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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.35 by root, Sat Apr 21 17:47:23 2012 UTC vs.
Revision 1.62 by root, Tue May 1 22:04:56 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h"
7#include "schmorp.h"
8
5#ifdef I_DLFCN 9#ifdef I_DLFCN
6 #include <dlfcn.h> 10 #include <dlfcn.h>
11#endif
12
13// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
14// but then fail to define the api functions unless you ALSO define
15// this. This breaks 100% of the opencl 1.1 apps, for what reason?
16// after all, the functions are deprecated, not removed.
17// in addition, you cannot test for this in any future-proof way.
18// each time a new opencl version comes out, you need to make a new
19// release.
20#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
21
22#ifndef PREFER_1_1
23 #define PREFER_1_1 1
24#endif
25
26#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
7#endif 28#endif
8 29
9#ifdef __APPLE__ 30#ifdef __APPLE__
10 #include <OpenCL/opencl.h> 31 #include <OpenCL/opencl.h>
11#else 32#else
12 #include <CL/opencl.h> 33 #include <CL/opencl.h>
34#endif
35
36#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1
38 #define PREFER_1_1 1
13#endif 39#endif
14 40
15typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
16typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
17typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
31typedef cl_program OpenCL__Program; 57typedef cl_program OpenCL__Program;
32typedef cl_kernel OpenCL__Kernel; 58typedef cl_kernel OpenCL__Kernel;
33typedef cl_event OpenCL__Event; 59typedef cl_event OpenCL__Event;
34typedef cl_event OpenCL__UserEvent; 60typedef cl_event OpenCL__UserEvent;
35 61
62typedef struct mapped * OpenCL__Mapped;
63
36typedef SV *FUTURE; 64typedef SV *FUTURE;
65
66static HV
67 *stash_platform,
68 *stash_device,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
37 89
38/*****************************************************************************/ 90/*****************************************************************************/
39 91
40// name must include a leading underscore 92// name must include a leading underscore
41// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
89 enum { buffers = 3 }; 141 enum { buffers = 3 };
90 static int idx; 142 static int idx;
91 static void *buf [buffers]; 143 static void *buf [buffers];
92 static size_t len [buffers]; 144 static size_t len [buffers];
93 145
94 idx = ++idx % buffers; 146 idx = (idx + 1) % buffers;
95 147
96 if (len [idx] < size) 148 if (len [idx] < size)
97 { 149 {
98 free (buf [idx]); 150 free (buf [idx]);
99 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
164 216
165#define NEED_SUCCESS_ARG(retdecl, name, args) \ 217#define NEED_SUCCESS_ARG(retdecl, name, args) \
166 retdecl = cl ## name args; \ 218 retdecl = cl ## name args; \
167 if (res) \ 219 if (res) \
168 FAIL (name); 220 FAIL (name);
169
170/*****************************************************************************/
171
172#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
173#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
174
175static void *
176SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
177{
178 if (SvROK (sv) && sv_derived_from (sv, pkg))
179 return (void *)SvIV (SvRV (sv));
180
181 croak ("%s: %s is not of type %s", func, svname, pkg);
182}
183 221
184/*****************************************************************************/ 222/*****************************************************************************/
185 223
186static cl_context_properties * 224static cl_context_properties *
187SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
248 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
249} 287}
250 288
251/*****************************************************************************/ 289/*****************************************************************************/
252 290
291static SV *
292new_clobj (HV *stash, IV id)
293{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
295}
296
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
299
300/* cl objects are either \$iv, or [$iv, ...] */
301/* they can be upgraded at runtime to the array form */
302static void *
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
304{
305 if (SvROK (sv) && sv_derived_from (sv, pkg))
306 return (void *)SvIV (SvRV (sv));
307
308 croak ("%s: %s is not of type %s", func, svname, pkg);
309}
310
311/*****************************************************************************/
312/* callback stuff */
313
314/* default context callback, log to stderr */
315static void CL_CALLBACK
316context_default_notify (const char *msg, const void *info, size_t cb, void *data)
317{
318 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
319}
320
321typedef struct
322{
323 int free_cb;
324 void (*push)(void *data1, void *data2, void *data3);
325} eq_vtbl;
326
327typedef struct eq_item
328{
329 struct eq_item *next;
330 eq_vtbl *vtbl;
331 SV *cb;
332 void *data1, *data2, *data3;
333} eq_item;
334
335static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail;
339
340static void
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{
343 eq_item *item = malloc (sizeof (eq_item));
344
345 item->next = 0;
346 item->vtbl = vtbl;
347 item->cb = cb;
348 item->data1 = data1;
349 item->data2 = data2;
350 item->data3 = data3;
351
352 X_LOCK (eq_lock);
353
354 *(eq_head ? &eq_tail->next : &eq_head) = item;
355 eq_tail = item;
356
357 X_UNLOCK (eq_lock);
358
359 eq_signal_func (eq_signal_arg, 0);
360}
361
362static eq_item *
363eq_dec (void)
364{
365 eq_item *res;
366
367 X_LOCK (eq_lock);
368
369 res = eq_head;
370 if (res)
371 eq_head = res->next;
372
373 X_UNLOCK (eq_lock);
374
375 return res;
376}
377
378static void
379eq_poll (void)
380{
381 eq_item *item;
382
383 while ((item = eq_dec ()))
384 {
385 ENTER;
386 SAVETMPS;
387
388 dSP;
389 PUSHMARK (SP);
390 EXTEND (SP, 2);
391
392 if (item->vtbl->free_cb)
393 sv_2mortal (item->cb);
394
395 PUTBACK;
396 item->vtbl->push (item->data1, item->data2, item->data3);
397
398 SV *cb = item->cb;
399 free (item);
400
401 call_sv (cb, G_DISCARD | G_VOID);
402
403 FREETMPS;
404 LEAVE;
405 }
406}
407
408static void
409eq_poll_interrupt (pTHX_ void *c_arg, int value)
410{
411 eq_poll ();
412}
413
414/*****************************************************************************/
415/* context notify */
416
417static void
418eq_context_push (void *data1, void *data2, void *data3)
419{
420 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
423 PUTBACK;
424
425 free (data1);
426 free (data2);
427}
428
429static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
430
431static void CL_CALLBACK
432eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
433{
434 void *pvt_copy = malloc (cb);
435 memcpy (pvt_copy, pvt, cb);
436 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
437}
438
439#define CONTEXT_NOTIFY_CALLBACK \
440 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
441 void *user_data = 0; \
442 \
443 if (SvOK (notify)) \
444 { \
445 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \
447 }
448
449static SV *
450new_clobj_context (cl_context ctx, void *user_data)
451{
452 SV *sv = new_clobj (stash_context, (IV)ctx);
453
454 if (user_data)
455 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
456
457 return sv;
458}
459
460#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
461
462/*****************************************************************************/
463/* build/compile/link notify */
464
465static void
466eq_program_push (void *data1, void *data2, void *data3)
467{
468 dSP;
469 PUSH_CLOBJ (stash_program, data1);
470 PUTBACK;
471}
472
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474
475static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data)
477{
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
479}
480
481struct build_args
482{
483 cl_program program;
484 char *options;
485 void *user_data;
486 cl_uint num_devices;
487};
488
489X_THREAD_PROC (build_program_thread)
490{
491 struct build_args *arg = thr_arg;
492
493 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
494
495 if (arg->user_data)
496 eq_program_notify (arg->program, arg->user_data);
497 else
498 clReleaseProgram (arg->program);
499
500 free (arg->options);
501 free (arg);
502}
503
504static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{
507 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
508
509 arg->program = program;
510 arg->options = strdup (options);
511 arg->user_data = user_data;
512 arg->num_devices = num_devices;
513 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
514
515 xthread_t id;
516 thread_create (&id, build_program_thread, arg);
517}
518
519/*****************************************************************************/
520/* event objects */
521
522static void
523eq_event_push (void *data1, void *data2, void *data3)
524{
525 dSP;
526 PUSH_CLOBJ (stash_event, data1);
527 PUSHs (sv_2mortal (newSViv ((IV)data2)));
528 PUTBACK;
529}
530
531static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
532
533static void CL_CALLBACK
534eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
535{
536 clRetainEvent (event);
537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
538}
539
540/*****************************************************************************/
541/* utilities for XS code */
542
253static size_t 543static size_t
254img_row_pitch (cl_mem img) 544img_row_pitch (cl_mem img)
255{ 545{
256 size_t res; 546 size_t res;
257 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
258 return res; 548 return res;
259} 549}
260 550
261static cl_event * 551static cl_event *
262event_list (SV **items, cl_uint *rcount) 552event_list (SV **items, cl_uint *rcount, cl_event extra)
263{ 553{
264 cl_uint count = *rcount; 554 cl_uint count = *rcount;
265 555
266 if (!count) 556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed
557 count = 0;
558
559 if (!count && !extra)
267 return 0; 560 return 0;
268 561
269 cl_event *list = tmpbuf (sizeof (cl_event) * count); 562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
270 int i = 0; 563 int i = 0;
271 564
272 do 565 while (count--)
273 {
274 --count;
275 if (SvOK (items [count])) 566 if (SvOK (items [count]))
276 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
277 } 568
278 while (count); 569 if (extra)
570 list [i++] = extra;
279 571
280 *rcount = i; 572 *rcount = i;
281 573
282 return i ? list : 0; 574 return i ? list : 0;
283} 575}
284 576
285#define EVENT_LIST(items,count) \ 577#define EVENT_LIST(skip) \
286 cl_uint event_list_count = (count); \ 578 cl_uint event_list_count = items - (skip); \
287 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
288 580
289#define INFO(class) \ 581#define INFO(class) \
290{ \ 582{ \
291 size_t size; \ 583 size_t size; \
292 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
296 SvCUR_set (sv, size); \ 588 SvCUR_set (sv, size); \
297 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 589 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
298 XPUSHs (sv); \ 590 XPUSHs (sv); \
299} 591}
300 592
593/*****************************************************************************/
594/* mapped_xxx */
595
596static OpenCL__Mapped
597SvMAPPED (SV *sv)
598{
599 // no typechecking atm., keep your fingers crossed
600 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
601}
602
603struct mapped
604{
605 cl_command_queue queue;
606 cl_mem memobj;
607 void *ptr;
608 size_t cb;
609 cl_event event;
610 size_t row_pitch;
611 size_t slice_pitch;
612};
613
614static SV *
615mapped_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)
616{
617 SV *data = newSV (0);
618 SvUPGRADE (data, SVt_PVMG);
619
620 OpenCL__Mapped mapped;
621 New (0, mapped, 1, struct mapped);
622
623 clRetainCommandQueue (queue);
624
625 mapped->queue = queue;
626 mapped->memobj = memobj;
627 mapped->ptr = ptr;
628 mapped->cb = cb;
629 mapped->event = ev;
630 mapped->row_pitch = row_pitch;
631 mapped->slice_pitch = slice_pitch;
632
633 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
634
635 if (SvLEN (data))
636 Safefree (data);
637
638 SvPVX (data) = (char *)ptr;
639 SvCUR_set (data, cb);
640 SvLEN_set (data, 0);
641 SvPOK_only (data);
642
643 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
644
645 if (!(flags & CL_MAP_WRITE))
646 SvREADONLY_on (data);
647
648 return obj;
649}
650
651static void
652mapped_detach (SV *sv, OpenCL__Mapped mapped)
653{
654 SV *data = SvRV (sv);
655
656 // the next check checks both before AND after detach, where SvPVX should STILL be 0
657 if (SvPVX (data) != (char *)mapped->ptr)
658 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
659 else
660 {
661 SvREADONLY_off (data);
662 SvCUR_set (data, 0);
663 SvPVX (data) = 0;
664 SvOK_off (data);
665 }
666
667 mapped->ptr = 0;
668}
669
670static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
674 cl_event ev;
675
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677
678 clReleaseEvent (mapped->event);
679 mapped->event = ev;
680
681 mapped_detach (self, mapped);
682}
683
684/*****************************************************************************/
685
301MODULE = OpenCL PACKAGE = OpenCL 686MODULE = OpenCL PACKAGE = OpenCL
302 687
303PROTOTYPES: ENABLE 688PROTOTYPES: ENABLE
689
690void
691poll ()
692 CODE:
693 eq_poll ();
694
695void
696_eq_initialise (IV func, IV arg)
697 CODE:
698 eq_signal_func = (void (*)(void *, int))func;
699 eq_signal_arg = (void*)arg;
304 700
305BOOT: 701BOOT:
306{ 702{
307 HV *stash = gv_stashpv ("OpenCL", 1); 703 HV *stash = gv_stashpv ("OpenCL", 1);
308 static const ivstr *civ, const_iv[] = { 704 static const ivstr *civ, const_iv[] = {
317 { sizeof (cl_half ), "SIZEOF_HALF" }, 713 { sizeof (cl_half ), "SIZEOF_HALF" },
318 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 714 { sizeof (cl_float ), "SIZEOF_FLOAT" },
319 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 715 { sizeof (cl_double), "SIZEOF_DOUBLE" },
320#include "constiv.h" 716#include "constiv.h"
321 }; 717 };
718
322 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
323 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
729 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
730 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
731 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
732 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
733 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
734 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
735 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
736 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
737 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
738 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
739 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
740 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
741 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
742 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
743 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
744
745 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
324} 746}
325 747
326cl_int 748cl_int
327errno () 749errno ()
328 CODE: 750 CODE:
329 errno = res; 751 RETVAL = res;
752 OUTPUT:
753 RETVAL
330 754
331const char * 755const char *
332err2str (cl_int err) 756err2str (cl_int err = res)
333 757
334const char * 758const char *
335enum2str (cl_uint value) 759enum2str (cl_uint value)
336 760
337void 761void
345 list = tmpbuf (sizeof (*list) * count); 769 list = tmpbuf (sizeof (*list) * count);
346 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
347 771
348 EXTEND (SP, count); 772 EXTEND (SP, count);
349 for (i = 0; i < count; ++i) 773 for (i = 0; i < count; ++i)
350 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 774 PUSH_CLOBJ (stash_platform, (IV)list [i]);
351 775
352void 776void
353context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
354 PPCODE: 778 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK;
355 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 780 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
356 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 781 XPUSH_CLOBJ_CONTEXT;
357 782
358void 783void
359context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 784context (FUTURE properties, FUTURE devices, FUTURE notify)
360 PPCODE: 785 PPCODE:
361 /* der Gipfel der Kunst */ 786 /* der Gipfel der Kunst */
362 787
363void 788void
364wait_for_events (...) 789wait_for_events (...)
