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 1.71 by root, Sat May 5 02:45:38 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
62typedef SV *FUTURE; 61typedef struct mapped * OpenCL__Mapped;
62
63static HV
64 *stash_platform,
65 *stash_device,
66 *stash_subdevice,
67 *stash_context,
68 *stash_queue,
69 *stash_program,
70 *stash_kernel,
71 *stash_sampler,
72 *stash_event,
73 *stash_userevent,
74 *stash_memory,
75 *stash_buffer,
76 *stash_bufferobj,
77 *stash_image,
78 *stash_image1d,
79 *stash_image1darray,
80 *stash_image1dbuffer,
81 *stash_image2d,
82 *stash_image2darray,
83 *stash_image3d,
84 *stash_mapped,
85 *stash_mappedbuffer,
86 *stash_mappedimage;
63 87
64/*****************************************************************************/ 88/*****************************************************************************/
65 89
66// name must include a leading underscore 90// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 91// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 131}
108 132
109/*****************************************************************************/ 133/*****************************************************************************/
110 134
111/* up to two temporary buffers */ 135/* up to two temporary buffers */
112static void * 136static void * ecb_noinline
113tmpbuf (size_t size) 137tmpbuf (size_t size)
114{ 138{
115 enum { buffers = 3 }; 139 enum { buffers = 4 };
116 static int idx; 140 static int idx;
117 static void *buf [buffers]; 141 static void *buf [buffers];
118 static size_t len [buffers]; 142 static size_t len [buffers];
119 143
120 idx = (idx + 1) % buffers; 144 idx = (idx + 1) % buffers;
125 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 149 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
126 buf [idx] = malloc (len [idx]); 150 buf [idx] = malloc (len [idx]);
127 } 151 }
128 152
129 return buf [idx]; 153 return buf [idx];
154}
155
156static const char * ecb_noinline
157cv_get_name (CV *cv)
158{
159 static char fullname [256];
160
161 GV *gv = CvGV (cv); // gv better be nonzero
162
163 HV *stash = GvSTASH (gv);
164 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
165 const char *gvname = GvNAME (gv);
166
167 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
168 return fullname;
130} 169}
131 170
132/*****************************************************************************/ 171/*****************************************************************************/
133 172
134typedef struct 173typedef struct
193 if (res) \ 232 if (res) \
194 FAIL (name); 233 FAIL (name);
195 234
196/*****************************************************************************/ 235/*****************************************************************************/
197 236
237static SV *
238new_clobj (HV *stash, IV id)
239{
240 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
241}
242
243#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
244#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
245
246/* cl objects are either \$iv, or [$iv, ...] */
247/* they can be upgraded at runtime to the array form */
248static void * ecb_noinline
249SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
250{
251 // sv_derived_from is quite slow :(
252 if (SvROK (sv) && sv_derived_from (sv, pkg))
253 return (void *)SvIV (SvRV (sv));
254
255 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
256}
257
258// the "no-inherit" version of the above
259static void * ecb_noinline
260SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
261{
262 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
263 return (void *)SvIV (SvRV (sv));
264
265 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
266}
267
268/*****************************************************************************/
269
198static cl_context_properties * 270static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 271SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 272{
201 if (!sv || !SvOK (sv)) 273 if (!sv || !SvOK (sv))
202 if (extra) 274 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 275 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 276 else
210 int i, len = av_len (av) + 1; 282 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 283 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 284 cl_context_properties *l = p;
213 285
214 if (len & 1) 286 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 287 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 288
217 while (extracount--) 289 while (extracount--)
218 *l++ = *extra++; 290 *l++ = *extra++;
219 291
220 for (i = 0; i < len; i += 2) 292 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 295 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 296 cl_context_properties v = SvIV (p_sv); // code below can override
225 297
226 switch (t) 298 switch (t)
227 { 299 {
300 case CL_CONTEXT_PLATFORM:
301 if (SvROK (p_sv))
302 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
303 break;
304
228 case CL_GLX_DISPLAY_KHR: 305 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 306 if (!SvOK (p_sv))
230 { 307 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 308 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 309 if (func)
255 *l = 0; 332 *l = 0;
256 333
257 return p; 334 return p;
258 } 335 }
259 336
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 337 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 338}
262 339
263/*****************************************************************************/ 340// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 341// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 342static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
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 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
275 return (void *)SvIV (SvRV (sv)); 346 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
347 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
276 348
277 croak ("%s: %s is not of type %s", func, svname, pkg); 349 AV *av = (AV *)SvRV (arg);
350 void **list = 0;
351 cl_uint count = av_len (av) + 1;
352
353 if (count)
354 {
355 list = tmpbuf (sizeof (*list) * count);
356 int i;
357 for (i = 0; i < count; ++i)
358 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
359 }
360
361 if (!count && !or_undef)
362 croak ("%s: '%s' must contain at least one %s object",
363 cv_get_name (cv), argname, klass);
364
365 *rcount = count;
366 return (void *)list;
278} 367}
279 368
280/*****************************************************************************/ 369/*****************************************************************************/
281/* callback stuff */ 370/* callback stuff */
282 371
304static void (*eq_signal_func)(void *signal_arg, int value); 393static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 394static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 395static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 396static eq_item *eq_head, *eq_tail;
308 397
309static void 398static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 399eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 400{
312 eq_item *item = malloc (sizeof (eq_item)); 401 eq_item *item = malloc (sizeof (eq_item));
313 402
314 item->next = 0; 403 item->next = 0;
381} 470}
382 471
383/*****************************************************************************/ 472/*****************************************************************************/
384/* context notify */ 473/* context notify */
385 474
386static void 475static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 476eq_context_push (void *data1, void *data2, void *data3)
388{ 477{
389 dSP; 478 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 479 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 480 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 502 { \
414 pfn_notify = eq_context_notify; \ 503 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 504 user_data = s_get_cv (notify); \
416 } 505 }
417 506
418static SV * 507static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 508new_clobj_context (cl_context ctx, void *user_data)
420{ 509{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 510 SV *sv = new_clobj (stash_context, (IV)ctx);
422 511
423 if (user_data) 512 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 513 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 514
426 return sv; 515 return sv;
433 522
434static void 523static void
435eq_program_push (void *data1, void *data2, void *data3) 524eq_program_push (void *data1, void *data2, void *data3)
436{ 525{
437 dSP; 526 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 527 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 528 PUTBACK;
440} 529}
441 530
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 531static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 532
444static void CL_CALLBACK 533static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 534eq_program_notify (cl_program program, void *user_data)
446{ 535{
536 clRetainProgram (program);
537
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 538 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
539}
540
541typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
542
543static program_callback ecb_noinline
544make_program_callback (SV *notify, void **ruser_data)
545{
546 if (SvOK (notify))
547 {
548 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
549 return eq_program_notify;
550 }
551 else
552 {
553 *ruser_data = 0;
554 return 0;
555 }
448} 556}
449 557
450struct build_args 558struct build_args
451{ 559{
452 cl_program program; 560 cl_program program;
466 else 574 else
467 clReleaseProgram (arg->program); 575 clReleaseProgram (arg->program);
468 576
469 free (arg->options); 577 free (arg->options);
470 free (arg); 578 free (arg);
579
580 return 0;
471} 581}
472 582
473static void 583static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 584build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 585{
490 600
491static void 601static void
492eq_event_push (void *data1, void *data2, void *data3) 602eq_event_push (void *data1, void *data2, void *data3)
493{ 603{
494 dSP; 604 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 605 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 606 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 607 PUTBACK;
498} 608}
499 609
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 610static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 615 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 616 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 617}
508 618
509/*****************************************************************************/ 619/*****************************************************************************/
620/* utilities for XS code */
510 621
511static size_t 622static size_t
512img_row_pitch (cl_mem img) 623img_row_pitch (cl_mem img)
513{ 624{
514 size_t res; 625 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 626 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 627 return res;
517} 628}
518 629
519static cl_event * 630static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 631event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 632{
522 cl_uint count = *rcount; 633 cl_uint count = *rcount;
523 634
524 if (!count) 635 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
636 *rcount = count = 0;
637
638 if (!count && !extra)
525 return 0; 639 return 0;
526 640
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 641 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 642 int i = 0;
529 643
530 do 644 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 645 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 646 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 647
536 while (count); 648 if (extra)
649 list [i++] = extra;
537 650
538 *rcount = i; 651 *rcount = i;
539 652
540 return i ? list : 0; 653 return i ? list : 0;
541} 654}
542 655
543#define EVENT_LIST(items,count) \ 656#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 657 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 658 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 659
547#define INFO(class) \ 660#define INFO(class) \
548{ \ 661{ \
549 size_t size; \ 662 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 663 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
553 SvPOK_only (sv); \ 666 SvPOK_only (sv); \
554 SvCUR_set (sv, size); \ 667 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 668 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 669 XPUSHs (sv); \
557} 670}
671
672/*****************************************************************************/
673/* mapped_xxx */
674
675static OpenCL__Mapped
676SvMAPPED (SV *sv)
677{
678 // no typechecking atm., keep your fingers crossed
679 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
680}
681
682struct mapped
683{
684 cl_command_queue queue;
685 cl_mem memobj;
686 void *ptr;
687 size_t cb;
688 cl_event event;
689 size_t row_pitch;
690 size_t slice_pitch;
691};
692
693static SV *
694mapped_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)
695{
696 SV *data = newSV (0);
697 SvUPGRADE (data, SVt_PVMG);
698
699 OpenCL__Mapped mapped;
700 New (0, mapped, 1, struct mapped);
701
702 clRetainCommandQueue (queue);
703
704 mapped->queue = queue;
705 mapped->memobj = memobj;
706 mapped->ptr = ptr;
707 mapped->cb = cb;
708 mapped->event = ev;
709 mapped->row_pitch = row_pitch;
710 mapped->slice_pitch = slice_pitch;
711
712 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
713
714 if (SvLEN (data))
715 Safefree (data);
716
717 SvPVX (data) = (char *)ptr;
718 SvCUR_set (data, cb);
719 SvLEN_set (data, 0);
720 SvPOK_only (data);
721
722 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
723
724 if (!(flags & CL_MAP_WRITE))
725 SvREADONLY_on (data);
726
727 return obj;
728}
729
730static void
731mapped_detach (SV *sv, OpenCL__Mapped mapped)
732{
733 SV *data = SvRV (sv);
734
735 // the next check checks both before AND after detach, where SvPVX should STILL be 0
736 if (SvPVX (data) != (char *)mapped->ptr)
737 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
738 else
739 {
740 SvREADONLY_off (data);
741 SvCUR_set (data, 0);
742 SvPVX (data) = 0;
743 SvOK_off (data);
744 }
745
746 mapped->ptr = 0;
747}
748
749static void
750mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
751{
752 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
753 cl_event ev;
754
755 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
756
757 clReleaseEvent (mapped->event);
758 mapped->event = ev;
759
760 mapped_detach (self, mapped);
761}
762
763/*****************************************************************************/
558 764
559MODULE = OpenCL PACKAGE = OpenCL 765MODULE = OpenCL PACKAGE = OpenCL
560 766
561PROTOTYPES: ENABLE 767PROTOTYPES: ENABLE
562 768
590 }; 796 };
591 797
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 798 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)); 799 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
594 800
801 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
802 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
803 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
804 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
805 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
806 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
807 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
808 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
809 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
810 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
811 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
812 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
813 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
814 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
815 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
816 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
817 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
818 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
819 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
820 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
821 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
822 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
823 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
824
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 825 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 826}
597 827
598cl_int 828cl_int
599errno () 829errno ()
601 RETVAL = res; 831 RETVAL = res;
602 OUTPUT: 832 OUTPUT:
603 RETVAL 833 RETVAL
604 834
605const char * 835const char *
606err2str (cl_int err) 836err2str (cl_int err = res)
607 837
608const char * 838const char *
609enum2str (cl_uint value) 839enum2str (cl_uint value)
610 840
611void 841void
619 list = tmpbuf (sizeof (*list) * count); 849 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 850 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 851
622 EXTEND (SP, count); 852 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 853 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 854 PUSH_CLOBJ (stash_platform, list [i]);
625 855
626void 856void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 857context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 858 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 859 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 860 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 861 XPUSH_CLOBJ_CONTEXT;
632 862
633void 863void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 864context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 865 PPCODE:
636 /* der Gipfel der Kunst */ 866 cl_uint device_count;
867 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
868
869 CONTEXT_NOTIFY_CALLBACK;
870 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
871 XPUSH_CLOBJ_CONTEXT;
637 872
638void 873void
639wait_for_events (...) 874wait_for_events (...)
640 CODE: 875 CODE:
641 EVENT_LIST (0, items); 876 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 877 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 878
644PROTOTYPES: DISABLE 879PROTOTYPES: DISABLE
645 880
646MODULE = OpenCL PACKAGE = OpenCL::Platform 881MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 924 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 925 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 926
692 EXTEND (SP, count); 927 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 929 PUSH_CLOBJ (stash_device, list [i]);
695 930
696void 931void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 932context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 933 PPCODE:
934 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
935 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
936
937 cl_uint device_count;
938 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
939
940 CONTEXT_NOTIFY_CALLBACK;
941 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
942 XPUSH_CLOBJ_CONTEXT;
943
944void
945context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
946 PPCODE:
947 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
948 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
949
950 CONTEXT_NOTIFY_CALLBACK;
951 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
952 XPUSH_CLOBJ_CONTEXT;
953
954MODULE = OpenCL PACKAGE = OpenCL::Device
955
956void
957info (OpenCL::Device self, cl_device_info name)
958 PPCODE:
959 INFO (Device)
960
961#if CL_VERSION_1_2
962
963void
964sub_devices (OpenCL::Device self, SV *properties)
965 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 966 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 967 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 968
702 AV *av = (AV *)SvRV (devices); 969 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 970
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 971 cl_uint count = av_len ((AV *)properties) + 1;
972 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 973
706 int i; 974 int i;
707 for (i = num_devices; i--; ) 975 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 976 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 977
710 CONTEXT_NOTIFY_CALLBACK; 978 props [count] = 0;
711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
712 XPUSH_CLOBJ_CONTEXT;
713 979
714void 980 cl_uint num_devices;
715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 981 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 982 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 983 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719 984
720 CONTEXT_NOTIFY_CALLBACK; 985 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 986 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 987 PUSH_CLOBJ (stash_subdevice, list [i]);
723 988
724MODULE = OpenCL PACKAGE = OpenCL::Device 989#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 990
731#BEGIN:device 991#BEGIN:device
732 992
733void 993void
734type (OpenCL::Device self) 994type (OpenCL::Device self)
735 PPCODE: 995 PPCODE:
736 cl_device_type value [1]; 996 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 997 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 998 EXTEND (SP, 1);
739 const int i = 0; 999 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1000 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1001
742void 1002void
743vendor_id (OpenCL::Device self) 1003vendor_id (OpenCL::Device self)
744 ALIAS: 1004 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1005 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1150 PPCODE:
891 cl_command_queue_properties value [1]; 1151 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1152 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1153 EXTEND (SP, 1);
894 const int i = 0; 1154 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1155 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1156
897void 1157void
898platform (OpenCL::Device self) 1158platform (OpenCL::Device self)
899 PPCODE: 1159 PPCODE:
900 cl_platform_id value [1]; 1160 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1161 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1162 EXTEND (SP, 1);
903 const int i = 0; 1163 const int i = 0;
904 { 1164 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1165 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1166 }
907 1167
908void 1168void
909name (OpenCL::Device self) 1169name (OpenCL::Device self)
910 ALIAS: 1170 ALIAS:
929 cl_device_id value [1]; 1189 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1190 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1191 EXTEND (SP, 1);
932 const int i = 0; 1192 const int i = 0;
933 { 1193 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1194 PUSH_CLOBJ (stash_device, value [i]);
935 } 1195 }
936 1196
937void 1197void
938partition_types_ext (OpenCL::Device self) 1198partition_types_ext (OpenCL::Device self)
939 ALIAS: 1199 ALIAS:
950 for (i = 0; i < n; ++i) 1210 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1211 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1212
953#END:device 1213#END:device
954 1214
1215MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1216
1217#if CL_VERSION_1_2
1218
1219void
1220DESTROY (OpenCL::SubDevice self)
1221 CODE:
1222 clReleaseDevice (self);
1223
1224#endif
1225
955MODULE = OpenCL PACKAGE = OpenCL::Context 1226MODULE = OpenCL PACKAGE = OpenCL::Context
956 1227
957void 1228void
958DESTROY (OpenCL::Context context) 1229DESTROY (OpenCL::Context self)
959 CODE: 1230 CODE:
960 clReleaseContext (context); 1231 clReleaseContext (self);
961 1232
962void 1233void
963info (OpenCL::Context self, cl_context_info name) 1234info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1235 PPCODE:
965 INFO (Context) 1236 INFO (Context)
966 1237
967void 1238void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1239queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1240 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1241 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1242 XPUSH_CLOBJ (stash_queue, queue);
972 1243
973void 1244void
974user_event (OpenCL::Context self) 1245user_event (OpenCL::Context self)
975 PPCODE: 1246 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1247 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1248 XPUSH_CLOBJ (stash_userevent, ev);
978 1249
979void 1250void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1251buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1252 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1253 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?"); 1254 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1255
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1256 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1257 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1258
988void 1259void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1260buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1261 PPCODE:
991 STRLEN len; 1262 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1263 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1264 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?"); 1265 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)); 1266 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1267 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1268
998#if CL_VERSION_1_2 1269#if CL_VERSION_1_2
999 1270
1000void 1271void
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) 1272image (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: 1273 PPCODE:
1003 STRLEN len; 1274 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1275 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1276 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1277 const cl_image_desc desc = {
1007 type, 1278 type,
1008 width, height, depth, 1279 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1280 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1281 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1282 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1283 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1284 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1285 HV *stash = stash_image;
1015 switch (type) 1286 switch (type)
1016 { 1287 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1288 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1289 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1290 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1291 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1292 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1293 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1294 }
1024 XPUSH_CLOBJ (klass, mem); 1295 XPUSH_CLOBJ (stash, mem);
1025 1296
1026#endif 1297#endif
1027 1298
1028void 1299void
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) 1300image2d (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)); 1306 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1307#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1308 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)); 1309 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1310#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1311 XPUSH_CLOBJ (stash_image2d, mem);
1041 1312
1042void 1313void
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) 1314image3d (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: 1315 PPCODE:
1045 STRLEN len; 1316 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1320 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1321#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1322 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)); 1323 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1324#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1325 XPUSH_CLOBJ (stash_image3d, mem);
1055 1326
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1327#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1328
1058void 1329void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1330gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1331 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1332 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1333 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1334
1064void 1335void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1336gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1337 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1339 XPUSH_CLOBJ (stash_image2d, mem);
1069 1340
1070#if CL_VERSION_1_2 1341#if CL_VERSION_1_2
1071 1342
1072void 1343void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1344gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1345 ALIAS:
1075 PPCODE: 1346 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1348 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1349 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1350 HV *stash = stash_memory;
1080 switch (type) 1351 switch (type)
1081 { 1352 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1353 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1354 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1355 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1356 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1357 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1358 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1359 }
1089 XPUSH_CLOBJ (klass, mem); 1360 XPUSH_CLOBJ (stash, mem);
1090 1361
1091#endif 1362#endif
1092 1363
1093void 1364void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1365gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1367#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1368 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1369#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1370 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1371#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1372 XPUSH_CLOBJ (stash_image2d, mem);
1102 1373
1103void 1374void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1375gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1376 PPCODE:
1106#if PREFER_1_1 1377#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1378 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1379#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1380 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1381#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1382 XPUSH_CLOBJ (stash_image3d, mem);
1112 1383
1113#endif 1384#endif
1114 1385
1115void 1386void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1387supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1407
1137void 1408void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1409sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1410 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1411 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1412 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1413
1143void 1414void
1144program_with_source (OpenCL::Context self, SV *program) 1415program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1416 PPCODE:
1146 STRLEN len; 1417 STRLEN len;
1147 size_t len2; 1418 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1419 const char *ptr = SvPVbyte (program, len);
1149 1420
1150 len2 = len; 1421 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1422 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1423 XPUSH_CLOBJ (stash_program, prog);
1424
1425void
1426program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1427 PPCODE:
1428 cl_uint device_count;
1429 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1430
1431 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1432 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1433
1434 binaries = SvRV (binaries);
1435
1436 if (device_count != av_len ((AV *)binaries) + 1)
1437 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1438
1439 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1440 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1441 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1442
1443 int i;
1444 for (i = 0; i < device_count; ++i)
1445 {
1446 STRLEN len;
1447 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1448 length_list [i] = len;
1449 }
1450
1451 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1452 length_list, binary_list,
1453 GIMME_V == G_ARRAY ? status_list : 0, &res));
1454
1455 EXTEND (SP, 2);
1456 PUSH_CLOBJ (stash_program, prog);
1457
1458 if (GIMME_V == G_ARRAY)
1459 {
1460 AV *av = newAV ();
1461 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1462
1463 for (i = device_count; i--; )
1464 av_store (av, i, newSViv (status_list [i]));
1465 }
1466
1467#if CL_VERSION_1_2
1468
1469void
1470program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1471 PPCODE:
1472 cl_uint device_count;
1473 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1474
1475 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1476
1477 XPUSH_CLOBJ (stash_program, prog);
1478
1479void
1480link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1481 CODE:
1482 cl_uint device_count = 0;
1483 cl_device_id *device_list = 0;
1484
1485 if (SvOK (devices))
1486 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 cl_uint program_count;
1489 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1490
1491 void *user_data;
1492 program_callback pfn_notify = make_program_callback (notify, &user_data);
1493
1494 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1495 program_count, program_list, pfn_notify, user_data, &res));
1496
1497 XPUSH_CLOBJ (stash_program, prog);
1498
1499#endif
1153 1500
1154#BEGIN:context 1501#BEGIN:context
1155 1502
1156void 1503void
1157reference_count (OpenCL::Context self) 1504reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1521 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1522 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1523 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1524 for (i = 0; i < n; ++i)
1178 { 1525 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1526 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1527 }
1181 1528
1182void 1529void
1183properties (OpenCL::Context self) 1530properties (OpenCL::Context self)
1184 PPCODE: 1531 PPCODE:
1199DESTROY (OpenCL::Queue self) 1546DESTROY (OpenCL::Queue self)
1200 CODE: 1547 CODE:
1201 clReleaseCommandQueue (self); 1548 clReleaseCommandQueue (self);
1202 1549
1203void 1550void
1204enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1551read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1552 ALIAS:
1553 enqueue_read_buffer = 0
1205 PPCODE: 1554 PPCODE:
1206 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1555 EVENT_LIST (6);
1208 1556
1209 SvUPGRADE (data, SVt_PV); 1557 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1558 SvGROW (data, len);
1211 SvPOK_only (data); 1559 SvPOK_only (data);
1212 SvCUR_set (data, len); 1560 SvCUR_set (data, len);
1561
1562 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)); 1563 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1214 1564
1215 if (ev) 1565 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1566 XPUSH_CLOBJ (stash_event, ev);
1217 1567
1218void 1568void
1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1569write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1570 ALIAS:
1571 enqueue_write_buffer = 0
1220 PPCODE: 1572 PPCODE:
1221 cl_event ev = 0; 1573 EVENT_LIST (5);
1574
1222 STRLEN len; 1575 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1576 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5);
1225 1577
1578 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)); 1579 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1227 1580
1228 if (ev) 1581 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1582 XPUSH_CLOBJ (stash_event, ev);
1230 1583
1231#if CL_VERSION_1_2 1584#if CL_VERSION_1_2
1232 1585
1233void 1586void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1587fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1588 ALIAS:
1589 enqueue_fill_buffer = 0
1235 PPCODE: 1590 PPCODE:
1236 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1237 STRLEN len; 1593 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240 1595
1596 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)); 1597 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1598
1243 if (ev) 1599 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1245 1601
1246void 1602void
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, ...) 1603fill_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, ...)
1604 ALIAS:
1605 enqueue_fill_image = 0
1248 PPCODE: 1606 PPCODE:
1249 cl_event ev = 0; 1607 EVENT_LIST (12);
1250 STRLEN len; 1608
1251 const size_t origin [3] = { x, y, z }; 1609 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1610 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254 1611
1255 const cl_float c_f [4] = { r, g, b, a }; 1612 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1613 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1614 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1615 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1618 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1619 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) 1620 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."); 1621 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1622
1623 cl_event ev = 0;
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1624 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)); 1625 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1626
1269 if (ev) 1627 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1628 XPUSH_CLOBJ (stash_event, ev);
1271 1629
1272#endif 1630#endif
1273 1631
1274void 1632void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1633copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1634 ALIAS:
1635 enqueue_copy_buffer = 0
1276 PPCODE: 1636 PPCODE:
1637 EVENT_LIST (6);
1638
1277 cl_event ev = 0; 1639 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)); 1640 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1641
1282 if (ev) 1642 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1643 XPUSH_CLOBJ (stash_event, ev);
1284 1644
1285void 1645void
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, ...) 1646read_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, ...)
1647 ALIAS:
1648 enqueue_read_buffer_rect = 0
1287 PPCODE: 1649 PPCODE:
1288 cl_event ev = 0; 1650 EVENT_LIST (17);
1651
1289 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1652 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 }; 1653 const size_t host_origin[3] = { host_x, host_y, host_z };
1291 const size_t region[3] = { width, height, depth }; 1654 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17);
1293 1655
1294 if (!buf_row_pitch) 1656 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1657 buf_row_pitch = region [0];
1296 1658
1297 if (!buf_slice_pitch) 1659 if (!buf_slice_pitch)
1307 1669
1308 SvUPGRADE (data, SVt_PV); 1670 SvUPGRADE (data, SVt_PV);
1309 SvGROW (data, len); 1671 SvGROW (data, len);
1310 SvPOK_only (data); 1672 SvPOK_only (data);
1311 SvCUR_set (data, len); 1673 SvCUR_set (data, len);
1674
1675 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)); 1676 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 1677
1314 if (ev) 1678 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1679 XPUSH_CLOBJ (stash_event, ev);
1316 1680
1317void 1681void
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, ...) 1682write_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, ...)
1683 ALIAS:
1684 enqueue_write_buffer_rect = 0
1319 PPCODE: 1685 PPCODE:
1320 cl_event ev = 0; 1686 EVENT_LIST (17);
1687
1321 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1688 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 }; 1689 const size_t host_origin[3] = { host_x, host_y, host_z };
1323 const size_t region[3] = { width, height, depth }; 1690 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1691 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1692 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17);
1327 1693
1328 if (!buf_row_pitch) 1694 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1695 buf_row_pitch = region [0];
1330 1696
1331 if (!buf_slice_pitch) 1697 if (!buf_slice_pitch)
1340 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1706 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1341 1707
1342 if (len < min_len) 1708 if (len < min_len)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1709 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1710
1711 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)); 1712 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 1713
1347 if (ev) 1714 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1715 XPUSH_CLOBJ (stash_event, ev);
1349 1716
1350void 1717void
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, ...) 1718copy_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, ...)
1719 ALIAS:
1720 enqueue_copy_buffer_rect = 0
1352 PPCODE: 1721 PPCODE:
1353 cl_event ev = 0; 1722 EVENT_LIST (16);
1723
1354 const size_t src_origin[3] = { src_x, src_y, src_z }; 1724 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 }; 1725 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1356 const size_t region[3] = { width, height, depth }; 1726 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16);
1358 1727
1728 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)); 1729 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 1730
1361 if (ev) 1731 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1732 XPUSH_CLOBJ (stash_event, ev);
1363 1733
1364void 1734void
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, ...) 1735read_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, ...)
1736 ALIAS:
1737 enqueue_read_image = 0
1366 PPCODE: 1738 PPCODE:
1367 cl_event ev = 0; 1739 EVENT_LIST (12);
1740
1368 const size_t src_origin[3] = { src_x, src_y, src_z }; 1741 const size_t src_origin[3] = { src_x, src_y, src_z };
1369 const size_t region[3] = { width, height, depth }; 1742 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12);
1371 1743
1372 if (!row_pitch) 1744 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1745 row_pitch = img_row_pitch (src);
1374 1746
1375 if (depth > 1 && !slice_pitch) 1747 if (depth > 1 && !slice_pitch)
1379 1751
1380 SvUPGRADE (data, SVt_PV); 1752 SvUPGRADE (data, SVt_PV);
1381 SvGROW (data, len); 1753 SvGROW (data, len);
1382 SvPOK_only (data); 1754 SvPOK_only (data);
1383 SvCUR_set (data, len); 1755 SvCUR_set (data, len);
1756
1757 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)); 1758 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 1759
1386 if (ev) 1760 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1761 XPUSH_CLOBJ (stash_event, ev);
1388 1762
1389void 1763void
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, ...) 1764write_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, ...)
1765 ALIAS:
1766 enqueue_write_image = 0
1391 PPCODE: 1767 PPCODE:
1392 cl_event ev = 0; 1768 EVENT_LIST (12);
1769
1393 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1770 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1394 const size_t region[3] = { width, height, depth }; 1771 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1772 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1773 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12);
1398 1774
1399 if (!row_pitch) 1775 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1776 row_pitch = img_row_pitch (dst);
1401 1777
1402 if (depth > 1 && !slice_pitch) 1778 if (depth > 1 && !slice_pitch)
1405 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1781 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1406 1782
1407 if (len < min_len) 1783 if (len < min_len)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1784 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1785
1786 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)); 1787 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 1788
1412 if (ev) 1789 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1790 XPUSH_CLOBJ (stash_event, ev);
1414 1791
1415void 1792void
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, ...) 1793copy_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, ...)
1794 ALIAS:
1795 enqueue_copy_image = 0
1417 PPCODE: 1796 PPCODE:
1418 cl_event ev = 0; 1797 EVENT_LIST (12);
1798
1419 const size_t src_origin[3] = { src_x, src_y, src_z }; 1799 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 }; 1800 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1421 const size_t region[3] = { width, height, depth }; 1801 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12);
1423 1802
1803 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)); 1804 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1805
1426 if (ev) 1806 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1807 XPUSH_CLOBJ (stash_event, ev);
1428 1808
1429void 1809void
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, ...) 1810copy_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, ...)
1811 ALIAS:
1812 enqueue_copy_image_to_buffer = 0
1431 PPCODE: 1813 PPCODE:
1814 EVENT_LIST (10);
1815
1816 const size_t src_origin[3] = { src_x, src_y, src_z };
1817 const size_t region [3] = { width, height, depth };
1818
1432 cl_event ev = 0; 1819 cl_event ev = 0;
1820 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1821
1822 if (ev)
1823 XPUSH_CLOBJ (stash_event, ev);
1824
1825void
1826copy_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, ...)
1827 ALIAS:
1828 enqueue_copy_buffer_to_image = 0
1829 PPCODE:
1830 EVENT_LIST (10);
1831
1433 const size_t src_origin[3] = { src_x, src_y, src_z }; 1832 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1833 const size_t region [3] = { width, height, depth };
1834
1835 cl_event ev = 0;
1836 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1837
1838 if (ev)
1839 XPUSH_CLOBJ (stash_event, ev);
1840
1841void
1842map_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, ...)
1843 ALIAS:
1844 enqueue_map_buffer = 0
1845 PPCODE:
1846 EVENT_LIST (6);
1847
1848 size_t cb = SvIV (cb_);
1849
1850 if (!SvOK (cb_))
1851 {
1852 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1853 cb -= offset;
1854 }
1855
1856 cl_event ev;
1857 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1858 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1859
1860void
1861map_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, ...)
1862 ALIAS:
1863 enqueue_map_image = 0
1864 PPCODE:
1865 size_t width = SvIV (width_);
1866 if (!SvOK (width_))
1867 {
1868 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1869 width -= x;
1870 }
1871
1872 size_t height = SvIV (width_);
1873 if (!SvOK (height_))
1874 {
1875 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1876 height -= y;
1877 }
1878
1879 size_t depth = SvIV (width_);
1880 if (!SvOK (depth_))
1881 {
1882 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1883 depth -= z;
1884
1885 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1886 if (!depth)
1887 depth = 1;
1888 }
1889
1890 const size_t origin[3] = { x, y, z };
1434 const size_t region[3] = { width, height, depth }; 1891 const size_t region[3] = { width, height, depth };
1892 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1893 EVENT_LIST (10);
1436 1894
1895 cl_event ev;
1896 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));
1897
1898 size_t cb = slice_pitch ? slice_pitch * region [2]
1899 : row_pitch ? row_pitch * region [1]
1900 : region [0];
1901
1902 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1903
1904void
1905unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1906 PPCODE:
1907 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1908 if (GIMME_V != G_VOID)
1909 {
1910 clRetainEvent (mapped->event);
1911 XPUSH_CLOBJ (stash_event, mapped->event);
1912 }
1913
1914void
1915task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1916 ALIAS:
1917 enqueue_task = 0
1918 PPCODE:
1919 EVENT_LIST (2);
1920
1921 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)); 1922 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1438 1923
1439 if (ev) 1924 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1925 XPUSH_CLOBJ (stash_event, ev);
1441 1926
1442void 1927void
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, ...) 1928nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1929 ALIAS:
1930 enqueue_nd_range_kernel = 0
1468 PPCODE: 1931 PPCODE:
1469 cl_event ev = 0; 1932 EVENT_LIST (5);
1933
1470 size_t *gwo = 0, *gws, *lws = 0; 1934 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1935 int gws_len;
1472 size_t *lists; 1936 size_t *lists;
1473 int i; 1937 int i;
1474 EVENT_LIST (5, items - 5);
1475 1938
1476 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1939 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1940 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1941
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1942 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1943
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1944 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1945
1483 gws = lists + gws_len * 0; 1946 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1947 for (i = 0; i < gws_len; ++i)
1948 {
1485 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1949 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1950 // at least nvidia crashes for 0-sized work group sizes, work around
1951 if (!gws [i])
1952 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1953 }
1486 1954
1487 if (SvOK (global_work_offset)) 1955 if (SvOK (global_work_offset))
1488 { 1956 {
1489 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1957 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"); 1958 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1498 } 1966 }
1499 1967
1500 if (SvOK (local_work_size)) 1968 if (SvOK (local_work_size))
1501 { 1969 {
1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1970 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"); 1971 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1504 1972
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1973 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"); 1974 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1507 1975
1508 lws = lists + gws_len * 2; 1976 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 1977 for (i = 0; i < gws_len; ++i)
1978 {
1510 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1979 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1980 // at least nvidia crashes for 0-sized work group sizes, work around
1981 if (!lws [i])
1982 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1983 }
1511 } 1984 }
1512 1985
1986 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)); 1987 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1514 1988
1515 if (ev) 1989 if (ev)
1516 XPUSH_CLOBJ ("OpenCL::Event", ev); 1990 XPUSH_CLOBJ (stash_event, ev);
1991
1992#if CL_VERSION_1_2
1993
1994void
1995migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1996 ALIAS:
1997 enqueue_migrate_mem_objects = 0
1998 PPCODE:
1999 EVENT_LIST (3);
2000
2001 cl_uint obj_count;
2002 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2003
2004 cl_event ev = 0;
2005 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2006
2007 if (ev)
2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#endif
1517 2011
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 2012#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 2013
1520void 2014void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2015acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 2016 ALIAS:
2017 release_gl_objects = 1
2018 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 2019 enqueue_release_gl_objects = 1
1524 PPCODE: 2020 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2021 EVENT_LIST (2);
1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2022
2023 cl_uint obj_count;
2024 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1527 2025
1528 cl_event ev = 0; 2026 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 2027
1538 if (ix) 2028 if (ix)
1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2029 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1540 else 2030 else
1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2031 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1542 2032
1543 if (ev) 2033 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 2034 XPUSH_CLOBJ (stash_event, ev);
1545 2035
1546#endif 2036#endif
1547 2037
1548void 2038void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 2039wait_for_events (OpenCL::Queue self, ...)
2040 ALIAS:
2041 enqueue_wait_for_events = 0
1550 CODE: 2042 CODE:
1551 EVENT_LIST (1, items - 1); 2043 EVENT_LIST (1);
1552#if PREFER_1_1 2044#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2045 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 2046#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2047 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 2048#endif
1557 2049
1558void 2050void
1559enqueue_marker (OpenCL::Queue self, ...) 2051marker (OpenCL::Queue self, ...)
2052 ALIAS:
2053 enqueue_marker = 0
1560 PPCODE: 2054 PPCODE:
2055 EVENT_LIST (1);
1561 cl_event ev = 0; 2056 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1 2057#if PREFER_1_1
1564 if (!event_list_count) 2058 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2059 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 2060 else
1567#if CL_VERSION_1_2 2061#if CL_VERSION_1_2
1574#endif 2068#endif
1575#else 2069#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2070 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif 2071#endif
1578 if (ev) 2072 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 2073 XPUSH_CLOBJ (stash_event, ev);
1580 2074
1581void 2075void
1582enqueue_barrier (OpenCL::Queue self, ...) 2076barrier (OpenCL::Queue self, ...)
2077 ALIAS:
2078 enqueue_barrier = 0
1583 PPCODE: 2079 PPCODE:
2080 EVENT_LIST (1);
1584 cl_event ev = 0; 2081 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1 2082#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 2083 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 2084 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 2085 else
1590#if CL_VERSION_1_2 2086#if CL_VERSION_1_2
1600#endif 2096#endif
1601#else 2097#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2098 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif 2099#endif
1604 if (ev) 2100 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 2101 XPUSH_CLOBJ (stash_event, ev);
1606 2102
1607void 2103void
1608flush (OpenCL::Queue self) 2104flush (OpenCL::Queue self)
1609 CODE: 2105 CODE:
1610 NEED_SUCCESS (Flush, (self)); 2106 NEED_SUCCESS (Flush, (self));
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2124 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 2125 EXTEND (SP, 1);
1630 const int i = 0; 2126 const int i = 0;
1631 { 2127 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 2128 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2129 PUSH_CLOBJ (stash_context, value [i]);
1634 } 2130 }
1635 2131
1636void 2132void
1637device (OpenCL::Queue self) 2133device (OpenCL::Queue self)
1638 PPCODE: 2134 PPCODE:
1639 cl_device_id value [1]; 2135 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2136 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 2137 EXTEND (SP, 1);
1642 const int i = 0; 2138 const int i = 0;
1643 { 2139 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2140 PUSH_CLOBJ (stash_device, value [i]);
1645 } 2141 }
1646 2142
1647void 2143void
1648reference_count (OpenCL::Queue self) 2144reference_count (OpenCL::Queue self)
1649 PPCODE: 2145 PPCODE:
1658 PPCODE: 2154 PPCODE:
1659 cl_command_queue_properties value [1]; 2155 cl_command_queue_properties value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2156 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 2157 EXTEND (SP, 1);
1662 const int i = 0; 2158 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 2159 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 2160
1665#END:command_queue 2161#END:command_queue
1666 2162
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 2163MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 2164
1683 PPCODE: 2179 PPCODE:
1684 cl_mem_object_type value [1]; 2180 cl_mem_object_type value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2181 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 2182 EXTEND (SP, 1);
1687 const int i = 0; 2183 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 2184 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 2185
1690void 2186void
1691flags (OpenCL::Memory self) 2187flags (OpenCL::Memory self)
1692 PPCODE: 2188 PPCODE:
1693 cl_mem_flags value [1]; 2189 cl_mem_flags value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2190 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2191 EXTEND (SP, 1);
1696 const int i = 0; 2192 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 2193 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 2194
1699void 2195void
1700size (OpenCL::Memory self) 2196size (OpenCL::Memory self)
1701 ALIAS: 2197 ALIAS:
1702 size = CL_MEM_SIZE 2198 size = CL_MEM_SIZE
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2232 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 2233 EXTEND (SP, 1);
1738 const int i = 0; 2234 const int i = 0;
1739 { 2235 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 2236 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2237 PUSH_CLOBJ (stash_context, value [i]);
1742 } 2238 }
1743 2239
1744void 2240void
1745associated_memobject (OpenCL::Memory self) 2241associated_memobject (OpenCL::Memory self)
1746 PPCODE: 2242 PPCODE:
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2244 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2245 EXTEND (SP, 1);
1750 const int i = 0; 2246 const int i = 0;
1751 { 2247 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 2248 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2249 PUSH_CLOBJ (stash_memory, value [i]);
1754 } 2250 }
1755 2251
1756#END:mem 2252#END:mem
1757 2253
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 2254#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!"); 2274 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1779 2275
1780 cl_buffer_region crdata = { origin, size }; 2276 cl_buffer_region crdata = { origin, size };
1781 2277
1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2278 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2279 XPUSH_CLOBJ (stash_buffer, mem);
1784 2280
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2281MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2282
1787void 2283void
1788image_info (OpenCL::Image self, cl_image_info name) 2284image_info (OpenCL::Image self, cl_image_info name)
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2370 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2371 EXTEND (SP, 1);
1876 const int i = 0; 2372 const int i = 0;
1877 { 2373 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2374 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2375 PUSH_CLOBJ (stash_context, value [i]);
1880 } 2376 }
1881 2377
1882void 2378void
1883normalized_coords (OpenCL::Sampler self) 2379normalized_coords (OpenCL::Sampler self)
1884 PPCODE: 2380 PPCODE:
1885 cl_addressing_mode value [1]; 2381 cl_addressing_mode value [1];
1886 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2382 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1887 EXTEND (SP, 1); 2383 EXTEND (SP, 1);
1888 const int i = 0; 2384 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i]))); 2385 PUSHs (sv_2mortal (newSVuv (value [i])));
1890 2386
1891void 2387void
1892addressing_mode (OpenCL::Sampler self) 2388addressing_mode (OpenCL::Sampler self)
1893 PPCODE: 2389 PPCODE:
1894 cl_filter_mode value [1]; 2390 cl_filter_mode value [1];
1895 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2391 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1896 EXTEND (SP, 1); 2392 EXTEND (SP, 1);
1897 const int i = 0; 2393 const int i = 0;
1898 PUSHs (sv_2mortal (newSViv (value [i]))); 2394 PUSHs (sv_2mortal (newSVuv (value [i])));
1899 2395
1900void 2396void
1901filter_mode (OpenCL::Sampler self) 2397filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2398 PPCODE:
1903 cl_bool value [1]; 2399 cl_bool value [1];
1918void 2414void
1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2415build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS: 2416 ALIAS:
1921 build_async = 1 2417 build_async = 1
1922 CODE: 2418 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; 2419 cl_uint device_count = 0;
1926 cl_device_id *device_list = 0; 2420 cl_device_id *device_list = 0;
1927 2421
1928 if (SvOK (devices)) 2422 if (SvOK (devices))
1929 { 2423 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2424
2425 void *user_data;
2426 program_callback pfn_notify = make_program_callback (notify, &user_data);
2427
2428 if (ix)
2429 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2430 else
2431 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2432
2433#if CL_VERSION_1_2
2434
2435void
2436compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2437 CODE:
2438 cl_uint device_count = 0;
2439 cl_device_id *device_list = 0;
2440
2441 if (SvOK (devices))
2442 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2443
2444 cl_uint header_count = 0;
2445 cl_program *header_list = 0;
2446 const char **header_name = 0;
2447
2448 if (SvOK (headers))
2449 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2450 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2451 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1932 2452
1933 AV *av = (AV *)SvRV (devices); 2453 HV *hv = (HV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935 2454
1936 if (num_devices) 2455 header_count = hv_iterinit (hv);
2456 header_list = tmpbuf (sizeof (*header_list) * header_count);
2457 header_name = tmpbuf (sizeof (*header_name) * header_count);
2458
2459 HE *he;
2460 int i = 0;
2461 while (he = hv_iternext (hv))
1937 { 2462 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2463 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2464 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1939 int count; 2465 ++i;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 } 2466 }
1943 } 2467 }
1944 2468
1945 if (SvOK (notify)) 2469 void *user_data;
1946 { 2470 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 2471
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)); 2472 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2473 header_count, header_list, header_name, pfn_notify, user_data));
2474
2475#endif
1956 2476
1957void 2477void
1958build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2478build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1959 PPCODE: 2479 PPCODE:
1960 size_t size; 2480 size_t size;
1995 2515
1996void 2516void
1997kernel (OpenCL::Program program, SV *function) 2517kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2518 PPCODE:
1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2519 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2520 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2521
2002void 2522void
2003kernels_in_program (OpenCL::Program program) 2523kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2524 PPCODE:
2005 cl_uint num_kernels; 2525 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2528 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2529
2010 int i; 2530 int i;
2011 EXTEND (SP, num_kernels); 2531 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2532 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2533 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2534
2015void 2535void
2016info (OpenCL::Program self, cl_program_info name) 2536info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2537 PPCODE:
2018 INFO (Program) 2538 INFO (Program)
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2586 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2587 EXTEND (SP, 1);
2068 const int i = 0; 2588 const int i = 0;
2069 { 2589 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2590 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2591 PUSH_CLOBJ (stash_context, value [i]);
2072 } 2592 }
2073 2593
2074void 2594void
2075devices (OpenCL::Program self) 2595devices (OpenCL::Program self)
2076 PPCODE: 2596 PPCODE:
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2600 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2601 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2602 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2603 for (i = 0; i < n; ++i)
2084 { 2604 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2605 PUSH_CLOBJ (stash_device, value [i]);
2086 } 2606 }
2087 2607
2088void 2608void
2089source (OpenCL::Program self) 2609source (OpenCL::Program self)
2090 PPCODE: 2610 PPCODE:
2116DESTROY (OpenCL::Kernel self) 2636DESTROY (OpenCL::Kernel self)
2117 CODE: 2637 CODE:
2118 clReleaseKernel (self); 2638 clReleaseKernel (self);
2119 2639
2120void 2640void
2641setf (OpenCL::Kernel self, const char *format, ...)
2642 CODE:
2643 int i;
2644 for (i = 2; ; ++i)
2645 {
2646 while (*format == ' ')
2647 ++format;
2648
2649 char type = *format++;
2650
2651 if (!type)
2652 break;
2653
2654 if (i >= items)
2655 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2656
2657 SV *sv = ST (i);
2658
2659 union
2660 {
2661 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2662 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2663 cl_half ch; cl_float cf; cl_double cd;
2664 cl_mem cm;
2665 cl_sampler ca;
2666 size_t cz;
2667 cl_event ce;
2668 } arg;
2669 size_t size;
2670 int nullarg = 0;
2671
2672 switch (type)
2673 {
2674 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2675 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2676 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2677 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2678 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2679 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2680 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2681 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2682
2683 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2684 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2685 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2686
2687 case 'z': nullarg = 1; size = SvIV (sv); break;
2688
2689 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2690 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2691 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2692
2693 default:
2694 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2695 }
2696
2697 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2698 if (res)
2699 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2700 }
2701
2702 if (i != items)
2703 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2704
2705void
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2706set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2707 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2708 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2709
2125void 2710void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2758 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2759
2175void 2760void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2761set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2762 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2764
2180void 2765void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2766set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2767 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2769
2185void 2770void
2186set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2771set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2772 ALIAS:
2773 set_image2d = 0
2774 set_image3d = 0
2187 CODE: 2775 CODE:
2188 clSetKernelArg (self, idx, sizeof (value), &value); 2776 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 2777
2195void 2778void
2196set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2779set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2197 CODE: 2780 CODE:
2198 clSetKernelArg (self, idx, sizeof (value), &value); 2781 clSetKernelArg (self, idx, sizeof (value), &value);
2244 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2827 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2245 EXTEND (SP, 1); 2828 EXTEND (SP, 1);
2246 const int i = 0; 2829 const int i = 0;
2247 { 2830 {
2248 NEED_SUCCESS (RetainContext, (value [i])); 2831 NEED_SUCCESS (RetainContext, (value [i]));
2249 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2832 PUSH_CLOBJ (stash_context, value [i]);
2250 } 2833 }
2251 2834
2252void 2835void
2253program (OpenCL::Kernel self) 2836program (OpenCL::Kernel self)
2254 PPCODE: 2837 PPCODE:
2256 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2839 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2257 EXTEND (SP, 1); 2840 EXTEND (SP, 1);
2258 const int i = 0; 2841 const int i = 0;
2259 { 2842 {
2260 NEED_SUCCESS (RetainProgram, (value [i])); 2843 NEED_SUCCESS (RetainProgram, (value [i]));
2261 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2844 PUSH_CLOBJ (stash_program, value [i]);
2262 } 2845 }
2263 2846
2264#END:kernel 2847#END:kernel
2265 2848
2266void 2849void
2313 const int i = 0; 2896 const int i = 0;
2314 PUSHs (sv_2mortal (newSVuv (value [i]))); 2897 PUSHs (sv_2mortal (newSVuv (value [i])));
2315 2898
2316#END:kernel_work_group 2899#END:kernel_work_group
2317 2900
2901#if CL_VERSION_1_2
2902
2903void
2904arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2905 PPCODE:
2906 size_t size;
2907 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2908 SV *sv = sv_2mortal (newSV (size));
2909 SvUPGRADE (sv, SVt_PV);
2910 SvPOK_only (sv);
2911 SvCUR_set (sv, size);
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2913 XPUSHs (sv);
2914
2915#BEGIN:kernel_arg
2916
2917void
2918arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2919 PPCODE:
2920 cl_kernel_arg_address_qualifier value [1];
2921 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2922 EXTEND (SP, 1);
2923 const int i = 0;
2924 PUSHs (sv_2mortal (newSVuv (value [i])));
2925
2926void
2927arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2928 PPCODE:
2929 cl_kernel_arg_access_qualifier value [1];
2930 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2931 EXTEND (SP, 1);
2932 const int i = 0;
2933 PUSHs (sv_2mortal (newSVuv (value [i])));
2934
2935void
2936arg_type_name (OpenCL::Kernel self, cl_uint idx)
2937 ALIAS:
2938 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2939 arg_name = CL_KERNEL_ARG_NAME
2940 PPCODE:
2941 size_t size;
2942 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2943 char *value = tmpbuf (size);
2944 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2945 EXTEND (SP, 1);
2946 const int i = 0;
2947 PUSHs (sv_2mortal (newSVpv (value, 0)));
2948
2949void
2950arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2951 PPCODE:
2952 cl_kernel_arg_type_qualifier value [1];
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2954 EXTEND (SP, 1);
2955 const int i = 0;
2956 PUSHs (sv_2mortal (newSVuv (value [i])));
2957
2958#END:kernel_arg
2959
2960#endif
2961
2318MODULE = OpenCL PACKAGE = OpenCL::Event 2962MODULE = OpenCL PACKAGE = OpenCL::Event
2319 2963
2320void 2964void
2321DESTROY (OpenCL::Event self) 2965DESTROY (OpenCL::Event self)
2322 CODE: 2966 CODE:
2346 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2990 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2347 EXTEND (SP, 1); 2991 EXTEND (SP, 1);
2348 const int i = 0; 2992 const int i = 0;
2349 { 2993 {
2350 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2994 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2351 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2995 PUSH_CLOBJ (stash_queue, value [i]);
2352 } 2996 }
2353 2997
2354void 2998void
2355command_type (OpenCL::Event self) 2999command_type (OpenCL::Event self)
2356 PPCODE: 3000 PPCODE:
2379 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3023 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2380 EXTEND (SP, 1); 3024 EXTEND (SP, 1);
2381 const int i = 0; 3025 const int i = 0;
2382 { 3026 {
2383 NEED_SUCCESS (RetainContext, (value [i])); 3027 NEED_SUCCESS (RetainContext, (value [i]));
2384 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3028 PUSH_CLOBJ (stash_context, value [i]);
2385 } 3029 }
2386 3030
2387#END:event 3031#END:event
2388 3032
2389void 3033void
2414void 3058void
2415set_status (OpenCL::UserEvent self, cl_int execution_status) 3059set_status (OpenCL::UserEvent self, cl_int execution_status)
2416 CODE: 3060 CODE:
2417 clSetUserEventStatus (self, execution_status); 3061 clSetUserEventStatus (self, execution_status);
2418 3062
3063MODULE = OpenCL PACKAGE = OpenCL::Mapped
3064
3065void
3066DESTROY (SV *self)
3067 CODE:
3068 OpenCL__Mapped mapped = SvMAPPED (self);
3069
3070 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3071 mapped_detach (self, mapped);
3072
3073 clReleaseCommandQueue (mapped->queue);
3074 clReleaseEvent (mapped->event);
3075 Safefree (mapped);
3076
3077void
3078unmap (OpenCL::Mapped self, ...)
3079 CODE:
3080 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3081
3082bool
3083mapped (OpenCL::Mapped self)
3084 CODE:
3085 RETVAL = !!self->ptr;
3086 OUTPUT:
3087 RETVAL
3088
3089void
3090wait (OpenCL::Mapped self)
3091 PPCODE:
3092 if (self->event)
3093 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3094
3095void
3096event (OpenCL::Mapped self)
3097 PPCODE:
3098 if (!self->event)
3099 XSRETURN_UNDEF;
3100
3101 clRetainEvent (self->event);
3102 XPUSH_CLOBJ (stash_event, self->event);
3103
3104size_t
3105size (OpenCL::Mapped self)
3106 CODE:
3107 RETVAL = self->cb;
3108 OUTPUT:
3109 RETVAL
3110
3111IV
3112ptr (OpenCL::Mapped self)
3113 CODE:
3114 RETVAL = PTR2IV (self->ptr);
3115 OUTPUT:
3116 RETVAL
3117
3118void
3119set (OpenCL::Mapped self, size_t offset, SV *data)
3120 CODE:
3121 STRLEN len;
3122 const char *ptr = SvPVbyte (data, len);
3123
3124 if (offset + len > self->cb)
3125 croak ("OpenCL::Mapped::set out of bound condition detected");
3126
3127 memcpy (offset + (char *)self->ptr, ptr, len);
3128
3129MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3130
3131MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3132
3133IV
3134row_pitch (OpenCL::Mapped self)
3135 ALIAS:
3136 slice_pitch = 1
3137 CODE:
3138 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3139 OUTPUT:
3140 RETVAL
3141
3142
3143

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines