945 2887
1946void 2888void
1947info (OpenCL::Event self, cl_event_info name) 2889info (OpenCL::Event self, cl_event_info name)
1948 PPCODE: 2890 PPCODE:
1949 INFO (Event) 2891 INFO (Event)
1957 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2899 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1958 EXTEND (SP, 1); 2900 EXTEND (SP, 1);
1959 const int i = 0; 2901 const int i = 0;
1960 { 2902 {
1961 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2903 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1962 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2904 PUSH_CLOBJ (stash_queue, value [i]);
1963 } 2905 }
1964 2906
1965void 2907void
1966command_type (OpenCL::Event self) 2908command_type (OpenCL::Event self)
1967 PPCODE: 2909 PPCODE:
1990 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2932 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1991 EXTEND (SP, 1); 2933 EXTEND (SP, 1);
1992 const int i = 0; 2934 const int i = 0;
1993 { 2935 {
1994 NEED_SUCCESS (RetainContext, (value [i])); 2936 NEED_SUCCESS (RetainContext, (value [i]));
1995 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2937 PUSH_CLOBJ (stash_context, value [i]);
1996 } 2938 }
1997 2939
1998#END:event 2940#END:event
1999 2941
2000void 2942void
2025void 2967void
2026set_status (OpenCL::UserEvent self, cl_int execution_status) 2968set_status (OpenCL::UserEvent self, cl_int execution_status)
2027 CODE: 2969 CODE:
2028 clSetUserEventStatus (self, execution_status); 2970 clSetUserEventStatus (self, execution_status);
2029 2971
2972MODULE = OpenCL PACKAGE = OpenCL::Mapped
2973
2974void
2975DESTROY (SV *self)
2976 CODE:
2977 OpenCL__Mapped mapped = SvMAPPED (self);
2978
2979 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2980 mapped_detach (self, mapped);
2981
2982 clReleaseCommandQueue (mapped->queue);
2983 clReleaseEvent (mapped->event);
2984 Safefree (mapped);
2985
2986void
2987unmap (OpenCL::Mapped self, ...)
2988 CODE:
2989 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2990
2991bool
2992mapped (OpenCL::Mapped self)
2993 CODE:
2994 RETVAL = !!self->ptr;
2995 OUTPUT:
2996 RETVAL
2997
2998void
2999wait (OpenCL::Mapped self)
3000 PPCODE:
3001 if (self->event)
3002 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3003
3004void
3005event (OpenCL::Mapped self)
3006 PPCODE:
3007 if (!self->event)
3008 XSRETURN_UNDEF;
3009
3010 clRetainEvent (self->event);
3011 XPUSH_CLOBJ (stash_event, self->event);
3012
3013size_t
3014size (OpenCL::Mapped self)
3015 CODE:
3016 RETVAL = self->cb;
3017 OUTPUT:
3018 RETVAL
3019
3020IV
3021ptr (OpenCL::Mapped self)
3022 CODE:
3023 RETVAL = PTR2IV (self->ptr);
3024 OUTPUT:
3025 RETVAL
3026
3027void
3028set (OpenCL::Mapped self, size_t offset, SV *data)
3029 CODE:
3030 STRLEN len;
3031 const char *ptr = SvPVbyte (data, len);
3032
3033 if (offset + len > self->cb)
3034 croak ("OpenCL::Mapped::set out of bound condition detected");
3035
3036 memcpy (offset + (char *)self->ptr, ptr, len);
3037
3038MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3039
3040MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3041
3042IV
3043row_pitch (OpenCL::Mapped self)
3044 ALIAS:
3045 slice_pitch = 1
3046 CODE:
3047 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3048 OUTPUT:
3049 RETVAL
3050
3051
3052

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines