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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.56 by root, Wed Apr 25 21:48:45 2012 UTC vs.
Revision 1.75 by root, Sat May 5 14:18:35 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
26#if PREFER_1_1 28#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 30#endif
29 31
30#ifdef __APPLE__ 32#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
32#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
33 #include <CL/opencl.h> 38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 50#endif
35 51
36#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 53 #undef PREFER_1_1
38 #define PREFER_1_1 1 54 #define PREFER_1_1 1
39#endif 55#endif
40 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
41typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
61 77
62typedef SV *FUTURE; 78typedef struct mapped * OpenCL__Mapped;
79
80static HV
81 *stash_platform,
82 *stash_device,
83 *stash_subdevice,
84 *stash_context,
85 *stash_queue,
86 *stash_program,
87 *stash_kernel,
88 *stash_sampler,
89 *stash_event,
90 *stash_userevent,
91 *stash_memory,
92 *stash_buffer,
93 *stash_bufferobj,
94 *stash_image,
95 *stash_image1d,
96 *stash_image1darray,
97 *stash_image1dbuffer,
98 *stash_image2d,
99 *stash_image2darray,
100 *stash_image3d,
101 *stash_mapped,
102 *stash_mappedbuffer,
103 *stash_mappedimage;
63 104
64/*****************************************************************************/ 105/*****************************************************************************/
65 106
66// name must include a leading underscore 107// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 148}
108 149
109/*****************************************************************************/ 150/*****************************************************************************/
110 151
111/* up to two temporary buffers */ 152/* up to two temporary buffers */
112static void * 153static void * ecb_noinline
113tmpbuf (size_t size) 154tmpbuf (size_t size)
114{ 155{
115 enum { buffers = 3 }; 156 enum { buffers = 4 };
116 static int idx; 157 static int idx;
117 static void *buf [buffers]; 158 static void *buf [buffers];
118 static size_t len [buffers]; 159 static size_t len [buffers];
119 160
120 idx = (idx + 1) % buffers; 161 idx = (idx + 1) % buffers;
125 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
126 buf [idx] = malloc (len [idx]); 167 buf [idx] = malloc (len [idx]);
127 } 168 }
128 169
129 return buf [idx]; 170 return buf [idx];
171}
172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
130} 186}
131 187
132/*****************************************************************************/ 188/*****************************************************************************/
133 189
134typedef struct 190typedef struct
193 if (res) \ 249 if (res) \
194 FAIL (name); 250 FAIL (name);
195 251
196/*****************************************************************************/ 252/*****************************************************************************/
197 253
254static SV *
255new_clobj (HV *stash, IV id)
256{
257 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
258}
259
260#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
261#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
262
263/* cl objects are either \$iv, or [$iv, ...] */
264/* they can be upgraded at runtime to the array form */
265static void * ecb_noinline
266SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
267{
268 // sv_derived_from is quite slow :(
269 if (SvROK (sv) && sv_derived_from (sv, pkg))
270 return (void *)SvIV (SvRV (sv));
271
272 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
273}
274
275// the "no-inherit" version of the above
276static void * ecb_noinline
277SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
278{
279 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
280 return (void *)SvIV (SvRV (sv));
281
282 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
283}
284
285/*****************************************************************************/
286
198static cl_context_properties * 287static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 288SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 289{
201 if (!sv || !SvOK (sv)) 290 if (!sv || !SvOK (sv))
202 if (extra) 291 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 292 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 293 else
210 int i, len = av_len (av) + 1; 299 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 300 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 301 cl_context_properties *l = p;
213 302
214 if (len & 1) 303 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 304 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 305
217 while (extracount--) 306 while (extracount--)
218 *l++ = *extra++; 307 *l++ = *extra++;
219 308
220 for (i = 0; i < len; i += 2) 309 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 312 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 313 cl_context_properties v = SvIV (p_sv); // code below can override
225 314
226 switch (t) 315 switch (t)
227 { 316 {
317 case CL_CONTEXT_PLATFORM:
318 if (SvROK (p_sv))
319 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
320 break;
321
228 case CL_GLX_DISPLAY_KHR: 322 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 323 if (!SvOK (p_sv))
230 { 324 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 325 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 326 if (func)
255 *l = 0; 349 *l = 0;
256 350
257 return p; 351 return p;
258 } 352 }
259 353
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 354 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 355}
262 356
263/*****************************************************************************/ 357// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 358// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 359static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 360object_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{ 361{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 362 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
275 return (void *)SvIV (SvRV (sv)); 363 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
364 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
276 365
277 croak ("%s: %s is not of type %s", func, svname, pkg); 366 AV *av = (AV *)SvRV (arg);
367 void **list = 0;
368 cl_uint count = av_len (av) + 1;
369
370 if (count)
371 {
372 list = tmpbuf (sizeof (*list) * count);
373 int i;
374 for (i = 0; i < count; ++i)
375 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
376 }
377
378 if (!count && !or_undef)
379 croak ("%s: '%s' must contain at least one %s object",
380 cv_get_name (cv), argname, klass);
381
382 *rcount = count;
383 return (void *)list;
278} 384}
279 385
280/*****************************************************************************/ 386/*****************************************************************************/
281/* callback stuff */ 387/* callback stuff */
282 388
304static void (*eq_signal_func)(void *signal_arg, int value); 410static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 411static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 412static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 413static eq_item *eq_head, *eq_tail;
308 414
309static void 415static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 416eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 417{
312 eq_item *item = malloc (sizeof (eq_item)); 418 eq_item *item = malloc (sizeof (eq_item));
313 419
314 item->next = 0; 420 item->next = 0;
381} 487}
382 488
383/*****************************************************************************/ 489/*****************************************************************************/
384/* context notify */ 490/* context notify */
385 491
386static void 492static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 493eq_context_push (void *data1, void *data2, void *data3)
388{ 494{
389 dSP; 495 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 496 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 497 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 519 { \
414 pfn_notify = eq_context_notify; \ 520 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 521 user_data = s_get_cv (notify); \
416 } 522 }
417 523
418static SV * 524static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 525new_clobj_context (cl_context ctx, void *user_data)
420{ 526{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 527 SV *sv = new_clobj (stash_context, (IV)ctx);
422 528
423 if (user_data) 529 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 530 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 531
426 return sv; 532 return sv;
433 539
434static void 540static void
435eq_program_push (void *data1, void *data2, void *data3) 541eq_program_push (void *data1, void *data2, void *data3)
436{ 542{
437 dSP; 543 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 544 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 545 PUTBACK;
440} 546}
441 547
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 548static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 549
444static void CL_CALLBACK 550static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 551eq_program_notify (cl_program program, void *user_data)
446{ 552{
553 clRetainProgram (program);
554
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 555 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
556}
557
558typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
559
560static program_callback ecb_noinline
561make_program_callback (SV *notify, void **ruser_data)
562{
563 if (SvOK (notify))
564 {
565 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
566 return eq_program_notify;
567 }
568 else
569 {
570 *ruser_data = 0;
571 return 0;
572 }
448} 573}
449 574
450struct build_args 575struct build_args
451{ 576{
452 cl_program program; 577 cl_program program;
466 else 591 else
467 clReleaseProgram (arg->program); 592 clReleaseProgram (arg->program);
468 593
469 free (arg->options); 594 free (arg->options);
470 free (arg); 595 free (arg);
596
597 return 0;
471} 598}
472 599
473static void 600static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 601build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 602{
490 617
491static void 618static void
492eq_event_push (void *data1, void *data2, void *data3) 619eq_event_push (void *data1, void *data2, void *data3)
493{ 620{
494 dSP; 621 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 622 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 623 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 624 PUTBACK;
498} 625}
499 626
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 627static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 632 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 633 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 634}
508 635
509/*****************************************************************************/ 636/*****************************************************************************/
637/* utilities for XS code */
510 638
511static size_t 639static size_t
512img_row_pitch (cl_mem img) 640img_row_pitch (cl_mem img)
513{ 641{
514 size_t res; 642 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 643 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 644 return res;
517} 645}
518 646
519static cl_event * 647static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 648event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 649{
522 cl_uint count = *rcount; 650 cl_uint count = *rcount;
523 651
524 if (!count) 652 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
653 *rcount = count = 0;
654
655 if (!count && !extra)
525 return 0; 656 return 0;
526 657
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 658 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 659 int i = 0;
529 660
530 do 661 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 662 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 663 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 664
536 while (count); 665 if (extra)
666 list [i++] = extra;
537 667
538 *rcount = i; 668 *rcount = i;
539 669
540 return i ? list : 0; 670 return i ? list : 0;
541} 671}
542 672
543#define EVENT_LIST(items,count) \ 673#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 674 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 675 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 676
547#define INFO(class) \ 677#define INFO(class) \
548{ \ 678{ \
549 size_t size; \ 679 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 680 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
553 SvPOK_only (sv); \ 683 SvPOK_only (sv); \
554 SvCUR_set (sv, size); \ 684 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 685 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 686 XPUSHs (sv); \
557} 687}
688
689/*****************************************************************************/
690/* mapped_xxx */
691
692static OpenCL__Mapped
693SvMAPPED (SV *sv)
694{
695 // no typechecking atm., keep your fingers crossed
696 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
697}
698
699struct mapped
700{
701 cl_command_queue queue;
702 cl_mem memobj;
703 void *ptr;
704 size_t cb;
705 cl_event event;
706 size_t row_pitch;
707 size_t slice_pitch;
708};
709
710static SV *
711mapped_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)
712{
713 SV *data = newSV (0);
714 SvUPGRADE (data, SVt_PVMG);
715
716 OpenCL__Mapped mapped;
717 New (0, mapped, 1, struct mapped);
718
719 clRetainCommandQueue (queue);
720
721 mapped->queue = queue;
722 mapped->memobj = memobj;
723 mapped->ptr = ptr;
724 mapped->cb = cb;
725 mapped->event = ev;
726 mapped->row_pitch = row_pitch;
727 mapped->slice_pitch = slice_pitch;
728
729 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
730
731 if (SvLEN (data))
732 Safefree (data);
733
734 SvPVX (data) = (char *)ptr;
735 SvCUR_set (data, cb);
736 SvLEN_set (data, 0);
737 SvPOK_only (data);
738
739 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
740
741 if (!(flags & CL_MAP_WRITE))
742 SvREADONLY_on (data);
743
744 return obj;
745}
746
747static void
748mapped_detach (SV *sv, OpenCL__Mapped mapped)
749{
750 SV *data = SvRV (sv);
751
752 // the next check checks both before AND after detach, where SvPVX should STILL be 0
753 if (SvPVX (data) != (char *)mapped->ptr)
754 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
755 else
756 {
757 SvREADONLY_off (data);
758 SvCUR_set (data, 0);
759 SvPVX (data) = 0;
760 SvOK_off (data);
761 }
762
763 mapped->ptr = 0;
764}
765
766static void
767mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
768{
769 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
770 cl_event ev;
771
772 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
773
774 clReleaseEvent (mapped->event);
775 mapped->event = ev;
776
777 mapped_detach (self, mapped);
778}
779
780/*****************************************************************************/
558 781
559MODULE = OpenCL PACKAGE = OpenCL 782MODULE = OpenCL PACKAGE = OpenCL
560 783
561PROTOTYPES: ENABLE 784PROTOTYPES: ENABLE
562 785
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 807 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 808 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 809 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 810 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 811 { sizeof (cl_double), "SIZEOF_DOUBLE" },
812 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 813#include "constiv.h"
590 }; 814 };
591 815
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 816 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)); 817 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
818
819 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
820 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
821 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
822 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
823 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
824 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
825 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
826 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
827 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
828 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
829 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
830 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
831 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
832 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
833 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
834 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
835 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
836 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
837 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
838 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
839 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
840 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
841 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 842
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 843 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 844}
597 845
598cl_int 846cl_int
601 RETVAL = res; 849 RETVAL = res;
602 OUTPUT: 850 OUTPUT:
603 RETVAL 851 RETVAL
604 852
605const char * 853const char *
606err2str (cl_int err) 854err2str (cl_int err = res)
607 855
608const char * 856const char *
609enum2str (cl_uint value) 857enum2str (cl_uint value)
610 858
611void 859void
619 list = tmpbuf (sizeof (*list) * count); 867 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 868 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 869
622 EXTEND (SP, count); 870 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 871 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 872 PUSH_CLOBJ (stash_platform, list [i]);
625 873
626void 874void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 875context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 876 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 877 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 878 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 879 XPUSH_CLOBJ_CONTEXT;
632 880
633void 881void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 882context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 883 PPCODE:
636 /* der Gipfel der Kunst */ 884 cl_uint device_count;
885 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
886
887 CONTEXT_NOTIFY_CALLBACK;
888 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
889 XPUSH_CLOBJ_CONTEXT;
637 890
638void 891void
639wait_for_events (...) 892wait_for_events (...)
640 CODE: 893 CODE:
641 EVENT_LIST (0, items); 894 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 895 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 896
644PROTOTYPES: DISABLE 897PROTOTYPES: DISABLE
645 898
646MODULE = OpenCL PACKAGE = OpenCL::Platform 899MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 942 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 943 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 944
692 EXTEND (SP, count); 945 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 946 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 947 PUSH_CLOBJ (stash_device, list [i]);
695 948
696void 949void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 950context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 951 PPCODE:
952 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
953 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
954
955 cl_uint device_count;
956 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
957
958 CONTEXT_NOTIFY_CALLBACK;
959 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
960 XPUSH_CLOBJ_CONTEXT;
961
962void
963context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
964 PPCODE:
965 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
966 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
967
968 CONTEXT_NOTIFY_CALLBACK;
969 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
970 XPUSH_CLOBJ_CONTEXT;
971
972MODULE = OpenCL PACKAGE = OpenCL::Device
973
974void
975info (OpenCL::Device self, cl_device_info name)
976 PPCODE:
977 INFO (Device)
978
979#if CL_VERSION_1_2
980
981void
982sub_devices (OpenCL::Device self, SV *properties)
983 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 984 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 985 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 986
702 AV *av = (AV *)SvRV (devices); 987 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 988
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 989 cl_uint count = av_len ((AV *)properties) + 1;
990 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 991
706 int i; 992 int i;
707 for (i = num_devices; i--; ) 993 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 994 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 995
710 CONTEXT_NOTIFY_CALLBACK; 996 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;
713 997
714void 998 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) 999 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 1000 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1001 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 1002
720 CONTEXT_NOTIFY_CALLBACK; 1003 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1004 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 1005 PUSH_CLOBJ (stash_subdevice, list [i]);
723 1006
724MODULE = OpenCL PACKAGE = OpenCL::Device 1007#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 1008
731#BEGIN:device 1009#BEGIN:device
732 1010
733void 1011void
734type (OpenCL::Device self) 1012type (OpenCL::Device self)
735 PPCODE: 1013 PPCODE:
736 cl_device_type value [1]; 1014 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1015 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 1016 EXTEND (SP, 1);
739 const int i = 0; 1017 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1018 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1019
742void 1020void
743vendor_id (OpenCL::Device self) 1021vendor_id (OpenCL::Device self)
744 ALIAS: 1022 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1023 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1168 PPCODE:
891 cl_command_queue_properties value [1]; 1169 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1170 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1171 EXTEND (SP, 1);
894 const int i = 0; 1172 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1173 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1174
897void 1175void
898platform (OpenCL::Device self) 1176platform (OpenCL::Device self)
899 PPCODE: 1177 PPCODE:
900 cl_platform_id value [1]; 1178 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1179 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1180 EXTEND (SP, 1);
903 const int i = 0; 1181 const int i = 0;
904 { 1182 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1183 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1184 }
907 1185
908void 1186void
909name (OpenCL::Device self) 1187name (OpenCL::Device self)
910 ALIAS: 1188 ALIAS:
929 cl_device_id value [1]; 1207 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1208 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1209 EXTEND (SP, 1);
932 const int i = 0; 1210 const int i = 0;
933 { 1211 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1212 PUSH_CLOBJ (stash_device, value [i]);
935 } 1213 }
936 1214
937void 1215void
938partition_types_ext (OpenCL::Device self) 1216partition_types_ext (OpenCL::Device self)
939 ALIAS: 1217 ALIAS:
950 for (i = 0; i < n; ++i) 1228 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1229 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1230
953#END:device 1231#END:device
954 1232
1233MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1234
1235#if CL_VERSION_1_2
1236
1237void
1238DESTROY (OpenCL::SubDevice self)
1239 CODE:
1240 clReleaseDevice (self);
1241
1242#endif
1243
955MODULE = OpenCL PACKAGE = OpenCL::Context 1244MODULE = OpenCL PACKAGE = OpenCL::Context
956 1245
957void 1246void
958DESTROY (OpenCL::Context context) 1247DESTROY (OpenCL::Context self)
959 CODE: 1248 CODE:
960 clReleaseContext (context); 1249 clReleaseContext (self);
961 1250
962void 1251void
963info (OpenCL::Context self, cl_context_info name) 1252info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1253 PPCODE:
965 INFO (Context) 1254 INFO (Context)
966 1255
967void 1256void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1257queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1258 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1259 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1260 XPUSH_CLOBJ (stash_queue, queue);
972 1261
973void 1262void
974user_event (OpenCL::Context self) 1263user_event (OpenCL::Context self)
975 PPCODE: 1264 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1265 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1266 XPUSH_CLOBJ (stash_userevent, ev);
978 1267
979void 1268void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1269buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1270 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1271 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?"); 1272 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1273
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1274 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1275 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1276
988void 1277void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1278buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1279 PPCODE:
991 STRLEN len; 1280 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1281 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1282 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?"); 1283 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)); 1284 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1285 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1286
998#if CL_VERSION_1_2 1287#if CL_VERSION_1_2
999 1288
1000void 1289void
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 = 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) 1290image (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)
1006 const cl_image_desc desc = { 1295 const cl_image_desc desc = {
1007 type, 1296 type,
1008 width, height, depth, 1297 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1298 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1299 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1300 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1301 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1302 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1303 HV *stash = stash_image;
1015 switch (type) 1304 switch (type)
1016 { 1305 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1306 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1307 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1308 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1309 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1310 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1311 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1312 }
1024 XPUSH_CLOBJ (klass, mem); 1313 XPUSH_CLOBJ (stash, mem);
1025 1314
1026#endif 1315#endif
1027 1316
1028void 1317void
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) 1318image2d (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)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1325#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1326 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)); 1327 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1328#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1329 XPUSH_CLOBJ (stash_image2d, mem);
1041 1330
1042void 1331void
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) 1332image3d (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: 1333 PPCODE:
1045 STRLEN len; 1334 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1339#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1340 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)); 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1342#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1343 XPUSH_CLOBJ (stash_image3d, mem);
1055 1344
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1345#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1346
1058void 1347void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1348gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1349 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1350 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1351 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1352
1064void 1353void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1354gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1355 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1356 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1357 XPUSH_CLOBJ (stash_image2d, mem);
1069 1358
1070#if CL_VERSION_1_2 1359#if CL_VERSION_1_2
1071 1360
1072void 1361void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1362gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1363 ALIAS:
1075 PPCODE: 1364 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1365 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1366 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1367 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1368 HV *stash = stash_memory;
1080 switch (type) 1369 switch (type)
1081 { 1370 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1371 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1372 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1373 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1374 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1375 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1376 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1377 }
1089 XPUSH_CLOBJ (klass, mem); 1378 XPUSH_CLOBJ (stash, mem);
1090 1379
1091#endif 1380#endif
1092 1381
1093void 1382void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1383gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1385#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1386 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1387#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1388 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1389#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1390 XPUSH_CLOBJ (stash_image2d, mem);
1102 1391
1103void 1392void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1393gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1394 PPCODE:
1106#if PREFER_1_1 1395#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1396 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1397#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1398 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1399#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1400 XPUSH_CLOBJ (stash_image3d, mem);
1112 1401
1113#endif 1402#endif
1114 1403
1115void 1404void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1405supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1425
1137void 1426void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1427sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1428 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1429 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1430 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1431
1143void 1432void
1144program_with_source (OpenCL::Context self, SV *program) 1433program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1434 PPCODE:
1146 STRLEN len; 1435 STRLEN len;
1147 size_t len2; 1436 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1437 const char *ptr = SvPVbyte (program, len);
1149 1438
1150 len2 = len; 1439 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1440 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1441 XPUSH_CLOBJ (stash_program, prog);
1442
1443void
1444program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1445 PPCODE:
1446 cl_uint device_count;
1447 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1448
1449 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1450 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1451
1452 binaries = SvRV (binaries);
1453
1454 if (device_count != av_len ((AV *)binaries) + 1)
1455 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1456
1457 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1458 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1459 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1460
1461 int i;
1462 for (i = 0; i < device_count; ++i)
1463 {
1464 STRLEN len;
1465 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1466 length_list [i] = len;
1467 }
1468
1469 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1470 length_list, binary_list,
1471 GIMME_V == G_ARRAY ? status_list : 0, &res));
1472
1473 EXTEND (SP, 2);
1474 PUSH_CLOBJ (stash_program, prog);
1475
1476 if (GIMME_V == G_ARRAY)
1477 {
1478 AV *av = newAV ();
1479 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1480
1481 for (i = device_count; i--; )
1482 av_store (av, i, newSViv (status_list [i]));
1483 }
1484
1485#if CL_VERSION_1_2
1486
1487void
1488program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1489 PPCODE:
1490 cl_uint device_count;
1491 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1492
1493 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1494
1495 XPUSH_CLOBJ (stash_program, prog);
1496
1497void
1498link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1499 CODE:
1500 cl_uint device_count = 0;
1501 cl_device_id *device_list = 0;
1502
1503 if (SvOK (devices))
1504 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1505
1506 cl_uint program_count;
1507 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1508
1509 void *user_data;
1510 program_callback pfn_notify = make_program_callback (notify, &user_data);
1511
1512 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1513 program_count, program_list, pfn_notify, user_data, &res));
1514
1515 XPUSH_CLOBJ (stash_program, prog);
1516
1517#endif
1153 1518
1154#BEGIN:context 1519#BEGIN:context
1155 1520
1156void 1521void
1157reference_count (OpenCL::Context self) 1522reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1539 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1540 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1541 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1542 for (i = 0; i < n; ++i)
1178 { 1543 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1544 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1545 }
1181 1546
1182void 1547void
1183properties (OpenCL::Context self) 1548properties (OpenCL::Context self)
1184 PPCODE: 1549 PPCODE:
1203void 1568void
1204read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1569read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1205 ALIAS: 1570 ALIAS:
1206 enqueue_read_buffer = 0 1571 enqueue_read_buffer = 0
1207 PPCODE: 1572 PPCODE:
1208 cl_event ev = 0;
1209 EVENT_LIST (6, items - 6); 1573 EVENT_LIST (6);
1210 1574
1211 SvUPGRADE (data, SVt_PV); 1575 SvUPGRADE (data, SVt_PV);
1212 SvGROW (data, len); 1576 SvGROW (data, len);
1213 SvPOK_only (data); 1577 SvPOK_only (data);
1214 SvCUR_set (data, len); 1578 SvCUR_set (data, len);
1579
1580 cl_event ev = 0;
1215 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1581 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1216 1582
1217 if (ev) 1583 if (ev)
1218 XPUSH_CLOBJ ("OpenCL::Event", ev); 1584 XPUSH_CLOBJ (stash_event, ev);
1219 1585
1220void 1586void
1221write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1587write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1222 ALIAS: 1588 ALIAS:
1223 enqueue_write_buffer = 0 1589 enqueue_write_buffer = 0
1224 PPCODE: 1590 PPCODE:
1225 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1226 STRLEN len; 1593 STRLEN len;
1227 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1228 EVENT_LIST (5, items - 5);
1229 1595
1596 cl_event ev = 0;
1230 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1597 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1231 1598
1232 if (ev) 1599 if (ev)
1233 XPUSH_CLOBJ ("OpenCL::Event", ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1234 1601
1235#if CL_VERSION_1_2 1602#if CL_VERSION_1_2
1236 1603
1237void 1604void
1238fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1605fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1239 ALIAS: 1606 ALIAS:
1240 enqueue_fill_buffer = 0 1607 enqueue_fill_buffer = 0
1241 PPCODE: 1608 PPCODE:
1242 cl_event ev = 0; 1609 EVENT_LIST (5);
1610
1243 STRLEN len; 1611 STRLEN len;
1244 char *ptr = SvPVbyte (data, len); 1612 char *ptr = SvPVbyte (data, len);
1245 EVENT_LIST (5, items - 5);
1246 1613
1614 cl_event ev = 0;
1247 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1615 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1248 1616
1249 if (ev) 1617 if (ev)
1250 XPUSH_CLOBJ ("OpenCL::Event", ev); 1618 XPUSH_CLOBJ (stash_event, ev);
1251 1619
1252void 1620void
1253fill_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, ...) 1621fill_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, ...)
1254 ALIAS: 1622 ALIAS:
1255 enqueue_fill_image = 0 1623 enqueue_fill_image = 0
1256 PPCODE: 1624 PPCODE:
1257 cl_event ev = 0; 1625 EVENT_LIST (12);
1258 STRLEN len; 1626
1259 const size_t origin [3] = { x, y, z }; 1627 const size_t origin [3] = { x, y, z };
1260 const size_t region [3] = { width, height, depth }; 1628 const size_t region [3] = { width, height, depth };
1261 EVENT_LIST (12, items - 12);
1262 1629
1263 const cl_float c_f [4] = { r, g, b, a }; 1630 const cl_float c_f [4] = { r, g, b, a };
1264 const cl_uint c_u [4] = { r, g, b, a }; 1631 const cl_uint c_u [4] = { r, g, b, a };
1265 const cl_int c_s [4] = { r, g, b, a }; 1632 const cl_int c_s [4] = { r, g, b, a };
1266 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1633 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1269 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1636 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1270 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1637 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1271 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1638 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1272 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1639 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1273 1640
1641 cl_event ev = 0;
1274 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1642 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1275 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1643 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1276 1644
1277 if (ev) 1645 if (ev)
1278 XPUSH_CLOBJ ("OpenCL::Event", ev); 1646 XPUSH_CLOBJ (stash_event, ev);
1279 1647
1280#endif 1648#endif
1281 1649
1282void 1650void
1283copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1651copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1284 ALIAS: 1652 ALIAS:
1285 enqueue_copy_buffer = 0 1653 enqueue_copy_buffer = 0
1286 PPCODE: 1654 PPCODE:
1655 EVENT_LIST (6);
1656
1287 cl_event ev = 0; 1657 cl_event ev = 0;
1288 EVENT_LIST (6, items - 6);
1289
1290 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1658 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1291 1659
1292 if (ev) 1660 if (ev)
1293 XPUSH_CLOBJ ("OpenCL::Event", ev); 1661 XPUSH_CLOBJ (stash_event, ev);
1294 1662
1295void 1663void
1296read_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, ...) 1664read_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, ...)
1297 ALIAS: 1665 ALIAS:
1298 enqueue_read_buffer_rect = 0 1666 enqueue_read_buffer_rect = 0
1299 PPCODE: 1667 PPCODE:
1300 cl_event ev = 0; 1668 EVENT_LIST (17);
1669
1301 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1670 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1302 const size_t host_origin[3] = { host_x, host_y, host_z }; 1671 const size_t host_origin[3] = { host_x, host_y, host_z };
1303 const size_t region[3] = { width, height, depth }; 1672 const size_t region[3] = { width, height, depth };
1304 EVENT_LIST (17, items - 17);
1305 1673
1306 if (!buf_row_pitch) 1674 if (!buf_row_pitch)
1307 buf_row_pitch = region [0]; 1675 buf_row_pitch = region [0];
1308 1676
1309 if (!buf_slice_pitch) 1677 if (!buf_slice_pitch)
1319 1687
1320 SvUPGRADE (data, SVt_PV); 1688 SvUPGRADE (data, SVt_PV);
1321 SvGROW (data, len); 1689 SvGROW (data, len);
1322 SvPOK_only (data); 1690 SvPOK_only (data);
1323 SvCUR_set (data, len); 1691 SvCUR_set (data, len);
1692
1693 cl_event ev = 0;
1324 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)); 1694 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));
1325 1695
1326 if (ev) 1696 if (ev)
1327 XPUSH_CLOBJ ("OpenCL::Event", ev); 1697 XPUSH_CLOBJ (stash_event, ev);
1328 1698
1329void 1699void
1330write_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, ...) 1700write_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, ...)
1331 ALIAS: 1701 ALIAS:
1332 enqueue_write_buffer_rect = 0 1702 enqueue_write_buffer_rect = 0
1333 PPCODE: 1703 PPCODE:
1334 cl_event ev = 0; 1704 EVENT_LIST (17);
1705
1335 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1706 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1336 const size_t host_origin[3] = { host_x, host_y, host_z }; 1707 const size_t host_origin[3] = { host_x, host_y, host_z };
1337 const size_t region[3] = { width, height, depth }; 1708 const size_t region[3] = { width, height, depth };
1338 STRLEN len; 1709 STRLEN len;
1339 char *ptr = SvPVbyte (data, len); 1710 char *ptr = SvPVbyte (data, len);
1340 EVENT_LIST (17, items - 17);
1341 1711
1342 if (!buf_row_pitch) 1712 if (!buf_row_pitch)
1343 buf_row_pitch = region [0]; 1713 buf_row_pitch = region [0];
1344 1714
1345 if (!buf_slice_pitch) 1715 if (!buf_slice_pitch)
1354 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1724 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1355 1725
1356 if (len < min_len) 1726 if (len < min_len)
1357 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1727 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1358 1728
1729 cl_event ev = 0;
1359 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)); 1730 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));
1360 1731
1361 if (ev) 1732 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1733 XPUSH_CLOBJ (stash_event, ev);
1363 1734
1364void 1735void
1365copy_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, ...) 1736copy_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, ...)
1366 ALIAS: 1737 ALIAS:
1367 enqueue_copy_buffer_rect = 0 1738 enqueue_copy_buffer_rect = 0
1368 PPCODE: 1739 PPCODE:
1369 cl_event ev = 0; 1740 EVENT_LIST (16);
1741
1370 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 };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1743 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth }; 1744 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16);
1374 1745
1746 cl_event ev = 0;
1375 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)); 1747 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));
1376 1748
1377 if (ev) 1749 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1379 1751
1380void 1752void
1381read_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, ...) 1753read_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, ...)
1382 ALIAS: 1754 ALIAS:
1383 enqueue_read_image = 0 1755 enqueue_read_image = 0
1384 PPCODE: 1756 PPCODE:
1385 cl_event ev = 0; 1757 EVENT_LIST (12);
1758
1386 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1387 const size_t region[3] = { width, height, depth }; 1760 const size_t region[3] = { width, height, depth };
1388 EVENT_LIST (12, items - 12);
1389 1761
1390 if (!row_pitch) 1762 if (!row_pitch)
1391 row_pitch = img_row_pitch (src); 1763 row_pitch = img_row_pitch (src);
1392 1764
1393 if (depth > 1 && !slice_pitch) 1765 if (depth > 1 && !slice_pitch)
1397 1769
1398 SvUPGRADE (data, SVt_PV); 1770 SvUPGRADE (data, SVt_PV);
1399 SvGROW (data, len); 1771 SvGROW (data, len);
1400 SvPOK_only (data); 1772 SvPOK_only (data);
1401 SvCUR_set (data, len); 1773 SvCUR_set (data, len);
1774
1775 cl_event ev = 0;
1402 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)); 1776 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));
1403 1777
1404 if (ev) 1778 if (ev)
1405 XPUSH_CLOBJ ("OpenCL::Event", ev); 1779 XPUSH_CLOBJ (stash_event, ev);
1406 1780
1407void 1781void
1408write_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, ...) 1782write_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, ...)
1409 ALIAS: 1783 ALIAS:
1410 enqueue_write_image = 0 1784 enqueue_write_image = 0
1411 PPCODE: 1785 PPCODE:
1412 cl_event ev = 0; 1786 EVENT_LIST (12);
1787
1413 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1788 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1414 const size_t region[3] = { width, height, depth }; 1789 const size_t region[3] = { width, height, depth };
1415 STRLEN len; 1790 STRLEN len;
1416 char *ptr = SvPVbyte (data, len); 1791 char *ptr = SvPVbyte (data, len);
1417 EVENT_LIST (12, items - 12);
1418 1792
1419 if (!row_pitch) 1793 if (!row_pitch)
1420 row_pitch = img_row_pitch (dst); 1794 row_pitch = img_row_pitch (dst);
1421 1795
1422 if (depth > 1 && !slice_pitch) 1796 if (depth > 1 && !slice_pitch)
1425 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1799 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1426 1800
1427 if (len < min_len) 1801 if (len < min_len)
1428 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1802 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1429 1803
1804 cl_event ev = 0;
1430 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)); 1805 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));
1431 1806
1432 if (ev) 1807 if (ev)
1433 XPUSH_CLOBJ ("OpenCL::Event", ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1434 1809
1435void 1810void
1436copy_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, ...) 1811copy_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, ...)
1437 ALIAS: 1812 ALIAS:
1438 enqueue_copy_image = 0 1813 enqueue_copy_image = 0
1439 PPCODE: 1814 PPCODE:
1440 cl_event ev = 0; 1815 EVENT_LIST (12);
1816
1441 const size_t src_origin[3] = { src_x, src_y, src_z }; 1817 const size_t src_origin[3] = { src_x, src_y, src_z };
1442 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1818 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1443 const size_t region[3] = { width, height, depth }; 1819 const size_t region[3] = { width, height, depth };
1444 EVENT_LIST (12, items - 12);
1445 1820
1821 cl_event ev = 0;
1446 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1822 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1447 1823
1448 if (ev) 1824 if (ev)
1449 XPUSH_CLOBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1450 1826
1451void 1827void
1452copy_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, ...) 1828copy_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, ...)
1453 ALIAS: 1829 ALIAS:
1454 enqueue_copy_image_to_buffer = 0 1830 enqueue_copy_image_to_buffer = 0
1455 PPCODE: 1831 PPCODE:
1832 EVENT_LIST (10);
1833
1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1835 const size_t region [3] = { width, height, depth };
1836
1456 cl_event ev = 0; 1837 cl_event ev = 0;
1457 const size_t src_origin[3] = { src_x, src_y, src_z };
1458 const size_t region[3] = { width, height, depth };
1459 EVENT_LIST (10, items - 10);
1460
1461 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1838 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1462 1839
1463 if (ev) 1840 if (ev)
1464 XPUSH_CLOBJ ("OpenCL::Event", ev); 1841 XPUSH_CLOBJ (stash_event, ev);
1465 1842
1466void 1843void
1467copy_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, ...) 1844copy_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, ...)
1468 ALIAS: 1845 ALIAS:
1469 enqueue_copy_buffer_to_image = 0 1846 enqueue_copy_buffer_to_image = 0
1470 PPCODE: 1847 PPCODE:
1848 EVENT_LIST (10);
1849
1850 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1851 const size_t region [3] = { width, height, depth };
1852
1471 cl_event ev = 0; 1853 cl_event ev = 0;
1854 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1855
1856 if (ev)
1857 XPUSH_CLOBJ (stash_event, ev);
1858
1859void
1860map_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, ...)
1861 ALIAS:
1862 enqueue_map_buffer = 0
1863 PPCODE:
1864 EVENT_LIST (6);
1865
1866 size_t cb = SvIV (cb_);
1867
1868 if (!SvOK (cb_))
1869 {
1870 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1871 cb -= offset;
1872 }
1873
1874 cl_event ev;
1875 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1876 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1877
1878void
1879map_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, ...)
1880 ALIAS:
1881 enqueue_map_image = 0
1882 PPCODE:
1883 size_t width = SvIV (width_);
1884 if (!SvOK (width_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1887 width -= x;
1888 }
1889
1890 size_t height = SvIV (width_);
1891 if (!SvOK (height_))
1892 {
1893 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1894 height -= y;
1895 }
1896
1897 size_t depth = SvIV (width_);
1898 if (!SvOK (depth_))
1899 {
1900 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1901 depth -= z;
1902
1903 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1904 if (!depth)
1905 depth = 1;
1906 }
1907
1472 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1908 const size_t origin[3] = { x, y, z };
1473 const size_t region[3] = { width, height, depth }; 1909 const size_t region[3] = { width, height, depth };
1910 size_t row_pitch, slice_pitch;
1474 EVENT_LIST (10, items - 10); 1911 EVENT_LIST (10);
1475 1912
1476 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1913 cl_event ev;
1914 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));
1477 1915
1478 if (ev) 1916 size_t cb = slice_pitch ? slice_pitch * region [2]
1479 XPUSH_CLOBJ ("OpenCL::Event", ev); 1917 : row_pitch ? row_pitch * region [1]
1918 : region [0];
1919
1920 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1921
1922void
1923unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1924 PPCODE:
1925 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1926 if (GIMME_V != G_VOID)
1927 {
1928 clRetainEvent (mapped->event);
1929 XPUSH_CLOBJ (stash_event, mapped->event);
1930 }
1480 1931
1481void 1932void
1482task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1933task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1483 ALIAS: 1934 ALIAS:
1484 enqueue_task = 0 1935 enqueue_task = 0
1485 PPCODE: 1936 PPCODE:
1937 EVENT_LIST (2);
1938
1486 cl_event ev = 0; 1939 cl_event ev = 0;
1487 EVENT_LIST (2, items - 2);
1488
1489 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1940 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1490 1941
1491 if (ev) 1942 if (ev)
1492 XPUSH_CLOBJ ("OpenCL::Event", ev); 1943 XPUSH_CLOBJ (stash_event, ev);
1493 1944
1494void 1945void
1495nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1946nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1496 ALIAS: 1947 ALIAS:
1497 enqueue_nd_range_kernel = 0 1948 enqueue_nd_range_kernel = 0
1498 PPCODE: 1949 PPCODE:
1499 cl_event ev = 0; 1950 EVENT_LIST (5);
1951
1500 size_t *gwo = 0, *gws, *lws = 0; 1952 size_t *gwo = 0, *gws, *lws = 0;
1501 int gws_len; 1953 int gws_len;
1502 size_t *lists; 1954 size_t *lists;
1503 int i; 1955 int i;
1504 EVENT_LIST (5, items - 5);
1505 1956
1506 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1957 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1507 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1958 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1508 1959
1509 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1960 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1510 1961
1511 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1962 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1512 1963
1513 gws = lists + gws_len * 0; 1964 gws = lists + gws_len * 0;
1514 for (i = 0; i < gws_len; ++i) 1965 for (i = 0; i < gws_len; ++i)
1966 {
1515 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1967 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1968 // at least nvidia crashes for 0-sized work group sizes, work around
1969 if (!gws [i])
1970 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1971 }
1516 1972
1517 if (SvOK (global_work_offset)) 1973 if (SvOK (global_work_offset))
1518 { 1974 {
1519 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1975 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1520 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1976 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1528 } 1984 }
1529 1985
1530 if (SvOK (local_work_size)) 1986 if (SvOK (local_work_size))
1531 { 1987 {
1532 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1988 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1533 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1989 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1534 1990
1535 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1991 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1536 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1992 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1537 1993
1538 lws = lists + gws_len * 2; 1994 lws = lists + gws_len * 2;
1539 for (i = 0; i < gws_len; ++i) 1995 for (i = 0; i < gws_len; ++i)
1996 {
1540 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1997 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1998 // at least nvidia crashes for 0-sized work group sizes, work around
1999 if (!lws [i])
2000 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2001 }
1541 } 2002 }
1542 2003
2004 cl_event ev = 0;
1543 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2005 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1544 2006
1545 if (ev) 2007 if (ev)
1546 XPUSH_CLOBJ ("OpenCL::Event", ev); 2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#if CL_VERSION_1_2
2011
2012void
2013migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2014 ALIAS:
2015 enqueue_migrate_mem_objects = 0
2016 PPCODE:
2017 EVENT_LIST (3);
2018
2019 cl_uint obj_count;
2020 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2021
2022 cl_event ev = 0;
2023 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2024
2025 if (ev)
2026 XPUSH_CLOBJ (stash_event, ev);
2027
2028#endif
1547 2029
1548#if cl_apple_gl_sharing || cl_khr_gl_sharing 2030#if cl_apple_gl_sharing || cl_khr_gl_sharing
1549 2031
1550void 2032void
1551acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2033acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1552 ALIAS: 2034 ALIAS:
2035 release_gl_objects = 1
1553 enqueue_acquire_gl_objects = 0 2036 enqueue_acquire_gl_objects = 0
1554 ALIAS:
1555 enqueue_release_gl_objects = 1 2037 enqueue_release_gl_objects = 1
1556 PPCODE: 2038 PPCODE:
1557 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2039 EVENT_LIST (2);
1558 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2040
2041 cl_uint obj_count;
2042 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1559 2043
1560 cl_event ev = 0; 2044 cl_event ev = 0;
1561 EVENT_LIST (2, items - 2);
1562 AV *av = (AV *)SvRV (objects);
1563 cl_uint num_objects = av_len (av) + 1;
1564 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1565 int i;
1566
1567 for (i = num_objects; i--; )
1568 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1569 2045
1570 if (ix) 2046 if (ix)
1571 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2047 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1572 else 2048 else
1573 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2049 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1574 2050
1575 if (ev) 2051 if (ev)
1576 XPUSH_CLOBJ ("OpenCL::Event", ev); 2052 XPUSH_CLOBJ (stash_event, ev);
1577 2053
1578#endif 2054#endif
1579 2055
1580void 2056void
1581wait_for_events (OpenCL::Queue self, ...) 2057wait_for_events (OpenCL::Queue self, ...)
1582 ALIAS: 2058 ALIAS:
1583 enqueue_wait_for_events = 0 2059 enqueue_wait_for_events = 0
1584 CODE: 2060 CODE:
1585 EVENT_LIST (1, items - 1); 2061 EVENT_LIST (1);
1586#if PREFER_1_1 2062#if PREFER_1_1
1587 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2063 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1588#else 2064#else
1589 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2065 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1590#endif 2066#endif
1592void 2068void
1593marker (OpenCL::Queue self, ...) 2069marker (OpenCL::Queue self, ...)
1594 ALIAS: 2070 ALIAS:
1595 enqueue_marker = 0 2071 enqueue_marker = 0
1596 PPCODE: 2072 PPCODE:
2073 EVENT_LIST (1);
1597 cl_event ev = 0; 2074 cl_event ev = 0;
1598 EVENT_LIST (1, items - 1);
1599#if PREFER_1_1 2075#if PREFER_1_1
1600 if (!event_list_count) 2076 if (!event_list_count)
1601 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2077 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1602 else 2078 else
1603#if CL_VERSION_1_2 2079#if CL_VERSION_1_2
1610#endif 2086#endif
1611#else 2087#else
1612 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2088 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1613#endif 2089#endif
1614 if (ev) 2090 if (ev)
1615 XPUSH_CLOBJ ("OpenCL::Event", ev); 2091 XPUSH_CLOBJ (stash_event, ev);
1616 2092
1617void 2093void
1618barrier (OpenCL::Queue self, ...) 2094barrier (OpenCL::Queue self, ...)
1619 ALIAS: 2095 ALIAS:
1620 enqueue_barrier = 0 2096 enqueue_barrier = 0
1621 PPCODE: 2097 PPCODE:
2098 EVENT_LIST (1);
1622 cl_event ev = 0; 2099 cl_event ev = 0;
1623 EVENT_LIST (1, items - 1);
1624#if PREFER_1_1 2100#if PREFER_1_1
1625 if (!event_list_count && GIMME_V == G_VOID) 2101 if (!event_list_count && GIMME_V == G_VOID)
1626 NEED_SUCCESS (EnqueueBarrier, (self)); 2102 NEED_SUCCESS (EnqueueBarrier, (self));
1627 else 2103 else
1628#if CL_VERSION_1_2 2104#if CL_VERSION_1_2
1638#endif 2114#endif
1639#else 2115#else
1640 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2116 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1641#endif 2117#endif
1642 if (ev) 2118 if (ev)
1643 XPUSH_CLOBJ ("OpenCL::Event", ev); 2119 XPUSH_CLOBJ (stash_event, ev);
1644 2120
1645void 2121void
1646flush (OpenCL::Queue self) 2122flush (OpenCL::Queue self)
1647 CODE: 2123 CODE:
1648 NEED_SUCCESS (Flush, (self)); 2124 NEED_SUCCESS (Flush, (self));
1666 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2142 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1667 EXTEND (SP, 1); 2143 EXTEND (SP, 1);
1668 const int i = 0; 2144 const int i = 0;
1669 { 2145 {
1670 NEED_SUCCESS (RetainContext, (value [i])); 2146 NEED_SUCCESS (RetainContext, (value [i]));
1671 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2147 PUSH_CLOBJ (stash_context, value [i]);
1672 } 2148 }
1673 2149
1674void 2150void
1675device (OpenCL::Queue self) 2151device (OpenCL::Queue self)
1676 PPCODE: 2152 PPCODE:
1677 cl_device_id value [1]; 2153 cl_device_id value [1];
1678 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2154 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1679 EXTEND (SP, 1); 2155 EXTEND (SP, 1);
1680 const int i = 0; 2156 const int i = 0;
1681 { 2157 {
1682 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2158 PUSH_CLOBJ (stash_device, value [i]);
1683 } 2159 }
1684 2160
1685void 2161void
1686reference_count (OpenCL::Queue self) 2162reference_count (OpenCL::Queue self)
1687 PPCODE: 2163 PPCODE:
1696 PPCODE: 2172 PPCODE:
1697 cl_command_queue_properties value [1]; 2173 cl_command_queue_properties value [1];
1698 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2174 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1699 EXTEND (SP, 1); 2175 EXTEND (SP, 1);
1700 const int i = 0; 2176 const int i = 0;
1701 PUSHs (sv_2mortal (newSViv (value [i]))); 2177 PUSHs (sv_2mortal (newSVuv (value [i])));
1702 2178
1703#END:command_queue 2179#END:command_queue
1704 2180
1705MODULE = OpenCL PACKAGE = OpenCL::Memory 2181MODULE = OpenCL PACKAGE = OpenCL::Memory
1706 2182
1721 PPCODE: 2197 PPCODE:
1722 cl_mem_object_type value [1]; 2198 cl_mem_object_type value [1];
1723 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2199 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1724 EXTEND (SP, 1); 2200 EXTEND (SP, 1);
1725 const int i = 0; 2201 const int i = 0;
1726 PUSHs (sv_2mortal (newSViv (value [i]))); 2202 PUSHs (sv_2mortal (newSVuv (value [i])));
1727 2203
1728void 2204void
1729flags (OpenCL::Memory self) 2205flags (OpenCL::Memory self)
1730 PPCODE: 2206 PPCODE:
1731 cl_mem_flags value [1]; 2207 cl_mem_flags value [1];
1732 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2208 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1733 EXTEND (SP, 1); 2209 EXTEND (SP, 1);
1734 const int i = 0; 2210 const int i = 0;
1735 PUSHs (sv_2mortal (newSViv (value [i]))); 2211 PUSHs (sv_2mortal (newSVuv (value [i])));
1736 2212
1737void 2213void
1738size (OpenCL::Memory self) 2214size (OpenCL::Memory self)
1739 ALIAS: 2215 ALIAS:
1740 size = CL_MEM_SIZE 2216 size = CL_MEM_SIZE
1774 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1775 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1776 const int i = 0; 2252 const int i = 0;
1777 { 2253 {
1778 NEED_SUCCESS (RetainContext, (value [i])); 2254 NEED_SUCCESS (RetainContext, (value [i]));
1779 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2255 PUSH_CLOBJ (stash_context, value [i]);
1780 } 2256 }
1781 2257
1782void 2258void
1783associated_memobject (OpenCL::Memory self) 2259associated_memobject (OpenCL::Memory self)
1784 PPCODE: 2260 PPCODE:
1786 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2262 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1787 EXTEND (SP, 1); 2263 EXTEND (SP, 1);
1788 const int i = 0; 2264 const int i = 0;
1789 { 2265 {
1790 NEED_SUCCESS (RetainMemObject, (value [i])); 2266 NEED_SUCCESS (RetainMemObject, (value [i]));
1791 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2267 PUSH_CLOBJ (stash_memory, value [i]);
1792 } 2268 }
1793 2269
1794#END:mem 2270#END:mem
1795 2271
1796#if cl_apple_gl_sharing || cl_khr_gl_sharing 2272#if cl_apple_gl_sharing || cl_khr_gl_sharing
1816 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2292 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1817 2293
1818 cl_buffer_region crdata = { origin, size }; 2294 cl_buffer_region crdata = { origin, size };
1819 2295
1820 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2296 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1821 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2297 XPUSH_CLOBJ (stash_buffer, mem);
1822 2298
1823MODULE = OpenCL PACKAGE = OpenCL::Image 2299MODULE = OpenCL PACKAGE = OpenCL::Image
1824 2300
1825void 2301void
1826image_info (OpenCL::Image self, cl_image_info name) 2302image_info (OpenCL::Image self, cl_image_info name)
1912 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1913 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1914 const int i = 0; 2390 const int i = 0;
1915 { 2391 {
1916 NEED_SUCCESS (RetainContext, (value [i])); 2392 NEED_SUCCESS (RetainContext, (value [i]));
1917 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2393 PUSH_CLOBJ (stash_context, value [i]);
1918 } 2394 }
1919 2395
1920void 2396void
1921normalized_coords (OpenCL::Sampler self) 2397normalized_coords (OpenCL::Sampler self)
1922 PPCODE: 2398 PPCODE:
1923 cl_addressing_mode value [1]; 2399 cl_addressing_mode value [1];
1924 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2400 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1925 EXTEND (SP, 1); 2401 EXTEND (SP, 1);
1926 const int i = 0; 2402 const int i = 0;
1927 PUSHs (sv_2mortal (newSViv (value [i]))); 2403 PUSHs (sv_2mortal (newSVuv (value [i])));
1928 2404
1929void 2405void
1930addressing_mode (OpenCL::Sampler self) 2406addressing_mode (OpenCL::Sampler self)
1931 PPCODE: 2407 PPCODE:
1932 cl_filter_mode value [1]; 2408 cl_filter_mode value [1];
1933 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2409 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1934 EXTEND (SP, 1); 2410 EXTEND (SP, 1);
1935 const int i = 0; 2411 const int i = 0;
1936 PUSHs (sv_2mortal (newSViv (value [i]))); 2412 PUSHs (sv_2mortal (newSVuv (value [i])));
1937 2413
1938void 2414void
1939filter_mode (OpenCL::Sampler self) 2415filter_mode (OpenCL::Sampler self)
1940 PPCODE: 2416 PPCODE:
1941 cl_bool value [1]; 2417 cl_bool value [1];
1956void 2432void
1957build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2433build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1958 ALIAS: 2434 ALIAS:
1959 build_async = 1 2435 build_async = 1
1960 CODE: 2436 CODE:
1961 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1962 void *user_data = 0;
1963 cl_uint num_devices = 0; 2437 cl_uint device_count = 0;
1964 cl_device_id *device_list = 0; 2438 cl_device_id *device_list = 0;
1965 2439
1966 if (SvOK (devices)) 2440 if (SvOK (devices))
1967 { 2441 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2442
2443 void *user_data;
2444 program_callback pfn_notify = make_program_callback (notify, &user_data);
2445
2446 if (ix)
2447 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2448 else
2449 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2450
2451#if CL_VERSION_1_2
2452
2453void
2454compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2455 CODE:
2456 cl_uint device_count = 0;
2457 cl_device_id *device_list = 0;
2458
2459 if (SvOK (devices))
2460 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2461
2462 cl_uint header_count = 0;
2463 cl_program *header_list = 0;
2464 const char **header_name = 0;
2465
2466 if (SvOK (headers))
2467 {
1968 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2468 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1969 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2469 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1970 2470
1971 AV *av = (AV *)SvRV (devices); 2471 HV *hv = (HV *)SvRV (devices);
1972 num_devices = av_len (av) + 1;
1973 2472
1974 if (num_devices) 2473 header_count = hv_iterinit (hv);
2474 header_list = tmpbuf (sizeof (*header_list) * header_count);
2475 header_name = tmpbuf (sizeof (*header_name) * header_count);
2476
2477 HE *he;
2478 int i = 0;
2479 while (he = hv_iternext (hv))
1975 { 2480 {
1976 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2481 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2482 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1977 int count; 2483 ++i;
1978 for (count = 0; count < num_devices; ++count)
1979 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1980 } 2484 }
1981 } 2485 }
1982 2486
1983 if (SvOK (notify)) 2487 void *user_data;
1984 { 2488 program_callback pfn_notify = make_program_callback (notify, &user_data);
1985 NEED_SUCCESS (RetainProgram, (self));
1986 pfn_notify = eq_program_notify;
1987 user_data = SvREFCNT_inc (s_get_cv (notify));
1988 }
1989 2489
1990 if (ix)
1991 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1992 else
1993 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2490 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2491 header_count, header_list, header_name, pfn_notify, user_data));
2492
2493#endif
1994 2494
1995void 2495void
1996build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2496build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1997 PPCODE: 2497 PPCODE:
1998 size_t size; 2498 size_t size;
2027 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2527 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2028 EXTEND (SP, 1); 2528 EXTEND (SP, 1);
2029 const int i = 0; 2529 const int i = 0;
2030 PUSHs (sv_2mortal (newSVpv (value, 0))); 2530 PUSHs (sv_2mortal (newSVpv (value, 0)));
2031 2531
2532void
2533binary_type (OpenCL::Program self, OpenCL::Device device)
2534 PPCODE:
2535 cl_program_binary_type value [1];
2536 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2537 EXTEND (SP, 1);
2538 const int i = 0;
2539 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2540
2032#END:program_build 2541#END:program_build
2033 2542
2034void 2543void
2035kernel (OpenCL::Program program, SV *function) 2544kernel (OpenCL::Program program, SV *function)
2036 PPCODE: 2545 PPCODE:
2037 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2546 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2038 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2547 XPUSH_CLOBJ (stash_kernel, kernel);
2039 2548
2040void 2549void
2041kernels_in_program (OpenCL::Program program) 2550kernels_in_program (OpenCL::Program program)
2042 PPCODE: 2551 PPCODE:
2043 cl_uint num_kernels; 2552 cl_uint num_kernels;
2046 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2555 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2047 2556
2048 int i; 2557 int i;
2049 EXTEND (SP, num_kernels); 2558 EXTEND (SP, num_kernels);
2050 for (i = 0; i < num_kernels; ++i) 2559 for (i = 0; i < num_kernels; ++i)
2051 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2560 PUSH_CLOBJ (stash_kernel, kernels [i]);
2052 2561
2053void 2562void
2054info (OpenCL::Program self, cl_program_info name) 2563info (OpenCL::Program self, cl_program_info name)
2055 PPCODE: 2564 PPCODE:
2056 INFO (Program) 2565 INFO (Program)
2104 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2613 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2105 EXTEND (SP, 1); 2614 EXTEND (SP, 1);
2106 const int i = 0; 2615 const int i = 0;
2107 { 2616 {
2108 NEED_SUCCESS (RetainContext, (value [i])); 2617 NEED_SUCCESS (RetainContext, (value [i]));
2109 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2618 PUSH_CLOBJ (stash_context, value [i]);
2110 } 2619 }
2111 2620
2112void 2621void
2113devices (OpenCL::Program self) 2622devices (OpenCL::Program self)
2114 PPCODE: 2623 PPCODE:
2118 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2627 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2119 int i, n = size / sizeof (*value); 2628 int i, n = size / sizeof (*value);
2120 EXTEND (SP, n); 2629 EXTEND (SP, n);
2121 for (i = 0; i < n; ++i) 2630 for (i = 0; i < n; ++i)
2122 { 2631 {
2123 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2632 PUSH_CLOBJ (stash_device, value [i]);
2124 } 2633 }
2125 2634
2126void 2635void
2127source (OpenCL::Program self) 2636source (OpenCL::Program self)
2128 PPCODE: 2637 PPCODE:
2183 cl_sampler ca; 2692 cl_sampler ca;
2184 size_t cz; 2693 size_t cz;
2185 cl_event ce; 2694 cl_event ce;
2186 } arg; 2695 } arg;
2187 size_t size; 2696 size_t size;
2697 int nullarg = 0;
2188 2698
2189 switch (type) 2699 switch (type)
2190 { 2700 {
2191 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break; 2701 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2192 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break; 2702 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2198 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break; 2708 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2199 2709
2200 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break; 2710 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2201 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2711 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2202 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2712 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2203 case 'z': arg.cz = SvUV (sv); size = sizeof (arg.cz); break;
2204 2713
2714 case 'z': nullarg = 1; size = SvIV (sv); break;
2715
2205 case 'm': arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2716 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2206 case 'a': arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2717 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2207 case 'e': arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2718 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2208 2719
2209 default: 2720 default:
2210 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2721 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2211 } 2722 }
2212 2723
2213 clSetKernelArg (self, i - 2, size, &arg); 2724 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2725 if (res)
2726 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2214 } 2727 }
2215 2728
2216 if (i != items) 2729 if (i != items)
2217 croak ("OpenCL::Kernel::setf format string too short (too many arguments)"); 2730 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2218 2731
2272 clSetKernelArg (self, idx, sizeof (value), &value); 2785 clSetKernelArg (self, idx, sizeof (value), &value);
2273 2786
2274void 2787void
2275set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2788set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2276 CODE: 2789 CODE:
2277 clSetKernelArg (self, idx, sizeof (value), &value); 2790 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2278 2791
2279void 2792void
2280set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2793set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2281 CODE: 2794 CODE:
2282 clSetKernelArg (self, idx, sizeof (value), &value); 2795 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2283 2796
2284void 2797void
2285set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2798set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2286 ALIAS:
2287 set_image2d = 0
2288 set_image3d = 0
2289 CODE: 2799 CODE:
2290 clSetKernelArg (self, idx, sizeof (value), &value); 2800 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2291 2801
2292void 2802void
2293set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2803set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2294 CODE: 2804 CODE:
2295 clSetKernelArg (self, idx, sizeof (value), &value); 2805 clSetKernelArg (self, idx, sizeof (value), &value);
2341 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2851 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2342 EXTEND (SP, 1); 2852 EXTEND (SP, 1);
2343 const int i = 0; 2853 const int i = 0;
2344 { 2854 {
2345 NEED_SUCCESS (RetainContext, (value [i])); 2855 NEED_SUCCESS (RetainContext, (value [i]));
2346 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2856 PUSH_CLOBJ (stash_context, value [i]);
2347 } 2857 }
2348 2858
2349void 2859void
2350program (OpenCL::Kernel self) 2860program (OpenCL::Kernel self)
2351 PPCODE: 2861 PPCODE:
2353 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2863 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2354 EXTEND (SP, 1); 2864 EXTEND (SP, 1);
2355 const int i = 0; 2865 const int i = 0;
2356 { 2866 {
2357 NEED_SUCCESS (RetainProgram, (value [i])); 2867 NEED_SUCCESS (RetainProgram, (value [i]));
2358 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2868 PUSH_CLOBJ (stash_program, value [i]);
2359 } 2869 }
2360 2870
2361#END:kernel 2871#END:kernel
2362 2872
2363void 2873void
2410 const int i = 0; 2920 const int i = 0;
2411 PUSHs (sv_2mortal (newSVuv (value [i]))); 2921 PUSHs (sv_2mortal (newSVuv (value [i])));
2412 2922
2413#END:kernel_work_group 2923#END:kernel_work_group
2414 2924
2925#if CL_VERSION_1_2
2926
2927void
2928arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2929 PPCODE:
2930 size_t size;
2931 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2932 SV *sv = sv_2mortal (newSV (size));
2933 SvUPGRADE (sv, SVt_PV);
2934 SvPOK_only (sv);
2935 SvCUR_set (sv, size);
2936 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2937 XPUSHs (sv);
2938
2939#BEGIN:kernel_arg
2940
2941void
2942arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2943 PPCODE:
2944 cl_kernel_arg_address_qualifier value [1];
2945 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2946 EXTEND (SP, 1);
2947 const int i = 0;
2948 PUSHs (sv_2mortal (newSVuv (value [i])));
2949
2950void
2951arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2952 PPCODE:
2953 cl_kernel_arg_access_qualifier value [1];
2954 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2955 EXTEND (SP, 1);
2956 const int i = 0;
2957 PUSHs (sv_2mortal (newSVuv (value [i])));
2958
2959void
2960arg_type_name (OpenCL::Kernel self, cl_uint idx)
2961 ALIAS:
2962 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2963 arg_name = CL_KERNEL_ARG_NAME
2964 PPCODE:
2965 size_t size;
2966 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2967 char *value = tmpbuf (size);
2968 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2969 EXTEND (SP, 1);
2970 const int i = 0;
2971 PUSHs (sv_2mortal (newSVpv (value, 0)));
2972
2973void
2974arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2975 PPCODE:
2976 cl_kernel_arg_type_qualifier value [1];
2977 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2978 EXTEND (SP, 1);
2979 const int i = 0;
2980 PUSHs (sv_2mortal (newSVuv (value [i])));
2981
2982#END:kernel_arg
2983
2984#endif
2985
2415MODULE = OpenCL PACKAGE = OpenCL::Event 2986MODULE = OpenCL PACKAGE = OpenCL::Event
2416 2987
2417void 2988void
2418DESTROY (OpenCL::Event self) 2989DESTROY (OpenCL::Event self)
2419 CODE: 2990 CODE:
2443 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3014 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2444 EXTEND (SP, 1); 3015 EXTEND (SP, 1);
2445 const int i = 0; 3016 const int i = 0;
2446 { 3017 {
2447 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3018 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2448 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 3019 PUSH_CLOBJ (stash_queue, value [i]);
2449 } 3020 }
2450 3021
2451void 3022void
2452command_type (OpenCL::Event self) 3023command_type (OpenCL::Event self)
2453 PPCODE: 3024 PPCODE:
2476 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3047 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2477 EXTEND (SP, 1); 3048 EXTEND (SP, 1);
2478 const int i = 0; 3049 const int i = 0;
2479 { 3050 {
2480 NEED_SUCCESS (RetainContext, (value [i])); 3051 NEED_SUCCESS (RetainContext, (value [i]));
2481 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3052 PUSH_CLOBJ (stash_context, value [i]);
2482 } 3053 }
2483 3054
2484#END:event 3055#END:event
2485 3056
2486void 3057void
2511void 3082void
2512set_status (OpenCL::UserEvent self, cl_int execution_status) 3083set_status (OpenCL::UserEvent self, cl_int execution_status)
2513 CODE: 3084 CODE:
2514 clSetUserEventStatus (self, execution_status); 3085 clSetUserEventStatus (self, execution_status);
2515 3086
3087MODULE = OpenCL PACKAGE = OpenCL::Mapped
3088
3089void
3090DESTROY (SV *self)
3091 CODE:
3092 OpenCL__Mapped mapped = SvMAPPED (self);
3093
3094 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3095 mapped_detach (self, mapped);
3096
3097 clReleaseCommandQueue (mapped->queue);
3098 clReleaseEvent (mapped->event);
3099 Safefree (mapped);
3100
3101void
3102unmap (OpenCL::Mapped self, ...)
3103 CODE:
3104 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3105
3106bool
3107mapped (OpenCL::Mapped self)
3108 CODE:
3109 RETVAL = !!self->ptr;
3110 OUTPUT:
3111 RETVAL
3112
3113void
3114wait (OpenCL::Mapped self)
3115 PPCODE:
3116 if (self->event)
3117 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3118
3119void
3120event (OpenCL::Mapped self)
3121 PPCODE:
3122 if (!self->event)
3123 XSRETURN_UNDEF;
3124
3125 clRetainEvent (self->event);
3126 XPUSH_CLOBJ (stash_event, self->event);
3127
3128size_t
3129size (OpenCL::Mapped self)
3130 CODE:
3131 RETVAL = self->cb;
3132 OUTPUT:
3133 RETVAL
3134
3135IV
3136ptr (OpenCL::Mapped self)
3137 CODE:
3138 RETVAL = PTR2IV (self->ptr);
3139 OUTPUT:
3140 RETVAL
3141
3142void
3143set (OpenCL::Mapped self, size_t offset, SV *data)
3144 CODE:
3145 STRLEN len;
3146 const char *ptr = SvPVbyte (data, len);
3147
3148 if (offset + len > self->cb)
3149 croak ("OpenCL::Mapped::set out of bound condition detected");
3150
3151 memcpy (offset + (char *)self->ptr, ptr, len);
3152
3153MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3154
3155MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3156
3157IV
3158row_pitch (OpenCL::Mapped self)
3159 ALIAS:
3160 slice_pitch = 1
3161 CODE:
3162 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3163 OUTPUT:
3164 RETVAL
3165
3166
3167

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines