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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.53 by root, Wed Apr 25 00:02:25 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
4 6
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 8#include "xthread.h"
7#include "schmorp.h" 9#include "schmorp.h"
8 10
38 #define PREFER_1_1 1 40 #define PREFER_1_1 1
39#endif 41#endif
40 42
41typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
61 60
61typedef struct mapped * OpenCL__Mapped;
62
62typedef SV *FUTURE; 63typedef SV *FUTURE;
64
65static HV
66 *stash_platform,
67 *stash_device,
68 *stash_subdevice,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
63 89
64/*****************************************************************************/ 90/*****************************************************************************/
65 91
66// name must include a leading underscore 92// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 133}
108 134
109/*****************************************************************************/ 135/*****************************************************************************/
110 136
111/* up to two temporary buffers */ 137/* up to two temporary buffers */
112static void * 138static void * ecb_noinline
113tmpbuf (size_t size) 139tmpbuf (size_t size)
114{ 140{
115 enum { buffers = 3 }; 141 enum { buffers = 4 };
116 static int idx; 142 static int idx;
117 static void *buf [buffers]; 143 static void *buf [buffers];
118 static size_t len [buffers]; 144 static size_t len [buffers];
119 145
120 idx = (idx + 1) % buffers; 146 idx = (idx + 1) % buffers;
125 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
126 buf [idx] = malloc (len [idx]); 152 buf [idx] = malloc (len [idx]);
127 } 153 }
128 154
129 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;
130} 171}
131 172
132/*****************************************************************************/ 173/*****************************************************************************/
133 174
134typedef struct 175typedef struct
193 if (res) \ 234 if (res) \
194 FAIL (name); 235 FAIL (name);
195 236
196/*****************************************************************************/ 237/*****************************************************************************/
197 238
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
251SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
252{
253 // sv_derived_from is quite slow :(
254 if (SvROK (sv) && sv_derived_from (sv, pkg))
255 return (void *)SvIV (SvRV (sv));
256
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));
268}
269
270/*****************************************************************************/
271
198static cl_context_properties * 272static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (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)
200{ 274{
201 if (!sv || !SvOK (sv)) 275 if (!sv || !SvOK (sv))
202 if (extra) 276 if (extra)
203 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
210 int i, len = av_len (av) + 1; 284 int i, len = av_len (av) + 1;
211 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));
212 cl_context_properties *l = p; 286 cl_context_properties *l = p;
213 287
214 if (len & 1) 288 if (len & 1)
215 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);
216 290
217 while (extracount--) 291 while (extracount--)
218 *l++ = *extra++; 292 *l++ = *extra++;
219 293
220 for (i = 0; i < len; i += 2) 294 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 297 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 298 cl_context_properties v = SvIV (p_sv); // code below can override
225 299
226 switch (t) 300 switch (t)
227 { 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
228 case CL_GLX_DISPLAY_KHR: 307 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 308 if (!SvOK (p_sv))
230 { 309 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 310 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 311 if (func)
258 } 337 }
259 338
260 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);
261} 340}
262 341
263/*****************************************************************************/ 342static void * ecb_noinline
264 343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr))
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{ 344{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 345 const char *funcname = cv_get_name (cv);
275 return (void *)SvIV (SvRV (sv));
276 346
277 croak ("%s: %s is not of type %s", func, svname, pkg); 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;
278} 369}
279 370
280/*****************************************************************************/ 371/*****************************************************************************/
281/* callback stuff */ 372/* callback stuff */
282 373
304static void (*eq_signal_func)(void *signal_arg, int value); 395static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 396static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 397static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 398static eq_item *eq_head, *eq_tail;
308 399
309static void 400static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 401eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 402{
312 eq_item *item = malloc (sizeof (eq_item)); 403 eq_item *item = malloc (sizeof (eq_item));
313 404
314 item->next = 0; 405 item->next = 0;
381} 472}
382 473
383/*****************************************************************************/ 474/*****************************************************************************/
384/* context notify */ 475/* context notify */
385 476
386static void 477static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 478eq_context_push (void *data1, void *data2, void *data3)
388{ 479{
389 dSP; 480 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 481 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 482 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 504 { \
414 pfn_notify = eq_context_notify; \ 505 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 506 user_data = s_get_cv (notify); \
416 } 507 }
417 508
418static SV * 509static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 510new_clobj_context (cl_context ctx, void *user_data)
420{ 511{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 512 SV *sv = new_clobj (stash_context, (IV)ctx);
422 513
423 if (user_data) 514 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 515 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 516
426 return sv; 517 return sv;
433 524
434static void 525static void
435eq_program_push (void *data1, void *data2, void *data3) 526eq_program_push (void *data1, void *data2, void *data3)
436{ 527{
437 dSP; 528 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 529 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 530 PUTBACK;
440} 531}
441 532
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 533static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 534
444static void CL_CALLBACK 535static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 536eq_program_notify (cl_program program, void *user_data)
446{ 537{
538 clRetainProgram (program);
539
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 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 }
448} 558}
449 559
450struct build_args 560struct build_args
451{ 561{
452 cl_program program; 562 cl_program program;
466 else 576 else
467 clReleaseProgram (arg->program); 577 clReleaseProgram (arg->program);
468 578
469 free (arg->options); 579 free (arg->options);
470 free (arg); 580 free (arg);
581
582 return 0;
471} 583}
472 584
473static void 585static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 586build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 587{
490 602
491static void 603static void
492eq_event_push (void *data1, void *data2, void *data3) 604eq_event_push (void *data1, void *data2, void *data3)
493{ 605{
494 dSP; 606 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 607 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 608 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 609 PUTBACK;
498} 610}
499 611
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 612static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 617 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 618 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 619}
508 620
509/*****************************************************************************/ 621/*****************************************************************************/
622/* utilities for XS code */
510 623
511static size_t 624static size_t
512img_row_pitch (cl_mem img) 625img_row_pitch (cl_mem img)
513{ 626{
514 size_t res; 627 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 628 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 629 return res;
517} 630}
518 631
519static cl_event * 632static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 633event_list (SV **items, cl_uint *rcount, cl_event extra)
521{ 634{
522 cl_uint count = *rcount; 635 cl_uint count = *rcount;
523 636
524 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)
525 return 0; 641 return 0;
526 642
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 643 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 644 int i = 0;
529 645
530 do 646 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 647 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 648 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
535 } 649
536 while (count); 650 if (extra)
651 list [i++] = extra;
537 652
538 *rcount = i; 653 *rcount = i;
539 654
540 return i ? list : 0; 655 return i ? list : 0;
541} 656}
542 657
543#define EVENT_LIST(items,count) \ 658#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 659 cl_uint event_list_count = items - (skip); \
545 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)
546 661
547#define INFO(class) \ 662#define INFO(class) \
548{ \ 663{ \
549 size_t size; \ 664 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 665 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
553 SvPOK_only (sv); \ 668 SvPOK_only (sv); \
554 SvCUR_set (sv, size); \ 669 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 670 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 671 XPUSHs (sv); \
557} 672}
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/*****************************************************************************/
558 766
559MODULE = OpenCL PACKAGE = OpenCL 767MODULE = OpenCL PACKAGE = OpenCL
560 768
561PROTOTYPES: ENABLE 769PROTOTYPES: ENABLE
562 770
590 }; 798 };
591 799
592 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--)
593 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 801 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
594 802
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
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 827 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 828}
597 829
598cl_int 830cl_int
599errno () 831errno ()
601 RETVAL = res; 833 RETVAL = res;
602 OUTPUT: 834 OUTPUT:
603 RETVAL 835 RETVAL
604 836
605const char * 837const char *
606err2str (cl_int err) 838err2str (cl_int err = res)
607 839
608const char * 840const char *
609enum2str (cl_uint value) 841enum2str (cl_uint value)
610 842
611void 843void
619 list = tmpbuf (sizeof (*list) * count); 851 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 852 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 853
622 EXTEND (SP, count); 854 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 855 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 856 PUSH_CLOBJ (stash_platform, list [i]);
625 857
626void 858void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 859context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 860 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 861 CONTEXT_NOTIFY_CALLBACK;
630 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));
631 XPUSH_CLOBJ_CONTEXT; 863 XPUSH_CLOBJ_CONTEXT;
632 864
633void 865void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 866context (FUTURE properties, FUTURE devices, FUTURE notify)
635 PPCODE: 867 PPCODE:
636 /* der Gipfel der Kunst */ 868 /* der Gipfel der Kunst */
637 869
638void 870void
639wait_for_events (...) 871wait_for_events (...)
640 CODE: 872 CODE:
641 EVENT_LIST (0, items); 873 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 874 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 875
644PROTOTYPES: DISABLE 876PROTOTYPES: DISABLE
645 877
646MODULE = OpenCL PACKAGE = OpenCL::Platform 878MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 921 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 922 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 923
692 EXTEND (SP, count); 924 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 925 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 926 PUSH_CLOBJ (stash_device, list [i]);
695 927
696void 928void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 929context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 930 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 931 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
700 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);
701 933
702 AV *av = (AV *)SvRV (devices); 934 cl_uint device_count = 0;
703 cl_uint num_devices = av_len (av) + 1; 935 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
705
706 int i;
707 for (i = num_devices; i--; )
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
709 936
710 CONTEXT_NOTIFY_CALLBACK; 937 CONTEXT_NOTIFY_CALLBACK;
711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res)); 938 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
712 XPUSH_CLOBJ_CONTEXT; 939 XPUSH_CLOBJ_CONTEXT;
713 940
714void 941void
715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 942context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
716 PPCODE: 943 PPCODE:
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 944 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
718 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);
719 946
720 CONTEXT_NOTIFY_CALLBACK; 947 CONTEXT_NOTIFY_CALLBACK;
721 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));
722 XPUSH_CLOBJ_CONTEXT; 949 XPUSH_CLOBJ_CONTEXT;
723 950
724MODULE = OpenCL PACKAGE = OpenCL::Device 951MODULE = OpenCL PACKAGE = OpenCL::Device
725 952
726void 953void
727info (OpenCL::Device self, cl_device_info name) 954info (OpenCL::Device self, cl_device_info name)
728 PPCODE: 955 PPCODE:
729 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
730 987
731#BEGIN:device 988#BEGIN:device
732 989
733void 990void
734type (OpenCL::Device self) 991type (OpenCL::Device self)
735 PPCODE: 992 PPCODE:
736 cl_device_type value [1]; 993 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 994 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 995 EXTEND (SP, 1);
739 const int i = 0; 996 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 997 PUSHs (sv_2mortal (newSVuv (value [i])));
741 998
742void 999void
743vendor_id (OpenCL::Device self) 1000vendor_id (OpenCL::Device self)
744 ALIAS: 1001 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1002 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1147 PPCODE:
891 cl_command_queue_properties value [1]; 1148 cl_command_queue_properties value [1];
892 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));
893 EXTEND (SP, 1); 1150 EXTEND (SP, 1);
894 const int i = 0; 1151 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1152 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1153
897void 1154void
898platform (OpenCL::Device self) 1155platform (OpenCL::Device self)
899 PPCODE: 1156 PPCODE:
900 cl_platform_id value [1]; 1157 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1158 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1159 EXTEND (SP, 1);
903 const int i = 0; 1160 const int i = 0;
904 { 1161 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1162 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1163 }
907 1164
908void 1165void
909name (OpenCL::Device self) 1166name (OpenCL::Device self)
910 ALIAS: 1167 ALIAS:
929 cl_device_id value [1]; 1186 cl_device_id value [1];
930 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));
931 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
932 const int i = 0; 1189 const int i = 0;
933 { 1190 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1191 PUSH_CLOBJ (stash_device, value [i]);
935 } 1192 }
936 1193
937void 1194void
938partition_types_ext (OpenCL::Device self) 1195partition_types_ext (OpenCL::Device self)
939 ALIAS: 1196 ALIAS:
950 for (i = 0; i < n; ++i) 1207 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1208 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1209
953#END:device 1210#END:device
954 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
955MODULE = OpenCL PACKAGE = OpenCL::Context 1223MODULE = OpenCL PACKAGE = OpenCL::Context
956 1224
957void 1225void
958DESTROY (OpenCL::Context context) 1226DESTROY (OpenCL::Context self)
959 CODE: 1227 CODE:
960 clReleaseContext (context); 1228 clReleaseContext (self);
961 1229
962void 1230void
963info (OpenCL::Context self, cl_context_info name) 1231info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1232 PPCODE:
965 INFO (Context) 1233 INFO (Context)
966 1234
967void 1235void
968queue (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)
969 PPCODE: 1237 PPCODE:
970 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));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1239 XPUSH_CLOBJ (stash_queue, queue);
972 1240
973void 1241void
974user_event (OpenCL::Context self) 1242user_event (OpenCL::Context self)
975 PPCODE: 1243 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1244 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1245 XPUSH_CLOBJ (stash_userevent, ev);
978 1246
979void 1247void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1248buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1249 PPCODE:
982 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))
983 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?");
984 1252
985 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));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1254 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1255
988void 1256void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1257buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1258 PPCODE:
991 STRLEN len; 1259 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1260 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 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)))
994 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?");
995 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));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1264 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1265
998#if CL_VERSION_1_2 1266#if CL_VERSION_1_2
999 1267
1000void 1268void
1001image (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)
1002 PPCODE: 1270 PPCODE:
1003 STRLEN len; 1271 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1272 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1273 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1274 const cl_image_desc desc = {
1009 array_size, row_pitch, slice_pitch, 1277 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1278 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("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
1012 }; 1280 };
1013 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));
1014 char *klass = "OpenCL::Image"; 1282 HV *stash = stash_image;
1015 switch (type) 1283 switch (type)
1016 { 1284 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1285 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1286 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1287 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1288 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1289 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1290 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1291 }
1024 XPUSH_CLOBJ (klass, mem); 1292 XPUSH_CLOBJ (stash, mem);
1025 1293
1026#endif 1294#endif
1027 1295
1028void 1296void
1029image2d (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)
1035 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));
1036#else 1304#else
1037 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 };
1038 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));
1039#endif 1307#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1308 XPUSH_CLOBJ (stash_image2d, mem);
1041 1309
1042void 1310void
1043image3d (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)
1044 PPCODE: 1312 PPCODE:
1045 STRLEN len; 1313 STRLEN len;
1049 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));
1050#else 1318#else
1051 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 };
1052 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));
1053#endif 1321#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1322 XPUSH_CLOBJ (stash_image3d, mem);
1055 1323
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1324#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1325
1058void 1326void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1327gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1328 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1329 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1330 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1331
1064void 1332void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1333gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1334 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1336 XPUSH_CLOBJ (stash_image2d, mem);
1069 1337
1070#if CL_VERSION_1_2 1338#if CL_VERSION_1_2
1071 1339
1072void 1340void
1073gl_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)
1074 ALIAS: 1342 ALIAS:
1075 PPCODE: 1343 PPCODE:
1076 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));
1077 cl_gl_object_type type; 1345 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1346 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1347 HV *stash = stash_memory;
1080 switch (type) 1348 switch (type)
1081 { 1349 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1350 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1351 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1352 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1353 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1354 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1355 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1356 }
1089 XPUSH_CLOBJ (klass, mem); 1357 XPUSH_CLOBJ (stash, mem);
1090 1358
1091#endif 1359#endif
1092 1360
1093void 1361void
1094gl_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)
1096#if PREFER_1_1 1364#if PREFER_1_1
1097 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));
1098#else 1366#else
1099 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));
1100#endif 1368#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1369 XPUSH_CLOBJ (stash_image2d, mem);
1102 1370
1103void 1371void
1104gl_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)
1105 PPCODE: 1373 PPCODE:
1106#if PREFER_1_1 1374#if PREFER_1_1
1107 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));
1108#else 1376#else
1109 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));
1110#endif 1378#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1379 XPUSH_CLOBJ (stash_image3d, mem);
1112 1380
1113#endif 1381#endif
1114 1382
1115void 1383void
1116supported_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)
1136 1404
1137void 1405void
1138sampler (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)
1139 PPCODE: 1407 PPCODE:
1140 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));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1409 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1410
1143void 1411void
1144program_with_source (OpenCL::Context self, SV *program) 1412program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1413 PPCODE:
1146 STRLEN len; 1414 STRLEN len;
1147 size_t len2; 1415 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1416 const char *ptr = SvPVbyte (program, len);
1149 1417
1150 len2 = len; 1418 len2 = len;
1151 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));
1152 XPUSH_CLOBJ ("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
1153 1497
1154#BEGIN:context 1498#BEGIN:context
1155 1499
1156void 1500void
1157reference_count (OpenCL::Context self) 1501reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1518 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1519 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1520 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1521 for (i = 0; i < n; ++i)
1178 { 1522 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1523 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1524 }
1181 1525
1182void 1526void
1183properties (OpenCL::Context self) 1527properties (OpenCL::Context self)
1184 PPCODE: 1528 PPCODE:
1199DESTROY (OpenCL::Queue self) 1543DESTROY (OpenCL::Queue self)
1200 CODE: 1544 CODE:
1201 clReleaseCommandQueue (self); 1545 clReleaseCommandQueue (self);
1202 1546
1203void 1547void
1204enqueue_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
1205 PPCODE: 1551 PPCODE:
1206 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1552 EVENT_LIST (6);
1208 1553
1209 SvUPGRADE (data, SVt_PV); 1554 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1555 SvGROW (data, len);
1211 SvPOK_only (data); 1556 SvPOK_only (data);
1212 SvCUR_set (data, len); 1557 SvCUR_set (data, len);
1558
1559 cl_event ev = 0;
1213 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));
1214 1561
1215 if (ev) 1562 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1563 XPUSH_CLOBJ (stash_event, ev);
1217 1564
1218void 1565void
1219enqueue_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
1220 PPCODE: 1569 PPCODE:
1221 cl_event ev = 0; 1570 EVENT_LIST (5);
1571
1222 STRLEN len; 1572 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1573 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5);
1225 1574
1575 cl_event ev = 0;
1226 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));
1227 1577
1228 if (ev) 1578 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
1230 1580
1231#if CL_VERSION_1_2 1581#if CL_VERSION_1_2
1232 1582
1233void 1583void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1584fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1585 ALIAS:
1586 enqueue_fill_buffer = 0
1235 PPCODE: 1587 PPCODE:
1236 cl_event ev = 0; 1588 EVENT_LIST (5);
1589
1237 STRLEN len; 1590 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1591 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240 1592
1593 cl_event ev = 0;
1241 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1594 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1595
1243 if (ev) 1596 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1597 XPUSH_CLOBJ (stash_event, ev);
1245 1598
1246void 1599void
1247enqueue_fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 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
1248 PPCODE: 1603 PPCODE:
1249 cl_event ev = 0; 1604 EVENT_LIST (12);
1250 STRLEN len; 1605
1251 const size_t origin [3] = { x, y, z }; 1606 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1607 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254 1608
1255 const cl_float c_f [4] = { r, g, b, a }; 1609 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1610 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1611 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1612 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1615 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1616 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1263 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1617 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1264 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1618 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1619
1620 cl_event ev = 0;
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1621 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1267 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1622 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1623
1269 if (ev) 1624 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1625 XPUSH_CLOBJ (stash_event, ev);
1271 1626
1272#endif 1627#endif
1273 1628
1274void 1629void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 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
1276 PPCODE: 1633 PPCODE:
1634 EVENT_LIST (6);
1635
1277 cl_event ev = 0; 1636 cl_event ev = 0;
1278 EVENT_LIST (6, items - 6);
1279
1280 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));
1281 1638
1282 if (ev) 1639 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1640 XPUSH_CLOBJ (stash_event, ev);
1284 1641
1285void 1642void
1286enqueue_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
1287 PPCODE: 1646 PPCODE:
1288 cl_event ev = 0; 1647 EVENT_LIST (17);
1648
1289 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 };
1290 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 };
1291 const size_t region[3] = { width, height, depth }; 1651 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17);
1293 1652
1294 if (!buf_row_pitch) 1653 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1654 buf_row_pitch = region [0];
1296 1655
1297 if (!buf_slice_pitch) 1656 if (!buf_slice_pitch)
1307 1666
1308 SvUPGRADE (data, SVt_PV); 1667 SvUPGRADE (data, SVt_PV);
1309 SvGROW (data, len); 1668 SvGROW (data, len);
1310 SvPOK_only (data); 1669 SvPOK_only (data);
1311 SvCUR_set (data, len); 1670 SvCUR_set (data, len);
1671
1672 cl_event ev = 0;
1312 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));
1313 1674
1314 if (ev) 1675 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1676 XPUSH_CLOBJ (stash_event, ev);
1316 1677
1317void 1678void
1318enqueue_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
1319 PPCODE: 1682 PPCODE:
1320 cl_event ev = 0; 1683 EVENT_LIST (17);
1684
1321 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 };
1322 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 };
1323 const size_t region[3] = { width, height, depth }; 1687 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1688 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1689 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17);
1327 1690
1328 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
1330 1693
1331 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
1340 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];
1341 1704
1342 if (len < min_len) 1705 if (len < min_len)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1706 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1707
1708 cl_event ev = 0;
1345 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));
1346 1710
1347 if (ev) 1711 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1712 XPUSH_CLOBJ (stash_event, ev);
1349 1713
1350void 1714void
1351enqueue_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
1352 PPCODE: 1718 PPCODE:
1353 cl_event ev = 0; 1719 EVENT_LIST (16);
1720
1354 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 };
1355 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 };
1356 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16);
1358 1724
1725 cl_event ev = 0;
1359 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));
1360 1727
1361 if (ev) 1728 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
1363 1730
1364void 1731void
1365enqueue_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
1366 PPCODE: 1735 PPCODE:
1367 cl_event ev = 0; 1736 EVENT_LIST (12);
1737
1368 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 };
1369 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12);
1371 1740
1372 if (!row_pitch) 1741 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1742 row_pitch = img_row_pitch (src);
1374 1743
1375 if (depth > 1 && !slice_pitch) 1744 if (depth > 1 && !slice_pitch)
1379 1748
1380 SvUPGRADE (data, SVt_PV); 1749 SvUPGRADE (data, SVt_PV);
1381 SvGROW (data, len); 1750 SvGROW (data, len);
1382 SvPOK_only (data); 1751 SvPOK_only (data);
1383 SvCUR_set (data, len); 1752 SvCUR_set (data, len);
1753
1754 cl_event ev = 0;
1384 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));
1385 1756
1386 if (ev) 1757 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1758 XPUSH_CLOBJ (stash_event, ev);
1388 1759
1389void 1760void
1390enqueue_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
1391 PPCODE: 1764 PPCODE:
1392 cl_event ev = 0; 1765 EVENT_LIST (12);
1766
1393 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 };
1394 const size_t region[3] = { width, height, depth }; 1768 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1769 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1770 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12);
1398 1771
1399 if (!row_pitch) 1772 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1773 row_pitch = img_row_pitch (dst);
1401 1774
1402 if (depth > 1 && !slice_pitch) 1775 if (depth > 1 && !slice_pitch)
1405 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;
1406 1779
1407 if (len < min_len) 1780 if (len < min_len)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1781 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1782
1783 cl_event ev = 0;
1410 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));
1411 1785
1412 if (ev) 1786 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1787 XPUSH_CLOBJ (stash_event, ev);
1414 1788
1415void 1789void
1416enqueue_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
1417 PPCODE: 1793 PPCODE:
1418 cl_event ev = 0; 1794 EVENT_LIST (12);
1795
1419 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 };
1420 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 };
1421 const size_t region[3] = { width, height, depth }; 1798 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12);
1423 1799
1800 cl_event ev = 0;
1424 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));
1425 1802
1426 if (ev) 1803 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1804 XPUSH_CLOBJ (stash_event, ev);
1428 1805
1429void 1806void
1430enqueue_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
1431 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
1432 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
1433 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 };
1434 const size_t region[3] = { width, height, depth }; 1888 const size_t region[3] = { width, height, depth };
1889 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1890 EVENT_LIST (10);
1436 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;
1437 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));
1438 1920
1439 if (ev) 1921 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1922 XPUSH_CLOBJ (stash_event, ev);
1441 1923
1442void 1924void
1443enqueue_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, ...)
1444 PPCODE:
1445 cl_event ev = 0;
1446 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1447 const size_t region[3] = { width, height, depth };
1448 EVENT_LIST (10, items - 10);
1449
1450 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1451
1452 if (ev)
1453 XPUSH_CLOBJ ("OpenCL::Event", ev);
1454
1455void
1456enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1457 PPCODE:
1458 cl_event ev = 0;
1459 EVENT_LIST (2, items - 2);
1460
1461 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1462
1463 if (ev)
1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
1465
1466void
1467enqueue_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
1468 PPCODE: 1928 PPCODE:
1469 cl_event ev = 0; 1929 EVENT_LIST (5);
1930
1470 size_t *gwo = 0, *gws, *lws = 0; 1931 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1932 int gws_len;
1472 size_t *lists; 1933 size_t *lists;
1473 int i; 1934 int i;
1474 EVENT_LIST (5, items - 5);
1475 1935
1476 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)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1937 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1938
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1939 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1940
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1941 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1942
1483 gws = lists + gws_len * 0; 1943 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1944 for (i = 0; i < gws_len; ++i)
1945 {
1485 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 }
1486 1951
1487 if (SvOK (global_work_offset)) 1952 if (SvOK (global_work_offset))
1488 { 1953 {
1489 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)
1490 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");
1498 } 1963 }
1499 1964
1500 if (SvOK (local_work_size)) 1965 if (SvOK (local_work_size))
1501 { 1966 {
1502 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)
1503 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");
1504 1969
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1970 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1506 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");
1507 1972
1508 lws = lists + gws_len * 2; 1973 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 1974 for (i = 0; i < gws_len; ++i)
1975 {
1510 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 }
1511 } 1981 }
1512 1982
1983 cl_event ev = 0;
1513 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));
1514 1985
1515 if (ev) 1986 if (ev)
1516 XPUSH_CLOBJ ("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
1517 2008
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 2009#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 2010
1520void 2011void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2012acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 2013 ALIAS:
2014 release_gl_objects = 1
2015 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 2016 enqueue_release_gl_objects = 1
1524 PPCODE: 2017 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2018 EVENT_LIST (2);
1526 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);
1527 2022
1528 cl_event ev = 0; 2023 cl_event ev = 0;
1529 EVENT_LIST (2, items - 2);
1530 AV *av = (AV *)SvRV (objects);
1531 cl_uint num_objects = av_len (av) + 1;
1532 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1533 int i;
1534
1535 for (i = num_objects; i--; )
1536 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1537 2024
1538 if (ix) 2025 if (ix)
1539 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));
1540 else 2027 else
1541 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));
1542 2029
1543 if (ev) 2030 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 2031 XPUSH_CLOBJ (stash_event, ev);
1545 2032
1546#endif 2033#endif
1547 2034
1548void 2035void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 2036wait_for_events (OpenCL::Queue self, ...)
2037 ALIAS:
2038 enqueue_wait_for_events = 0
1550 CODE: 2039 CODE:
1551 EVENT_LIST (1, items - 1); 2040 EVENT_LIST (1);
1552#if PREFER_1_1 2041#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2042 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 2043#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2044 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 2045#endif
1557 2046
1558void 2047void
1559enqueue_marker (OpenCL::Queue self, ...) 2048marker (OpenCL::Queue self, ...)
2049 ALIAS:
2050 enqueue_marker = 0
1560 PPCODE: 2051 PPCODE:
2052 EVENT_LIST (1);
1561 cl_event ev = 0; 2053 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1 2054#if PREFER_1_1
1564 if (!event_list_count) 2055 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2056 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 2057 else
1567#if CL_VERSION_1_2 2058#if CL_VERSION_1_2
1574#endif 2065#endif
1575#else 2066#else
1576 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));
1577#endif 2068#endif
1578 if (ev) 2069 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 2070 XPUSH_CLOBJ (stash_event, ev);
1580 2071
1581void 2072void
1582enqueue_barrier (OpenCL::Queue self, ...) 2073barrier (OpenCL::Queue self, ...)
2074 ALIAS:
2075 enqueue_barrier = 0
1583 PPCODE: 2076 PPCODE:
2077 EVENT_LIST (1);
1584 cl_event ev = 0; 2078 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1 2079#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 2080 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 2081 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 2082 else
1590#if CL_VERSION_1_2 2083#if CL_VERSION_1_2
1600#endif 2093#endif
1601#else 2094#else
1602 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));
1603#endif 2096#endif
1604 if (ev) 2097 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 2098 XPUSH_CLOBJ (stash_event, ev);
1606 2099
1607void 2100void
1608flush (OpenCL::Queue self) 2101flush (OpenCL::Queue self)
1609 CODE: 2102 CODE:
1610 NEED_SUCCESS (Flush, (self)); 2103 NEED_SUCCESS (Flush, (self));
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 2122 EXTEND (SP, 1);
1630 const int i = 0; 2123 const int i = 0;
1631 { 2124 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 2125 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2126 PUSH_CLOBJ (stash_context, value [i]);
1634 } 2127 }
1635 2128
1636void 2129void
1637device (OpenCL::Queue self) 2130device (OpenCL::Queue self)
1638 PPCODE: 2131 PPCODE:
1639 cl_device_id value [1]; 2132 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2133 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 2134 EXTEND (SP, 1);
1642 const int i = 0; 2135 const int i = 0;
1643 { 2136 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2137 PUSH_CLOBJ (stash_device, value [i]);
1645 } 2138 }
1646 2139
1647void 2140void
1648reference_count (OpenCL::Queue self) 2141reference_count (OpenCL::Queue self)
1649 PPCODE: 2142 PPCODE:
1658 PPCODE: 2151 PPCODE:
1659 cl_command_queue_properties value [1]; 2152 cl_command_queue_properties value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2153 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 2154 EXTEND (SP, 1);
1662 const int i = 0; 2155 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 2156 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 2157
1665#END:command_queue 2158#END:command_queue
1666 2159
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 2160MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 2161
1683 PPCODE: 2176 PPCODE:
1684 cl_mem_object_type value [1]; 2177 cl_mem_object_type value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2178 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 2179 EXTEND (SP, 1);
1687 const int i = 0; 2180 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 2181 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 2182
1690void 2183void
1691flags (OpenCL::Memory self) 2184flags (OpenCL::Memory self)
1692 PPCODE: 2185 PPCODE:
1693 cl_mem_flags value [1]; 2186 cl_mem_flags value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2187 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2188 EXTEND (SP, 1);
1696 const int i = 0; 2189 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 2190 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 2191
1699void 2192void
1700size (OpenCL::Memory self) 2193size (OpenCL::Memory self)
1701 ALIAS: 2194 ALIAS:
1702 size = CL_MEM_SIZE 2195 size = CL_MEM_SIZE
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 2230 EXTEND (SP, 1);
1738 const int i = 0; 2231 const int i = 0;
1739 { 2232 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 2233 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2234 PUSH_CLOBJ (stash_context, value [i]);
1742 } 2235 }
1743 2236
1744void 2237void
1745associated_memobject (OpenCL::Memory self) 2238associated_memobject (OpenCL::Memory self)
1746 PPCODE: 2239 PPCODE:
1748 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));
1749 EXTEND (SP, 1); 2242 EXTEND (SP, 1);
1750 const int i = 0; 2243 const int i = 0;
1751 { 2244 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 2245 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2246 PUSH_CLOBJ (stash_memory, value [i]);
1754 } 2247 }
1755 2248
1756#END:mem 2249#END:mem
1757 2250
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 2251#if cl_apple_gl_sharing || cl_khr_gl_sharing
1778 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!");
1779 2272
1780 cl_buffer_region crdata = { origin, size }; 2273 cl_buffer_region crdata = { origin, size };
1781 2274
1782 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));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2276 XPUSH_CLOBJ (stash_buffer, mem);
1784 2277
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2278MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2279
1787void 2280void
1788image_info (OpenCL::Image self, cl_image_info name) 2281image_info (OpenCL::Image self, cl_image_info name)
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2367 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2368 EXTEND (SP, 1);
1876 const int i = 0; 2369 const int i = 0;
1877 { 2370 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2371 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2372 PUSH_CLOBJ (stash_context, value [i]);
1880 } 2373 }
1881 2374
1882void 2375void
1883normalized_coords (OpenCL::Sampler self) 2376normalized_coords (OpenCL::Sampler self)
1884 PPCODE: 2377 PPCODE:
1885 cl_addressing_mode value [1]; 2378 cl_addressing_mode value [1];
1886 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));
1887 EXTEND (SP, 1); 2380 EXTEND (SP, 1);
1888 const int i = 0; 2381 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i]))); 2382 PUSHs (sv_2mortal (newSVuv (value [i])));
1890 2383
1891void 2384void
1892addressing_mode (OpenCL::Sampler self) 2385addressing_mode (OpenCL::Sampler self)
1893 PPCODE: 2386 PPCODE:
1894 cl_filter_mode value [1]; 2387 cl_filter_mode value [1];
1895 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));
1896 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1897 const int i = 0; 2390 const int i = 0;
1898 PUSHs (sv_2mortal (newSViv (value [i]))); 2391 PUSHs (sv_2mortal (newSVuv (value [i])));
1899 2392
1900void 2393void
1901filter_mode (OpenCL::Sampler self) 2394filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2395 PPCODE:
1903 cl_bool value [1]; 2396 cl_bool value [1];
1918void 2411void
1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2412build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS: 2413 ALIAS:
1921 build_async = 1 2414 build_async = 1
1922 CODE: 2415 CODE:
1923 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1924 void *user_data = 0;
1925 cl_uint num_devices = 0; 2416 cl_uint device_count = 0;
1926 cl_device_id *device_list = 0; 2417 cl_device_id *device_list = 0;
1927 2418
1928 if (SvOK (devices)) 2419 if (SvOK (devices))
1929 { 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 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2447 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2448 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1932 2449
1933 AV *av = (AV *)SvRV (devices); 2450 HV *hv = (HV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935 2451
1936 if (num_devices) 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))
1937 { 2459 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2460 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2461 header_list [i] = SvCLOBJ ("clCompileProgram", "headers", HeVAL (he), "OpenCL::Program");
1939 int count; 2462 ++i;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 } 2463 }
1943 } 2464 }
1944 2465
1945 if (SvOK (notify)) 2466 void *user_data;
1946 { 2467 program_callback pfn_notify = make_program_callback (notify, &user_data);
1947 NEED_SUCCESS (RetainProgram, (self));
1948 pfn_notify = eq_program_notify;
1949 user_data = SvREFCNT_inc (s_get_cv (notify));
1950 }
1951 2468
1952 if (ix)
1953 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1954 else
1955 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 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
1956 2473
1957void 2474void
1958build_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)
1959 PPCODE: 2476 PPCODE:
1960 size_t size; 2477 size_t size;
1995 2512
1996void 2513void
1997kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2515 PPCODE:
1999 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));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2518
2002void 2519void
2003kernels_in_program (OpenCL::Program program) 2520kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2521 PPCODE:
2005 cl_uint num_kernels; 2522 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2526
2010 int i; 2527 int i;
2011 EXTEND (SP, num_kernels); 2528 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2529 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2531
2015void 2532void
2016info (OpenCL::Program self, cl_program_info name) 2533info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2534 PPCODE:
2018 INFO (Program) 2535 INFO (Program)
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
2068 const int i = 0; 2585 const int i = 0;
2069 { 2586 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2587 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2588 PUSH_CLOBJ (stash_context, value [i]);
2072 } 2589 }
2073 2590
2074void 2591void
2075devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
2076 PPCODE: 2593 PPCODE:
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2599 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
2084 { 2601 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2602 PUSH_CLOBJ (stash_device, value [i]);
2086 } 2603 }
2087 2604
2088void 2605void
2089source (OpenCL::Program self) 2606source (OpenCL::Program self)
2090 PPCODE: 2607 PPCODE:
2116DESTROY (OpenCL::Kernel self) 2633DESTROY (OpenCL::Kernel self)
2117 CODE: 2634 CODE:
2118 clReleaseKernel (self); 2635 clReleaseKernel (self);
2119 2636
2120void 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
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2703set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2704 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2705 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2706
2125void 2707void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2755 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2756
2175void 2757void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2758set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2759 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2760 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2761
2180void 2762void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2763set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2764 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2765 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2766
2185void 2767void
2186set_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
2187 CODE: 2772 CODE:
2188 clSetKernelArg (self, idx, sizeof (value), &value); 2773 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2189
2190void
2191set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
2192 CODE:
2193 clSetKernelArg (self, idx, sizeof (value), &value);
2194 2774
2195void 2775void
2196set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2776set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2197 CODE: 2777 CODE:
2198 clSetKernelArg (self, idx, sizeof (value), &value); 2778 clSetKernelArg (self, idx, sizeof (value), &value);
2244 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2824 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2245 EXTEND (SP, 1); 2825 EXTEND (SP, 1);
2246 const int i = 0; 2826 const int i = 0;
2247 { 2827 {
2248 NEED_SUCCESS (RetainContext, (value [i])); 2828 NEED_SUCCESS (RetainContext, (value [i]));
2249 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2829 PUSH_CLOBJ (stash_context, value [i]);
2250 } 2830 }
2251 2831
2252void 2832void
2253program (OpenCL::Kernel self) 2833program (OpenCL::Kernel self)
2254 PPCODE: 2834 PPCODE:
2256 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2836 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2257 EXTEND (SP, 1); 2837 EXTEND (SP, 1);
2258 const int i = 0; 2838 const int i = 0;
2259 { 2839 {
2260 NEED_SUCCESS (RetainProgram, (value [i])); 2840 NEED_SUCCESS (RetainProgram, (value [i]));
2261 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2841 PUSH_CLOBJ (stash_program, value [i]);
2262 } 2842 }
2263 2843
2264#END:kernel 2844#END:kernel
2265 2845
2266void 2846void
2313 const int i = 0; 2893 const int i = 0;
2314 PUSHs (sv_2mortal (newSVuv (value [i]))); 2894 PUSHs (sv_2mortal (newSVuv (value [i])));
2315 2895
2316#END:kernel_work_group 2896#END:kernel_work_group
2317 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
2318MODULE = OpenCL PACKAGE = OpenCL::Event 2959MODULE = OpenCL PACKAGE = OpenCL::Event
2319 2960
2320void 2961void
2321DESTROY (OpenCL::Event self) 2962DESTROY (OpenCL::Event self)
2322 CODE: 2963 CODE:
2346 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));
2347 EXTEND (SP, 1); 2988 EXTEND (SP, 1);
2348 const int i = 0; 2989 const int i = 0;
2349 { 2990 {
2350 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2991 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2351 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2992 PUSH_CLOBJ (stash_queue, value [i]);
2352 } 2993 }
2353 2994
2354void 2995void
2355command_type (OpenCL::Event self) 2996command_type (OpenCL::Event self)
2356 PPCODE: 2997 PPCODE:
2379 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3020 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2380 EXTEND (SP, 1); 3021 EXTEND (SP, 1);
2381 const int i = 0; 3022 const int i = 0;
2382 { 3023 {
2383 NEED_SUCCESS (RetainContext, (value [i])); 3024 NEED_SUCCESS (RetainContext, (value [i]));
2384 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3025 PUSH_CLOBJ (stash_context, value [i]);
2385 } 3026 }
2386 3027
2387#END:event 3028#END:event
2388 3029
2389void 3030void
2414void 3055void
2415set_status (OpenCL::UserEvent self, cl_int execution_status) 3056set_status (OpenCL::UserEvent self, cl_int execution_status)
2416 CODE: 3057 CODE:
2417 clSetUserEventStatus (self, execution_status); 3058 clSetUserEventStatus (self, execution_status);
2418 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