365 CODE: 790 CODE:
366 EVENT_LIST (0, items); 791 EVENT_LIST (0);
367 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 792 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
368 793
369PROTOTYPES: DISABLE 794PROTOTYPES: DISABLE
370 795
371MODULE = OpenCL PACKAGE = OpenCL::Platform 796MODULE = OpenCL PACKAGE = OpenCL::Platform
372 797
373void 798void
374info (OpenCL::Platform self, cl_platform_info name) 799info (OpenCL::Platform self, cl_platform_info name)
375 PPCODE: 800 PPCODE:
376 INFO (Platform) 801 INFO (Platform)
802
803void
804unload_compiler (OpenCL::Platform self)
805 CODE:
806#if CL_VERSION_1_2
807 clUnloadPlatformCompiler (self);
808#endif
377 809
378#BEGIN:platform 810#BEGIN:platform
379 811
380void 812void
381profile (OpenCL::Platform self) 813profile (OpenCL::Platform self)
407 list = tmpbuf (sizeof (*list) * count); 839 list = tmpbuf (sizeof (*list) * count);
408 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 840 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
409 841
410 EXTEND (SP, count); 842 EXTEND (SP, count);
411 for (i = 0; i < count; ++i) 843 for (i = 0; i < count; ++i)
412 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 844 PUSH_CLOBJ (stash_device, list [i]);
413 845
414void 846void
415context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
416 PPCODE: 848 PPCODE:
417 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
418 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 850 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
419 851
420 AV *av = (AV *)SvRV (devices); 852 AV *av = (AV *)SvRV (devices);
421 cl_uint num_devices = av_len (av) + 1; 853 cl_uint num_devices = av_len (av) + 1;
422 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
855
423 int i; 856 int i;
424
425 for (i = num_devices; i--; ) 857 for (i = num_devices; i--; )
426 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
427 859
860 CONTEXT_NOTIFY_CALLBACK;
428 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 861 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
429 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 862 XPUSH_CLOBJ_CONTEXT;
430 863
431void 864void
432context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 865context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
433 PPCODE: 866 PPCODE:
434 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
435 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 868 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
869
870 CONTEXT_NOTIFY_CALLBACK;
436 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 871 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
437 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 872 XPUSH_CLOBJ_CONTEXT;
438 873
439MODULE = OpenCL PACKAGE = OpenCL::Device 874MODULE = OpenCL PACKAGE = OpenCL::Device
440 875
441void 876void
442info (OpenCL::Device self, cl_device_info name) 877info (OpenCL::Device self, cl_device_info name)
450 PPCODE: 885 PPCODE:
451 cl_device_type value [1]; 886 cl_device_type value [1];
452 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 887 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
453 EXTEND (SP, 1); 888 EXTEND (SP, 1);
454 const int i = 0; 889 const int i = 0;
455 PUSHs (sv_2mortal (newSViv (value [i]))); 890 PUSHs (sv_2mortal (newSVuv (value [i])));
456 891
457void 892void
458vendor_id (OpenCL::Device self) 893vendor_id (OpenCL::Device self)
459 ALIAS: 894 ALIAS:
460 vendor_id = CL_DEVICE_VENDOR_ID 895 vendor_id = CL_DEVICE_VENDOR_ID
481 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 916 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
482 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 917 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
483 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 918 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
484 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 919 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
485 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 920 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
486 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 921 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
487 PPCODE: 922 PPCODE:
488 cl_uint value [1]; 923 cl_uint value [1];
489 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 924 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
490 EXTEND (SP, 1); 925 EXTEND (SP, 1);
491 const int i = 0; 926 const int i = 0;
605 PPCODE: 1040 PPCODE:
606 cl_command_queue_properties value [1]; 1041 cl_command_queue_properties value [1];
607 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1042 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
608 EXTEND (SP, 1); 1043 EXTEND (SP, 1);
609 const int i = 0; 1044 const int i = 0;
610 PUSHs (sv_2mortal (newSViv (value [i]))); 1045 PUSHs (sv_2mortal (newSVuv (value [i])));
611 1046
612void 1047void
613platform (OpenCL::Device self) 1048platform (OpenCL::Device self)
614 PPCODE: 1049 PPCODE:
615 cl_platform_id value [1]; 1050 cl_platform_id value [1];
616 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1051 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
617 EXTEND (SP, 1); 1052 EXTEND (SP, 1);
618 const int i = 0; 1053 const int i = 0;
619 { 1054 {
620 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1055 PUSH_CLOBJ (stash_platform, value [i]);
621 } 1056 }
622 1057
623void 1058void
624name (OpenCL::Device self) 1059name (OpenCL::Device self)
625 ALIAS: 1060 ALIAS:
644 cl_device_id value [1]; 1079 cl_device_id value [1];
645 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1080 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
646 EXTEND (SP, 1); 1081 EXTEND (SP, 1);
647 const int i = 0; 1082 const int i = 0;
648 { 1083 {
649 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1084 PUSH_CLOBJ (stash_device, value [i]);
650 } 1085 }
651 1086
652void 1087void
653partition_types_ext (OpenCL::Device self) 1088partition_types_ext (OpenCL::Device self)
654 ALIAS: 1089 ALIAS:
681 1116
682void 1117void
683queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1118queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
684 PPCODE: 1119 PPCODE:
685 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1120 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
686 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1121 XPUSH_CLOBJ (stash_queue, queue);
687 1122
688void 1123void
689user_event (OpenCL::Context self) 1124user_event (OpenCL::Context self)
690 PPCODE: 1125 PPCODE:
691 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1126 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
692 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1127 XPUSH_CLOBJ (stash_userevent, ev);
693 1128
694void 1129void
695buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1130buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
696 PPCODE: 1131 PPCODE:
697 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1132 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
698 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1133 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
699 1134
700 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1135 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
701 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1136 XPUSH_CLOBJ (stash_bufferobj, mem);
702 1137
703void 1138void
704buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1139buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
705 PPCODE: 1140 PPCODE:
706 STRLEN len; 1141 STRLEN len;
707 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1142 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
708 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1143 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
709 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1144 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
710 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1145 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
711 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1146 XPUSH_CLOBJ (stash_bufferobj, mem);
712 1147
1148#if CL_VERSION_1_2
1149
713void 1150void
714image2d (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) 1151image (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)
715 PPCODE: 1152 PPCODE:
716 STRLEN len; 1153 STRLEN len;
717 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1154 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
718 const cl_image_format format = { channel_order, channel_type }; 1155 const cl_image_format format = { channel_order, channel_type };
1156 const cl_image_desc desc = {
1157 type,
1158 width, height, depth,
1159 array_size, row_pitch, slice_pitch,
1160 num_mip_level, num_samples,
1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1162 };
719 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
720 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1164 HV *stash = stash_image;
1165 switch (type)
1166 {
1167 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1168 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1169 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1170 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1171 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1172 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1173 }
1174 XPUSH_CLOBJ (stash, mem);
721 1175
1176#endif
1177
722void 1178void
723image3d (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) 1179image2d (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)
724 PPCODE: 1180 PPCODE:
725 STRLEN len; 1181 STRLEN len;
726 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1182 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
727 const cl_image_format format = { channel_order, channel_type }; 1183 const cl_image_format format = { channel_order, channel_type };
1184#if PREFER_1_1
1185 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1186#else
1187 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1188 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1189#endif
1190 XPUSH_CLOBJ (stash_image2d, mem);
1191
1192void
1193image3d (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)
1194 PPCODE:
1195 STRLEN len;
1196 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1197 const cl_image_format format = { channel_order, channel_type };
1198#if PREFER_1_1
728 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1199 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
729 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1200#else
1201 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1202 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1203#endif
1204 XPUSH_CLOBJ (stash_image3d, mem);
730 1205
731#if cl_apple_gl_sharing || cl_khr_gl_sharing 1206#if cl_apple_gl_sharing || cl_khr_gl_sharing
732 1207
733void 1208void
734gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1209gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
735 PPCODE: 1210 PPCODE:
736 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1211 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
737 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1212 XPUSH_CLOBJ (stash_bufferobj, mem);
1213
1214void
1215gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1216 PPCODE:
1217 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1218 XPUSH_CLOBJ (stash_image2d, mem);
1219
1220#if CL_VERSION_1_2
1221
1222void
1223gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1224 ALIAS:
1225 PPCODE:
1226 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1227 cl_gl_object_type type;
1228 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1229 HV *stash = stash_memory;
1230 switch (type)
1231 {
1232 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1233 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1234 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1235 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1236 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1237 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1238 }
1239 XPUSH_CLOBJ (stash, mem);
1240
1241#endif
738 1242
739void 1243void
740gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1244gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
741 PPCODE: 1245 PPCODE:
1246#if PREFER_1_1
742 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1247 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
743 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1248#else
1249 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1250#endif
1251 XPUSH_CLOBJ (stash_image2d, mem);
744 1252
745void 1253void
746gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1254gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
747 PPCODE: 1255 PPCODE:
1256#if PREFER_1_1
748 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1257 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
749 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1258#else
750 1259 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
751void 1260#endif
752gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1261 XPUSH_CLOBJ (stash_image3d, mem);
753 PPCODE:
754 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
755 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
756 1262
757#endif 1263#endif
758 1264
759void 1265void
760supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1266supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
780 1286
781void 1287void
782sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1288sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
783 PPCODE: 1289 PPCODE:
784 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1290 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
785 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1291 XPUSH_CLOBJ (stash_sampler, sampler);
786 1292
787void 1293void
788program_with_source (OpenCL::Context self, SV *program) 1294program_with_source (OpenCL::Context self, SV *program)
789 PPCODE: 1295 PPCODE:
790 STRLEN len; 1296 STRLEN len;
791 size_t len2; 1297 size_t len2;
792 const char *ptr = SvPVbyte (program, len); 1298 const char *ptr = SvPVbyte (program, len);
793 1299
794 len2 = len; 1300 len2 = len;
795 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
796 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1302 XPUSH_CLOBJ (stash_program, prog);
797 1303
798#BEGIN:context 1304#BEGIN:context
799 1305
800void 1306void
801reference_count (OpenCL::Context self) 1307reference_count (OpenCL::Context self)
818 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1324 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
819 int i, n = size / sizeof (*value); 1325 int i, n = size / sizeof (*value);
820 EXTEND (SP, n); 1326 EXTEND (SP, n);
821 for (i = 0; i < n; ++i) 1327 for (i = 0; i < n; ++i)
822 { 1328 {
823 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1329 PUSH_CLOBJ (stash_device, value [i]);
824 } 1330 }
825 1331
826void 1332void
827properties (OpenCL::Context self) 1333properties (OpenCL::Context self)
828 PPCODE: 1334 PPCODE:
843DESTROY (OpenCL::Queue self) 1349DESTROY (OpenCL::Queue self)
844 CODE: 1350 CODE:
845 clReleaseCommandQueue (self); 1351 clReleaseCommandQueue (self);
846 1352
847void 1353void
848enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS:
1356 enqueue_read_buffer = 0
849 PPCODE: 1357 PPCODE:
850 cl_event ev = 0; 1358 cl_event ev = 0;
851 EVENT_LIST (6, items - 6); 1359 EVENT_LIST (6);
852 1360
853 SvUPGRADE (data, SVt_PV); 1361 SvUPGRADE (data, SVt_PV);
854 SvGROW (data, len); 1362 SvGROW (data, len);
855 SvPOK_only (data); 1363 SvPOK_only (data);
856 SvCUR_set (data, len); 1364 SvCUR_set (data, len);
857 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1365 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
858 1366
859 if (ev) 1367 if (ev)
860 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1368 XPUSH_CLOBJ (stash_event, ev);
861 1369
862void 1370void
863enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS:
1373 enqueue_write_buffer = 0
864 PPCODE: 1374 PPCODE:
865 cl_event ev = 0; 1375 cl_event ev = 0;
866 STRLEN len; 1376 STRLEN len;
867 char *ptr = SvPVbyte (data, len); 1377 char *ptr = SvPVbyte (data, len);
868 EVENT_LIST (5, items - 5); 1378 EVENT_LIST (5);
869 1379
870 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1380 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
871 1381
872 if (ev) 1382 if (ev)
873 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1383 XPUSH_CLOBJ (stash_event, ev);
874 1384
1385#if CL_VERSION_1_2
1386
875void 1387void
876enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS:
1390 enqueue_fill_buffer = 0
877 PPCODE: 1391 PPCODE:
878 cl_event ev = 0; 1392 cl_event ev = 0;
1393 STRLEN len;
1394 char *ptr = SvPVbyte (data, len);
1395 EVENT_LIST (5);
1396
1397 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1398
1399 if (ev)
1400 XPUSH_CLOBJ (stash_event, ev);
1401
1402void
1403fill_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, ...)
1404 ALIAS:
1405 enqueue_fill_image = 0
1406 PPCODE:
1407 cl_event ev = 0;
1408 STRLEN len;
1409 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412
1413 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1417 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1418 cl_image_format format;
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1421 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1422 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1425 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426
1427 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev);
1429
1430#endif
1431
1432void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS:
1435 enqueue_copy_buffer = 0
1436 PPCODE:
1437 cl_event ev = 0;
879 EVENT_LIST (6, items - 6); 1438 EVENT_LIST (6);
880 1439
881 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1440 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
882 1441
883 if (ev) 1442 if (ev)
884 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1443 XPUSH_CLOBJ (stash_event, ev);
885 1444
886void 1445void
887enqueue_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, ...) 1446read_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, ...)
1447 ALIAS:
1448 enqueue_read_buffer_rect = 0
888 PPCODE: 1449 PPCODE:
889 cl_event ev = 0; 1450 cl_event ev = 0;
890 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
891 const size_t host_origin[3] = { host_x, host_y, host_z }; 1452 const size_t host_origin[3] = { host_x, host_y, host_z };
892 const size_t region[3] = { width, height, depth }; 1453 const size_t region[3] = { width, height, depth };
893 EVENT_LIST (17, items - 17); 1454 EVENT_LIST (17);
894 1455
895 if (!buf_row_pitch) 1456 if (!buf_row_pitch)
896 buf_row_pitch = region [0]; 1457 buf_row_pitch = region [0];
897 1458
898 if (!buf_slice_pitch) 1459 if (!buf_slice_pitch)
911 SvPOK_only (data); 1472 SvPOK_only (data);
912 SvCUR_set (data, len); 1473 SvCUR_set (data, len);
913 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)); 1474 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));
914 1475
915 if (ev) 1476 if (ev)
916 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1477 XPUSH_CLOBJ (stash_event, ev);
917 1478
918void 1479void
919enqueue_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, ...) 1480write_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, ...)
1481 ALIAS:
1482 enqueue_write_buffer_rect = 0
920 PPCODE: 1483 PPCODE:
921 cl_event ev = 0; 1484 cl_event ev = 0;
922 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
923 const size_t host_origin[3] = { host_x, host_y, host_z }; 1486 const size_t host_origin[3] = { host_x, host_y, host_z };
924 const size_t region[3] = { width, height, depth }; 1487 const size_t region[3] = { width, height, depth };
925 STRLEN len; 1488 STRLEN len;
926 char *ptr = SvPVbyte (data, len); 1489 char *ptr = SvPVbyte (data, len);
927 EVENT_LIST (17, items - 17); 1490 EVENT_LIST (17);
928 1491
929 if (!buf_row_pitch) 1492 if (!buf_row_pitch)
930 buf_row_pitch = region [0]; 1493 buf_row_pitch = region [0];
931 1494
932 if (!buf_slice_pitch) 1495 if (!buf_slice_pitch)
941 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1504 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
942 1505
943 if (len < min_len) 1506 if (len < min_len)
944 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
945 1508
946 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1509 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));
947 1510
948 if (ev) 1511 if (ev)
949 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1512 XPUSH_CLOBJ (stash_event, ev);
950 1513
951void 1514void
952enqueue_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, ...) 1515copy_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, ...)
1516 ALIAS:
1517 enqueue_copy_buffer_rect = 0
953 PPCODE: 1518 PPCODE:
954 cl_event ev = 0; 1519 cl_event ev = 0;
955 const size_t src_origin[3] = { src_x, src_y, src_z }; 1520 const size_t src_origin[3] = { src_x, src_y, src_z };
956 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1521 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
957 const size_t region[3] = { width, height, depth }; 1522 const size_t region[3] = { width, height, depth };
958 EVENT_LIST (16, items - 16); 1523 EVENT_LIST (16);
959 1524
960 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)); 1525 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));
961 1526
962 if (ev) 1527 if (ev)
963 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1528 XPUSH_CLOBJ (stash_event, ev);
964 1529
965void 1530void
966enqueue_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, ...) 1531read_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, ...)
1532 ALIAS:
1533 enqueue_read_image = 0
967 PPCODE: 1534 PPCODE:
968 cl_event ev = 0; 1535 cl_event ev = 0;
969 const size_t src_origin[3] = { src_x, src_y, src_z }; 1536 const size_t src_origin[3] = { src_x, src_y, src_z };
970 const size_t region[3] = { width, height, depth }; 1537 const size_t region[3] = { width, height, depth };
971 EVENT_LIST (12, items - 12); 1538 EVENT_LIST (12);
972 1539
973 if (!row_pitch) 1540 if (!row_pitch)
974 row_pitch = img_row_pitch (src); 1541 row_pitch = img_row_pitch (src);
975 1542
976 if (depth > 1 && !slice_pitch) 1543 if (depth > 1 && !slice_pitch)
983 SvPOK_only (data); 1550 SvPOK_only (data);
984 SvCUR_set (data, len); 1551 SvCUR_set (data, len);
985 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)); 1552 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));
986 1553
987 if (ev) 1554 if (ev)
988 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1555 XPUSH_CLOBJ (stash_event, ev);
989 1556
990void 1557void
991enqueue_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, ...) 1558write_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, ...)
1559 ALIAS:
1560 enqueue_write_image = 0
992 PPCODE: 1561 PPCODE:
993 cl_event ev = 0; 1562 cl_event ev = 0;
994 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
995 const size_t region[3] = { width, height, depth }; 1564 const size_t region[3] = { width, height, depth };
996 STRLEN len; 1565 STRLEN len;
997 char *ptr = SvPVbyte (data, len); 1566 char *ptr = SvPVbyte (data, len);
998 EVENT_LIST (12, items - 12); 1567 EVENT_LIST (12);
999 1568
1000 if (!row_pitch) 1569 if (!row_pitch)
1001 row_pitch = img_row_pitch (dst); 1570 row_pitch = img_row_pitch (dst);
1002 1571
1003 if (depth > 1 && !slice_pitch) 1572 if (depth > 1 && !slice_pitch)
1006 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1575 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1007 1576
1008 if (len < min_len) 1577 if (len < min_len)
1009 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1010 1579
1011 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1580 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));
1012 1581
1013 if (ev) 1582 if (ev)
1014 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1583 XPUSH_CLOBJ (stash_event, ev);
1015 1584
1016void 1585void
1017enqueue_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, ...) 1586copy_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, ...)
1587 ALIAS:
1588 enqueue_copy_image = 0
1018 PPCODE: 1589 PPCODE:
1019 cl_event ev = 0; 1590 cl_event ev = 0;
1020 const size_t src_origin[3] = { src_x, src_y, src_z }; 1591 const size_t src_origin[3] = { src_x, src_y, src_z };
1021 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1592 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1022 const size_t region[3] = { width, height, depth }; 1593 const size_t region[3] = { width, height, depth };
1023 EVENT_LIST (12, items - 12); 1594 EVENT_LIST (12);
1024 1595
1025 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1596 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1026 1597
1027 if (ev) 1598 if (ev)
1028 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1599 XPUSH_CLOBJ (stash_event, ev);
1029 1600
1030void 1601void
1031enqueue_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, ...) 1602copy_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, ...)
1603 ALIAS:
1604 enqueue_copy_image_to_buffer = 0
1032 PPCODE: 1605 PPCODE:
1033 cl_event ev = 0; 1606 cl_event ev = 0;
1034 const size_t src_origin[3] = { src_x, src_y, src_z }; 1607 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610
1611 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612
1613 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev);
1615
1616void
1617copy_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, ...)
1618 ALIAS:
1619 enqueue_copy_buffer_to_image = 0
1620 PPCODE:
1621 cl_event ev = 0;
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625
1626 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627
1628 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev);
1630
1631void
1632map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, size_t cb = 0, ...)
1633 ALIAS:
1634 enqueue_map_buffer = 0
1635 PPCODE:
1636 cl_event ev;
1637 EVENT_LIST (6);
1638
1639 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1640 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1641
1642void
1643map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, size_t width = 0, size_t height = 0, size_t depth = 0, ...)
1644 ALIAS:
1645 enqueue_map_image = 0
1646 PPCODE:
1647 cl_event ev;
1648 const size_t origin[3] = { x, y, z };
1035 const size_t region[3] = { width, height, depth }; 1649 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch;
1036 EVENT_LIST (10, items - 10); 1651 EVENT_LIST (10);
1037 1652
1653 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));
1654
1655 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1]
1657 : region [0];
1658
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1660
1661void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID)
1666 {
1667 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event);
1669 }
1670
1671void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS:
1674 enqueue_task = 0
1675 PPCODE:
1676 cl_event ev = 0;
1677 EVENT_LIST (2);
1678
1038 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1679 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1039 1680
1040 if (ev) 1681 if (ev)
1041 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1682 XPUSH_CLOBJ (stash_event, ev);
1042 1683
1043void 1684void
1044enqueue_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, ...)
1045 PPCODE:
1046 cl_event ev = 0;
1047 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1048 const size_t region[3] = { width, height, depth };
1049 EVENT_LIST (10, items - 10);
1050
1051 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1052
1053 if (ev)
1054 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1055
1056void
1057enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1058 PPCODE:
1059 cl_event ev = 0;
1060 EVENT_LIST (2, items - 2);
1061
1062 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1063
1064 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1066
1067void
1068enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1686 ALIAS:
1687 enqueue_nd_range_kernel = 0
1069 PPCODE: 1688 PPCODE:
1070 cl_event ev = 0; 1689 cl_event ev = 0;
1071 size_t *gwo = 0, *gws, *lws = 0; 1690 size_t *gwo = 0, *gws, *lws = 0;
1072 int gws_len; 1691 int gws_len;
1073 size_t *lists; 1692 size_t *lists;
1074 int i; 1693 int i;
1075 EVENT_LIST (5, items - 5); 1694 EVENT_LIST (5);
1076 1695
1077 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1078 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1079 1698
1080 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1081 1700
1082 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1701 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1083 1702
1084 gws = lists + gws_len * 0; 1703 gws = lists + gws_len * 0;
1085 for (i = 0; i < gws_len; ++i) 1704 for (i = 0; i < gws_len; ++i)
1705 {
1086 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1706 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1707 // at least nvidia crashes for 0-sized work group sizes, work around
1708 if (!gws [i])
1709 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1710 }
1087 1711
1088 if (SvOK (global_work_offset)) 1712 if (SvOK (global_work_offset))
1089 { 1713 {
1090 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1714 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1091 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1715 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1098 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1722 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1099 } 1723 }
1100 1724
1101 if (SvOK (local_work_size)) 1725 if (SvOK (local_work_size))
1102 { 1726 {
1103 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1727 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1104 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1728 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1105 1729
1106 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1730 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1107 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1731 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1108 1732
1109 lws = lists + gws_len * 2; 1733 lws = lists + gws_len * 2;
1110 for (i = 0; i < gws_len; ++i) 1734 for (i = 0; i < gws_len; ++i)
1735 {
1111 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1736 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1737 // at least nvidia crashes for 0-sized work group sizes, work around
1738 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 }
1112 } 1741 }
1113 1742
1114 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1743 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1115 1744
1116 if (ev) 1745 if (ev)
1117 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1746 XPUSH_CLOBJ (stash_event, ev);
1118 1747
1119#if cl_apple_gl_sharing || cl_khr_gl_sharing 1748#if cl_apple_gl_sharing || cl_khr_gl_sharing
1120 1749
1121void 1750void
1122enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1123 ALIAS: 1752 ALIAS:
1753 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0
1124 enqueue_release_gl_objects = 1 1755 enqueue_release_gl_objects = 1
1125 CODE: 1756 PPCODE:
1126 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1127 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1128 1759
1129 cl_event ev = 0; 1760 cl_event ev = 0;
1130 EVENT_LIST (2, items - 2); 1761 EVENT_LIST (2);
1131 AV *av = (AV *)SvRV (objects); 1762 AV *av = (AV *)SvRV (objects);
1132 cl_uint num_objects = av_len (av) + 1; 1763 cl_uint num_objects = av_len (av) + 1;
1133 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1764 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1134 int i; 1765 int i;
1135 1766
1136 for (i = num_objects; i--; ) 1767 for (i = num_objects; i--; )
1137 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1768 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1138 1769
1139 if (ix) 1770 if (ix)
1140 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1141 else 1772 else
1142 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1143 1774
1144 if (ev) 1775 if (ev)
1145 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1776 XPUSH_CLOBJ (stash_event, ev);
1146 1777
1147#endif 1778#endif
1148 1779
1149void 1780void
1150enqueue_marker (OpenCL::Queue self)
1151 PPCODE:
1152 cl_event ev;
1153 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1154 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1155
1156void
1157enqueue_wait_for_events (OpenCL::Queue self, ...) 1781wait_for_events (OpenCL::Queue self, ...)
1782 ALIAS:
1783 enqueue_wait_for_events = 0
1158 CODE: 1784 CODE:
1159 EVENT_LIST (1, items - 1); 1785 EVENT_LIST (1);
1786#if PREFER_1_1
1160 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1787 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1788#else
1789 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1790#endif
1161 1791
1162void 1792void
1163enqueue_barrier (OpenCL::Queue self) 1793marker (OpenCL::Queue self, ...)
1794 ALIAS:
1795 enqueue_marker = 0
1164 CODE: 1796 PPCODE:
1797 cl_event ev = 0;
1798 EVENT_LIST (1);
1799#if PREFER_1_1
1800 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else
1803#if CL_VERSION_1_2
1804 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1805#else
1806 {
1807 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1808 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1809 }
1810#endif
1811#else
1812 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1813#endif
1814 if (ev)
1815 XPUSH_CLOBJ (stash_event, ev);
1816
1817void
1818barrier (OpenCL::Queue self, ...)
1819 ALIAS:
1820 enqueue_barrier = 0
1821 PPCODE:
1822 cl_event ev = 0;
1823 EVENT_LIST (1);
1824#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID)
1165 NEED_SUCCESS (EnqueueBarrier, (self)); 1826 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else
1828#if CL_VERSION_1_2
1829 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1830#else
1831 {
1832 if (event_list_count)
1833 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1834
1835 if (GIMME_V != G_VOID)
1836 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1837 }
1838#endif
1839#else
1840 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1841#endif
1842 if (ev)
1843 XPUSH_CLOBJ (stash_event, ev);
1166 1844
1167void 1845void
1168flush (OpenCL::Queue self) 1846flush (OpenCL::Queue self)
1169 CODE: 1847 CODE:
1170 NEED_SUCCESS (Flush, (self)); 1848 NEED_SUCCESS (Flush, (self));
1188 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1866 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1189 EXTEND (SP, 1); 1867 EXTEND (SP, 1);
1190 const int i = 0; 1868 const int i = 0;
1191 { 1869 {
1192 NEED_SUCCESS (RetainContext, (value [i])); 1870 NEED_SUCCESS (RetainContext, (value [i]));
1193 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1871 PUSH_CLOBJ (stash_context, value [i]);
1194 } 1872 }
1195 1873
1196void 1874void
1197device (OpenCL::Queue self) 1875device (OpenCL::Queue self)
1198 PPCODE: 1876 PPCODE:
1199 cl_device_id value [1]; 1877 cl_device_id value [1];
1200 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1878 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1201 EXTEND (SP, 1); 1879 EXTEND (SP, 1);
1202 const int i = 0; 1880 const int i = 0;
1203 { 1881 {
1204 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1882 PUSH_CLOBJ (stash_device, value [i]);
1205 } 1883 }
1206 1884
1207void 1885void
1208reference_count (OpenCL::Queue self) 1886reference_count (OpenCL::Queue self)
1209 PPCODE: 1887 PPCODE:
1218 PPCODE: 1896 PPCODE:
1219 cl_command_queue_properties value [1]; 1897 cl_command_queue_properties value [1];
1220 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1898 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1221 EXTEND (SP, 1); 1899 EXTEND (SP, 1);
1222 const int i = 0; 1900 const int i = 0;
1223 PUSHs (sv_2mortal (newSViv (value [i]))); 1901 PUSHs (sv_2mortal (newSVuv (value [i])));
1224 1902
1225#END:command_queue 1903#END:command_queue
1226 1904
1227MODULE = OpenCL PACKAGE = OpenCL::Memory 1905MODULE = OpenCL PACKAGE = OpenCL::Memory
1228 1906
1243 PPCODE: 1921 PPCODE:
1244 cl_mem_object_type value [1]; 1922 cl_mem_object_type value [1];
1245 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 1923 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1246 EXTEND (SP, 1); 1924 EXTEND (SP, 1);
1247 const int i = 0; 1925 const int i = 0;
1248 PUSHs (sv_2mortal (newSViv (value [i]))); 1926 PUSHs (sv_2mortal (newSVuv (value [i])));
1249 1927
1250void 1928void
1251flags (OpenCL::Memory self) 1929flags (OpenCL::Memory self)
1252 PPCODE: 1930 PPCODE:
1253 cl_mem_flags value [1]; 1931 cl_mem_flags value [1];
1254 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 1932 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1255 EXTEND (SP, 1); 1933 EXTEND (SP, 1);
1256 const int i = 0; 1934 const int i = 0;
1257 PUSHs (sv_2mortal (newSViv (value [i]))); 1935 PUSHs (sv_2mortal (newSVuv (value [i])));
1258 1936
1259void 1937void
1260size (OpenCL::Memory self) 1938size (OpenCL::Memory self)
1261 ALIAS: 1939 ALIAS:
1262 size = CL_MEM_SIZE 1940 size = CL_MEM_SIZE
1296 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1974 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1297 EXTEND (SP, 1); 1975 EXTEND (SP, 1);
1298 const int i = 0; 1976 const int i = 0;
1299 { 1977 {
1300 NEED_SUCCESS (RetainContext, (value [i])); 1978 NEED_SUCCESS (RetainContext, (value [i]));
1301 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1979 PUSH_CLOBJ (stash_context, value [i]);
1302 } 1980 }
1303 1981
1304void 1982void
1305associated_memobject (OpenCL::Memory self) 1983associated_memobject (OpenCL::Memory self)
1306 PPCODE: 1984 PPCODE:
1308 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1986 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1309 EXTEND (SP, 1); 1987 EXTEND (SP, 1);
1310 const int i = 0; 1988 const int i = 0;
1311 { 1989 {
1312 NEED_SUCCESS (RetainMemObject, (value [i])); 1990 NEED_SUCCESS (RetainMemObject, (value [i]));
1313 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1991 PUSH_CLOBJ (stash_memory, value [i]);
1314 } 1992 }
1315 1993
1316#END:mem 1994#END:mem
1317 1995
1318#if cl_apple_gl_sharing || cl_khr_gl_sharing 1996#if cl_apple_gl_sharing || cl_khr_gl_sharing
1338 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2016 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1339 2017
1340 cl_buffer_region crdata = { origin, size }; 2018 cl_buffer_region crdata = { origin, size };
1341 2019
1342 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2020 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1343 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2021 XPUSH_CLOBJ (stash_buffer, mem);
1344 2022
1345MODULE = OpenCL PACKAGE = OpenCL::Image 2023MODULE = OpenCL PACKAGE = OpenCL::Image
1346 2024
1347void 2025void
1348image_info (OpenCL::Image self, cl_image_info name) 2026image_info (OpenCL::Image self, cl_image_info name)
1349 PPCODE: 2027 PPCODE:
1350 INFO (Image) 2028 INFO (Image)
2029
2030void
2031format (OpenCL::Image self)
2032 PPCODE:
2033 cl_image_format format;
2034 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2035 EXTEND (SP, 2);
2036 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2037 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1351 2038
1352#BEGIN:image 2039#BEGIN:image
1353 2040
1354void 2041void
1355element_size (OpenCL::Image self) 2042element_size (OpenCL::Image self)
1425 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2112 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1426 EXTEND (SP, 1); 2113 EXTEND (SP, 1);
1427 const int i = 0; 2114 const int i = 0;
1428 { 2115 {
1429 NEED_SUCCESS (RetainContext, (value [i])); 2116 NEED_SUCCESS (RetainContext, (value [i]));
1430 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2117 PUSH_CLOBJ (stash_context, value [i]);
1431 } 2118 }
1432 2119
1433void 2120void
1434normalized_coords (OpenCL::Sampler self) 2121normalized_coords (OpenCL::Sampler self)
1435 PPCODE: 2122 PPCODE:
1436 cl_addressing_mode value [1]; 2123 cl_addressing_mode value [1];
1437 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2124 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1438 EXTEND (SP, 1); 2125 EXTEND (SP, 1);
1439 const int i = 0; 2126 const int i = 0;
1440 PUSHs (sv_2mortal (newSViv (value [i]))); 2127 PUSHs (sv_2mortal (newSVuv (value [i])));
1441 2128
1442void 2129void
1443addressing_mode (OpenCL::Sampler self) 2130addressing_mode (OpenCL::Sampler self)
1444 PPCODE: 2131 PPCODE:
1445 cl_filter_mode value [1]; 2132 cl_filter_mode value [1];
1446 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2133 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1447 EXTEND (SP, 1); 2134 EXTEND (SP, 1);
1448 const int i = 0; 2135 const int i = 0;
1449 PUSHs (sv_2mortal (newSViv (value [i]))); 2136 PUSHs (sv_2mortal (newSVuv (value [i])));
1450 2137
1451void 2138void
1452filter_mode (OpenCL::Sampler self) 2139filter_mode (OpenCL::Sampler self)
1453 PPCODE: 2140 PPCODE:
1454 cl_bool value [1]; 2141 cl_bool value [1];
1465DESTROY (OpenCL::Program self) 2152DESTROY (OpenCL::Program self)
1466 CODE: 2153 CODE:
1467 clReleaseProgram (self); 2154 clReleaseProgram (self);
1468 2155
1469void 2156void
1470build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS:
2159 build_async = 1
1471 CODE: 2160 CODE:
2161 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2162 void *user_data = 0;
2163 cl_uint num_devices = 0;
2164 cl_device_id *device_list = 0;
2165
2166 if (SvOK (devices))
2167 {
2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2170
2171 AV *av = (AV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173
2174 if (num_devices)
2175 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2177 int count;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 }
2181 }
2182
2183 if (SvOK (notify))
2184 {
2185 NEED_SUCCESS (RetainProgram, (self));
2186 pfn_notify = eq_program_notify;
2187 user_data = SvREFCNT_inc (s_get_cv (notify));
2188 }
2189
2190 if (ix)
2191 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2192 else
1472 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2193 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1473 2194
1474void 2195void
1475build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1476 PPCODE: 2197 PPCODE:
1477 size_t size; 2198 size_t size;
1512 2233
1513void 2234void
1514kernel (OpenCL::Program program, SV *function) 2235kernel (OpenCL::Program program, SV *function)
1515 PPCODE: 2236 PPCODE:
1516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2237 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1517 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2238 XPUSH_CLOBJ (stash_kernel, kernel);
2239
2240void
2241kernels_in_program (OpenCL::Program program)
2242 PPCODE:
2243 cl_uint num_kernels;
2244 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2245 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2246 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2247
2248 int i;
2249 EXTEND (SP, num_kernels);
2250 for (i = 0; i < num_kernels; ++i)
2251 PUSH_CLOBJ (stash_kernel, kernels [i]);
1518 2252
1519void 2253void
1520info (OpenCL::Program self, cl_program_info name) 2254info (OpenCL::Program self, cl_program_info name)
1521 PPCODE: 2255 PPCODE:
1522 INFO (Program) 2256 INFO (Program)
1540 { 2274 {
1541 SV *sv = sv_2mortal (newSV (sizes [i])); 2275 SV *sv = sv_2mortal (newSV (sizes [i]));
1542 SvUPGRADE (sv, SVt_PV); 2276 SvUPGRADE (sv, SVt_PV);
1543 SvPOK_only (sv); 2277 SvPOK_only (sv);
1544 SvCUR_set (sv, sizes [i]); 2278 SvCUR_set (sv, sizes [i]);
1545 ptrs [i] = SvPVX (sv); 2279 ptrs [i] = (void *)SvPVX (sv);
1546 PUSHs (sv); 2280 PUSHs (sv);
1547 } 2281 }
1548 2282
1549 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2283 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1550 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2284 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1570 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2304 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1571 EXTEND (SP, 1); 2305 EXTEND (SP, 1);
1572 const int i = 0; 2306 const int i = 0;
1573 { 2307 {
1574 NEED_SUCCESS (RetainContext, (value [i])); 2308 NEED_SUCCESS (RetainContext, (value [i]));
1575 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2309 PUSH_CLOBJ (stash_context, value [i]);
1576 } 2310 }
1577 2311
1578void 2312void
1579devices (OpenCL::Program self) 2313devices (OpenCL::Program self)
1580 PPCODE: 2314 PPCODE:
1584 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2318 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1585 int i, n = size / sizeof (*value); 2319 int i, n = size / sizeof (*value);
1586 EXTEND (SP, n); 2320 EXTEND (SP, n);
1587 for (i = 0; i < n; ++i) 2321 for (i = 0; i < n; ++i)
1588 { 2322 {
1589 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2323 PUSH_CLOBJ (stash_device, value [i]);
1590 } 2324 }
1591 2325
1592void 2326void
1593source (OpenCL::Program self) 2327source (OpenCL::Program self)
1594 PPCODE: 2328 PPCODE:
1620DESTROY (OpenCL::Kernel self) 2354DESTROY (OpenCL::Kernel self)
1621 CODE: 2355 CODE:
1622 clReleaseKernel (self); 2356 clReleaseKernel (self);
1623 2357
1624void 2358void
2359setf (OpenCL::Kernel self, const char *format, ...)
2360 CODE:
2361 int i;
2362 for (i = 2; ; ++i)
2363 {
2364 while (*format == ' ')
2365 ++format;
2366
2367 char type = *format++;
2368
2369 if (!type)
2370 break;
2371
2372 if (i >= items)
2373 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2374
2375 SV *sv = ST (i);
2376
2377 union
2378 {
2379 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2380 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2381 cl_half ch; cl_float cf; cl_double cd;
2382 cl_mem cm;
2383 cl_sampler ca;
2384 size_t cz;
2385 cl_event ce;
2386 } arg;
2387 size_t size;
2388 int nullarg = 0;
2389
2390 switch (type)
2391 {
2392 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2393 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2394 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2395 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2396 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2397 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2398 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2399 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2400
2401 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404
2405 case 'z': nullarg = 1; size = SvIV (sv); break;
2406
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2408 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2409 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410
2411 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 }
2414
2415 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2416 if (res)
2417 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2418 }
2419
2420 if (i != items)
2421 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2422
2423void
1625set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2424set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1626 CODE: 2425 CODE:
1627 clSetKernelArg (self, idx, sizeof (value), &value); 2426 clSetKernelArg (self, idx, sizeof (value), &value);
1628 2427
1629void 2428void
1677 clSetKernelArg (self, idx, sizeof (value), &value); 2476 clSetKernelArg (self, idx, sizeof (value), &value);
1678 2477
1679void 2478void
1680set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2479set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1681 CODE: 2480 CODE:
1682 clSetKernelArg (self, idx, sizeof (value), &value); 2481 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1683 2482
1684void 2483void
1685set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2484set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1686 CODE: 2485 CODE:
1687 clSetKernelArg (self, idx, sizeof (value), &value); 2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1688 2487
1689void 2488void
1690set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
1691 CODE: 2493 CODE:
1692 clSetKernelArg (self, idx, sizeof (value), &value); 2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1693
1694void
1695set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1696 CODE:
1697 clSetKernelArg (self, idx, sizeof (value), &value);
1698 2495
1699void 2496void
1700set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1701 CODE: 2498 CODE:
1702 clSetKernelArg (self, idx, sizeof (value), &value); 2499 clSetKernelArg (self, idx, sizeof (value), &value);
1748 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2545 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2546 EXTEND (SP, 1);
1750 const int i = 0; 2547 const int i = 0;
1751 { 2548 {
1752 NEED_SUCCESS (RetainContext, (value [i])); 2549 NEED_SUCCESS (RetainContext, (value [i]));
1753 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2550 PUSH_CLOBJ (stash_context, value [i]);
1754 } 2551 }
1755 2552
1756void 2553void
1757program (OpenCL::Kernel self) 2554program (OpenCL::Kernel self)
1758 PPCODE: 2555 PPCODE:
1760 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2557 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1761 EXTEND (SP, 1); 2558 EXTEND (SP, 1);
1762 const int i = 0; 2559 const int i = 0;
1763 { 2560 {
1764 NEED_SUCCESS (RetainProgram, (value [i])); 2561 NEED_SUCCESS (RetainProgram, (value [i]));
1765 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2562 PUSH_CLOBJ (stash_program, value [i]);
1766 } 2563 }
1767 2564
1768#END:kernel 2565#END:kernel
1769 2566
1770void 2567void
1830wait (OpenCL::Event self) 2627wait (OpenCL::Event self)
1831 CODE: 2628 CODE:
1832 clWaitForEvents (1, &self); 2629 clWaitForEvents (1, &self);
1833 2630
1834void 2631void
2632cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2633 CODE:
2634 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2635
2636void
1835info (OpenCL::Event self, cl_event_info name) 2637info (OpenCL::Event self, cl_event_info name)
1836 PPCODE: 2638 PPCODE:
1837 INFO (Event) 2639 INFO (Event)
1838 2640
1839#BEGIN:event 2641#BEGIN:event
1845 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1846 EXTEND (SP, 1); 2648 EXTEND (SP, 1);
1847 const int i = 0; 2649 const int i = 0;
1848 { 2650 {
1849 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2651 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1850 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2652 PUSH_CLOBJ (stash_queue, value [i]);
1851 } 2653 }
1852 2654
1853void 2655void
1854command_type (OpenCL::Event self) 2656command_type (OpenCL::Event self)
1855 PPCODE: 2657 PPCODE:
1878 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2680 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1879 EXTEND (SP, 1); 2681 EXTEND (SP, 1);
1880 const int i = 0; 2682 const int i = 0;
1881 { 2683 {
1882 NEED_SUCCESS (RetainContext, (value [i])); 2684 NEED_SUCCESS (RetainContext, (value [i]));
1883 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2685 PUSH_CLOBJ (stash_context, value [i]);
1884 } 2686 }
1885 2687
1886#END:event 2688#END:event
1887 2689
1888void 2690void
1913void 2715void
1914set_status (OpenCL::UserEvent self, cl_int execution_status) 2716set_status (OpenCL::UserEvent self, cl_int execution_status)
1915 CODE: 2717 CODE:
1916 clSetUserEventStatus (self, execution_status); 2718 clSetUserEventStatus (self, execution_status);
1917 2719
2720MODULE = OpenCL PACKAGE = OpenCL::Mapped
2721
2722void
2723DESTROY (SV *self)
2724 CODE:
2725 OpenCL__Mapped mapped = SvMAPPED (self);
2726
2727 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2728 mapped_detach (self, mapped);
2729
2730 clReleaseCommandQueue (mapped->queue);
2731 clReleaseEvent (mapped->event);
2732 Safefree (mapped);
2733
2734void
2735unmap (OpenCL::Mapped self, ...)
2736 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2738
2739bool
2740mapped (OpenCL::Mapped self)
2741 CODE:
2742 RETVAL = !!self->ptr;
2743 OUTPUT:
2744 RETVAL
2745
2746void
2747wait (OpenCL::Mapped self)
2748 PPCODE:
2749 if (self->event)
2750 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2751
2752void
2753event (OpenCL::Mapped self)
2754 PPCODE:
2755 if (!self->event)
2756 XSRETURN_UNDEF;
2757
2758 clRetainEvent (self->event);
2759 XPUSH_CLOBJ (stash_event, self->event);
2760
2761size_t
2762size (OpenCL::Mapped self)
2763 CODE:
2764 RETVAL = self->cb;
2765 OUTPUT:
2766 RETVAL
2767
2768IV
2769ptr (OpenCL::Mapped self)
2770 CODE:
2771 RETVAL = PTR2IV (self->ptr);
2772 OUTPUT:
2773 RETVAL
2774
2775MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2776
2777MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2778
2779IV
2780row_pitch (OpenCL::Mapped self)
2781 ALIAS:
2782 slice_pitch = 1
2783 CODE:
2784 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2785 OUTPUT:
2786 RETVAL
2787
2788
2789

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines