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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.47 by root, Tue Apr 24 13:30:49 2012 UTC vs.
Revision 1.70 by root, Sat May 5 02:33:55 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
4 10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
7#endif 13#endif
8 14
34 #define PREFER_1_1 1 40 #define PREFER_1_1 1
35#endif 41#endif
36 42
37typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
38typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
39typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
40typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
41typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
42typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
43typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
44typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
45typedef cl_mem OpenCL__Image2D;
46typedef cl_mem OpenCL__Image3D;
47typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
48typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
49typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
50typedef cl_mem OpenCL__Image2D_ornull;
51typedef cl_mem OpenCL__Image3D_ornull;
52typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
53typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
54typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
55typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
56typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
57 60
61typedef struct mapped * OpenCL__Mapped;
62
58typedef 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;
59 89
60/*****************************************************************************/ 90/*****************************************************************************/
61 91
62// name must include a leading underscore 92// name must include a leading underscore
63// 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
103} 133}
104 134
105/*****************************************************************************/ 135/*****************************************************************************/
106 136
107/* up to two temporary buffers */ 137/* up to two temporary buffers */
108static void * 138static void * ecb_noinline
109tmpbuf (size_t size) 139tmpbuf (size_t size)
110{ 140{
111 enum { buffers = 3 }; 141 enum { buffers = 4 };
112 static int idx; 142 static int idx;
113 static void *buf [buffers]; 143 static void *buf [buffers];
114 static size_t len [buffers]; 144 static size_t len [buffers];
115 145
116 idx = (idx + 1) % buffers; 146 idx = (idx + 1) % buffers;
121 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
122 buf [idx] = malloc (len [idx]); 152 buf [idx] = malloc (len [idx]);
123 } 153 }
124 154
125 return buf [idx]; 155 return buf [idx];
156}
157
158static const char * ecb_noinline
159cv_get_name (CV *cv)
160{
161 static char fullname [256];
162
163 GV *gv = CvGV (cv); // gv better be nonzero
164
165 HV *stash = GvSTASH (gv);
166 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
167 const char *gvname = GvNAME (gv);
168
169 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
170 return fullname;
126} 171}
127 172
128/*****************************************************************************/ 173/*****************************************************************************/
129 174
130typedef struct 175typedef struct
189 if (res) \ 234 if (res) \
190 FAIL (name); 235 FAIL (name);
191 236
192/*****************************************************************************/ 237/*****************************************************************************/
193 238
194#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
195#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
196
197static void * 239static SV *
240new_clobj (HV *stash, IV id)
241{
242 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
243}
244
245#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
246#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
247
248/* cl objects are either \$iv, or [$iv, ...] */
249/* they can be upgraded at runtime to the array form */
250static void * ecb_noinline
198SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 251SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
199{ 252{
253 // sv_derived_from is quite slow :(
200 if (SvROK (sv) && sv_derived_from (sv, pkg)) 254 if (SvROK (sv) && sv_derived_from (sv, pkg))
201 return (void *)SvIV (SvRV (sv)); 255 return (void *)SvIV (SvRV (sv));
202 256
203 croak ("%s: %s is not of type %s", func, svname, pkg); 257 croak ("%s: %s is not of type %s", func, svname, pkg);
258}
259
260// the "no-inherit" version of the above
261static void * ecb_noinline
262SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
263{
264 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
265 return (void *)SvIV (SvRV (sv));
266
267 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
204} 268}
205 269
206/*****************************************************************************/ 270/*****************************************************************************/
207 271
208static cl_context_properties * 272static cl_context_properties * ecb_noinline
209SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 273SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
210{ 274{
211 if (!sv || !SvOK (sv)) 275 if (!sv || !SvOK (sv))
212 if (extra) 276 if (extra)
213 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 277 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
220 int i, len = av_len (av) + 1; 284 int i, len = av_len (av) + 1;
221 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 285 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
222 cl_context_properties *l = p; 286 cl_context_properties *l = p;
223 287
224 if (len & 1) 288 if (len & 1)
225 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 289 croak ("%s: %s is not a property list (must contain an even number of elements)", func, svname);
226 290
227 while (extracount--) 291 while (extracount--)
228 *l++ = *extra++; 292 *l++ = *extra++;
229 293
230 for (i = 0; i < len; i += 2) 294 for (i = 0; i < len; i += 2)
233 SV *p_sv = *av_fetch (av, i + 1, 0); 297 SV *p_sv = *av_fetch (av, i + 1, 0);
234 cl_context_properties v = SvIV (p_sv); // code below can override 298 cl_context_properties v = SvIV (p_sv); // code below can override
235 299
236 switch (t) 300 switch (t)
237 { 301 {
302 case CL_CONTEXT_PLATFORM:
303 if (SvROK (p_sv))
304 v = (cl_context_properties)SvCLOBJ (func, svname, p_sv, "OpenCL::Platform");
305 break;
306
238 case CL_GLX_DISPLAY_KHR: 307 case CL_GLX_DISPLAY_KHR:
239 if (!SvOK (p_sv)) 308 if (!SvOK (p_sv))
240 { 309 {
241 void *func = glsym ("_glXGetCurrentDisplay"); 310 void *func = glsym ("_glXGetCurrentDisplay");
242 if (func) 311 if (func)
268 } 337 }
269 338
270 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 339 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
271} 340}
272 341
342static void * ecb_noinline
343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
344{
345 const char *funcname = cv_get_name (cv);
346
347 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
348 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
349 funcname, argname, or_undef ? "undef or " : "", klass);
350
351 AV *av = (AV *)SvRV (arg);
352 void **list = 0;
353 cl_uint count = av_len (av) + 1;
354
355 if (count)
356 {
357 list = tmpbuf (sizeof (*list) * count);
358 int i;
359 for (i = 0; i < count; ++i)
360 list [i] = SvCLOBJ (funcname, argname, *av_fetch (av, i, 1), klass);
361 }
362
363 if (!count && !or_undef)
364 croak ("%s: '%s' must contain at least one %s object",
365 funcname, argname, klass);
366
367 *rcount = count;
368 return (void *)list;
369}
370
273/*****************************************************************************/ 371/*****************************************************************************/
372/* callback stuff */
373
374/* default context callback, log to stderr */
375static void CL_CALLBACK
376context_default_notify (const char *msg, const void *info, size_t cb, void *data)
377{
378 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
379}
380
381typedef struct
382{
383 int free_cb;
384 void (*push)(void *data1, void *data2, void *data3);
385} eq_vtbl;
386
387typedef struct eq_item
388{
389 struct eq_item *next;
390 eq_vtbl *vtbl;
391 SV *cb;
392 void *data1, *data2, *data3;
393} eq_item;
394
395static void (*eq_signal_func)(void *signal_arg, int value);
396static void *eq_signal_arg;
397static xmutex_t eq_lock = X_MUTEX_INIT;
398static eq_item *eq_head, *eq_tail;
399
400static void ecb_noinline
401eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
402{
403 eq_item *item = malloc (sizeof (eq_item));
404
405 item->next = 0;
406 item->vtbl = vtbl;
407 item->cb = cb;
408 item->data1 = data1;
409 item->data2 = data2;
410 item->data3 = data3;
411
412 X_LOCK (eq_lock);
413
414 *(eq_head ? &eq_tail->next : &eq_head) = item;
415 eq_tail = item;
416
417 X_UNLOCK (eq_lock);
418
419 eq_signal_func (eq_signal_arg, 0);
420}
421
422static eq_item *
423eq_dec (void)
424{
425 eq_item *res;
426
427 X_LOCK (eq_lock);
428
429 res = eq_head;
430 if (res)
431 eq_head = res->next;
432
433 X_UNLOCK (eq_lock);
434
435 return res;
436}
437
438static void
439eq_poll (void)
440{
441 eq_item *item;
442
443 while ((item = eq_dec ()))
444 {
445 ENTER;
446 SAVETMPS;
447
448 dSP;
449 PUSHMARK (SP);
450 EXTEND (SP, 2);
451
452 if (item->vtbl->free_cb)
453 sv_2mortal (item->cb);
454
455 PUTBACK;
456 item->vtbl->push (item->data1, item->data2, item->data3);
457
458 SV *cb = item->cb;
459 free (item);
460
461 call_sv (cb, G_DISCARD | G_VOID);
462
463 FREETMPS;
464 LEAVE;
465 }
466}
467
468static void
469eq_poll_interrupt (pTHX_ void *c_arg, int value)
470{
471 eq_poll ();
472}
473
474/*****************************************************************************/
475/* context notify */
476
477static void ecb_noinline
478eq_context_push (void *data1, void *data2, void *data3)
479{
480 dSP;
481 PUSHs (sv_2mortal (newSVpv (data1, 0)));
482 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
483 PUTBACK;
484
485 free (data1);
486 free (data2);
487}
488
489static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
490
491static void CL_CALLBACK
492eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
493{
494 void *pvt_copy = malloc (cb);
495 memcpy (pvt_copy, pvt, cb);
496 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
497}
498
499#define CONTEXT_NOTIFY_CALLBACK \
500 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
501 void *user_data = 0; \
502 \
503 if (SvOK (notify)) \
504 { \
505 pfn_notify = eq_context_notify; \
506 user_data = s_get_cv (notify); \
507 }
508
509static SV * ecb_noinline
510new_clobj_context (cl_context ctx, void *user_data)
511{
512 SV *sv = new_clobj (stash_context, (IV)ctx);
513
514 if (user_data)
515 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
516
517 return sv;
518}
519
520#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
521
522/*****************************************************************************/
523/* build/compile/link notify */
524
525static void
526eq_program_push (void *data1, void *data2, void *data3)
527{
528 dSP;
529 PUSH_CLOBJ (stash_program, data1);
530 PUTBACK;
531}
532
533static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
534
535static void CL_CALLBACK
536eq_program_notify (cl_program program, void *user_data)
537{
538 clRetainProgram (program);
539
540 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
541}
542
543typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
544
545static program_callback ecb_noinline
546make_program_callback (SV *notify, void **ruser_data)
547{
548 if (SvOK (notify))
549 {
550 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
551 return eq_program_notify;
552 }
553 else
554 {
555 *ruser_data = 0;
556 return 0;
557 }
558}
559
560struct build_args
561{
562 cl_program program;
563 char *options;
564 void *user_data;
565 cl_uint num_devices;
566};
567
568X_THREAD_PROC (build_program_thread)
569{
570 struct build_args *arg = thr_arg;
571
572 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
573
574 if (arg->user_data)
575 eq_program_notify (arg->program, arg->user_data);
576 else
577 clReleaseProgram (arg->program);
578
579 free (arg->options);
580 free (arg);
581
582 return 0;
583}
584
585static void
586build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
587{
588 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
589
590 arg->program = program;
591 arg->options = strdup (options);
592 arg->user_data = user_data;
593 arg->num_devices = num_devices;
594 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
595
596 xthread_t id;
597 thread_create (&id, build_program_thread, arg);
598}
599
600/*****************************************************************************/
601/* event objects */
602
603static void
604eq_event_push (void *data1, void *data2, void *data3)
605{
606 dSP;
607 PUSH_CLOBJ (stash_event, data1);
608 PUSHs (sv_2mortal (newSViv ((IV)data2)));
609 PUTBACK;
610}
611
612static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
613
614static void CL_CALLBACK
615eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
616{
617 clRetainEvent (event);
618 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
619}
620
621/*****************************************************************************/
622/* utilities for XS code */
274 623
275static size_t 624static size_t
276img_row_pitch (cl_mem img) 625img_row_pitch (cl_mem img)
277{ 626{
278 size_t res; 627 size_t res;
279 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 628 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
280 return res; 629 return res;
281} 630}
282 631
283static cl_event * 632static cl_event * ecb_noinline
284event_list (SV **items, cl_uint *rcount) 633event_list (SV **items, cl_uint *rcount, cl_event extra)
285{ 634{
286 cl_uint count = *rcount; 635 cl_uint count = *rcount;
287 636
288 if (!count) 637 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
638 *rcount = count = 0;
639
640 if (!count && !extra)
289 return 0; 641 return 0;
290 642
291 cl_event *list = tmpbuf (sizeof (cl_event) * count); 643 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
292 int i = 0; 644 int i = 0;
293 645
294 do 646 while (count--)
295 {
296 --count;
297 if (SvOK (items [count])) 647 if (SvOK (items [count]))
298 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 648 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
299 } 649
300 while (count); 650 if (extra)
651 list [i++] = extra;
301 652
302 *rcount = i; 653 *rcount = i;
303 654
304 return i ? list : 0; 655 return i ? list : 0;
305} 656}
306 657
307#define EVENT_LIST(items,count) \ 658#define EVENT_LIST(skip) \
308 cl_uint event_list_count = (count); \ 659 cl_uint event_list_count = items - (skip); \
309 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 660 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
310 661
311#define INFO(class) \ 662#define INFO(class) \
312{ \ 663{ \
313 size_t size; \ 664 size_t size; \
314 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 665 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
318 SvCUR_set (sv, size); \ 669 SvCUR_set (sv, size); \
319 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 670 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
320 XPUSHs (sv); \ 671 XPUSHs (sv); \
321} 672}
322 673
674/*****************************************************************************/
675/* mapped_xxx */
676
677static OpenCL__Mapped
678SvMAPPED (SV *sv)
679{
680 // no typechecking atm., keep your fingers crossed
681 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
682}
683
684struct mapped
685{
686 cl_command_queue queue;
687 cl_mem memobj;
688 void *ptr;
689 size_t cb;
690 cl_event event;
691 size_t row_pitch;
692 size_t slice_pitch;
693};
694
695static SV *
696mapped_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)
697{
698 SV *data = newSV (0);
699 SvUPGRADE (data, SVt_PVMG);
700
701 OpenCL__Mapped mapped;
702 New (0, mapped, 1, struct mapped);
703
704 clRetainCommandQueue (queue);
705
706 mapped->queue = queue;
707 mapped->memobj = memobj;
708 mapped->ptr = ptr;
709 mapped->cb = cb;
710 mapped->event = ev;
711 mapped->row_pitch = row_pitch;
712 mapped->slice_pitch = slice_pitch;
713
714 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
715
716 if (SvLEN (data))
717 Safefree (data);
718
719 SvPVX (data) = (char *)ptr;
720 SvCUR_set (data, cb);
721 SvLEN_set (data, 0);
722 SvPOK_only (data);
723
724 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
725
726 if (!(flags & CL_MAP_WRITE))
727 SvREADONLY_on (data);
728
729 return obj;
730}
731
732static void
733mapped_detach (SV *sv, OpenCL__Mapped mapped)
734{
735 SV *data = SvRV (sv);
736
737 // the next check checks both before AND after detach, where SvPVX should STILL be 0
738 if (SvPVX (data) != (char *)mapped->ptr)
739 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
740 else
741 {
742 SvREADONLY_off (data);
743 SvCUR_set (data, 0);
744 SvPVX (data) = 0;
745 SvOK_off (data);
746 }
747
748 mapped->ptr = 0;
749}
750
751static void
752mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
753{
754 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
755 cl_event ev;
756
757 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
758
759 clReleaseEvent (mapped->event);
760 mapped->event = ev;
761
762 mapped_detach (self, mapped);
763}
764
765/*****************************************************************************/
766
323MODULE = OpenCL PACKAGE = OpenCL 767MODULE = OpenCL PACKAGE = OpenCL
324 768
325PROTOTYPES: ENABLE 769PROTOTYPES: ENABLE
770
771void
772poll ()
773 CODE:
774 eq_poll ();
775
776void
777_eq_initialise (IV func, IV arg)
778 CODE:
779 eq_signal_func = (void (*)(void *, int))func;
780 eq_signal_arg = (void*)arg;
326 781
327BOOT: 782BOOT:
328{ 783{
329 HV *stash = gv_stashpv ("OpenCL", 1); 784 HV *stash = gv_stashpv ("OpenCL", 1);
330 static const ivstr *civ, const_iv[] = { 785 static const ivstr *civ, const_iv[] = {
339 { sizeof (cl_half ), "SIZEOF_HALF" }, 794 { sizeof (cl_half ), "SIZEOF_HALF" },
340 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 795 { sizeof (cl_float ), "SIZEOF_FLOAT" },
341 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 796 { sizeof (cl_double), "SIZEOF_DOUBLE" },
342#include "constiv.h" 797#include "constiv.h"
343 }; 798 };
799
344 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 800 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
345 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 801 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
802
803 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
804 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
805 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
806 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
807 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
808 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
809 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
810 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
811 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
812 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
813 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
814 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
815 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
816 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
817 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
818 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
819 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
820 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
821 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
822 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
823 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
824 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
825 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
826
827 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
346} 828}
347 829
348cl_int 830cl_int
349errno () 831errno ()
350 CODE: 832 CODE:
351 RETVAL = res; 833 RETVAL = res;
352 OUTPUT: 834 OUTPUT:
353 RETVAL 835 RETVAL
354 836
355const char * 837const char *
356err2str (cl_int err) 838err2str (cl_int err = res)
357 839
358const char * 840const char *
359enum2str (cl_uint value) 841enum2str (cl_uint value)
360 842
361void 843void
369 list = tmpbuf (sizeof (*list) * count); 851 list = tmpbuf (sizeof (*list) * count);
370 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 852 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
371 853
372 EXTEND (SP, count); 854 EXTEND (SP, count);
373 for (i = 0; i < count; ++i) 855 for (i = 0; i < count; ++i)
374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 856 PUSH_CLOBJ (stash_platform, list [i]);
375 857
376void 858void
377context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 859context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
378 PPCODE: 860 PPCODE:
861 CONTEXT_NOTIFY_CALLBACK;
379 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 862 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
380 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 863 XPUSH_CLOBJ_CONTEXT;
381 864
382#if 0
383
384void 865void
385context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 866context (FUTURE properties, FUTURE devices, FUTURE notify)
386 PPCODE: 867 PPCODE:
387 /* der Gipfel der Kunst */ 868 /* der Gipfel der Kunst */
388 869
389#endif
390
391void 870void
392wait_for_events (...) 871wait_for_events (...)
393 CODE: 872 CODE:
394 EVENT_LIST (0, items); 873 EVENT_LIST (0);
395 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 874 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
396 875
397PROTOTYPES: DISABLE 876PROTOTYPES: DISABLE
398 877
399MODULE = OpenCL PACKAGE = OpenCL::Platform 878MODULE = OpenCL PACKAGE = OpenCL::Platform
442 list = tmpbuf (sizeof (*list) * count); 921 list = tmpbuf (sizeof (*list) * count);
443 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 922 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
444 923
445 EXTEND (SP, count); 924 EXTEND (SP, count);
446 for (i = 0; i < count; ++i) 925 for (i = 0; i < count; ++i)
447 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 926 PUSH_CLOBJ (stash_device, list [i]);
448 927
449void 928void
450context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 929context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
451 PPCODE: 930 PPCODE:
452 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 931 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
453 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 932 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
454 933
455 AV *av = (AV *)SvRV (devices); 934 cl_uint device_count = 0;
456 cl_uint num_devices = av_len (av) + 1; 935 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
457 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
458 int i;
459 936
460 for (i = num_devices; i--; ) 937 CONTEXT_NOTIFY_CALLBACK;
461 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
462
463 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 938 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
464 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 939 XPUSH_CLOBJ_CONTEXT;
465 940
466void 941void
467context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 942context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
468 PPCODE: 943 PPCODE:
469 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 944 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
470 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 945 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
946
947 CONTEXT_NOTIFY_CALLBACK;
471 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 948 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
472 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 949 XPUSH_CLOBJ_CONTEXT;
473 950
474MODULE = OpenCL PACKAGE = OpenCL::Device 951MODULE = OpenCL PACKAGE = OpenCL::Device
475 952
476void 953void
477info (OpenCL::Device self, cl_device_info name) 954info (OpenCL::Device self, cl_device_info name)
478 PPCODE: 955 PPCODE:
479 INFO (Device) 956 INFO (Device)
957
958#if CL_VERSION_1_2
959
960void
961sub_devices (OpenCL::Device self, SV *properties)
962 PPCODE:
963 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
964 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
965
966 properties = SvRV (properties);
967
968 cl_uint count = av_len ((AV *)properties) + 1;
969 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
970
971 int i;
972 for (i = 0; i < count; ++i)
973 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
974
975 props [count] = 0;
976
977 cl_uint num_devices;
978 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
979 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
980 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
981
982 EXTEND (SP, num_devices);
983 for (i = 0; i < count; ++i)
984 PUSH_CLOBJ (stash_subdevice, list [i]);
985
986#endif
480 987
481#BEGIN:device 988#BEGIN:device
482 989
483void 990void
484type (OpenCL::Device self) 991type (OpenCL::Device self)
485 PPCODE: 992 PPCODE:
486 cl_device_type value [1]; 993 cl_device_type value [1];
487 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 994 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
488 EXTEND (SP, 1); 995 EXTEND (SP, 1);
489 const int i = 0; 996 const int i = 0;
490 PUSHs (sv_2mortal (newSViv (value [i]))); 997 PUSHs (sv_2mortal (newSVuv (value [i])));
491 998
492void 999void
493vendor_id (OpenCL::Device self) 1000vendor_id (OpenCL::Device self)
494 ALIAS: 1001 ALIAS:
495 vendor_id = CL_DEVICE_VENDOR_ID 1002 vendor_id = CL_DEVICE_VENDOR_ID
640 PPCODE: 1147 PPCODE:
641 cl_command_queue_properties value [1]; 1148 cl_command_queue_properties value [1];
642 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1149 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
643 EXTEND (SP, 1); 1150 EXTEND (SP, 1);
644 const int i = 0; 1151 const int i = 0;
645 PUSHs (sv_2mortal (newSViv (value [i]))); 1152 PUSHs (sv_2mortal (newSVuv (value [i])));
646 1153
647void 1154void
648platform (OpenCL::Device self) 1155platform (OpenCL::Device self)
649 PPCODE: 1156 PPCODE:
650 cl_platform_id value [1]; 1157 cl_platform_id value [1];
651 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1158 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
652 EXTEND (SP, 1); 1159 EXTEND (SP, 1);
653 const int i = 0; 1160 const int i = 0;
654 { 1161 {
655 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1162 PUSH_CLOBJ (stash_platform, value [i]);
656 } 1163 }
657 1164
658void 1165void
659name (OpenCL::Device self) 1166name (OpenCL::Device self)
660 ALIAS: 1167 ALIAS:
679 cl_device_id value [1]; 1186 cl_device_id value [1];
680 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
681 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
682 const int i = 0; 1189 const int i = 0;
683 { 1190 {
684 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1191 PUSH_CLOBJ (stash_device, value [i]);
685 } 1192 }
686 1193
687void 1194void
688partition_types_ext (OpenCL::Device self) 1195partition_types_ext (OpenCL::Device self)
689 ALIAS: 1196 ALIAS:
700 for (i = 0; i < n; ++i) 1207 for (i = 0; i < n; ++i)
701 PUSHs (sv_2mortal (newSVuv (value [i]))); 1208 PUSHs (sv_2mortal (newSVuv (value [i])));
702 1209
703#END:device 1210#END:device
704 1211
1212MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1213
1214#if CL_VERSION_1_2
1215
1216void
1217DESTROY (OpenCL::SubDevice self)
1218 CODE:
1219 clReleaseDevice (self);
1220
1221#endif
1222
705MODULE = OpenCL PACKAGE = OpenCL::Context 1223MODULE = OpenCL PACKAGE = OpenCL::Context
706 1224
707void 1225void
708DESTROY (OpenCL::Context context) 1226DESTROY (OpenCL::Context self)
709 CODE: 1227 CODE:
710 clReleaseContext (context); 1228 clReleaseContext (self);
711 1229
712void 1230void
713info (OpenCL::Context self, cl_context_info name) 1231info (OpenCL::Context self, cl_context_info name)
714 PPCODE: 1232 PPCODE:
715 INFO (Context) 1233 INFO (Context)
716 1234
717void 1235void
718queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1236queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
719 PPCODE: 1237 PPCODE:
720 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1238 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
721 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1239 XPUSH_CLOBJ (stash_queue, queue);
722 1240
723void 1241void
724user_event (OpenCL::Context self) 1242user_event (OpenCL::Context self)
725 PPCODE: 1243 PPCODE:
726 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1244 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
727 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1245 XPUSH_CLOBJ (stash_userevent, ev);
728 1246
729void 1247void
730buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1248buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
731 PPCODE: 1249 PPCODE:
732 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1250 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
733 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1251 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
734 1252
735 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1253 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
736 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1254 XPUSH_CLOBJ (stash_bufferobj, mem);
737 1255
738void 1256void
739buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1257buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
740 PPCODE: 1258 PPCODE:
741 STRLEN len; 1259 STRLEN len;
742 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1260 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
743 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1261 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
744 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1262 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
745 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1263 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
746 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1264 XPUSH_CLOBJ (stash_bufferobj, mem);
747 1265
748#if CL_VERSION_1_2 1266#if CL_VERSION_1_2
749 1267
750void 1268void
751image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1269image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
752 PPCODE: 1270 PPCODE:
753 STRLEN len; 1271 STRLEN len;
754 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1272 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
755 const cl_image_format format = { channel_order, channel_type }; 1273 const cl_image_format format = { channel_order, channel_type };
756 const cl_image_desc desc = { 1274 const cl_image_desc desc = {
757 type, 1275 type,
758 width, height, depth, 1276 width, height, depth,
759 array_size, row_pitch, slice_pitch, 1277 array_size, row_pitch, slice_pitch,
760 num_mip_level, num_samples, 1278 num_mip_level, num_samples,
761 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvPTROBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1279 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
762 }; 1280 };
763 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
764 char *klass = "OpenCL::Image"; 1282 HV *stash = stash_image;
765 switch (type) 1283 switch (type)
766 { 1284 {
767 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1285 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
768 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1286 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
769 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1287 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
770 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1288 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
771 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1289 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
772 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1290 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
773 } 1291 }
774 XPUSH_NEW_OBJ (klass, mem); 1292 XPUSH_CLOBJ (stash, mem);
775 1293
776#endif 1294#endif
777 1295
778void 1296void
779image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1297image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
785 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1303 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
786#else 1304#else
787 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1305 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
788 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1306 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
789#endif 1307#endif
790 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1308 XPUSH_CLOBJ (stash_image2d, mem);
791 1309
792void 1310void
793image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1311image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
794 PPCODE: 1312 PPCODE:
795 STRLEN len; 1313 STRLEN len;
799 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1317 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
800#else 1318#else
801 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1319 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
802 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1320 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
803#endif 1321#endif
804 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1322 XPUSH_CLOBJ (stash_image3d, mem);
805 1323
806#if cl_apple_gl_sharing || cl_khr_gl_sharing 1324#if cl_apple_gl_sharing || cl_khr_gl_sharing
807 1325
808void 1326void
809gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1327gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
810 PPCODE: 1328 PPCODE:
811 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1329 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
812 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1330 XPUSH_CLOBJ (stash_bufferobj, mem);
813 1331
814void 1332void
815gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1333gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
816 PPCODE: 1334 PPCODE:
817 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
818 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1336 XPUSH_CLOBJ (stash_image2d, mem);
819 1337
820#if CL_VERSION_1_2 1338#if CL_VERSION_1_2
821 1339
822void 1340void
823gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1341gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
824 ALIAS: 1342 ALIAS:
825 PPCODE: 1343 PPCODE:
826 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1344 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
827 cl_gl_object_type type; 1345 cl_gl_object_type type;
828 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1346 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
829 char *klass = "OpenCL::Memory"; 1347 HV *stash = stash_memory;
830 switch (type) 1348 switch (type)
831 { 1349 {
832 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1350 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
833 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1351 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
834 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1352 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
835 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1353 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
836 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1354 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
837 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1355 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
838 } 1356 }
839 XPUSH_NEW_OBJ (klass, mem); 1357 XPUSH_CLOBJ (stash, mem);
840 1358
841#endif 1359#endif
842 1360
843void 1361void
844gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1362gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
846#if PREFER_1_1 1364#if PREFER_1_1
847 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1365 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
848#else 1366#else
849 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
850#endif 1368#endif
851 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1369 XPUSH_CLOBJ (stash_image2d, mem);
852 1370
853void 1371void
854gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1372gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
855 PPCODE: 1373 PPCODE:
856#if PREFER_1_1 1374#if PREFER_1_1
857 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1375 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
858#else 1376#else
859 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1377 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
860#endif 1378#endif
861 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1379 XPUSH_CLOBJ (stash_image3d, mem);
862 1380
863#endif 1381#endif
864 1382
865void 1383void
866supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1384supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
886 1404
887void 1405void
888sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1406sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
889 PPCODE: 1407 PPCODE:
890 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1408 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
891 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1409 XPUSH_CLOBJ (stash_sampler, sampler);
892 1410
893void 1411void
894program_with_source (OpenCL::Context self, SV *program) 1412program_with_source (OpenCL::Context self, SV *program)
895 PPCODE: 1413 PPCODE:
896 STRLEN len; 1414 STRLEN len;
897 size_t len2; 1415 size_t len2;
898 const char *ptr = SvPVbyte (program, len); 1416 const char *ptr = SvPVbyte (program, len);
899 1417
900 len2 = len; 1418 len2 = len;
901 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1419 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
902 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1420 XPUSH_CLOBJ (stash_program, prog);
1421
1422void
1423program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1424 PPCODE:
1425 cl_uint device_count;
1426 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1427
1428 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1429 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1430
1431 binaries = SvRV (binaries);
1432
1433 if (device_count != av_len ((AV *)binaries) + 1)
1434 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1435
1436 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1437 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1438 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1439
1440 int i;
1441 for (i = 0; i < device_count; ++i)
1442 {
1443 STRLEN len;
1444 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1445 length_list [i] = len;
1446 }
1447
1448 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1449 length_list, binary_list,
1450 GIMME_V == G_ARRAY ? status_list : 0, &res));
1451
1452 EXTEND (SP, 2);
1453 PUSH_CLOBJ (stash_program, prog);
1454
1455 if (GIMME_V == G_ARRAY)
1456 {
1457 AV *av = newAV ();
1458 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1459
1460 for (i = device_count; i--; )
1461 av_store (av, i, newSViv (status_list [i]));
1462 }
1463
1464#if CL_VERSION_1_2
1465
1466void
1467program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1468 PPCODE:
1469 cl_uint device_count = 0;
1470 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1471
1472 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1473
1474 XPUSH_CLOBJ (stash_program, prog);
1475
1476void
1477link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1478 CODE:
1479 cl_uint device_count = 0;
1480 cl_device_id *device_list = 0;
1481
1482 if (SvOK (devices))
1483 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1484
1485 cl_uint program_count;
1486 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1487
1488 void *user_data;
1489 program_callback pfn_notify = make_program_callback (notify, &user_data);
1490
1491 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1492 program_count, program_list, pfn_notify, user_data, &res));
1493
1494 XPUSH_CLOBJ (stash_program, prog);
1495
1496#endif
903 1497
904#BEGIN:context 1498#BEGIN:context
905 1499
906void 1500void
907reference_count (OpenCL::Context self) 1501reference_count (OpenCL::Context self)
924 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1518 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
925 int i, n = size / sizeof (*value); 1519 int i, n = size / sizeof (*value);
926 EXTEND (SP, n); 1520 EXTEND (SP, n);
927 for (i = 0; i < n; ++i) 1521 for (i = 0; i < n; ++i)
928 { 1522 {
929 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1523 PUSH_CLOBJ (stash_device, value [i]);
930 } 1524 }
931 1525
932void 1526void
933properties (OpenCL::Context self) 1527properties (OpenCL::Context self)
934 PPCODE: 1528 PPCODE:
949DESTROY (OpenCL::Queue self) 1543DESTROY (OpenCL::Queue self)
950 CODE: 1544 CODE:
951 clReleaseCommandQueue (self); 1545 clReleaseCommandQueue (self);
952 1546
953void 1547void
954enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1548read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1549 ALIAS:
1550 enqueue_read_buffer = 0
955 PPCODE: 1551 PPCODE:
956 cl_event ev = 0;
957 EVENT_LIST (6, items - 6); 1552 EVENT_LIST (6);
958 1553
959 SvUPGRADE (data, SVt_PV); 1554 SvUPGRADE (data, SVt_PV);
960 SvGROW (data, len); 1555 SvGROW (data, len);
961 SvPOK_only (data); 1556 SvPOK_only (data);
962 SvCUR_set (data, len); 1557 SvCUR_set (data, len);
1558
1559 cl_event ev = 0;
963 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1560 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
964 1561
965 if (ev) 1562 if (ev)
966 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1563 XPUSH_CLOBJ (stash_event, ev);
967 1564
968void 1565void
969enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1566write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1567 ALIAS:
1568 enqueue_write_buffer = 0
970 PPCODE: 1569 PPCODE:
971 cl_event ev = 0; 1570 EVENT_LIST (5);
1571
972 STRLEN len; 1572 STRLEN len;
973 char *ptr = SvPVbyte (data, len); 1573 char *ptr = SvPVbyte (data, len);
974 EVENT_LIST (5, items - 5);
975 1574
1575 cl_event ev = 0;
976 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
977 1577
978 if (ev) 1578 if (ev)
979 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
980 1580
1581#if CL_VERSION_1_2
1582
981void 1583void
982enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1584fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1585 ALIAS:
1586 enqueue_fill_buffer = 0
983 PPCODE: 1587 PPCODE:
1588 EVENT_LIST (5);
1589
1590 STRLEN len;
1591 char *ptr = SvPVbyte (data, len);
1592
984 cl_event ev = 0; 1593 cl_event ev = 0;
1594 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1595
1596 if (ev)
1597 XPUSH_CLOBJ (stash_event, ev);
1598
1599void
1600fill_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, ...)
1601 ALIAS:
1602 enqueue_fill_image = 0
1603 PPCODE:
1604 EVENT_LIST (12);
1605
1606 const size_t origin [3] = { x, y, z };
1607 const size_t region [3] = { width, height, depth };
1608
1609 const cl_float c_f [4] = { r, g, b, a };
1610 const cl_uint c_u [4] = { r, g, b, a };
1611 const cl_int c_s [4] = { r, g, b, a };
1612 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1613 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1614 cl_image_format format;
1615 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1616 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1617 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1618 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1619
1620 cl_event ev = 0;
1621 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1622 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1623
1624 if (ev)
1625 XPUSH_CLOBJ (stash_event, ev);
1626
1627#endif
1628
1629void
1630copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1631 ALIAS:
1632 enqueue_copy_buffer = 0
1633 PPCODE:
985 EVENT_LIST (6, items - 6); 1634 EVENT_LIST (6);
986 1635
1636 cl_event ev = 0;
987 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1637 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
988 1638
989 if (ev) 1639 if (ev)
990 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1640 XPUSH_CLOBJ (stash_event, ev);
991 1641
992void 1642void
993enqueue_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, ...) 1643read_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, ...)
1644 ALIAS:
1645 enqueue_read_buffer_rect = 0
994 PPCODE: 1646 PPCODE:
995 cl_event ev = 0; 1647 EVENT_LIST (17);
1648
996 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1649 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
997 const size_t host_origin[3] = { host_x, host_y, host_z }; 1650 const size_t host_origin[3] = { host_x, host_y, host_z };
998 const size_t region[3] = { width, height, depth }; 1651 const size_t region[3] = { width, height, depth };
999 EVENT_LIST (17, items - 17);
1000 1652
1001 if (!buf_row_pitch) 1653 if (!buf_row_pitch)
1002 buf_row_pitch = region [0]; 1654 buf_row_pitch = region [0];
1003 1655
1004 if (!buf_slice_pitch) 1656 if (!buf_slice_pitch)
1014 1666
1015 SvUPGRADE (data, SVt_PV); 1667 SvUPGRADE (data, SVt_PV);
1016 SvGROW (data, len); 1668 SvGROW (data, len);
1017 SvPOK_only (data); 1669 SvPOK_only (data);
1018 SvCUR_set (data, len); 1670 SvCUR_set (data, len);
1671
1672 cl_event ev = 0;
1019 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)); 1673 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));
1020 1674
1021 if (ev) 1675 if (ev)
1022 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1676 XPUSH_CLOBJ (stash_event, ev);
1023 1677
1024void 1678void
1025enqueue_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, ...) 1679write_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, ...)
1680 ALIAS:
1681 enqueue_write_buffer_rect = 0
1026 PPCODE: 1682 PPCODE:
1027 cl_event ev = 0; 1683 EVENT_LIST (17);
1684
1028 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1685 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1029 const size_t host_origin[3] = { host_x, host_y, host_z }; 1686 const size_t host_origin[3] = { host_x, host_y, host_z };
1030 const size_t region[3] = { width, height, depth }; 1687 const size_t region[3] = { width, height, depth };
1031 STRLEN len; 1688 STRLEN len;
1032 char *ptr = SvPVbyte (data, len); 1689 char *ptr = SvPVbyte (data, len);
1033 EVENT_LIST (17, items - 17);
1034 1690
1035 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
1036 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
1037 1693
1038 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
1047 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1703 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1048 1704
1049 if (len < min_len) 1705 if (len < min_len)
1050 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1706 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1051 1707
1708 cl_event ev = 0;
1052 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)); 1709 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));
1053 1710
1054 if (ev) 1711 if (ev)
1055 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1712 XPUSH_CLOBJ (stash_event, ev);
1056 1713
1057void 1714void
1058enqueue_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, ...) 1715copy_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, ...)
1716 ALIAS:
1717 enqueue_copy_buffer_rect = 0
1059 PPCODE: 1718 PPCODE:
1060 cl_event ev = 0; 1719 EVENT_LIST (16);
1720
1061 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 };
1062 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 };
1063 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
1064 EVENT_LIST (16, items - 16);
1065 1724
1725 cl_event ev = 0;
1066 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)); 1726 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));
1067 1727
1068 if (ev) 1728 if (ev)
1069 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
1070 1730
1071void 1731void
1072enqueue_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, ...) 1732read_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, ...)
1733 ALIAS:
1734 enqueue_read_image = 0
1073 PPCODE: 1735 PPCODE:
1074 cl_event ev = 0; 1736 EVENT_LIST (12);
1737
1075 const size_t src_origin[3] = { src_x, src_y, src_z }; 1738 const size_t src_origin[3] = { src_x, src_y, src_z };
1076 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
1077 EVENT_LIST (12, items - 12);
1078 1740
1079 if (!row_pitch) 1741 if (!row_pitch)
1080 row_pitch = img_row_pitch (src); 1742 row_pitch = img_row_pitch (src);
1081 1743
1082 if (depth > 1 && !slice_pitch) 1744 if (depth > 1 && !slice_pitch)
1086 1748
1087 SvUPGRADE (data, SVt_PV); 1749 SvUPGRADE (data, SVt_PV);
1088 SvGROW (data, len); 1750 SvGROW (data, len);
1089 SvPOK_only (data); 1751 SvPOK_only (data);
1090 SvCUR_set (data, len); 1752 SvCUR_set (data, len);
1753
1754 cl_event ev = 0;
1091 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)); 1755 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));
1092 1756
1093 if (ev) 1757 if (ev)
1094 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1758 XPUSH_CLOBJ (stash_event, ev);
1095 1759
1096void 1760void
1097enqueue_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, ...) 1761write_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, ...)
1762 ALIAS:
1763 enqueue_write_image = 0
1098 PPCODE: 1764 PPCODE:
1099 cl_event ev = 0; 1765 EVENT_LIST (12);
1766
1100 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1767 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1101 const size_t region[3] = { width, height, depth }; 1768 const size_t region[3] = { width, height, depth };
1102 STRLEN len; 1769 STRLEN len;
1103 char *ptr = SvPVbyte (data, len); 1770 char *ptr = SvPVbyte (data, len);
1104 EVENT_LIST (12, items - 12);
1105 1771
1106 if (!row_pitch) 1772 if (!row_pitch)
1107 row_pitch = img_row_pitch (dst); 1773 row_pitch = img_row_pitch (dst);
1108 1774
1109 if (depth > 1 && !slice_pitch) 1775 if (depth > 1 && !slice_pitch)
1112 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1778 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1113 1779
1114 if (len < min_len) 1780 if (len < min_len)
1115 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1781 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1116 1782
1783 cl_event ev = 0;
1117 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)); 1784 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));
1118 1785
1119 if (ev) 1786 if (ev)
1120 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1787 XPUSH_CLOBJ (stash_event, ev);
1121 1788
1122void 1789void
1123enqueue_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, ...) 1790copy_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, ...)
1791 ALIAS:
1792 enqueue_copy_image = 0
1124 PPCODE: 1793 PPCODE:
1125 cl_event ev = 0; 1794 EVENT_LIST (12);
1795
1126 const size_t src_origin[3] = { src_x, src_y, src_z }; 1796 const size_t src_origin[3] = { src_x, src_y, src_z };
1127 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1797 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1128 const size_t region[3] = { width, height, depth }; 1798 const size_t region[3] = { width, height, depth };
1129 EVENT_LIST (12, items - 12);
1130 1799
1800 cl_event ev = 0;
1131 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1801 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1132 1802
1133 if (ev) 1803 if (ev)
1134 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1804 XPUSH_CLOBJ (stash_event, ev);
1135 1805
1136void 1806void
1137enqueue_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, ...) 1807copy_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, ...)
1808 ALIAS:
1809 enqueue_copy_image_to_buffer = 0
1138 PPCODE: 1810 PPCODE:
1811 EVENT_LIST (10);
1812
1813 const size_t src_origin[3] = { src_x, src_y, src_z };
1814 const size_t region [3] = { width, height, depth };
1815
1139 cl_event ev = 0; 1816 cl_event ev = 0;
1817 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1818
1819 if (ev)
1820 XPUSH_CLOBJ (stash_event, ev);
1821
1822void
1823copy_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, ...)
1824 ALIAS:
1825 enqueue_copy_buffer_to_image = 0
1826 PPCODE:
1827 EVENT_LIST (10);
1828
1140 const size_t src_origin[3] = { src_x, src_y, src_z }; 1829 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1830 const size_t region [3] = { width, height, depth };
1831
1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839map_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, ...)
1840 ALIAS:
1841 enqueue_map_buffer = 0
1842 PPCODE:
1843 EVENT_LIST (6);
1844
1845 size_t cb = SvIV (cb_);
1846
1847 if (!SvOK (cb_))
1848 {
1849 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1850 cb -= offset;
1851 }
1852
1853 cl_event ev;
1854 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1855 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1856
1857void
1858map_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, ...)
1859 ALIAS:
1860 enqueue_map_image = 0
1861 PPCODE:
1862 size_t width = SvIV (width_);
1863 if (!SvOK (width_))
1864 {
1865 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1866 width -= x;
1867 }
1868
1869 size_t height = SvIV (width_);
1870 if (!SvOK (height_))
1871 {
1872 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1873 height -= y;
1874 }
1875
1876 size_t depth = SvIV (width_);
1877 if (!SvOK (depth_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1880 depth -= z;
1881
1882 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1883 if (!depth)
1884 depth = 1;
1885 }
1886
1887 const size_t origin[3] = { x, y, z };
1141 const size_t region[3] = { width, height, depth }; 1888 const size_t region[3] = { width, height, depth };
1889 size_t row_pitch, slice_pitch;
1142 EVENT_LIST (10, items - 10); 1890 EVENT_LIST (10);
1143 1891
1892 cl_event ev;
1893 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));
1894
1895 size_t cb = slice_pitch ? slice_pitch * region [2]
1896 : row_pitch ? row_pitch * region [1]
1897 : region [0];
1898
1899 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1900
1901void
1902unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1903 PPCODE:
1904 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1905 if (GIMME_V != G_VOID)
1906 {
1907 clRetainEvent (mapped->event);
1908 XPUSH_CLOBJ (stash_event, mapped->event);
1909 }
1910
1911void
1912task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1913 ALIAS:
1914 enqueue_task = 0
1915 PPCODE:
1916 EVENT_LIST (2);
1917
1918 cl_event ev = 0;
1144 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1919 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1145 1920
1146 if (ev) 1921 if (ev)
1147 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1922 XPUSH_CLOBJ (stash_event, ev);
1148 1923
1149void 1924void
1150enqueue_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, ...)
1151 PPCODE:
1152 cl_event ev = 0;
1153 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1154 const size_t region[3] = { width, height, depth };
1155 EVENT_LIST (10, items - 10);
1156
1157 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1158
1159 if (ev)
1160 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1161
1162void
1163enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1164 PPCODE:
1165 cl_event ev = 0;
1166 EVENT_LIST (2, items - 2);
1167
1168 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1169
1170 if (ev)
1171 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1172
1173void
1174enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1925nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1926 ALIAS:
1927 enqueue_nd_range_kernel = 0
1175 PPCODE: 1928 PPCODE:
1176 cl_event ev = 0; 1929 EVENT_LIST (5);
1930
1177 size_t *gwo = 0, *gws, *lws = 0; 1931 size_t *gwo = 0, *gws, *lws = 0;
1178 int gws_len; 1932 int gws_len;
1179 size_t *lists; 1933 size_t *lists;
1180 int i; 1934 int i;
1181 EVENT_LIST (5, items - 5);
1182 1935
1183 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1936 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1184 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1937 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1185 1938
1186 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1939 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1187 1940
1188 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1941 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1189 1942
1190 gws = lists + gws_len * 0; 1943 gws = lists + gws_len * 0;
1191 for (i = 0; i < gws_len; ++i) 1944 for (i = 0; i < gws_len; ++i)
1945 {
1192 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1946 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1947 // at least nvidia crashes for 0-sized work group sizes, work around
1948 if (!gws [i])
1949 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1950 }
1193 1951
1194 if (SvOK (global_work_offset)) 1952 if (SvOK (global_work_offset))
1195 { 1953 {
1196 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1954 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1197 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1205 } 1963 }
1206 1964
1207 if (SvOK (local_work_size)) 1965 if (SvOK (local_work_size))
1208 { 1966 {
1209 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1967 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1210 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1968 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1211 1969
1212 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1970 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1213 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1971 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1214 1972
1215 lws = lists + gws_len * 2; 1973 lws = lists + gws_len * 2;
1216 for (i = 0; i < gws_len; ++i) 1974 for (i = 0; i < gws_len; ++i)
1975 {
1217 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1976 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1977 // at least nvidia crashes for 0-sized work group sizes, work around
1978 if (!lws [i])
1979 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1980 }
1218 } 1981 }
1219 1982
1983 cl_event ev = 0;
1220 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1984 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1221 1985
1222 if (ev) 1986 if (ev)
1223 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1987 XPUSH_CLOBJ (stash_event, ev);
1988
1989#if CL_VERSION_1_2
1990
1991void
1992migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1993 ALIAS:
1994 enqueue_migrate_mem_objects = 0
1995 PPCODE:
1996 EVENT_LIST (3);
1997
1998 cl_uint obj_count;
1999 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2000
2001 cl_event ev = 0;
2002 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2003
2004 if (ev)
2005 XPUSH_CLOBJ (stash_event, ev);
2006
2007#endif
1224 2008
1225#if cl_apple_gl_sharing || cl_khr_gl_sharing 2009#if cl_apple_gl_sharing || cl_khr_gl_sharing
1226 2010
1227void 2011void
1228enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2012acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1229 ALIAS: 2013 ALIAS:
2014 release_gl_objects = 1
2015 enqueue_acquire_gl_objects = 0
1230 enqueue_release_gl_objects = 1 2016 enqueue_release_gl_objects = 1
1231 PPCODE: 2017 PPCODE:
1232 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2018 EVENT_LIST (2);
1233 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2019
2020 cl_uint obj_count;
2021 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1234 2022
1235 cl_event ev = 0; 2023 cl_event ev = 0;
1236 EVENT_LIST (2, items - 2);
1237 AV *av = (AV *)SvRV (objects);
1238 cl_uint num_objects = av_len (av) + 1;
1239 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1240 int i;
1241
1242 for (i = num_objects; i--; )
1243 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1244 2024
1245 if (ix) 2025 if (ix)
1246 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2026 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1247 else 2027 else
1248 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2028 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1249 2029
1250 if (ev) 2030 if (ev)
1251 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2031 XPUSH_CLOBJ (stash_event, ev);
1252 2032
1253#endif 2033#endif
1254 2034
1255void 2035void
1256enqueue_wait_for_events (OpenCL::Queue self, ...) 2036wait_for_events (OpenCL::Queue self, ...)
2037 ALIAS:
2038 enqueue_wait_for_events = 0
1257 CODE: 2039 CODE:
1258 EVENT_LIST (1, items - 1); 2040 EVENT_LIST (1);
1259#if PREFER_1_1 2041#if PREFER_1_1
1260 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2042 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1261#else 2043#else
1262 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2044 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1263#endif 2045#endif
1264 2046
1265void 2047void
1266enqueue_marker (OpenCL::Queue self, ...) 2048marker (OpenCL::Queue self, ...)
2049 ALIAS:
2050 enqueue_marker = 0
1267 PPCODE: 2051 PPCODE:
2052 EVENT_LIST (1);
1268 cl_event ev = 0; 2053 cl_event ev = 0;
1269 EVENT_LIST (1, items - 1);
1270#if PREFER_1_1 2054#if PREFER_1_1
1271 if (!event_list_count) 2055 if (!event_list_count)
1272 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2056 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1273 else 2057 else
1274#if CL_VERSION_1_2 2058#if CL_VERSION_1_2
1275 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2059 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1276#else 2060#else
2061 {
1277 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); // also a barrier 2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2063 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2064 }
1278#endif 2065#endif
1279#else 2066#else
1280 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2067 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281#endif 2068#endif
1282 if (ev) 2069 if (ev)
1283 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2070 XPUSH_CLOBJ (stash_event, ev);
1284 2071
1285void 2072void
1286enqueue_barrier (OpenCL::Queue self, ...) 2073barrier (OpenCL::Queue self, ...)
2074 ALIAS:
2075 enqueue_barrier = 0
1287 PPCODE: 2076 PPCODE:
2077 EVENT_LIST (1);
1288 cl_event ev = 0; 2078 cl_event ev = 0;
1289 EVENT_LIST (1, items - 1);
1290#if PREFER_1_1 2079#if PREFER_1_1
1291 if (!event_list_count && GIMME_V == G_VOID) 2080 if (!event_list_count && GIMME_V == G_VOID)
1292 NEED_SUCCESS (EnqueueBarrier, (self)); 2081 NEED_SUCCESS (EnqueueBarrier, (self));
1293 else 2082 else
1294#if CL_VERSION_1_2 2083#if CL_VERSION_1_2
1295 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2084 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1296#else 2085#else
1297 { 2086 {
1298 if (event_list_count) 2087 if (event_list_count)
1299 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2088 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1300 2089
1301 if (GIMME_V != G_VOID) 2090 if (GIMME_V != G_VOID)
1302 NEED_SUCCESS (EnqueueMarker, (self, &ev)); 2091 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1303 } 2092 }
1304#endif 2093#endif
1305#else 2094#else
1306 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2095 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1307#endif 2096#endif
1308 if (ev) 2097 if (ev)
1309 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2098 XPUSH_CLOBJ (stash_event, ev);
1310 2099
1311void 2100void
1312flush (OpenCL::Queue self) 2101flush (OpenCL::Queue self)
1313 CODE: 2102 CODE:
1314 NEED_SUCCESS (Flush, (self)); 2103 NEED_SUCCESS (Flush, (self));
1332 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1333 EXTEND (SP, 1); 2122 EXTEND (SP, 1);
1334 const int i = 0; 2123 const int i = 0;
1335 { 2124 {
1336 NEED_SUCCESS (RetainContext, (value [i])); 2125 NEED_SUCCESS (RetainContext, (value [i]));
1337 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2126 PUSH_CLOBJ (stash_context, value [i]);
1338 } 2127 }
1339 2128
1340void 2129void
1341device (OpenCL::Queue self) 2130device (OpenCL::Queue self)
1342 PPCODE: 2131 PPCODE:
1343 cl_device_id value [1]; 2132 cl_device_id value [1];
1344 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2133 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1345 EXTEND (SP, 1); 2134 EXTEND (SP, 1);
1346 const int i = 0; 2135 const int i = 0;
1347 { 2136 {
1348 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2137 PUSH_CLOBJ (stash_device, value [i]);
1349 } 2138 }
1350 2139
1351void 2140void
1352reference_count (OpenCL::Queue self) 2141reference_count (OpenCL::Queue self)
1353 PPCODE: 2142 PPCODE:
1362 PPCODE: 2151 PPCODE:
1363 cl_command_queue_properties value [1]; 2152 cl_command_queue_properties value [1];
1364 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2153 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1365 EXTEND (SP, 1); 2154 EXTEND (SP, 1);
1366 const int i = 0; 2155 const int i = 0;
1367 PUSHs (sv_2mortal (newSViv (value [i]))); 2156 PUSHs (sv_2mortal (newSVuv (value [i])));
1368 2157
1369#END:command_queue 2158#END:command_queue
1370 2159
1371MODULE = OpenCL PACKAGE = OpenCL::Memory 2160MODULE = OpenCL PACKAGE = OpenCL::Memory
1372 2161
1387 PPCODE: 2176 PPCODE:
1388 cl_mem_object_type value [1]; 2177 cl_mem_object_type value [1];
1389 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2178 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1390 EXTEND (SP, 1); 2179 EXTEND (SP, 1);
1391 const int i = 0; 2180 const int i = 0;
1392 PUSHs (sv_2mortal (newSViv (value [i]))); 2181 PUSHs (sv_2mortal (newSVuv (value [i])));
1393 2182
1394void 2183void
1395flags (OpenCL::Memory self) 2184flags (OpenCL::Memory self)
1396 PPCODE: 2185 PPCODE:
1397 cl_mem_flags value [1]; 2186 cl_mem_flags value [1];
1398 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2187 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1399 EXTEND (SP, 1); 2188 EXTEND (SP, 1);
1400 const int i = 0; 2189 const int i = 0;
1401 PUSHs (sv_2mortal (newSViv (value [i]))); 2190 PUSHs (sv_2mortal (newSVuv (value [i])));
1402 2191
1403void 2192void
1404size (OpenCL::Memory self) 2193size (OpenCL::Memory self)
1405 ALIAS: 2194 ALIAS:
1406 size = CL_MEM_SIZE 2195 size = CL_MEM_SIZE
1440 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1441 EXTEND (SP, 1); 2230 EXTEND (SP, 1);
1442 const int i = 0; 2231 const int i = 0;
1443 { 2232 {
1444 NEED_SUCCESS (RetainContext, (value [i])); 2233 NEED_SUCCESS (RetainContext, (value [i]));
1445 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2234 PUSH_CLOBJ (stash_context, value [i]);
1446 } 2235 }
1447 2236
1448void 2237void
1449associated_memobject (OpenCL::Memory self) 2238associated_memobject (OpenCL::Memory self)
1450 PPCODE: 2239 PPCODE:
1452 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2241 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1453 EXTEND (SP, 1); 2242 EXTEND (SP, 1);
1454 const int i = 0; 2243 const int i = 0;
1455 { 2244 {
1456 NEED_SUCCESS (RetainMemObject, (value [i])); 2245 NEED_SUCCESS (RetainMemObject, (value [i]));
1457 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2246 PUSH_CLOBJ (stash_memory, value [i]);
1458 } 2247 }
1459 2248
1460#END:mem 2249#END:mem
1461 2250
1462#if cl_apple_gl_sharing || cl_khr_gl_sharing 2251#if cl_apple_gl_sharing || cl_khr_gl_sharing
1482 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2271 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1483 2272
1484 cl_buffer_region crdata = { origin, size }; 2273 cl_buffer_region crdata = { origin, size };
1485 2274
1486 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2275 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1487 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2276 XPUSH_CLOBJ (stash_buffer, mem);
1488 2277
1489MODULE = OpenCL PACKAGE = OpenCL::Image 2278MODULE = OpenCL PACKAGE = OpenCL::Image
1490 2279
1491void 2280void
1492image_info (OpenCL::Image self, cl_image_info name) 2281image_info (OpenCL::Image self, cl_image_info name)
1493 PPCODE: 2282 PPCODE:
1494 INFO (Image) 2283 INFO (Image)
2284
2285void
2286format (OpenCL::Image self)
2287 PPCODE:
2288 cl_image_format format;
2289 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2290 EXTEND (SP, 2);
2291 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2292 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1495 2293
1496#BEGIN:image 2294#BEGIN:image
1497 2295
1498void 2296void
1499element_size (OpenCL::Image self) 2297element_size (OpenCL::Image self)
1569 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2367 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1570 EXTEND (SP, 1); 2368 EXTEND (SP, 1);
1571 const int i = 0; 2369 const int i = 0;
1572 { 2370 {
1573 NEED_SUCCESS (RetainContext, (value [i])); 2371 NEED_SUCCESS (RetainContext, (value [i]));
1574 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2372 PUSH_CLOBJ (stash_context, value [i]);
1575 } 2373 }
1576 2374
1577void 2375void
1578normalized_coords (OpenCL::Sampler self) 2376normalized_coords (OpenCL::Sampler self)
1579 PPCODE: 2377 PPCODE:
1580 cl_addressing_mode value [1]; 2378 cl_addressing_mode value [1];
1581 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2379 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1582 EXTEND (SP, 1); 2380 EXTEND (SP, 1);
1583 const int i = 0; 2381 const int i = 0;
1584 PUSHs (sv_2mortal (newSViv (value [i]))); 2382 PUSHs (sv_2mortal (newSVuv (value [i])));
1585 2383
1586void 2384void
1587addressing_mode (OpenCL::Sampler self) 2385addressing_mode (OpenCL::Sampler self)
1588 PPCODE: 2386 PPCODE:
1589 cl_filter_mode value [1]; 2387 cl_filter_mode value [1];
1590 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1591 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1592 const int i = 0; 2390 const int i = 0;
1593 PUSHs (sv_2mortal (newSViv (value [i]))); 2391 PUSHs (sv_2mortal (newSVuv (value [i])));
1594 2392
1595void 2393void
1596filter_mode (OpenCL::Sampler self) 2394filter_mode (OpenCL::Sampler self)
1597 PPCODE: 2395 PPCODE:
1598 cl_bool value [1]; 2396 cl_bool value [1];
1609DESTROY (OpenCL::Program self) 2407DESTROY (OpenCL::Program self)
1610 CODE: 2408 CODE:
1611 clReleaseProgram (self); 2409 clReleaseProgram (self);
1612 2410
1613void 2411void
1614build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2412build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2413 ALIAS:
2414 build_async = 1
1615 CODE: 2415 CODE:
2416 cl_uint device_count = 0;
2417 cl_device_id *device_list = 0;
2418
2419 if (SvOK (devices))
2420 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2421
2422 void *user_data;
2423 program_callback pfn_notify = make_program_callback (notify, &user_data);
2424
2425 if (ix)
2426 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2427 else
2428 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2429
2430#if CL_VERSION_1_2
2431
2432void
2433compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2434 CODE:
2435 cl_uint device_count = 0;
2436 cl_device_id *device_list = 0;
2437
2438 if (SvOK (devices))
2439 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2440
2441 cl_uint header_count = 0;
2442 cl_program *header_list = 0;
2443 const char **header_name = 0;
2444
2445 if (SvOK (headers))
2446 {
2447 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2448 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2449
2450 HV *hv = (HV *)SvRV (devices);
2451
2452 header_count = hv_iterinit (hv);
2453 header_list = tmpbuf (sizeof (*header_list) * header_count);
2454 header_name = tmpbuf (sizeof (*header_name) * header_count);
2455
2456 HE *he;
2457 int i = 0;
2458 while (he = hv_iternext (hv))
2459 {
2460 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2461 header_list [i] = SvCLOBJ ("clCompileProgram", "headers", HeVAL (he), "OpenCL::Program");
2462 ++i;
2463 }
2464 }
2465
2466 void *user_data;
2467 program_callback pfn_notify = make_program_callback (notify, &user_data);
2468
1616 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2469 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2470 header_count, header_list, header_name, pfn_notify, user_data));
2471
2472#endif
1617 2473
1618void 2474void
1619build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2475build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1620 PPCODE: 2476 PPCODE:
1621 size_t size; 2477 size_t size;
1656 2512
1657void 2513void
1658kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
1659 PPCODE: 2515 PPCODE:
1660 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1661 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
1662 2518
1663void 2519void
1664kernels_in_program (OpenCL::Program program) 2520kernels_in_program (OpenCL::Program program)
1665 PPCODE: 2521 PPCODE:
1666 cl_uint num_kernels; 2522 cl_uint num_kernels;
1669 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
1670 2526
1671 int i; 2527 int i;
1672 EXTEND (SP, num_kernels); 2528 EXTEND (SP, num_kernels);
1673 for (i = 0; i < num_kernels; ++i) 2529 for (i = 0; i < num_kernels; ++i)
1674 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Kernel", kernels [i])); 2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
1675 2531
1676void 2532void
1677info (OpenCL::Program self, cl_program_info name) 2533info (OpenCL::Program self, cl_program_info name)
1678 PPCODE: 2534 PPCODE:
1679 INFO (Program) 2535 INFO (Program)
1727 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1728 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
1729 const int i = 0; 2585 const int i = 0;
1730 { 2586 {
1731 NEED_SUCCESS (RetainContext, (value [i])); 2587 NEED_SUCCESS (RetainContext, (value [i]));
1732 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2588 PUSH_CLOBJ (stash_context, value [i]);
1733 } 2589 }
1734 2590
1735void 2591void
1736devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
1737 PPCODE: 2593 PPCODE:
1741 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1742 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
1743 EXTEND (SP, n); 2599 EXTEND (SP, n);
1744 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
1745 { 2601 {
1746 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2602 PUSH_CLOBJ (stash_device, value [i]);
1747 } 2603 }
1748 2604
1749void 2605void
1750source (OpenCL::Program self) 2606source (OpenCL::Program self)
1751 PPCODE: 2607 PPCODE:
1777DESTROY (OpenCL::Kernel self) 2633DESTROY (OpenCL::Kernel self)
1778 CODE: 2634 CODE:
1779 clReleaseKernel (self); 2635 clReleaseKernel (self);
1780 2636
1781void 2637void
2638setf (OpenCL::Kernel self, const char *format, ...)
2639 CODE:
2640 int i;
2641 for (i = 2; ; ++i)
2642 {
2643 while (*format == ' ')
2644 ++format;
2645
2646 char type = *format++;
2647
2648 if (!type)
2649 break;
2650
2651 if (i >= items)
2652 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2653
2654 SV *sv = ST (i);
2655
2656 union
2657 {
2658 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2659 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2660 cl_half ch; cl_float cf; cl_double cd;
2661 cl_mem cm;
2662 cl_sampler ca;
2663 size_t cz;
2664 cl_event ce;
2665 } arg;
2666 size_t size;
2667 int nullarg = 0;
2668
2669 switch (type)
2670 {
2671 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2672 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2673 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2674 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2675 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2676 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2677 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2678 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2679
2680 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2681 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2682 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2683
2684 case 'z': nullarg = 1; size = SvIV (sv); break;
2685
2686 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2687 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2688 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2689
2690 default:
2691 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2692 }
2693
2694 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2695 if (res)
2696 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2697 }
2698
2699 if (i != items)
2700 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2701
2702void
1782set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2703set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1783 CODE: 2704 CODE:
1784 clSetKernelArg (self, idx, sizeof (value), &value); 2705 clSetKernelArg (self, idx, sizeof (value), &value);
1785 2706
1786void 2707void
1834 clSetKernelArg (self, idx, sizeof (value), &value); 2755 clSetKernelArg (self, idx, sizeof (value), &value);
1835 2756
1836void 2757void
1837set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2758set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1838 CODE: 2759 CODE:
1839 clSetKernelArg (self, idx, sizeof (value), &value); 2760 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1840 2761
1841void 2762void
1842set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2763set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1843 CODE: 2764 CODE:
1844 clSetKernelArg (self, idx, sizeof (value), &value); 2765 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1845 2766
1846void 2767void
1847set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2768set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2769 ALIAS:
2770 set_image2d = 0
2771 set_image3d = 0
1848 CODE: 2772 CODE:
1849 clSetKernelArg (self, idx, sizeof (value), &value); 2773 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1850
1851void
1852set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1853 CODE:
1854 clSetKernelArg (self, idx, sizeof (value), &value);
1855 2774
1856void 2775void
1857set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2776set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1858 CODE: 2777 CODE:
1859 clSetKernelArg (self, idx, sizeof (value), &value); 2778 clSetKernelArg (self, idx, sizeof (value), &value);
1905 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2824 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1906 EXTEND (SP, 1); 2825 EXTEND (SP, 1);
1907 const int i = 0; 2826 const int i = 0;
1908 { 2827 {
1909 NEED_SUCCESS (RetainContext, (value [i])); 2828 NEED_SUCCESS (RetainContext, (value [i]));
1910 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2829 PUSH_CLOBJ (stash_context, value [i]);
1911 } 2830 }
1912 2831
1913void 2832void
1914program (OpenCL::Kernel self) 2833program (OpenCL::Kernel self)
1915 PPCODE: 2834 PPCODE:
1917 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2836 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1918 EXTEND (SP, 1); 2837 EXTEND (SP, 1);
1919 const int i = 0; 2838 const int i = 0;
1920 { 2839 {
1921 NEED_SUCCESS (RetainProgram, (value [i])); 2840 NEED_SUCCESS (RetainProgram, (value [i]));
1922 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2841 PUSH_CLOBJ (stash_program, value [i]);
1923 } 2842 }
1924 2843
1925#END:kernel 2844#END:kernel
1926 2845
1927void 2846void
1974 const int i = 0; 2893 const int i = 0;
1975 PUSHs (sv_2mortal (newSVuv (value [i]))); 2894 PUSHs (sv_2mortal (newSVuv (value [i])));
1976 2895
1977#END:kernel_work_group 2896#END:kernel_work_group
1978 2897
2898#if CL_VERSION_1_2
2899
2900void
2901arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2902 PPCODE:
2903 size_t size;
2904 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2905 SV *sv = sv_2mortal (newSV (size));
2906 SvUPGRADE (sv, SVt_PV);
2907 SvPOK_only (sv);
2908 SvCUR_set (sv, size);
2909 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2910 XPUSHs (sv);
2911
2912#BEGIN:kernel_arg
2913
2914void
2915arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2916 PPCODE:
2917 cl_kernel_arg_address_qualifier value [1];
2918 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2919 EXTEND (SP, 1);
2920 const int i = 0;
2921 PUSHs (sv_2mortal (newSVuv (value [i])));
2922
2923void
2924arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2925 PPCODE:
2926 cl_kernel_arg_access_qualifier value [1];
2927 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2928 EXTEND (SP, 1);
2929 const int i = 0;
2930 PUSHs (sv_2mortal (newSVuv (value [i])));
2931
2932void
2933arg_type_name (OpenCL::Kernel self, cl_uint idx)
2934 ALIAS:
2935 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2936 arg_name = CL_KERNEL_ARG_NAME
2937 PPCODE:
2938 size_t size;
2939 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2940 char *value = tmpbuf (size);
2941 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2942 EXTEND (SP, 1);
2943 const int i = 0;
2944 PUSHs (sv_2mortal (newSVpv (value, 0)));
2945
2946void
2947arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2948 PPCODE:
2949 cl_kernel_arg_type_qualifier value [1];
2950 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2951 EXTEND (SP, 1);
2952 const int i = 0;
2953 PUSHs (sv_2mortal (newSVuv (value [i])));
2954
2955#END:kernel_arg
2956
2957#endif
2958
1979MODULE = OpenCL PACKAGE = OpenCL::Event 2959MODULE = OpenCL PACKAGE = OpenCL::Event
1980 2960
1981void 2961void
1982DESTROY (OpenCL::Event self) 2962DESTROY (OpenCL::Event self)
1983 CODE: 2963 CODE:
1985 2965
1986void 2966void
1987wait (OpenCL::Event self) 2967wait (OpenCL::Event self)
1988 CODE: 2968 CODE:
1989 clWaitForEvents (1, &self); 2969 clWaitForEvents (1, &self);
2970
2971void
2972cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2973 CODE:
2974 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1990 2975
1991void 2976void
1992info (OpenCL::Event self, cl_event_info name) 2977info (OpenCL::Event self, cl_event_info name)
1993 PPCODE: 2978 PPCODE:
1994 INFO (Event) 2979 INFO (Event)
2002 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2987 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2003 EXTEND (SP, 1); 2988 EXTEND (SP, 1);
2004 const int i = 0; 2989 const int i = 0;
2005 { 2990 {
2006 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2991 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2007 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2992 PUSH_CLOBJ (stash_queue, value [i]);
2008 } 2993 }
2009 2994
2010void 2995void
2011command_type (OpenCL::Event self) 2996command_type (OpenCL::Event self)
2012 PPCODE: 2997 PPCODE:
2035 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3020 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2036 EXTEND (SP, 1); 3021 EXTEND (SP, 1);
2037 const int i = 0; 3022 const int i = 0;
2038 { 3023 {
2039 NEED_SUCCESS (RetainContext, (value [i])); 3024 NEED_SUCCESS (RetainContext, (value [i]));
2040 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 3025 PUSH_CLOBJ (stash_context, value [i]);
2041 } 3026 }
2042 3027
2043#END:event 3028#END:event
2044 3029
2045void 3030void
2070void 3055void
2071set_status (OpenCL::UserEvent self, cl_int execution_status) 3056set_status (OpenCL::UserEvent self, cl_int execution_status)
2072 CODE: 3057 CODE:
2073 clSetUserEventStatus (self, execution_status); 3058 clSetUserEventStatus (self, execution_status);
2074 3059
3060MODULE = OpenCL PACKAGE = OpenCL::Mapped
3061
3062void
3063DESTROY (SV *self)
3064 CODE:
3065 OpenCL__Mapped mapped = SvMAPPED (self);
3066
3067 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3068 mapped_detach (self, mapped);
3069
3070 clReleaseCommandQueue (mapped->queue);
3071 clReleaseEvent (mapped->event);
3072 Safefree (mapped);
3073
3074void
3075unmap (OpenCL::Mapped self, ...)
3076 CODE:
3077 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
3078
3079bool
3080mapped (OpenCL::Mapped self)
3081 CODE:
3082 RETVAL = !!self->ptr;
3083 OUTPUT:
3084 RETVAL
3085
3086void
3087wait (OpenCL::Mapped self)
3088 PPCODE:
3089 if (self->event)
3090 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3091
3092void
3093event (OpenCL::Mapped self)
3094 PPCODE:
3095 if (!self->event)
3096 XSRETURN_UNDEF;
3097
3098 clRetainEvent (self->event);
3099 XPUSH_CLOBJ (stash_event, self->event);
3100
3101size_t
3102size (OpenCL::Mapped self)
3103 CODE:
3104 RETVAL = self->cb;
3105 OUTPUT:
3106 RETVAL
3107
3108IV
3109ptr (OpenCL::Mapped self)
3110 CODE:
3111 RETVAL = PTR2IV (self->ptr);
3112 OUTPUT:
3113 RETVAL
3114
3115void
3116set (OpenCL::Mapped self, size_t offset, SV *data)
3117 CODE:
3118 STRLEN len;
3119 const char *ptr = SvPVbyte (data, len);
3120
3121 if (offset + len > self->cb)
3122 croak ("OpenCL::Mapped::set out of bound condition detected");
3123
3124 memcpy (offset + (char *)self->ptr, ptr, len);
3125
3126MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3127
3128MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3129
3130IV
3131row_pitch (OpenCL::Mapped self)
3132 ALIAS:
3133 slice_pitch = 1
3134 CODE:
3135 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3136 OUTPUT:
3137 RETVAL
3138
3139
3140

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines