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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.32 by root, Thu Apr 19 19:36:24 2012 UTC vs.
Revision 1.66 by root, Fri May 4 14:46:02 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 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"
10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
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
7#endif 30#endif
8 31
9#ifdef __APPLE__ 32#ifdef __APPLE__
10 #include <OpenCL/opencl.h> 33 #include <OpenCL/opencl.h>
11#else 34#else
12 #include <CL/opencl.h> 35 #include <CL/opencl.h>
13#endif 36#endif
14 37
38#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1
40 #define PREFER_1_1 1
41#endif
42
15typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
16typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
17typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
18typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
19typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
20typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
21typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
22typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
23typedef cl_mem OpenCL__Image2D;
24typedef cl_mem OpenCL__Image3D;
25typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
26typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
27typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
28typedef cl_mem OpenCL__Image2D_ornull;
29typedef cl_mem OpenCL__Image3D_ornull;
30typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
31typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
32typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
33typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
34typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
35 60
61typedef struct mapped * OpenCL__Mapped;
62
36typedef 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;
37 89
38/*****************************************************************************/ 90/*****************************************************************************/
39 91
40// name must include a leading underscore 92// name must include a leading underscore
41// 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
81} 133}
82 134
83/*****************************************************************************/ 135/*****************************************************************************/
84 136
85/* up to two temporary buffers */ 137/* up to two temporary buffers */
86static void * 138static void * ecb_noinline
87tmpbuf (size_t size) 139tmpbuf (size_t size)
88{ 140{
89 enum { buffers = 3 }; 141 enum { buffers = 4 };
90 static int idx; 142 static int idx;
91 static void *buf [buffers]; 143 static void *buf [buffers];
92 static size_t len [buffers]; 144 static size_t len [buffers];
93 145
94 idx = ++idx % buffers; 146 idx = (idx + 1) % buffers;
95 147
96 if (len [idx] < size) 148 if (len [idx] < size)
97 { 149 {
98 free (buf [idx]); 150 free (buf [idx]);
99 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
167 if (res) \ 219 if (res) \
168 FAIL (name); 220 FAIL (name);
169 221
170/*****************************************************************************/ 222/*****************************************************************************/
171 223
172#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
173#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
174
175static 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
176SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 236SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
177{ 237{
238 // sv_derived_from is quite slow :(
178 if (SvROK (sv) && sv_derived_from (sv, pkg)) 239 if (SvROK (sv) && sv_derived_from (sv, pkg))
179 return (void *)SvIV (SvRV (sv)); 240 return (void *)SvIV (SvRV (sv));
180 241
181 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));
182} 253}
183 254
184/*****************************************************************************/ 255/*****************************************************************************/
185 256
186static cl_context_properties * 257static cl_context_properties * ecb_noinline
187SvCONTEXTPROPERTIES (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)
188{ 259{
189 if (!sv || !SvOK (sv)) 260 if (!sv || !SvOK (sv))
190 if (extra) 261 if (extra)
191 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
198 int i, len = av_len (av) + 1; 269 int i, len = av_len (av) + 1;
199 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));
200 cl_context_properties *l = p; 271 cl_context_properties *l = p;
201 272
202 if (len & 1) 273 if (len & 1)
203 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);
204 275
205 while (extracount--) 276 while (extracount--)
206 *l++ = *extra++; 277 *l++ = *extra++;
207 278
208 for (i = 0; i < len; i += 2) 279 for (i = 0; i < len; i += 2)
211 SV *p_sv = *av_fetch (av, i + 1, 0); 282 SV *p_sv = *av_fetch (av, i + 1, 0);
212 cl_context_properties v = SvIV (p_sv); // code below can override 283 cl_context_properties v = SvIV (p_sv); // code below can override
213 284
214 switch (t) 285 switch (t)
215 { 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
216 case CL_GLX_DISPLAY_KHR: 292 case CL_GLX_DISPLAY_KHR:
217 if (!SvOK (p_sv)) 293 if (!SvOK (p_sv))
218 { 294 {
219 void *func = glsym ("_glXGetCurrentDisplay"); 295 void *func = glsym ("_glXGetCurrentDisplay");
220 if (func) 296 if (func)
247 323
248 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);
249} 325}
250 326
251/*****************************************************************************/ 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 */
252 560
253static size_t 561static size_t
254img_row_pitch (cl_mem img) 562img_row_pitch (cl_mem img)
255{ 563{
256 size_t res; 564 size_t res;
257 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 565 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
258 return res; 566 return res;
259} 567}
260 568
261static cl_event * 569static cl_event * ecb_noinline
262event_list (SV **items, int count) 570event_list (SV **items, cl_uint *rcount, cl_event extra)
263{ 571{
264 if (!count) 572 cl_uint count = *rcount;
573
574 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
575 *rcount = count = 0;
576
577 if (!count && !extra)
265 return 0; 578 return 0;
266 579
267 cl_event *list = tmpbuf (sizeof (cl_event) * count); 580 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
581 int i = 0;
268 582
269 while (count--) 583 while (count--)
584 if (SvOK (items [count]))
270 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 585 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
271 586
587 if (extra)
588 list [i++] = extra;
589
590 *rcount = i;
591
272 return list; 592 return i ? list : 0;
273} 593}
274 594
275#define EVENT_LIST(items,count) \ 595#define EVENT_LIST(skip) \
276 cl_uint event_list_count = (count); \ 596 cl_uint event_list_count = items - (skip); \
277 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)
278 598
279#define INFO(class) \ 599#define INFO(class) \
280{ \ 600{ \
281 size_t size; \ 601 size_t size; \
282 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 602 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
286 SvCUR_set (sv, size); \ 606 SvCUR_set (sv, size); \
287 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 607 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
288 XPUSHs (sv); \ 608 XPUSHs (sv); \
289} 609}
290 610
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
291MODULE = OpenCL PACKAGE = OpenCL 704MODULE = OpenCL PACKAGE = OpenCL
292 705
293PROTOTYPES: 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;
294 718
295BOOT: 719BOOT:
296{ 720{
297 HV *stash = gv_stashpv ("OpenCL", 1); 721 HV *stash = gv_stashpv ("OpenCL", 1);
298 static const ivstr *civ, const_iv[] = { 722 static const ivstr *civ, const_iv[] = {
307 { sizeof (cl_half ), "SIZEOF_HALF" }, 731 { sizeof (cl_half ), "SIZEOF_HALF" },
308 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 732 { sizeof (cl_float ), "SIZEOF_FLOAT" },
309 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 733 { sizeof (cl_double), "SIZEOF_DOUBLE" },
310#include "constiv.h" 734#include "constiv.h"
311 }; 735 };
736
312 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--)
313 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);
314} 765}
315 766
316cl_int 767cl_int
317errno () 768errno ()
318 CODE: 769 CODE:
319 errno = res; 770 RETVAL = res;
771 OUTPUT:
772 RETVAL
320 773
321const char * 774const char *
322err2str (cl_int err) 775err2str (cl_int err = res)
323 776
324const char * 777const char *
325enum2str (cl_uint value) 778enum2str (cl_uint value)
326 779
327void 780void
335 list = tmpbuf (sizeof (*list) * count); 788 list = tmpbuf (sizeof (*list) * count);
336 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 789 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
337 790
338 EXTEND (SP, count); 791 EXTEND (SP, count);
339 for (i = 0; i < count; ++i) 792 for (i = 0; i < count; ++i)
340 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 793 PUSH_CLOBJ (stash_platform, list [i]);
341 794
342void 795void
343context_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)
344 PPCODE: 797 PPCODE:
798 CONTEXT_NOTIFY_CALLBACK;
345 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));
346 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 800 XPUSH_CLOBJ_CONTEXT;
347 801
348void 802void
349context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 803context (FUTURE properties, FUTURE devices, FUTURE notify)
350 PPCODE: 804 PPCODE:
351 /* der Gipfel der Kunst */ 805 /* der Gipfel der Kunst */
352 806
353void 807void
354wait_for_events (...) 808wait_for_events (...)
355 CODE: 809 CODE:
356 EVENT_LIST (0, items); 810 EVENT_LIST (0);
357 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 811 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
358 812
359PROTOTYPES: DISABLE 813PROTOTYPES: DISABLE
360 814
361MODULE = OpenCL PACKAGE = OpenCL::Platform 815MODULE = OpenCL PACKAGE = OpenCL::Platform
362 816
363void 817void
364info (OpenCL::Platform self, cl_platform_info name) 818info (OpenCL::Platform self, cl_platform_info name)
365 PPCODE: 819 PPCODE:
366 INFO (Platform) 820 INFO (Platform)
821
822void
823unload_compiler (OpenCL::Platform self)
824 CODE:
825#if CL_VERSION_1_2
826 clUnloadPlatformCompiler (self);
827#endif
367 828
368#BEGIN:platform 829#BEGIN:platform
369 830
370void 831void
371profile (OpenCL::Platform self) 832profile (OpenCL::Platform self)
397 list = tmpbuf (sizeof (*list) * count); 858 list = tmpbuf (sizeof (*list) * count);
398 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 859 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
399 860
400 EXTEND (SP, count); 861 EXTEND (SP, count);
401 for (i = 0; i < count; ++i) 862 for (i = 0; i < count; ++i)
402 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 863 PUSH_CLOBJ (stash_device, list [i]);
403 864
404void 865void
405context (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)
406 PPCODE: 867 PPCODE:
407 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 868 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
408 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);
409 873
410 AV *av = (AV *)SvRV (devices); 874 AV *av = (AV *)SvRV (devices);
411 cl_uint num_devices = av_len (av) + 1; 875 cl_uint num_devices = av_len (av) + 1;
412 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
413 int i; 878 int i;
414
415 for (i = num_devices; i--; ) 879 for (i = num_devices; i--; )
416 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");
417 881
882 CONTEXT_NOTIFY_CALLBACK;
418 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));
419 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 884 XPUSH_CLOBJ_CONTEXT;
420 885
421void 886void
422context_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)
423 PPCODE: 888 PPCODE:
424 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 889 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
425 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;
426 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));
427 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 894 XPUSH_CLOBJ_CONTEXT;
428 895
429MODULE = OpenCL PACKAGE = OpenCL::Device 896MODULE = OpenCL PACKAGE = OpenCL::Device
430 897
431void 898void
432info (OpenCL::Device self, cl_device_info name) 899info (OpenCL::Device self, cl_device_info name)
433 PPCODE: 900 PPCODE:
434 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
435 932
436#BEGIN:device 933#BEGIN:device
437 934
438void 935void
439type (OpenCL::Device self) 936type (OpenCL::Device self)
440 PPCODE: 937 PPCODE:
441 cl_device_type value [1]; 938 cl_device_type value [1];
442 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 939 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
443 EXTEND (SP, 1); 940 EXTEND (SP, 1);
444 const int i = 0; 941 const int i = 0;
445 PUSHs (sv_2mortal (newSViv (value [i]))); 942 PUSHs (sv_2mortal (newSVuv (value [i])));
446 943
447void 944void
448vendor_id (OpenCL::Device self) 945vendor_id (OpenCL::Device self)
449 ALIAS: 946 ALIAS:
450 vendor_id = CL_DEVICE_VENDOR_ID 947 vendor_id = CL_DEVICE_VENDOR_ID
471 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 968 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
472 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 969 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
473 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 970 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
474 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 971 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
475 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 972 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
476 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 973 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
477 PPCODE: 974 PPCODE:
478 cl_uint value [1]; 975 cl_uint value [1];
479 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 976 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
480 EXTEND (SP, 1); 977 EXTEND (SP, 1);
481 const int i = 0; 978 const int i = 0;
595 PPCODE: 1092 PPCODE:
596 cl_command_queue_properties value [1]; 1093 cl_command_queue_properties value [1];
597 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));
598 EXTEND (SP, 1); 1095 EXTEND (SP, 1);
599 const int i = 0; 1096 const int i = 0;
600 PUSHs (sv_2mortal (newSViv (value [i]))); 1097 PUSHs (sv_2mortal (newSVuv (value [i])));
601 1098
602void 1099void
603platform (OpenCL::Device self) 1100platform (OpenCL::Device self)
604 PPCODE: 1101 PPCODE:
605 cl_platform_id value [1]; 1102 cl_platform_id value [1];
606 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1103 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
607 EXTEND (SP, 1); 1104 EXTEND (SP, 1);
608 const int i = 0; 1105 const int i = 0;
609 { 1106 {
610 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1107 PUSH_CLOBJ (stash_platform, value [i]);
611 } 1108 }
612 1109
613void 1110void
614name (OpenCL::Device self) 1111name (OpenCL::Device self)
615 ALIAS: 1112 ALIAS:
634 cl_device_id value [1]; 1131 cl_device_id value [1];
635 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));
636 EXTEND (SP, 1); 1133 EXTEND (SP, 1);
637 const int i = 0; 1134 const int i = 0;
638 { 1135 {
639 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1136 PUSH_CLOBJ (stash_device, value [i]);
640 } 1137 }
641 1138
642void 1139void
643partition_types_ext (OpenCL::Device self) 1140partition_types_ext (OpenCL::Device self)
644 ALIAS: 1141 ALIAS:
655 for (i = 0; i < n; ++i) 1152 for (i = 0; i < n; ++i)
656 PUSHs (sv_2mortal (newSVuv (value [i]))); 1153 PUSHs (sv_2mortal (newSVuv (value [i])));
657 1154
658#END:device 1155#END:device
659 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
660MODULE = OpenCL PACKAGE = OpenCL::Context 1168MODULE = OpenCL PACKAGE = OpenCL::Context
661 1169
662void 1170void
663DESTROY (OpenCL::Context context) 1171DESTROY (OpenCL::Context self)
664 CODE: 1172 CODE:
665 clReleaseContext (context); 1173 clReleaseContext (self);
666 1174
667void 1175void
668info (OpenCL::Context self, cl_context_info name) 1176info (OpenCL::Context self, cl_context_info name)
669 PPCODE: 1177 PPCODE:
670 INFO (Context) 1178 INFO (Context)
671 1179
672void 1180void
673queue (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)
674 PPCODE: 1182 PPCODE:
675 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));
676 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1184 XPUSH_CLOBJ (stash_queue, queue);
677 1185
678void 1186void
679user_event (OpenCL::Context self) 1187user_event (OpenCL::Context self)
680 PPCODE: 1188 PPCODE:
681 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1189 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
682 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1190 XPUSH_CLOBJ (stash_userevent, ev);
683 1191
684void 1192void
685buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1193buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
686 PPCODE: 1194 PPCODE:
687 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))
688 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?");
689 1197
690 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));
691 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1199 XPUSH_CLOBJ (stash_bufferobj, mem);
692 1200
693void 1201void
694buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1202buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
695 PPCODE: 1203 PPCODE:
696 STRLEN len; 1204 STRLEN len;
697 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1205 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
698 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)))
699 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?");
700 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));
701 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1209 XPUSH_CLOBJ (stash_bufferobj, mem);
702 1210
1211#if CL_VERSION_1_2
1212
703void 1213void
704image2d (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) 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)
705 PPCODE: 1215 PPCODE:
706 STRLEN len; 1216 STRLEN len;
707 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1217 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
708 const cl_image_format format = { channel_order, channel_type }; 1218 const cl_image_format format = { channel_order, channel_type };
1219 const cl_image_desc desc = {
1220 type,
1221 width, height, depth,
1222 array_size, row_pitch, slice_pitch,
1223 num_mip_level, num_samples,
1224 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1225 };
709 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1226 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
710 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1227 HV *stash = stash_image;
1228 switch (type)
1229 {
1230 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1231 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1232 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1233 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1234 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1235 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1236 }
1237 XPUSH_CLOBJ (stash, mem);
711 1238
1239#endif
1240
712void 1241void
713image3d (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) 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)
714 PPCODE: 1243 PPCODE:
715 STRLEN len; 1244 STRLEN len;
716 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1245 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
717 const cl_image_format format = { channel_order, channel_type }; 1246 const cl_image_format format = { channel_order, channel_type };
1247#if PREFER_1_1
1248 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1249#else
1250 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1251 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1252#endif
1253 XPUSH_CLOBJ (stash_image2d, mem);
1254
1255void
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)
1257 PPCODE:
1258 STRLEN len;
1259 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1260 const cl_image_format format = { channel_order, channel_type };
1261#if PREFER_1_1
718 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1262 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
719 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1263#else
1264 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1265 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1266#endif
1267 XPUSH_CLOBJ (stash_image3d, mem);
720 1268
721#if cl_apple_gl_sharing || cl_khr_gl_sharing 1269#if cl_apple_gl_sharing || cl_khr_gl_sharing
722 1270
723void 1271void
724gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1272gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
725 PPCODE: 1273 PPCODE:
726 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1274 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
727 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1275 XPUSH_CLOBJ (stash_bufferobj, mem);
1276
1277void
1278gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1279 PPCODE:
1280 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1281 XPUSH_CLOBJ (stash_image2d, mem);
1282
1283#if CL_VERSION_1_2
1284
1285void
1286gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1287 ALIAS:
1288 PPCODE:
1289 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1290 cl_gl_object_type type;
1291 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1292 HV *stash = stash_memory;
1293 switch (type)
1294 {
1295 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1296 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1297 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1298 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1299 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1300 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1301 }
1302 XPUSH_CLOBJ (stash, mem);
1303
1304#endif
728 1305
729void 1306void
730gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1307gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
731 PPCODE: 1308 PPCODE:
1309#if PREFER_1_1
732 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1310 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
733 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1311#else
1312 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1313#endif
1314 XPUSH_CLOBJ (stash_image2d, mem);
734 1315
735void 1316void
736gl_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)
737 PPCODE: 1318 PPCODE:
1319#if PREFER_1_1
738 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));
739 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1321#else
740 1322 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
741void 1323#endif
742gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1324 XPUSH_CLOBJ (stash_image3d, mem);
743 PPCODE:
744 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
745 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
746 1325
747#endif 1326#endif
748 1327
749void 1328void
750supported_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)
770 1349
771void 1350void
772sampler (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)
773 PPCODE: 1352 PPCODE:
774 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));
775 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1354 XPUSH_CLOBJ (stash_sampler, sampler);
776 1355
777void 1356void
778program_with_source (OpenCL::Context self, SV *program) 1357program_with_source (OpenCL::Context self, SV *program)
779 PPCODE: 1358 PPCODE:
780 STRLEN len; 1359 STRLEN len;
781 size_t len2; 1360 size_t len2;
782 const char *ptr = SvPVbyte (program, len); 1361 const char *ptr = SvPVbyte (program, len);
783 1362
784 len2 = len; 1363 len2 = len;
785 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));
786 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
787 1434
788#BEGIN:context 1435#BEGIN:context
789 1436
790void 1437void
791reference_count (OpenCL::Context self) 1438reference_count (OpenCL::Context self)
808 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1455 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
809 int i, n = size / sizeof (*value); 1456 int i, n = size / sizeof (*value);
810 EXTEND (SP, n); 1457 EXTEND (SP, n);
811 for (i = 0; i < n; ++i) 1458 for (i = 0; i < n; ++i)
812 { 1459 {
813 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1460 PUSH_CLOBJ (stash_device, value [i]);
814 } 1461 }
815 1462
816void 1463void
817properties (OpenCL::Context self) 1464properties (OpenCL::Context self)
818 PPCODE: 1465 PPCODE:
833DESTROY (OpenCL::Queue self) 1480DESTROY (OpenCL::Queue self)
834 CODE: 1481 CODE:
835 clReleaseCommandQueue (self); 1482 clReleaseCommandQueue (self);
836 1483
837void 1484void
838enqueue_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
839 PPCODE: 1488 PPCODE:
840 cl_event ev = 0; 1489 cl_event ev = 0;
841 EVENT_LIST (6, items - 6); 1490 EVENT_LIST (6);
842 1491
843 SvUPGRADE (data, SVt_PV); 1492 SvUPGRADE (data, SVt_PV);
844 SvGROW (data, len); 1493 SvGROW (data, len);
845 SvPOK_only (data); 1494 SvPOK_only (data);
846 SvCUR_set (data, len); 1495 SvCUR_set (data, len);
847 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));
848 1497
849 if (ev) 1498 if (ev)
850 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1499 XPUSH_CLOBJ (stash_event, ev);
851 1500
852void 1501void
853enqueue_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
854 PPCODE: 1505 PPCODE:
855 cl_event ev = 0; 1506 cl_event ev = 0;
856 STRLEN len; 1507 STRLEN len;
857 char *ptr = SvPVbyte (data, len); 1508 char *ptr = SvPVbyte (data, len);
858 EVENT_LIST (5, items - 5); 1509 EVENT_LIST (5);
859 1510
860 NEED_SUCCESS (EnqueueReadBuffer, (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));
861 1512
862 if (ev) 1513 if (ev)
863 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1514 XPUSH_CLOBJ (stash_event, ev);
864 1515
1516#if CL_VERSION_1_2
1517
865void 1518void
866enqueue_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
867 PPCODE: 1522 PPCODE:
868 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;
869 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
870 1569
871 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));
872 1571
873 if (ev) 1572 if (ev)
874 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1573 XPUSH_CLOBJ (stash_event, ev);
875 1574
876void 1575void
877enqueue_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
878 PPCODE: 1579 PPCODE:
879 cl_event ev = 0; 1580 cl_event ev = 0;
880 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 };
881 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 };
882 const size_t region[3] = { width, height, depth }; 1583 const size_t region[3] = { width, height, depth };
883 EVENT_LIST (17, items - 17); 1584 EVENT_LIST (17);
884 1585
885 if (!buf_row_pitch) 1586 if (!buf_row_pitch)
886 buf_row_pitch = region [0]; 1587 buf_row_pitch = region [0];
887 1588
888 if (!buf_slice_pitch) 1589 if (!buf_slice_pitch)
901 SvPOK_only (data); 1602 SvPOK_only (data);
902 SvCUR_set (data, len); 1603 SvCUR_set (data, len);
903 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));
904 1605
905 if (ev) 1606 if (ev)
906 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1607 XPUSH_CLOBJ (stash_event, ev);
907 1608
908void 1609void
909enqueue_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
910 PPCODE: 1613 PPCODE:
911 cl_event ev = 0; 1614 cl_event ev = 0;
912 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 };
913 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 };
914 const size_t region[3] = { width, height, depth }; 1617 const size_t region[3] = { width, height, depth };
915 STRLEN len; 1618 STRLEN len;
916 char *ptr = SvPVbyte (data, len); 1619 char *ptr = SvPVbyte (data, len);
917 EVENT_LIST (17, items - 17); 1620 EVENT_LIST (17);
918 1621
919 if (!buf_row_pitch) 1622 if (!buf_row_pitch)
920 buf_row_pitch = region [0]; 1623 buf_row_pitch = region [0];
921 1624
922 if (!buf_slice_pitch) 1625 if (!buf_slice_pitch)
931 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1634 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
932 1635
933 if (len < min_len) 1636 if (len < min_len)
934 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1637 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
935 1638
936 NEED_SUCCESS (EnqueueWriteBufferRect, (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)); 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));
937 1640
938 if (ev) 1641 if (ev)
939 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1642 XPUSH_CLOBJ (stash_event, ev);
940 1643
941void 1644void
942enqueue_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
943 PPCODE: 1648 PPCODE:
944 cl_event ev = 0; 1649 cl_event ev = 0;
945 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 };
946 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 };
947 const size_t region[3] = { width, height, depth }; 1652 const size_t region[3] = { width, height, depth };
948 EVENT_LIST (16, items - 16); 1653 EVENT_LIST (16);
949 1654
950 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));
951 1656
952 if (ev) 1657 if (ev)
953 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1658 XPUSH_CLOBJ (stash_event, ev);
954 1659
955void 1660void
956enqueue_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
957 PPCODE: 1664 PPCODE:
958 cl_event ev = 0; 1665 cl_event ev = 0;
959 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 };
960 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
961 EVENT_LIST (12, items - 12); 1668 EVENT_LIST (12);
962 1669
963 if (!row_pitch) 1670 if (!row_pitch)
964 row_pitch = img_row_pitch (src); 1671 row_pitch = img_row_pitch (src);
965 1672
966 if (depth > 1 && !slice_pitch) 1673 if (depth > 1 && !slice_pitch)
973 SvPOK_only (data); 1680 SvPOK_only (data);
974 SvCUR_set (data, len); 1681 SvCUR_set (data, len);
975 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));
976 1683
977 if (ev) 1684 if (ev)
978 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1685 XPUSH_CLOBJ (stash_event, ev);
979 1686
980void 1687void
981enqueue_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
982 PPCODE: 1691 PPCODE:
983 cl_event ev = 0; 1692 cl_event ev = 0;
984 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 };
985 const size_t region[3] = { width, height, depth }; 1694 const size_t region[3] = { width, height, depth };
986 STRLEN len; 1695 STRLEN len;
987 char *ptr = SvPVbyte (data, len); 1696 char *ptr = SvPVbyte (data, len);
988 EVENT_LIST (12, items - 12); 1697 EVENT_LIST (12);
989 1698
990 if (!row_pitch) 1699 if (!row_pitch)
991 row_pitch = img_row_pitch (dst); 1700 row_pitch = img_row_pitch (dst);
992 1701
993 if (depth > 1 && !slice_pitch) 1702 if (depth > 1 && !slice_pitch)
996 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1705 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
997 1706
998 if (len < min_len) 1707 if (len < min_len)
999 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1708 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1000 1709
1001 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), 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));
1002 1711
1003 if (ev) 1712 if (ev)
1004 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1713 XPUSH_CLOBJ (stash_event, ev);
1005 1714
1006void 1715void
1007enqueue_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
1008 PPCODE: 1719 PPCODE:
1009 cl_event ev = 0; 1720 cl_event ev = 0;
1010 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 };
1011 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 };
1012 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
1013 EVENT_LIST (12, items - 12); 1724 EVENT_LIST (12);
1014 1725
1015 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));
1016 1727
1017 if (ev) 1728 if (ev)
1018 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
1019 1730
1020void 1731void
1021enqueue_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
1022 PPCODE: 1735 PPCODE:
1023 cl_event ev = 0; 1736 cl_event ev = 0;
1024 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 };
1025 const size_t region[3] = { width, height, depth }; 1810 const size_t region[3] = { width, height, depth };
1811 size_t row_pitch, slice_pitch;
1026 EVENT_LIST (10, items - 10); 1812 EVENT_LIST (10);
1027 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
1028 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));
1029 1842
1030 if (ev) 1843 if (ev)
1031 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1844 XPUSH_CLOBJ (stash_event, ev);
1032 1845
1033void 1846void
1034enqueue_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, ...)
1035 PPCODE:
1036 cl_event ev = 0;
1037 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1038 const size_t region[3] = { width, height, depth };
1039 EVENT_LIST (10, items - 10);
1040
1041 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1042
1043 if (ev)
1044 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1045
1046void
1047enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1048 PPCODE:
1049 cl_event ev = 0;
1050 EVENT_LIST (2, items - 2);
1051
1052 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1053
1054 if (ev)
1055 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1056
1057void
1058enqueue_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
1059 PPCODE: 1850 PPCODE:
1060 cl_event ev = 0; 1851 cl_event ev = 0;
1061 size_t *gwo = 0, *gws, *lws = 0; 1852 size_t *gwo = 0, *gws, *lws = 0;
1062 int gws_len; 1853 int gws_len;
1063 size_t *lists; 1854 size_t *lists;
1064 int i; 1855 int i;
1065 EVENT_LIST (5, items - 5); 1856 EVENT_LIST (5);
1066 1857
1067 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)
1068 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1859 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1069 1860
1070 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1861 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1071 1862
1072 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1863 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1073 1864
1074 gws = lists + gws_len * 0; 1865 gws = lists + gws_len * 0;
1075 for (i = 0; i < gws_len; ++i) 1866 for (i = 0; i < gws_len; ++i)
1867 {
1076 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 }
1077 1873
1078 if (SvOK (global_work_offset)) 1874 if (SvOK (global_work_offset))
1079 { 1875 {
1080 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)
1081 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");
1088 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1884 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1089 } 1885 }
1090 1886
1091 if (SvOK (local_work_size)) 1887 if (SvOK (local_work_size))
1092 { 1888 {
1093 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)
1094 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");
1095 1891
1096 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1892 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1097 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");
1098 1894
1099 lws = lists + gws_len * 2; 1895 lws = lists + gws_len * 2;
1100 for (i = 0; i < gws_len; ++i) 1896 for (i = 0; i < gws_len; ++i)
1897 {
1101 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 }
1102 } 1903 }
1103 1904
1104 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));
1105 1906
1106 if (ev) 1907 if (ev)
1107 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
1108 1938
1109#if cl_apple_gl_sharing || cl_khr_gl_sharing 1939#if cl_apple_gl_sharing || cl_khr_gl_sharing
1110 1940
1111void 1941void
1112enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1942acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1113 ALIAS: 1943 ALIAS:
1944 release_gl_objects = 1
1945 enqueue_acquire_gl_objects = 0
1114 enqueue_release_gl_objects = 1 1946 enqueue_release_gl_objects = 1
1115 CODE: 1947 PPCODE:
1116 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1948 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1117 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");
1118 1950
1119 cl_event ev = 0; 1951 cl_event ev = 0;
1120 EVENT_LIST (2, items - 2); 1952 EVENT_LIST (2);
1121 AV *av = (AV *)SvRV (objects); 1953 AV *av = (AV *)SvRV (objects);
1122 cl_uint num_objects = av_len (av) + 1; 1954 cl_uint num_objects = av_len (av) + 1;
1123 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1955 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1956
1124 int i; 1957 int i;
1125
1126 for (i = num_objects; i--; ) 1958 for (i = num_objects; i--; )
1127 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");
1128 1960
1129 if (ix) 1961 if (ix)
1130 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));
1131 else 1963 else
1132 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));
1133 1965
1134 if (ev) 1966 if (ev)
1135 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1967 XPUSH_CLOBJ (stash_event, ev);
1136 1968
1137#endif 1969#endif
1138 1970
1139void 1971void
1140enqueue_marker (OpenCL::Queue self)
1141 PPCODE:
1142 cl_event ev;
1143 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1144 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1145
1146void
1147enqueue_wait_for_events (OpenCL::Queue self, ...) 1972wait_for_events (OpenCL::Queue self, ...)
1973 ALIAS:
1974 enqueue_wait_for_events = 0
1148 CODE: 1975 CODE:
1149 EVENT_LIST (1, items - 1); 1976 EVENT_LIST (1);
1977#if PREFER_1_1
1150 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1978 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1979#else
1980 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1981#endif
1151 1982
1152void 1983void
1153enqueue_barrier (OpenCL::Queue self) 1984marker (OpenCL::Queue self, ...)
1985 ALIAS:
1986 enqueue_marker = 0
1154 CODE: 1987 PPCODE:
1988 cl_event ev = 0;
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
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
2003 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2004#endif
2005 if (ev)
2006 XPUSH_CLOBJ (stash_event, ev);
2007
2008void
2009barrier (OpenCL::Queue self, ...)
2010 ALIAS:
2011 enqueue_barrier = 0
2012 PPCODE:
2013 cl_event ev = 0;
2014 EVENT_LIST (1);
2015#if PREFER_1_1
2016 if (!event_list_count && GIMME_V == G_VOID)
1155 NEED_SUCCESS (EnqueueBarrier, (self)); 2017 NEED_SUCCESS (EnqueueBarrier, (self));
2018 else
2019#if CL_VERSION_1_2
2020 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2021#else
2022 {
2023 if (event_list_count)
2024 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2025
2026 if (GIMME_V != G_VOID)
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));
2032#endif
2033 if (ev)
2034 XPUSH_CLOBJ (stash_event, ev);
1156 2035
1157void 2036void
1158flush (OpenCL::Queue self) 2037flush (OpenCL::Queue self)
1159 CODE: 2038 CODE:
1160 NEED_SUCCESS (Flush, (self)); 2039 NEED_SUCCESS (Flush, (self));
1178 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2057 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1179 EXTEND (SP, 1); 2058 EXTEND (SP, 1);
1180 const int i = 0; 2059 const int i = 0;
1181 { 2060 {
1182 NEED_SUCCESS (RetainContext, (value [i])); 2061 NEED_SUCCESS (RetainContext, (value [i]));
1183 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2062 PUSH_CLOBJ (stash_context, value [i]);
1184 } 2063 }
1185 2064
1186void 2065void
1187device (OpenCL::Queue self) 2066device (OpenCL::Queue self)
1188 PPCODE: 2067 PPCODE:
1189 cl_device_id value [1]; 2068 cl_device_id value [1];
1190 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2069 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1191 EXTEND (SP, 1); 2070 EXTEND (SP, 1);
1192 const int i = 0; 2071 const int i = 0;
1193 { 2072 {
1194 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2073 PUSH_CLOBJ (stash_device, value [i]);
1195 } 2074 }
1196 2075
1197void 2076void
1198reference_count (OpenCL::Queue self) 2077reference_count (OpenCL::Queue self)
1199 PPCODE: 2078 PPCODE:
1208 PPCODE: 2087 PPCODE:
1209 cl_command_queue_properties value [1]; 2088 cl_command_queue_properties value [1];
1210 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2089 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1211 EXTEND (SP, 1); 2090 EXTEND (SP, 1);
1212 const int i = 0; 2091 const int i = 0;
1213 PUSHs (sv_2mortal (newSViv (value [i]))); 2092 PUSHs (sv_2mortal (newSVuv (value [i])));
1214 2093
1215#END:command_queue 2094#END:command_queue
1216 2095
1217MODULE = OpenCL PACKAGE = OpenCL::Memory 2096MODULE = OpenCL PACKAGE = OpenCL::Memory
1218 2097
1233 PPCODE: 2112 PPCODE:
1234 cl_mem_object_type value [1]; 2113 cl_mem_object_type value [1];
1235 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2114 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1236 EXTEND (SP, 1); 2115 EXTEND (SP, 1);
1237 const int i = 0; 2116 const int i = 0;
1238 PUSHs (sv_2mortal (newSViv (value [i]))); 2117 PUSHs (sv_2mortal (newSVuv (value [i])));
1239 2118
1240void 2119void
1241flags (OpenCL::Memory self) 2120flags (OpenCL::Memory self)
1242 PPCODE: 2121 PPCODE:
1243 cl_mem_flags value [1]; 2122 cl_mem_flags value [1];
1244 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2123 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1245 EXTEND (SP, 1); 2124 EXTEND (SP, 1);
1246 const int i = 0; 2125 const int i = 0;
1247 PUSHs (sv_2mortal (newSViv (value [i]))); 2126 PUSHs (sv_2mortal (newSVuv (value [i])));
1248 2127
1249void 2128void
1250size (OpenCL::Memory self) 2129size (OpenCL::Memory self)
1251 ALIAS: 2130 ALIAS:
1252 size = CL_MEM_SIZE 2131 size = CL_MEM_SIZE
1286 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2165 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1287 EXTEND (SP, 1); 2166 EXTEND (SP, 1);
1288 const int i = 0; 2167 const int i = 0;
1289 { 2168 {
1290 NEED_SUCCESS (RetainContext, (value [i])); 2169 NEED_SUCCESS (RetainContext, (value [i]));
1291 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2170 PUSH_CLOBJ (stash_context, value [i]);
1292 } 2171 }
1293 2172
1294void 2173void
1295associated_memobject (OpenCL::Memory self) 2174associated_memobject (OpenCL::Memory self)
1296 PPCODE: 2175 PPCODE:
1298 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));
1299 EXTEND (SP, 1); 2178 EXTEND (SP, 1);
1300 const int i = 0; 2179 const int i = 0;
1301 { 2180 {
1302 NEED_SUCCESS (RetainMemObject, (value [i])); 2181 NEED_SUCCESS (RetainMemObject, (value [i]));
1303 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2182 PUSH_CLOBJ (stash_memory, value [i]);
1304 } 2183 }
1305 2184
1306#END:mem 2185#END:mem
1307 2186
1308#if cl_apple_gl_sharing || cl_khr_gl_sharing 2187#if cl_apple_gl_sharing || cl_khr_gl_sharing
1328 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!");
1329 2208
1330 cl_buffer_region crdata = { origin, size }; 2209 cl_buffer_region crdata = { origin, size };
1331 2210
1332 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));
1333 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2212 XPUSH_CLOBJ (stash_buffer, mem);
1334 2213
1335MODULE = OpenCL PACKAGE = OpenCL::Image 2214MODULE = OpenCL PACKAGE = OpenCL::Image
1336 2215
1337void 2216void
1338image_info (OpenCL::Image self, cl_image_info name) 2217image_info (OpenCL::Image self, cl_image_info name)
1339 PPCODE: 2218 PPCODE:
1340 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)));
1341 2229
1342#BEGIN:image 2230#BEGIN:image
1343 2231
1344void 2232void
1345element_size (OpenCL::Image self) 2233element_size (OpenCL::Image self)
1415 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2303 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1416 EXTEND (SP, 1); 2304 EXTEND (SP, 1);
1417 const int i = 0; 2305 const int i = 0;
1418 { 2306 {
1419 NEED_SUCCESS (RetainContext, (value [i])); 2307 NEED_SUCCESS (RetainContext, (value [i]));
1420 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2308 PUSH_CLOBJ (stash_context, value [i]);
1421 } 2309 }
1422 2310
1423void 2311void
1424normalized_coords (OpenCL::Sampler self) 2312normalized_coords (OpenCL::Sampler self)
1425 PPCODE: 2313 PPCODE:
1426 cl_addressing_mode value [1]; 2314 cl_addressing_mode value [1];
1427 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));
1428 EXTEND (SP, 1); 2316 EXTEND (SP, 1);
1429 const int i = 0; 2317 const int i = 0;
1430 PUSHs (sv_2mortal (newSViv (value [i]))); 2318 PUSHs (sv_2mortal (newSVuv (value [i])));
1431 2319
1432void 2320void
1433addressing_mode (OpenCL::Sampler self) 2321addressing_mode (OpenCL::Sampler self)
1434 PPCODE: 2322 PPCODE:
1435 cl_filter_mode value [1]; 2323 cl_filter_mode value [1];
1436 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));
1437 EXTEND (SP, 1); 2325 EXTEND (SP, 1);
1438 const int i = 0; 2326 const int i = 0;
1439 PUSHs (sv_2mortal (newSViv (value [i]))); 2327 PUSHs (sv_2mortal (newSVuv (value [i])));
1440 2328
1441void 2329void
1442filter_mode (OpenCL::Sampler self) 2330filter_mode (OpenCL::Sampler self)
1443 PPCODE: 2331 PPCODE:
1444 cl_bool value [1]; 2332 cl_bool value [1];
1455DESTROY (OpenCL::Program self) 2343DESTROY (OpenCL::Program self)
1456 CODE: 2344 CODE:
1457 clReleaseProgram (self); 2345 clReleaseProgram (self);
1458 2346
1459void 2347void
1460build (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
1461 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
1462 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));
1463 2385
1464void 2386void
1465build_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)
1466 PPCODE: 2388 PPCODE:
1467 size_t size; 2389 size_t size;
1502 2424
1503void 2425void
1504kernel (OpenCL::Program program, SV *function) 2426kernel (OpenCL::Program program, SV *function)
1505 PPCODE: 2427 PPCODE:
1506 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));
1507 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]);
1508 2443
1509void 2444void
1510info (OpenCL::Program self, cl_program_info name) 2445info (OpenCL::Program self, cl_program_info name)
1511 PPCODE: 2446 PPCODE:
1512 INFO (Program) 2447 INFO (Program)
1530 { 2465 {
1531 SV *sv = sv_2mortal (newSV (sizes [i])); 2466 SV *sv = sv_2mortal (newSV (sizes [i]));
1532 SvUPGRADE (sv, SVt_PV); 2467 SvUPGRADE (sv, SVt_PV);
1533 SvPOK_only (sv); 2468 SvPOK_only (sv);
1534 SvCUR_set (sv, sizes [i]); 2469 SvCUR_set (sv, sizes [i]);
1535 ptrs [i] = SvPVX (sv); 2470 ptrs [i] = (void *)SvPVX (sv);
1536 PUSHs (sv); 2471 PUSHs (sv);
1537 } 2472 }
1538 2473
1539 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2474 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1540 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2475 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1560 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2495 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1561 EXTEND (SP, 1); 2496 EXTEND (SP, 1);
1562 const int i = 0; 2497 const int i = 0;
1563 { 2498 {
1564 NEED_SUCCESS (RetainContext, (value [i])); 2499 NEED_SUCCESS (RetainContext, (value [i]));
1565 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2500 PUSH_CLOBJ (stash_context, value [i]);
1566 } 2501 }
1567 2502
1568void 2503void
1569devices (OpenCL::Program self) 2504devices (OpenCL::Program self)
1570 PPCODE: 2505 PPCODE:
1574 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2509 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1575 int i, n = size / sizeof (*value); 2510 int i, n = size / sizeof (*value);
1576 EXTEND (SP, n); 2511 EXTEND (SP, n);
1577 for (i = 0; i < n; ++i) 2512 for (i = 0; i < n; ++i)
1578 { 2513 {
1579 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2514 PUSH_CLOBJ (stash_device, value [i]);
1580 } 2515 }
1581 2516
1582void 2517void
1583source (OpenCL::Program self) 2518source (OpenCL::Program self)
1584 PPCODE: 2519 PPCODE:
1610DESTROY (OpenCL::Kernel self) 2545DESTROY (OpenCL::Kernel self)
1611 CODE: 2546 CODE:
1612 clReleaseKernel (self); 2547 clReleaseKernel (self);
1613 2548
1614void 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
1615set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2615set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1616 CODE: 2616 CODE:
1617 clSetKernelArg (self, idx, sizeof (value), &value); 2617 clSetKernelArg (self, idx, sizeof (value), &value);
1618 2618
1619void 2619void
1667 clSetKernelArg (self, idx, sizeof (value), &value); 2667 clSetKernelArg (self, idx, sizeof (value), &value);
1668 2668
1669void 2669void
1670set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2670set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1671 CODE: 2671 CODE:
2672 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2673
2674void
2675set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2676 CODE:
2677 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2678
2679void
2680set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2681 ALIAS:
2682 set_image2d = 0
2683 set_image3d = 0
2684 CODE:
2685 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2686
2687void
2688set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2689 CODE:
1672 clSetKernelArg (self, idx, sizeof (value), &value); 2690 clSetKernelArg (self, idx, sizeof (value), &value);
1673 2691
1674void 2692void
1675set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2693set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1676 CODE: 2694 CODE:
1677 clSetKernelArg (self, idx, sizeof (value), &value); 2695 clSetKernelArg (self, idx, size, 0);
1678
1679void
1680set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1681 CODE:
1682 clSetKernelArg (self, idx, sizeof (value), &value);
1683
1684void
1685set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1686 CODE:
1687 clSetKernelArg (self, idx, sizeof (value), &value);
1688
1689void
1690set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1691 CODE:
1692 clSetKernelArg (self, idx, sizeof (value), &value);
1693 2696
1694void 2697void
1695set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2698set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1696 CODE: 2699 CODE:
1697 clSetKernelArg (self, idx, sizeof (value), &value); 2700 clSetKernelArg (self, idx, sizeof (value), &value);
1733 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2736 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1734 EXTEND (SP, 1); 2737 EXTEND (SP, 1);
1735 const int i = 0; 2738 const int i = 0;
1736 { 2739 {
1737 NEED_SUCCESS (RetainContext, (value [i])); 2740 NEED_SUCCESS (RetainContext, (value [i]));
1738 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2741 PUSH_CLOBJ (stash_context, value [i]);
1739 } 2742 }
1740 2743
1741void 2744void
1742program (OpenCL::Kernel self) 2745program (OpenCL::Kernel self)
1743 PPCODE: 2746 PPCODE:
1745 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2748 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1746 EXTEND (SP, 1); 2749 EXTEND (SP, 1);
1747 const int i = 0; 2750 const int i = 0;
1748 { 2751 {
1749 NEED_SUCCESS (RetainProgram, (value [i])); 2752 NEED_SUCCESS (RetainProgram, (value [i]));
1750 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2753 PUSH_CLOBJ (stash_program, value [i]);
1751 } 2754 }
1752 2755
1753#END:kernel 2756#END:kernel
1754 2757
1755void 2758void
1802 const int i = 0; 2805 const int i = 0;
1803 PUSHs (sv_2mortal (newSVuv (value [i]))); 2806 PUSHs (sv_2mortal (newSVuv (value [i])));
1804 2807
1805#END:kernel_work_group 2808#END:kernel_work_group
1806 2809
2810#if CL_VERSION_1_2
2811
2812void
2813arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2814 PPCODE:
2815 size_t size;
2816 NEED_SUCCESS (GetKernelArgInfo, (self, name, 0, 0, &size));
2817 SV *sv = sv_2mortal (newSV (size));
2818 SvUPGRADE (sv, SVt_PV);
2819 SvPOK_only (sv);
2820 SvCUR_set (sv, size);
2821 NEED_SUCCESS (GetKernelArgInfo, (self, name, size, SvPVX (sv), 0));
2822 XPUSHs (sv);
2823
2824#BEGIN:kernel_arg
2825
2826#END:kernel_arg
2827
2828#endif
2829
1807MODULE = OpenCL PACKAGE = OpenCL::Event 2830MODULE = OpenCL PACKAGE = OpenCL::Event
1808 2831
1809void 2832void
1810DESTROY (OpenCL::Event self) 2833DESTROY (OpenCL::Event self)
1811 CODE: 2834 CODE:
1813 2836
1814void 2837void
1815wait (OpenCL::Event self) 2838wait (OpenCL::Event self)
1816 CODE: 2839 CODE:
1817 clWaitForEvents (1, &self); 2840 clWaitForEvents (1, &self);
2841
2842void
2843cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2844 CODE:
2845 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1818 2846
1819void 2847void
1820info (OpenCL::Event self, cl_event_info name) 2848info (OpenCL::Event self, cl_event_info name)
1821 PPCODE: 2849 PPCODE:
1822 INFO (Event) 2850 INFO (Event)
1830 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2858 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1831 EXTEND (SP, 1); 2859 EXTEND (SP, 1);
1832 const int i = 0; 2860 const int i = 0;
1833 { 2861 {
1834 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2862 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1835 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2863 PUSH_CLOBJ (stash_queue, value [i]);
1836 } 2864 }
1837 2865
1838void 2866void
1839command_type (OpenCL::Event self) 2867command_type (OpenCL::Event self)
1840 PPCODE: 2868 PPCODE:
1863 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2891 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1864 EXTEND (SP, 1); 2892 EXTEND (SP, 1);
1865 const int i = 0; 2893 const int i = 0;
1866 { 2894 {
1867 NEED_SUCCESS (RetainContext, (value [i])); 2895 NEED_SUCCESS (RetainContext, (value [i]));
1868 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2896 PUSH_CLOBJ (stash_context, value [i]);
1869 } 2897 }
1870 2898
1871#END:event 2899#END:event
1872 2900
1873void 2901void
1898void 2926void
1899set_status (OpenCL::UserEvent self, cl_int execution_status) 2927set_status (OpenCL::UserEvent self, cl_int execution_status)
1900 CODE: 2928 CODE:
1901 clSetUserEventStatus (self, execution_status); 2929 clSetUserEventStatus (self, execution_status);
1902 2930
2931MODULE = OpenCL PACKAGE = OpenCL::Mapped
2932
2933void
2934DESTROY (SV *self)
2935 CODE:
2936 OpenCL__Mapped mapped = SvMAPPED (self);
2937
2938 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2939 mapped_detach (self, mapped);
2940
2941 clReleaseCommandQueue (mapped->queue);
2942 clReleaseEvent (mapped->event);
2943 Safefree (mapped);
2944
2945void
2946unmap (OpenCL::Mapped self, ...)
2947 CODE:
2948 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2949
2950bool
2951mapped (OpenCL::Mapped self)
2952 CODE:
2953 RETVAL = !!self->ptr;
2954 OUTPUT:
2955 RETVAL
2956
2957void
2958wait (OpenCL::Mapped self)
2959 PPCODE:
2960 if (self->event)
2961 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2962
2963void
2964event (OpenCL::Mapped self)
2965 PPCODE:
2966 if (!self->event)
2967 XSRETURN_UNDEF;
2968
2969 clRetainEvent (self->event);
2970 XPUSH_CLOBJ (stash_event, self->event);
2971
2972size_t
2973size (OpenCL::Mapped self)
2974 CODE:
2975 RETVAL = self->cb;
2976 OUTPUT:
2977 RETVAL
2978
2979IV
2980ptr (OpenCL::Mapped self)
2981 CODE:
2982 RETVAL = PTR2IV (self->ptr);
2983 OUTPUT:
2984 RETVAL
2985
2986void
2987set (OpenCL::Mapped self, size_t offset, SV *data)
2988 CODE:
2989 STRLEN len;
2990 const char *ptr = SvPVbyte (data, len);
2991
2992 if (offset + len > self->cb)
2993 croak ("OpenCL::Mapped::set out of bound condition detected");
2994
2995 memcpy (offset + (char *)self->ptr, ptr, len);
2996
2997MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2998
2999MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3000
3001IV
3002row_pitch (OpenCL::Mapped self)
3003 ALIAS:
3004 slice_pitch = 1
3005 CODE:
3006 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3007 OUTPUT:
3008 RETVAL
3009
3010
3011

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines