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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.54 by root, Wed Apr 25 20:29:03 2012 UTC vs.
Revision 1.72 by root, Sat May 5 12:54:04 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
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 790 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 791 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 792 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 793 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 794 { sizeof (cl_double), "SIZEOF_DOUBLE" },
795 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 796#include "constiv.h"
590 }; 797 };
591 798
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 799 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)); 800 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
801
802 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
803 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
804 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
805 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
806 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
807 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
808 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
809 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
810 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
811 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
812 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
813 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
814 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
815 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
816 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
817 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
818 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
819 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
820 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
821 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
822 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
823 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
824 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 825
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 826 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 827}
597 828
598cl_int 829cl_int
601 RETVAL = res; 832 RETVAL = res;
602 OUTPUT: 833 OUTPUT:
603 RETVAL 834 RETVAL
604 835
605const char * 836const char *
606err2str (cl_int err) 837err2str (cl_int err = res)
607 838
608const char * 839const char *
609enum2str (cl_uint value) 840enum2str (cl_uint value)
610 841
611void 842void
619 list = tmpbuf (sizeof (*list) * count); 850 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 851 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 852
622 EXTEND (SP, count); 853 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 854 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 855 PUSH_CLOBJ (stash_platform, list [i]);
625 856
626void 857void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 858context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 859 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 860 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 861 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 862 XPUSH_CLOBJ_CONTEXT;
632 863
633void 864void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 865context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 866 PPCODE:
636 /* der Gipfel der Kunst */ 867 cl_uint device_count;
868 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
869
870 CONTEXT_NOTIFY_CALLBACK;
871 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
872 XPUSH_CLOBJ_CONTEXT;
637 873
638void 874void
639wait_for_events (...) 875wait_for_events (...)
640 CODE: 876 CODE:
641 EVENT_LIST (0, items); 877 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 878 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 879
644PROTOTYPES: DISABLE 880PROTOTYPES: DISABLE
645 881
646MODULE = OpenCL PACKAGE = OpenCL::Platform 882MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 925 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 926 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 927
692 EXTEND (SP, count); 928 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 929 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 930 PUSH_CLOBJ (stash_device, list [i]);
695 931
696void 932void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 933context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 934 PPCODE:
935 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
936 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
937
938 cl_uint device_count;
939 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
940
941 CONTEXT_NOTIFY_CALLBACK;
942 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
943 XPUSH_CLOBJ_CONTEXT;
944
945void
946context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
947 PPCODE:
948 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
949 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
950
951 CONTEXT_NOTIFY_CALLBACK;
952 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
953 XPUSH_CLOBJ_CONTEXT;
954
955MODULE = OpenCL PACKAGE = OpenCL::Device
956
957void
958info (OpenCL::Device self, cl_device_info name)
959 PPCODE:
960 INFO (Device)
961
962#if CL_VERSION_1_2
963
964void
965sub_devices (OpenCL::Device self, SV *properties)
966 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 967 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 968 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 969
702 AV *av = (AV *)SvRV (devices); 970 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 971
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 972 cl_uint count = av_len ((AV *)properties) + 1;
973 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 974
706 int i; 975 int i;
707 for (i = num_devices; i--; ) 976 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 977 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 978
710 CONTEXT_NOTIFY_CALLBACK; 979 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;
713 980
714void 981 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) 982 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 983 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 984 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719 985
720 CONTEXT_NOTIFY_CALLBACK; 986 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 987 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 988 PUSH_CLOBJ (stash_subdevice, list [i]);
723 989
724MODULE = OpenCL PACKAGE = OpenCL::Device 990#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 991
731#BEGIN:device 992#BEGIN:device
732 993
733void 994void
734type (OpenCL::Device self) 995type (OpenCL::Device self)
735 PPCODE: 996 PPCODE:
736 cl_device_type value [1]; 997 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 998 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 999 EXTEND (SP, 1);
739 const int i = 0; 1000 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1001 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1002
742void 1003void
743vendor_id (OpenCL::Device self) 1004vendor_id (OpenCL::Device self)
744 ALIAS: 1005 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1006 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1151 PPCODE:
891 cl_command_queue_properties value [1]; 1152 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1153 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1154 EXTEND (SP, 1);
894 const int i = 0; 1155 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1156 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1157
897void 1158void
898platform (OpenCL::Device self) 1159platform (OpenCL::Device self)
899 PPCODE: 1160 PPCODE:
900 cl_platform_id value [1]; 1161 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1162 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1163 EXTEND (SP, 1);
903 const int i = 0; 1164 const int i = 0;
904 { 1165 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1166 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1167 }
907 1168
908void 1169void
909name (OpenCL::Device self) 1170name (OpenCL::Device self)
910 ALIAS: 1171 ALIAS:
929 cl_device_id value [1]; 1190 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1191 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1192 EXTEND (SP, 1);
932 const int i = 0; 1193 const int i = 0;
933 { 1194 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1195 PUSH_CLOBJ (stash_device, value [i]);
935 } 1196 }
936 1197
937void 1198void
938partition_types_ext (OpenCL::Device self) 1199partition_types_ext (OpenCL::Device self)
939 ALIAS: 1200 ALIAS:
950 for (i = 0; i < n; ++i) 1211 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1212 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1213
953#END:device 1214#END:device
954 1215
1216MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1217
1218#if CL_VERSION_1_2
1219
1220void
1221DESTROY (OpenCL::SubDevice self)
1222 CODE:
1223 clReleaseDevice (self);
1224
1225#endif
1226
955MODULE = OpenCL PACKAGE = OpenCL::Context 1227MODULE = OpenCL PACKAGE = OpenCL::Context
956 1228
957void 1229void
958DESTROY (OpenCL::Context context) 1230DESTROY (OpenCL::Context self)
959 CODE: 1231 CODE:
960 clReleaseContext (context); 1232 clReleaseContext (self);
961 1233
962void 1234void
963info (OpenCL::Context self, cl_context_info name) 1235info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1236 PPCODE:
965 INFO (Context) 1237 INFO (Context)
966 1238
967void 1239void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1240queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1241 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1242 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1243 XPUSH_CLOBJ (stash_queue, queue);
972 1244
973void 1245void
974user_event (OpenCL::Context self) 1246user_event (OpenCL::Context self)
975 PPCODE: 1247 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1248 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1249 XPUSH_CLOBJ (stash_userevent, ev);
978 1250
979void 1251void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1252buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1253 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1254 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?"); 1255 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1256
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1257 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1258 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1259
988void 1260void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1261buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1262 PPCODE:
991 STRLEN len; 1263 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1264 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1265 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?"); 1266 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)); 1267 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1268 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1269
998#if CL_VERSION_1_2 1270#if CL_VERSION_1_2
999 1271
1000void 1272void
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) 1273image (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: 1274 PPCODE:
1003 STRLEN len; 1275 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1276 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1277 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1278 const cl_image_desc desc = {
1007 type, 1279 type,
1008 width, height, depth, 1280 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1281 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1282 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1283 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1284 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1285 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1286 HV *stash = stash_image;
1015 switch (type) 1287 switch (type)
1016 { 1288 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1289 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1290 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1291 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1292 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1293 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1294 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1295 }
1024 XPUSH_CLOBJ (klass, mem); 1296 XPUSH_CLOBJ (stash, mem);
1025 1297
1026#endif 1298#endif
1027 1299
1028void 1300void
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) 1301image2d (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)); 1307 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1308#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1309 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)); 1310 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1311#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1312 XPUSH_CLOBJ (stash_image2d, mem);
1041 1313
1042void 1314void
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) 1315image3d (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: 1316 PPCODE:
1045 STRLEN len; 1317 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1322#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1323 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)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1325#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1326 XPUSH_CLOBJ (stash_image3d, mem);
1055 1327
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1328#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1329
1058void 1330void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1331gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1332 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1333 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1334 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1335
1064void 1336void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1337gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1338 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1339 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1340 XPUSH_CLOBJ (stash_image2d, mem);
1069 1341
1070#if CL_VERSION_1_2 1342#if CL_VERSION_1_2
1071 1343
1072void 1344void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1345gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1346 ALIAS:
1075 PPCODE: 1347 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1348 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1349 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1350 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1351 HV *stash = stash_memory;
1080 switch (type) 1352 switch (type)
1081 { 1353 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1354 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1355 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1356 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1357 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1358 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1359 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1360 }
1089 XPUSH_CLOBJ (klass, mem); 1361 XPUSH_CLOBJ (stash, mem);
1090 1362
1091#endif 1363#endif
1092 1364
1093void 1365void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1366gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1368#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1369 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1370#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1371 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1372#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1373 XPUSH_CLOBJ (stash_image2d, mem);
1102 1374
1103void 1375void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1376gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1377 PPCODE:
1106#if PREFER_1_1 1378#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1379 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1380#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1381 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1382#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1383 XPUSH_CLOBJ (stash_image3d, mem);
1112 1384
1113#endif 1385#endif
1114 1386
1115void 1387void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1388supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1408
1137void 1409void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1410sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1411 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1412 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1413 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1414
1143void 1415void
1144program_with_source (OpenCL::Context self, SV *program) 1416program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1417 PPCODE:
1146 STRLEN len; 1418 STRLEN len;
1147 size_t len2; 1419 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1420 const char *ptr = SvPVbyte (program, len);
1149 1421
1150 len2 = len; 1422 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1423 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1424 XPUSH_CLOBJ (stash_program, prog);
1425
1426void
1427program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1428 PPCODE:
1429 cl_uint device_count;
1430 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1431
1432 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1433 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1434
1435 binaries = SvRV (binaries);
1436
1437 if (device_count != av_len ((AV *)binaries) + 1)
1438 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1439
1440 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1441 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1442 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1443
1444 int i;
1445 for (i = 0; i < device_count; ++i)
1446 {
1447 STRLEN len;
1448 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1449 length_list [i] = len;
1450 }
1451
1452 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1453 length_list, binary_list,
1454 GIMME_V == G_ARRAY ? status_list : 0, &res));
1455
1456 EXTEND (SP, 2);
1457 PUSH_CLOBJ (stash_program, prog);
1458
1459 if (GIMME_V == G_ARRAY)
1460 {
1461 AV *av = newAV ();
1462 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1463
1464 for (i = device_count; i--; )
1465 av_store (av, i, newSViv (status_list [i]));
1466 }
1467
1468#if CL_VERSION_1_2
1469
1470void
1471program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1472 PPCODE:
1473 cl_uint device_count;
1474 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1475
1476 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1477
1478 XPUSH_CLOBJ (stash_program, prog);
1479
1480void
1481link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1482 CODE:
1483 cl_uint device_count = 0;
1484 cl_device_id *device_list = 0;
1485
1486 if (SvOK (devices))
1487 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1488
1489 cl_uint program_count;
1490 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1491
1492 void *user_data;
1493 program_callback pfn_notify = make_program_callback (notify, &user_data);
1494
1495 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1496 program_count, program_list, pfn_notify, user_data, &res));
1497
1498 XPUSH_CLOBJ (stash_program, prog);
1499
1500#endif
1153 1501
1154#BEGIN:context 1502#BEGIN:context
1155 1503
1156void 1504void
1157reference_count (OpenCL::Context self) 1505reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1522 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1523 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1524 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1525 for (i = 0; i < n; ++i)
1178 { 1526 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1527 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1528 }
1181 1529
1182void 1530void
1183properties (OpenCL::Context self) 1531properties (OpenCL::Context self)
1184 PPCODE: 1532 PPCODE:
1199DESTROY (OpenCL::Queue self) 1547DESTROY (OpenCL::Queue self)
1200 CODE: 1548 CODE:
1201 clReleaseCommandQueue (self); 1549 clReleaseCommandQueue (self);
1202 1550
1203void 1551void
1204enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1552read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1553 ALIAS:
1554 enqueue_read_buffer = 0
1205 PPCODE: 1555 PPCODE:
1206 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1556 EVENT_LIST (6);
1208 1557
1209 SvUPGRADE (data, SVt_PV); 1558 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1559 SvGROW (data, len);
1211 SvPOK_only (data); 1560 SvPOK_only (data);
1212 SvCUR_set (data, len); 1561 SvCUR_set (data, len);
1562
1563 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)); 1564 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1214 1565
1215 if (ev) 1566 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1567 XPUSH_CLOBJ (stash_event, ev);
1217 1568
1218void 1569void
1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1570write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1571 ALIAS:
1572 enqueue_write_buffer = 0
1220 PPCODE: 1573 PPCODE:
1221 cl_event ev = 0; 1574 EVENT_LIST (5);
1575
1222 STRLEN len; 1576 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1577 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5);
1225 1578
1579 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)); 1580 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1227 1581
1228 if (ev) 1582 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1583 XPUSH_CLOBJ (stash_event, ev);
1230 1584
1231#if CL_VERSION_1_2 1585#if CL_VERSION_1_2
1232 1586
1233void 1587void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1588fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1589 ALIAS:
1590 enqueue_fill_buffer = 0
1235 PPCODE: 1591 PPCODE:
1236 cl_event ev = 0; 1592 EVENT_LIST (5);
1593
1237 STRLEN len; 1594 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1595 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240 1596
1597 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)); 1598 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1599
1243 if (ev) 1600 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1601 XPUSH_CLOBJ (stash_event, ev);
1245 1602
1246void 1603void
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, ...) 1604fill_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, ...)
1605 ALIAS:
1606 enqueue_fill_image = 0
1248 PPCODE: 1607 PPCODE:
1249 cl_event ev = 0; 1608 EVENT_LIST (12);
1250 STRLEN len; 1609
1251 const size_t origin [3] = { x, y, z }; 1610 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1611 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254 1612
1255 const cl_float c_f [4] = { r, g, b, a }; 1613 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1614 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1615 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1616 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1619 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1620 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) 1621 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."); 1622 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1623
1624 cl_event ev = 0;
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1625 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)); 1626 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1627
1269 if (ev) 1628 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1629 XPUSH_CLOBJ (stash_event, ev);
1271 1630
1272#endif 1631#endif
1273 1632
1274void 1633void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1634copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1635 ALIAS:
1636 enqueue_copy_buffer = 0
1276 PPCODE: 1637 PPCODE:
1638 EVENT_LIST (6);
1639
1277 cl_event ev = 0; 1640 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)); 1641 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1642
1282 if (ev) 1643 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1644 XPUSH_CLOBJ (stash_event, ev);
1284 1645
1285void 1646void
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, ...) 1647read_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, ...)
1648 ALIAS:
1649 enqueue_read_buffer_rect = 0
1287 PPCODE: 1650 PPCODE:
1288 cl_event ev = 0; 1651 EVENT_LIST (17);
1652
1289 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1653 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 }; 1654 const size_t host_origin[3] = { host_x, host_y, host_z };
1291 const size_t region[3] = { width, height, depth }; 1655 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17);
1293 1656
1294 if (!buf_row_pitch) 1657 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1658 buf_row_pitch = region [0];
1296 1659
1297 if (!buf_slice_pitch) 1660 if (!buf_slice_pitch)
1307 1670
1308 SvUPGRADE (data, SVt_PV); 1671 SvUPGRADE (data, SVt_PV);
1309 SvGROW (data, len); 1672 SvGROW (data, len);
1310 SvPOK_only (data); 1673 SvPOK_only (data);
1311 SvCUR_set (data, len); 1674 SvCUR_set (data, len);
1675
1676 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)); 1677 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 1678
1314 if (ev) 1679 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1680 XPUSH_CLOBJ (stash_event, ev);
1316 1681
1317void 1682void
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, ...) 1683write_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, ...)
1684 ALIAS:
1685 enqueue_write_buffer_rect = 0
1319 PPCODE: 1686 PPCODE:
1320 cl_event ev = 0; 1687 EVENT_LIST (17);
1688
1321 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1689 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 }; 1690 const size_t host_origin[3] = { host_x, host_y, host_z };
1323 const size_t region[3] = { width, height, depth }; 1691 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1692 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1693 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17);
1327 1694
1328 if (!buf_row_pitch) 1695 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1696 buf_row_pitch = region [0];
1330 1697
1331 if (!buf_slice_pitch) 1698 if (!buf_slice_pitch)
1340 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1707 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1341 1708
1342 if (len < min_len) 1709 if (len < min_len)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1710 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1711
1712 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)); 1713 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 1714
1347 if (ev) 1715 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1716 XPUSH_CLOBJ (stash_event, ev);
1349 1717
1350void 1718void
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, ...) 1719copy_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, ...)
1720 ALIAS:
1721 enqueue_copy_buffer_rect = 0
1352 PPCODE: 1722 PPCODE:
1353 cl_event ev = 0; 1723 EVENT_LIST (16);
1724
1354 const size_t src_origin[3] = { src_x, src_y, src_z }; 1725 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 }; 1726 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1356 const size_t region[3] = { width, height, depth }; 1727 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16);
1358 1728
1729 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)); 1730 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 1731
1361 if (ev) 1732 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1733 XPUSH_CLOBJ (stash_event, ev);
1363 1734
1364void 1735void
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, ...) 1736read_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, ...)
1737 ALIAS:
1738 enqueue_read_image = 0
1366 PPCODE: 1739 PPCODE:
1367 cl_event ev = 0; 1740 EVENT_LIST (12);
1741
1368 const size_t src_origin[3] = { src_x, src_y, src_z }; 1742 const size_t src_origin[3] = { src_x, src_y, src_z };
1369 const size_t region[3] = { width, height, depth }; 1743 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12);
1371 1744
1372 if (!row_pitch) 1745 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1746 row_pitch = img_row_pitch (src);
1374 1747
1375 if (depth > 1 && !slice_pitch) 1748 if (depth > 1 && !slice_pitch)
1379 1752
1380 SvUPGRADE (data, SVt_PV); 1753 SvUPGRADE (data, SVt_PV);
1381 SvGROW (data, len); 1754 SvGROW (data, len);
1382 SvPOK_only (data); 1755 SvPOK_only (data);
1383 SvCUR_set (data, len); 1756 SvCUR_set (data, len);
1757
1758 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)); 1759 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 1760
1386 if (ev) 1761 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1762 XPUSH_CLOBJ (stash_event, ev);
1388 1763
1389void 1764void
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, ...) 1765write_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, ...)
1766 ALIAS:
1767 enqueue_write_image = 0
1391 PPCODE: 1768 PPCODE:
1392 cl_event ev = 0; 1769 EVENT_LIST (12);
1770
1393 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1771 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1394 const size_t region[3] = { width, height, depth }; 1772 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1773 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1774 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12);
1398 1775
1399 if (!row_pitch) 1776 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1777 row_pitch = img_row_pitch (dst);
1401 1778
1402 if (depth > 1 && !slice_pitch) 1779 if (depth > 1 && !slice_pitch)
1405 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1782 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1406 1783
1407 if (len < min_len) 1784 if (len < min_len)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1785 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1786
1787 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)); 1788 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 1789
1412 if (ev) 1790 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1791 XPUSH_CLOBJ (stash_event, ev);
1414 1792
1415void 1793void
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, ...) 1794copy_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, ...)
1795 ALIAS:
1796 enqueue_copy_image = 0
1417 PPCODE: 1797 PPCODE:
1418 cl_event ev = 0; 1798 EVENT_LIST (12);
1799
1419 const size_t src_origin[3] = { src_x, src_y, src_z }; 1800 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 }; 1801 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1421 const size_t region[3] = { width, height, depth }; 1802 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12);
1423 1803
1804 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)); 1805 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1806
1426 if (ev) 1807 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1428 1809
1429void 1810void
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, ...) 1811copy_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, ...)
1812 ALIAS:
1813 enqueue_copy_image_to_buffer = 0
1431 PPCODE: 1814 PPCODE:
1815 EVENT_LIST (10);
1816
1817 const size_t src_origin[3] = { src_x, src_y, src_z };
1818 const size_t region [3] = { width, height, depth };
1819
1432 cl_event ev = 0; 1820 cl_event ev = 0;
1821 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1822
1823 if (ev)
1824 XPUSH_CLOBJ (stash_event, ev);
1825
1826void
1827copy_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, ...)
1828 ALIAS:
1829 enqueue_copy_buffer_to_image = 0
1830 PPCODE:
1831 EVENT_LIST (10);
1832
1433 const size_t src_origin[3] = { src_x, src_y, src_z }; 1833 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1834 const size_t region [3] = { width, height, depth };
1835
1836 cl_event ev = 0;
1837 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1838
1839 if (ev)
1840 XPUSH_CLOBJ (stash_event, ev);
1841
1842void
1843map_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, ...)
1844 ALIAS:
1845 enqueue_map_buffer = 0
1846 PPCODE:
1847 EVENT_LIST (6);
1848
1849 size_t cb = SvIV (cb_);
1850
1851 if (!SvOK (cb_))
1852 {
1853 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1854 cb -= offset;
1855 }
1856
1857 cl_event ev;
1858 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1859 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1860
1861void
1862map_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, ...)
1863 ALIAS:
1864 enqueue_map_image = 0
1865 PPCODE:
1866 size_t width = SvIV (width_);
1867 if (!SvOK (width_))
1868 {
1869 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1870 width -= x;
1871 }
1872
1873 size_t height = SvIV (width_);
1874 if (!SvOK (height_))
1875 {
1876 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1877 height -= y;
1878 }
1879
1880 size_t depth = SvIV (width_);
1881 if (!SvOK (depth_))
1882 {
1883 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1884 depth -= z;
1885
1886 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1887 if (!depth)
1888 depth = 1;
1889 }
1890
1891 const size_t origin[3] = { x, y, z };
1434 const size_t region[3] = { width, height, depth }; 1892 const size_t region[3] = { width, height, depth };
1893 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1894 EVENT_LIST (10);
1436 1895
1896 cl_event ev;
1897 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));
1898
1899 size_t cb = slice_pitch ? slice_pitch * region [2]
1900 : row_pitch ? row_pitch * region [1]
1901 : region [0];
1902
1903 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1904
1905void
1906unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1907 PPCODE:
1908 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1909 if (GIMME_V != G_VOID)
1910 {
1911 clRetainEvent (mapped->event);
1912 XPUSH_CLOBJ (stash_event, mapped->event);
1913 }
1914
1915void
1916task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1917 ALIAS:
1918 enqueue_task = 0
1919 PPCODE:
1920 EVENT_LIST (2);
1921
1922 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)); 1923 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1438 1924
1439 if (ev) 1925 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1926 XPUSH_CLOBJ (stash_event, ev);
1441 1927
1442void 1928void
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, ...) 1929nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1930 ALIAS:
1931 enqueue_nd_range_kernel = 0
1468 PPCODE: 1932 PPCODE:
1469 cl_event ev = 0; 1933 EVENT_LIST (5);
1934
1470 size_t *gwo = 0, *gws, *lws = 0; 1935 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1936 int gws_len;
1472 size_t *lists; 1937 size_t *lists;
1473 int i; 1938 int i;
1474 EVENT_LIST (5, items - 5);
1475 1939
1476 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1940 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1941 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1942
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1943 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1944
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1945 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1946
1483 gws = lists + gws_len * 0; 1947 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1948 for (i = 0; i < gws_len; ++i)
1949 {
1485 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1950 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1951 // at least nvidia crashes for 0-sized work group sizes, work around
1952 if (!gws [i])
1953 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1954 }
1486 1955
1487 if (SvOK (global_work_offset)) 1956 if (SvOK (global_work_offset))
1488 { 1957 {
1489 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1958 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"); 1959 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1498 } 1967 }
1499 1968
1500 if (SvOK (local_work_size)) 1969 if (SvOK (local_work_size))
1501 { 1970 {
1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1971 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"); 1972 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1504 1973
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1974 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"); 1975 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1507 1976
1508 lws = lists + gws_len * 2; 1977 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 1978 for (i = 0; i < gws_len; ++i)
1979 {
1510 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1980 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1981 // at least nvidia crashes for 0-sized work group sizes, work around
1982 if (!lws [i])
1983 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1984 }
1511 } 1985 }
1512 1986
1987 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)); 1988 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1514 1989
1515 if (ev) 1990 if (ev)
1516 XPUSH_CLOBJ ("OpenCL::Event", ev); 1991 XPUSH_CLOBJ (stash_event, ev);
1992
1993#if CL_VERSION_1_2
1994
1995void
1996migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1997 ALIAS:
1998 enqueue_migrate_mem_objects = 0
1999 PPCODE:
2000 EVENT_LIST (3);
2001
2002 cl_uint obj_count;
2003 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2004
2005 cl_event ev = 0;
2006 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2007
2008 if (ev)
2009 XPUSH_CLOBJ (stash_event, ev);
2010
2011#endif
1517 2012
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 2013#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 2014
1520void 2015void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2016acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 2017 ALIAS:
2018 release_gl_objects = 1
2019 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 2020 enqueue_release_gl_objects = 1
1524 PPCODE: 2021 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2022 EVENT_LIST (2);
1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2023
2024 cl_uint obj_count;
2025 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1527 2026
1528 cl_event ev = 0; 2027 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 2028
1538 if (ix) 2029 if (ix)
1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2030 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1540 else 2031 else
1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2032 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1542 2033
1543 if (ev) 2034 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 2035 XPUSH_CLOBJ (stash_event, ev);
1545 2036
1546#endif 2037#endif
1547 2038
1548void 2039void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 2040wait_for_events (OpenCL::Queue self, ...)
2041 ALIAS:
2042 enqueue_wait_for_events = 0
1550 CODE: 2043 CODE:
1551 EVENT_LIST (1, items - 1); 2044 EVENT_LIST (1);
1552#if PREFER_1_1 2045#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2046 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 2047#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2048 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 2049#endif
1557 2050
1558void 2051void
1559enqueue_marker (OpenCL::Queue self, ...) 2052marker (OpenCL::Queue self, ...)
2053 ALIAS:
2054 enqueue_marker = 0
1560 PPCODE: 2055 PPCODE:
2056 EVENT_LIST (1);
1561 cl_event ev = 0; 2057 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1 2058#if PREFER_1_1
1564 if (!event_list_count) 2059 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2060 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 2061 else
1567#if CL_VERSION_1_2 2062#if CL_VERSION_1_2
1574#endif 2069#endif
1575#else 2070#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2071 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif 2072#endif
1578 if (ev) 2073 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 2074 XPUSH_CLOBJ (stash_event, ev);
1580 2075
1581void 2076void
1582enqueue_barrier (OpenCL::Queue self, ...) 2077barrier (OpenCL::Queue self, ...)
2078 ALIAS:
2079 enqueue_barrier = 0
1583 PPCODE: 2080 PPCODE:
2081 EVENT_LIST (1);
1584 cl_event ev = 0; 2082 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1 2083#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 2084 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 2085 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 2086 else
1590#if CL_VERSION_1_2 2087#if CL_VERSION_1_2
1600#endif 2097#endif
1601#else 2098#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2099 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif 2100#endif
1604 if (ev) 2101 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 2102 XPUSH_CLOBJ (stash_event, ev);
1606 2103
1607void 2104void
1608flush (OpenCL::Queue self) 2105flush (OpenCL::Queue self)
1609 CODE: 2106 CODE:
1610 NEED_SUCCESS (Flush, (self)); 2107 NEED_SUCCESS (Flush, (self));
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2125 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 2126 EXTEND (SP, 1);
1630 const int i = 0; 2127 const int i = 0;
1631 { 2128 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 2129 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2130 PUSH_CLOBJ (stash_context, value [i]);
1634 } 2131 }
1635 2132
1636void 2133void
1637device (OpenCL::Queue self) 2134device (OpenCL::Queue self)
1638 PPCODE: 2135 PPCODE:
1639 cl_device_id value [1]; 2136 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2137 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 2138 EXTEND (SP, 1);
1642 const int i = 0; 2139 const int i = 0;
1643 { 2140 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2141 PUSH_CLOBJ (stash_device, value [i]);
1645 } 2142 }
1646 2143
1647void 2144void
1648reference_count (OpenCL::Queue self) 2145reference_count (OpenCL::Queue self)
1649 PPCODE: 2146 PPCODE:
1658 PPCODE: 2155 PPCODE:
1659 cl_command_queue_properties value [1]; 2156 cl_command_queue_properties value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2157 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 2158 EXTEND (SP, 1);
1662 const int i = 0; 2159 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 2160 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 2161
1665#END:command_queue 2162#END:command_queue
1666 2163
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 2164MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 2165
1683 PPCODE: 2180 PPCODE:
1684 cl_mem_object_type value [1]; 2181 cl_mem_object_type value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2182 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 2183 EXTEND (SP, 1);
1687 const int i = 0; 2184 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 2185 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 2186
1690void 2187void
1691flags (OpenCL::Memory self) 2188flags (OpenCL::Memory self)
1692 PPCODE: 2189 PPCODE:
1693 cl_mem_flags value [1]; 2190 cl_mem_flags value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2191 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2192 EXTEND (SP, 1);
1696 const int i = 0; 2193 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 2194 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 2195
1699void 2196void
1700size (OpenCL::Memory self) 2197size (OpenCL::Memory self)
1701 ALIAS: 2198 ALIAS:
1702 size = CL_MEM_SIZE 2199 size = CL_MEM_SIZE
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2233 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1738 const int i = 0; 2235 const int i = 0;
1739 { 2236 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 2237 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2238 PUSH_CLOBJ (stash_context, value [i]);
1742 } 2239 }
1743 2240
1744void 2241void
1745associated_memobject (OpenCL::Memory self) 2242associated_memobject (OpenCL::Memory self)
1746 PPCODE: 2243 PPCODE:
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2245 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2246 EXTEND (SP, 1);
1750 const int i = 0; 2247 const int i = 0;
1751 { 2248 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 2249 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2250 PUSH_CLOBJ (stash_memory, value [i]);
1754 } 2251 }
1755 2252
1756#END:mem 2253#END:mem
1757 2254
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 2255#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!"); 2275 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1779 2276
1780 cl_buffer_region crdata = { origin, size }; 2277 cl_buffer_region crdata = { origin, size };
1781 2278
1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2279 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2280 XPUSH_CLOBJ (stash_buffer, mem);
1784 2281
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2282MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2283
1787void 2284void
1788image_info (OpenCL::Image self, cl_image_info name) 2285image_info (OpenCL::Image self, cl_image_info name)
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
1876 const int i = 0; 2373 const int i = 0;
1877 { 2374 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2375 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2376 PUSH_CLOBJ (stash_context, value [i]);
1880 } 2377 }
1881 2378
1882void 2379void
1883normalized_coords (OpenCL::Sampler self) 2380normalized_coords (OpenCL::Sampler self)
1884 PPCODE: 2381 PPCODE:
1885 cl_addressing_mode value [1]; 2382 cl_addressing_mode value [1];
1886 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2383 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1887 EXTEND (SP, 1); 2384 EXTEND (SP, 1);
1888 const int i = 0; 2385 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i]))); 2386 PUSHs (sv_2mortal (newSVuv (value [i])));
1890 2387
1891void 2388void
1892addressing_mode (OpenCL::Sampler self) 2389addressing_mode (OpenCL::Sampler self)
1893 PPCODE: 2390 PPCODE:
1894 cl_filter_mode value [1]; 2391 cl_filter_mode value [1];
1895 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2392 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1896 EXTEND (SP, 1); 2393 EXTEND (SP, 1);
1897 const int i = 0; 2394 const int i = 0;
1898 PUSHs (sv_2mortal (newSViv (value [i]))); 2395 PUSHs (sv_2mortal (newSVuv (value [i])));
1899 2396
1900void 2397void
1901filter_mode (OpenCL::Sampler self) 2398filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2399 PPCODE:
1903 cl_bool value [1]; 2400 cl_bool value [1];
1918void 2415void
1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2416build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS: 2417 ALIAS:
1921 build_async = 1 2418 build_async = 1
1922 CODE: 2419 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; 2420 cl_uint device_count = 0;
1926 cl_device_id *device_list = 0; 2421 cl_device_id *device_list = 0;
1927 2422
1928 if (SvOK (devices)) 2423 if (SvOK (devices))
1929 { 2424 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2425
2426 void *user_data;
2427 program_callback pfn_notify = make_program_callback (notify, &user_data);
2428
2429 if (ix)
2430 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2431 else
2432 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2433
2434#if CL_VERSION_1_2
2435
2436void
2437compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2438 CODE:
2439 cl_uint device_count = 0;
2440 cl_device_id *device_list = 0;
2441
2442 if (SvOK (devices))
2443 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2444
2445 cl_uint header_count = 0;
2446 cl_program *header_list = 0;
2447 const char **header_name = 0;
2448
2449 if (SvOK (headers))
2450 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2451 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2452 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1932 2453
1933 AV *av = (AV *)SvRV (devices); 2454 HV *hv = (HV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935 2455
1936 if (num_devices) 2456 header_count = hv_iterinit (hv);
2457 header_list = tmpbuf (sizeof (*header_list) * header_count);
2458 header_name = tmpbuf (sizeof (*header_name) * header_count);
2459
2460 HE *he;
2461 int i = 0;
2462 while (he = hv_iternext (hv))
1937 { 2463 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2464 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2465 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1939 int count; 2466 ++i;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 } 2467 }
1943 } 2468 }
1944 2469
1945 if (SvOK (notify)) 2470 void *user_data;
1946 { 2471 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 2472
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)); 2473 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2474 header_count, header_list, header_name, pfn_notify, user_data));
2475
2476#endif
1956 2477
1957void 2478void
1958build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2479build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1959 PPCODE: 2480 PPCODE:
1960 size_t size; 2481 size_t size;
1995 2516
1996void 2517void
1997kernel (OpenCL::Program program, SV *function) 2518kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2519 PPCODE:
1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2520 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2521 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2522
2002void 2523void
2003kernels_in_program (OpenCL::Program program) 2524kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2525 PPCODE:
2005 cl_uint num_kernels; 2526 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2529 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2530
2010 int i; 2531 int i;
2011 EXTEND (SP, num_kernels); 2532 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2533 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2534 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2535
2015void 2536void
2016info (OpenCL::Program self, cl_program_info name) 2537info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2538 PPCODE:
2018 INFO (Program) 2539 INFO (Program)
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2587 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2588 EXTEND (SP, 1);
2068 const int i = 0; 2589 const int i = 0;
2069 { 2590 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2591 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2592 PUSH_CLOBJ (stash_context, value [i]);
2072 } 2593 }
2073 2594
2074void 2595void
2075devices (OpenCL::Program self) 2596devices (OpenCL::Program self)
2076 PPCODE: 2597 PPCODE:
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2601 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2602 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2603 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2604 for (i = 0; i < n; ++i)
2084 { 2605 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2606 PUSH_CLOBJ (stash_device, value [i]);
2086 } 2607 }
2087 2608
2088void 2609void
2089source (OpenCL::Program self) 2610source (OpenCL::Program self)
2090 PPCODE: 2611 PPCODE:
2116DESTROY (OpenCL::Kernel self) 2637DESTROY (OpenCL::Kernel self)
2117 CODE: 2638 CODE:
2118 clReleaseKernel (self); 2639 clReleaseKernel (self);
2119 2640
2120void 2641void
2642setf (OpenCL::Kernel self, const char *format, ...)
2643 CODE:
2644 int i;
2645 for (i = 2; ; ++i)
2646 {
2647 while (*format == ' ')
2648 ++format;
2649
2650 char type = *format++;
2651
2652 if (!type)
2653 break;
2654
2655 if (i >= items)
2656 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2657
2658 SV *sv = ST (i);
2659
2660 union
2661 {
2662 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2663 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2664 cl_half ch; cl_float cf; cl_double cd;
2665 cl_mem cm;
2666 cl_sampler ca;
2667 size_t cz;
2668 cl_event ce;
2669 } arg;
2670 size_t size;
2671 int nullarg = 0;
2672
2673 switch (type)
2674 {
2675 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2676 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2677 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2678 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2679 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2680 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2681 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2682 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2683
2684 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2685 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2686 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2687
2688 case 'z': nullarg = 1; size = SvIV (sv); break;
2689
2690 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2691 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2692 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2693
2694 default:
2695 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2696 }
2697
2698 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2699 if (res)
2700 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2701 }
2702
2703 if (i != items)
2704 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2705
2706void
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2707set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2708 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2709 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2710
2125void 2711void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2759 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2760
2175void 2761void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2762set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2763 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2764 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2765
2180void 2766void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2767set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2768 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2769 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2770
2185void 2771void
2186set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2772set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2187 ALIAS: 2773 ALIAS:
2188 set_image2d = 0 2774 set_image2d = 0
2189 set_image3d = 0 2775 set_image3d = 0
2190 CODE: 2776 CODE:
2191 clSetKernelArg (self, idx, sizeof (value), &value); 2777 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2192 2778
2193void 2779void
2194set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2780set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2195 CODE: 2781 CODE:
2196 clSetKernelArg (self, idx, sizeof (value), &value); 2782 clSetKernelArg (self, idx, sizeof (value), &value);
2242 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2828 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2243 EXTEND (SP, 1); 2829 EXTEND (SP, 1);
2244 const int i = 0; 2830 const int i = 0;
2245 { 2831 {
2246 NEED_SUCCESS (RetainContext, (value [i])); 2832 NEED_SUCCESS (RetainContext, (value [i]));
2247 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2833 PUSH_CLOBJ (stash_context, value [i]);
2248 } 2834 }
2249 2835
2250void 2836void
2251program (OpenCL::Kernel self) 2837program (OpenCL::Kernel self)
2252 PPCODE: 2838 PPCODE:
2254 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2840 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2255 EXTEND (SP, 1); 2841 EXTEND (SP, 1);
2256 const int i = 0; 2842 const int i = 0;
2257 { 2843 {
2258 NEED_SUCCESS (RetainProgram, (value [i])); 2844 NEED_SUCCESS (RetainProgram, (value [i]));
2259 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2845 PUSH_CLOBJ (stash_program, value [i]);
2260 } 2846 }
2261 2847
2262#END:kernel 2848#END:kernel
2263 2849
2264void 2850void
2311 const int i = 0; 2897 const int i = 0;
2312 PUSHs (sv_2mortal (newSVuv (value [i]))); 2898 PUSHs (sv_2mortal (newSVuv (value [i])));
2313 2899
2314#END:kernel_work_group 2900#END:kernel_work_group
2315 2901
2902#if CL_VERSION_1_2
2903
2904void
2905arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2906 PPCODE:
2907 size_t size;
2908 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2909 SV *sv = sv_2mortal (newSV (size));
2910 SvUPGRADE (sv, SVt_PV);
2911 SvPOK_only (sv);
2912 SvCUR_set (sv, size);
2913 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2914 XPUSHs (sv);
2915
2916#BEGIN:kernel_arg
2917
2918void
2919arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2920 PPCODE:
2921 cl_kernel_arg_address_qualifier value [1];
2922 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2923 EXTEND (SP, 1);
2924 const int i = 0;
2925 PUSHs (sv_2mortal (newSVuv (value [i])));
2926
2927void
2928arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2929 PPCODE:
2930 cl_kernel_arg_access_qualifier value [1];
2931 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2932 EXTEND (SP, 1);
2933 const int i = 0;
2934 PUSHs (sv_2mortal (newSVuv (value [i])));
2935
2936void
2937arg_type_name (OpenCL::Kernel self, cl_uint idx)
2938 ALIAS:
2939 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2940 arg_name = CL_KERNEL_ARG_NAME
2941 PPCODE:
2942 size_t size;
2943 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2944 char *value = tmpbuf (size);
2945 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2946 EXTEND (SP, 1);
2947 const int i = 0;
2948 PUSHs (sv_2mortal (newSVpv (value, 0)));
2949
2950void
2951arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2952 PPCODE:
2953 cl_kernel_arg_type_qualifier value [1];
2954 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2955 EXTEND (SP, 1);
2956 const int i = 0;
2957 PUSHs (sv_2mortal (newSVuv (value [i])));
2958
2959#END:kernel_arg
2960
2961#endif
2962
2316MODULE = OpenCL PACKAGE = OpenCL::Event 2963MODULE = OpenCL PACKAGE = OpenCL::Event
2317 2964
2318void 2965void
2319DESTROY (OpenCL::Event self) 2966DESTROY (OpenCL::Event self)
2320 CODE: 2967 CODE:
2344 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2991 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2345 EXTEND (SP, 1); 2992 EXTEND (SP, 1);
2346 const int i = 0; 2993 const int i = 0;
2347 { 2994 {
2348 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2995 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2349 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2996 PUSH_CLOBJ (stash_queue, value [i]);
2350 } 2997 }
2351 2998
2352void 2999void
2353command_type (OpenCL::Event self) 3000command_type (OpenCL::Event self)
2354 PPCODE: 3001 PPCODE:
2377 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3024 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2378 EXTEND (SP, 1); 3025 EXTEND (SP, 1);
2379 const int i = 0; 3026 const int i = 0;
2380 { 3027 {
2381 NEED_SUCCESS (RetainContext, (value [i])); 3028 NEED_SUCCESS (RetainContext, (value [i]));
2382 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3029 PUSH_CLOBJ (stash_context, value [i]);
2383 } 3030 }
2384 3031
2385#END:event 3032#END:event
2386 3033
2387void 3034void
2412void 3059void
2413set_status (OpenCL::UserEvent self, cl_int execution_status) 3060set_status (OpenCL::UserEvent self, cl_int execution_status)
2414 CODE: 3061 CODE:
2415 clSetUserEventStatus (self, execution_status); 3062 clSetUserEventStatus (self, execution_status);
2416 3063
3064MODULE = OpenCL PACKAGE = OpenCL::Mapped
3065
3066void
3067DESTROY (SV *self)
3068 CODE:
3069 OpenCL__Mapped mapped = SvMAPPED (self);
3070
3071 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3072 mapped_detach (self, mapped);
3073
3074 clReleaseCommandQueue (mapped->queue);
3075 clReleaseEvent (mapped->event);
3076 Safefree (mapped);
3077
3078void
3079unmap (OpenCL::Mapped self, ...)
3080 CODE:
3081 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3082
3083bool
3084mapped (OpenCL::Mapped self)
3085 CODE:
3086 RETVAL = !!self->ptr;
3087 OUTPUT:
3088 RETVAL
3089
3090void
3091wait (OpenCL::Mapped self)
3092 PPCODE:
3093 if (self->event)
3094 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3095
3096void
3097event (OpenCL::Mapped self)
3098 PPCODE:
3099 if (!self->event)
3100 XSRETURN_UNDEF;
3101
3102 clRetainEvent (self->event);
3103 XPUSH_CLOBJ (stash_event, self->event);
3104
3105size_t
3106size (OpenCL::Mapped self)
3107 CODE:
3108 RETVAL = self->cb;
3109 OUTPUT:
3110 RETVAL
3111
3112IV
3113ptr (OpenCL::Mapped self)
3114 CODE:
3115 RETVAL = PTR2IV (self->ptr);
3116 OUTPUT:
3117 RETVAL
3118
3119void
3120set (OpenCL::Mapped self, size_t offset, SV *data)
3121 CODE:
3122 STRLEN len;
3123 const char *ptr = SvPVbyte (data, len);
3124
3125 if (offset + len > self->cb)
3126 croak ("OpenCL::Mapped::set out of bound condition detected");
3127
3128 memcpy (offset + (char *)self->ptr, ptr, len);
3129
3130MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3131
3132MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3133
3134IV
3135row_pitch (OpenCL::Mapped self)
3136 ALIAS:
3137 slice_pitch = 1
3138 CODE:
3139 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3140 OUTPUT:
3141 RETVAL
3142
3143
3144

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines