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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.54 by root, Wed Apr 25 20:29:03 2012 UTC vs.
Revision 1.77 by root, Sat May 5 14:57:32 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;
127 } 168 }
128 169
129 return buf [idx]; 170 return buf [idx];
130} 171}
131 172
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;
186}
187
132/*****************************************************************************/ 188/*****************************************************************************/
133 189
134typedef struct 190typedef struct
135{ 191{
136 IV iv; 192 IV iv;
137 const char *name; 193 const char *name;
138 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
139} ivstr; 195} ivstr;
196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
140 203
141static const char * 204static const char *
142iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
143{ 206{
144 int i; 207 int i;
193 if (res) \ 256 if (res) \
194 FAIL (name); 257 FAIL (name);
195 258
196/*****************************************************************************/ 259/*****************************************************************************/
197 260
261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
274{
275 // sv_derived_from is quite slow :(
276 if (SvROK (sv) && sv_derived_from (sv, pkg))
277 return (void *)SvIV (SvRV (sv));
278
279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
290}
291
292/*****************************************************************************/
293
198static cl_context_properties * 294static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 296{
201 if (!sv || !SvOK (sv)) 297 if (!sv || !SvOK (sv))
202 if (extra) 298 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 300 else
210 int i, len = av_len (av) + 1; 306 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 308 cl_context_properties *l = p;
213 309
214 if (len & 1) 310 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 312
217 while (extracount--) 313 while (extracount--)
218 *l++ = *extra++; 314 *l++ = *extra++;
219 315
220 for (i = 0; i < len; i += 2) 316 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 319 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 320 cl_context_properties v = SvIV (p_sv); // code below can override
225 321
226 switch (t) 322 switch (t)
227 { 323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
228 case CL_GLX_DISPLAY_KHR: 329 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 330 if (!SvOK (p_sv))
230 { 331 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 332 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 333 if (func)
255 *l = 0; 356 *l = 0;
256 357
257 return p; 358 return p;
258 } 359 }
259 360
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 362}
262 363
263/*****************************************************************************/ 364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 365// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 366static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 367object_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{ 368{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
275 return (void *)SvIV (SvRV (sv)); 370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
276 372
277 croak ("%s: %s is not of type %s", func, svname, pkg); 373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
378 {
379 list = tmpbuf (sizeof (*list) * count);
380 int i;
381 for (i = 0; i < count; ++i)
382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
383 }
384
385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
278} 391}
279 392
280/*****************************************************************************/ 393/*****************************************************************************/
281/* callback stuff */ 394/* callback stuff */
282 395
304static void (*eq_signal_func)(void *signal_arg, int value); 417static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 418static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 419static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 420static eq_item *eq_head, *eq_tail;
308 421
309static void 422static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 424{
312 eq_item *item = malloc (sizeof (eq_item)); 425 eq_item *item = malloc (sizeof (eq_item));
313 426
314 item->next = 0; 427 item->next = 0;
381} 494}
382 495
383/*****************************************************************************/ 496/*****************************************************************************/
384/* context notify */ 497/* context notify */
385 498
386static void 499static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 500eq_context_push (void *data1, void *data2, void *data3)
388{ 501{
389 dSP; 502 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 526 { \
414 pfn_notify = eq_context_notify; \ 527 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 528 user_data = s_get_cv (notify); \
416 } 529 }
417 530
418static SV * 531static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 532new_clobj_context (cl_context ctx, void *user_data)
420{ 533{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 534 SV *sv = new_clobj (stash_context, (IV)ctx);
422 535
423 if (user_data) 536 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 537 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 538
426 return sv; 539 return sv;
433 546
434static void 547static void
435eq_program_push (void *data1, void *data2, void *data3) 548eq_program_push (void *data1, void *data2, void *data3)
436{ 549{
437 dSP; 550 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 551 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 552 PUTBACK;
440} 553}
441 554
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 556
444static void CL_CALLBACK 557static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 558eq_program_notify (cl_program program, void *user_data)
446{ 559{
560 clRetainProgram (program);
561
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
448} 580}
449 581
450struct build_args 582struct build_args
451{ 583{
452 cl_program program; 584 cl_program program;
466 else 598 else
467 clReleaseProgram (arg->program); 599 clReleaseProgram (arg->program);
468 600
469 free (arg->options); 601 free (arg->options);
470 free (arg); 602 free (arg);
603
604 return 0;
471} 605}
472 606
473static void 607static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 609{
490 624
491static void 625static void
492eq_event_push (void *data1, void *data2, void *data3) 626eq_event_push (void *data1, void *data2, void *data3)
493{ 627{
494 dSP; 628 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 629 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 630 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 631 PUTBACK;
498} 632}
499 633
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 634static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 639 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 640 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 641}
508 642
509/*****************************************************************************/ 643/*****************************************************************************/
644/* utilities for XS code */
510 645
511static size_t 646static size_t
512img_row_pitch (cl_mem img) 647img_row_pitch (cl_mem img)
513{ 648{
514 size_t res; 649 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 650 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 651 return res;
517} 652}
518 653
519static cl_event * 654static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 655event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 656{
522 cl_uint count = *rcount; 657 cl_uint count = *rcount;
523 658
524 if (!count) 659 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
660 *rcount = count = 0;
661
662 if (!count && !extra)
525 return 0; 663 return 0;
526 664
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 665 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 666 int i = 0;
529 667
530 do 668 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 669 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 670 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 671
536 while (count); 672 if (extra)
673 list [i++] = extra;
537 674
538 *rcount = i; 675 *rcount = i;
539 676
540 return i ? list : 0; 677 return i ? list : 0;
541} 678}
542 679
543#define EVENT_LIST(items,count) \ 680#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 681 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 682 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 683
547#define INFO(class) \ 684#define INFO(class) \
548{ \ 685{ \
549 size_t size; \ 686 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 687 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
554 SvCUR_set (sv, size); \ 691 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 692 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 693 XPUSHs (sv); \
557} 694}
558 695
696/*****************************************************************************/
697/* mapped_xxx */
698
699static OpenCL__Mapped
700SvMAPPED (SV *sv)
701{
702 // no typechecking atm., keep your fingers crossed
703 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
704}
705
706struct mapped
707{
708 cl_command_queue queue;
709 cl_mem memobj;
710 void *ptr;
711 size_t cb;
712 cl_event event;
713 size_t row_pitch;
714 size_t slice_pitch;
715};
716
717static SV *
718mapped_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)
719{
720 SV *data = newSV (0);
721 SvUPGRADE (data, SVt_PVMG);
722
723 OpenCL__Mapped mapped;
724 New (0, mapped, 1, struct mapped);
725
726 clRetainCommandQueue (queue);
727
728 mapped->queue = queue;
729 mapped->memobj = memobj;
730 mapped->ptr = ptr;
731 mapped->cb = cb;
732 mapped->event = ev;
733 mapped->row_pitch = row_pitch;
734 mapped->slice_pitch = slice_pitch;
735
736 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
737
738 if (SvLEN (data))
739 Safefree (data);
740
741 SvPVX (data) = (char *)ptr;
742 SvCUR_set (data, cb);
743 SvLEN_set (data, 0);
744 SvPOK_only (data);
745
746 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
747
748 if (!(flags & CL_MAP_WRITE))
749 SvREADONLY_on (data);
750
751 return obj;
752}
753
754static void
755mapped_detach (SV *sv, OpenCL__Mapped mapped)
756{
757 SV *data = SvRV (sv);
758
759 // the next check checks both before AND after detach, where SvPVX should STILL be 0
760 if (SvPVX (data) != (char *)mapped->ptr)
761 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
762 else
763 {
764 SvREADONLY_off (data);
765 SvCUR_set (data, 0);
766 SvPVX (data) = 0;
767 SvOK_off (data);
768 }
769
770 mapped->ptr = 0;
771}
772
773static void
774mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
775{
776 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
777 cl_event ev;
778
779 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
780
781 clReleaseEvent (mapped->event);
782 mapped->event = ev;
783
784 mapped_detach (self, mapped);
785}
786
787/*****************************************************************************/
788
559MODULE = OpenCL PACKAGE = OpenCL 789MODULE = OpenCL PACKAGE = OpenCL
560 790
561PROTOTYPES: ENABLE 791PROTOTYPES: ENABLE
562 792
563void 793void
572 eq_signal_arg = (void*)arg; 802 eq_signal_arg = (void*)arg;
573 803
574BOOT: 804BOOT:
575{ 805{
576 HV *stash = gv_stashpv ("OpenCL", 1); 806 HV *stash = gv_stashpv ("OpenCL", 1);
807
577 static const ivstr *civ, const_iv[] = { 808 static const ivstr *civ, const_iv[] = {
578 { sizeof (cl_char ), "SIZEOF_CHAR" }, 809 { sizeof (cl_char ), "SIZEOF_CHAR" },
579 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 810 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
580 { sizeof (cl_short ), "SIZEOF_SHORT" }, 811 { sizeof (cl_short ), "SIZEOF_SHORT" },
581 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 812 { sizeof (cl_ushort), "SIZEOF_USHORT" },
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 815 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 816 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 817 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 818 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 819 { sizeof (cl_double), "SIZEOF_DOUBLE" },
820 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 821#include "constiv.h"
590 }; 822 };
591 823
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 824 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)); 825 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
826
827 static const nvstr *cnv, const_nv[] = {
828#include "constnv.h"
829 };
830
831 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
832 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
833
834 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
835
836 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
837 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
838 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
839 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
840 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
841 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
842 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
843 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
844 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
845 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
846 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
847 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
848 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
849 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
850 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
851 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
852 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
853 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
854 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
855 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
856 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
857 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
858 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 859
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 860 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 861}
597 862
598cl_int 863cl_int
601 RETVAL = res; 866 RETVAL = res;
602 OUTPUT: 867 OUTPUT:
603 RETVAL 868 RETVAL
604 869
605const char * 870const char *
606err2str (cl_int err) 871err2str (cl_int err = res)
607 872
608const char * 873const char *
609enum2str (cl_uint value) 874enum2str (cl_uint value)
610 875
611void 876void
619 list = tmpbuf (sizeof (*list) * count); 884 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 885 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 886
622 EXTEND (SP, count); 887 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 888 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 889 PUSH_CLOBJ (stash_platform, list [i]);
625 890
626void 891void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 892context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 893 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 894 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 895 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 896 XPUSH_CLOBJ_CONTEXT;
632 897
633void 898void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 899context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 900 PPCODE:
636 /* der Gipfel der Kunst */ 901 cl_uint device_count;
902 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
903
904 CONTEXT_NOTIFY_CALLBACK;
905 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
906 XPUSH_CLOBJ_CONTEXT;
637 907
638void 908void
639wait_for_events (...) 909wait_for_events (...)
640 CODE: 910 CODE:
641 EVENT_LIST (0, items); 911 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 912 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 913
644PROTOTYPES: DISABLE 914PROTOTYPES: DISABLE
645 915
646MODULE = OpenCL PACKAGE = OpenCL::Platform 916MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 959 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 960 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 961
692 EXTEND (SP, count); 962 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 963 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 964 PUSH_CLOBJ (stash_device, list [i]);
695 965
696void 966void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 967context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 968 PPCODE:
969 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
970 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
971
972 cl_uint device_count;
973 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
974
975 CONTEXT_NOTIFY_CALLBACK;
976 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
977 XPUSH_CLOBJ_CONTEXT;
978
979void
980context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
981 PPCODE:
982 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
983 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
984
985 CONTEXT_NOTIFY_CALLBACK;
986 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
987 XPUSH_CLOBJ_CONTEXT;
988
989MODULE = OpenCL PACKAGE = OpenCL::Device
990
991void
992info (OpenCL::Device self, cl_device_info name)
993 PPCODE:
994 INFO (Device)
995
996#if CL_VERSION_1_2
997
998void
999sub_devices (OpenCL::Device self, SV *properties)
1000 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1001 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1002 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 1003
702 AV *av = (AV *)SvRV (devices); 1004 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 1005
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1006 cl_uint count = av_len ((AV *)properties) + 1;
1007 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 1008
706 int i; 1009 int i;
707 for (i = num_devices; i--; ) 1010 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1011 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 1012
710 CONTEXT_NOTIFY_CALLBACK; 1013 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 1014
714void 1015 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) 1016 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 1017 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1018 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 1019
720 CONTEXT_NOTIFY_CALLBACK; 1020 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1021 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 1022 PUSH_CLOBJ (stash_subdevice, list [i]);
723 1023
724MODULE = OpenCL PACKAGE = OpenCL::Device 1024#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 1025
731#BEGIN:device 1026#BEGIN:device
732 1027
733void 1028void
734type (OpenCL::Device self) 1029type (OpenCL::Device self)
735 PPCODE: 1030 PPCODE:
736 cl_device_type value [1]; 1031 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1032 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 1033 EXTEND (SP, 1);
739 const int i = 0; 1034 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1035 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1036
742void 1037void
743vendor_id (OpenCL::Device self) 1038vendor_id (OpenCL::Device self)
744 ALIAS: 1039 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1040 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1185 PPCODE:
891 cl_command_queue_properties value [1]; 1186 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
894 const int i = 0; 1189 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1190 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1191
897void 1192void
898platform (OpenCL::Device self) 1193platform (OpenCL::Device self)
899 PPCODE: 1194 PPCODE:
900 cl_platform_id value [1]; 1195 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1196 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1197 EXTEND (SP, 1);
903 const int i = 0; 1198 const int i = 0;
904 { 1199 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1200 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1201 }
907 1202
908void 1203void
909name (OpenCL::Device self) 1204name (OpenCL::Device self)
910 ALIAS: 1205 ALIAS:
929 cl_device_id value [1]; 1224 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1225 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1226 EXTEND (SP, 1);
932 const int i = 0; 1227 const int i = 0;
933 { 1228 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1229 PUSH_CLOBJ (stash_device, value [i]);
935 } 1230 }
936 1231
937void 1232void
938partition_types_ext (OpenCL::Device self) 1233partition_types_ext (OpenCL::Device self)
939 ALIAS: 1234 ALIAS:
950 for (i = 0; i < n; ++i) 1245 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1246 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1247
953#END:device 1248#END:device
954 1249
1250MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1251
1252#if CL_VERSION_1_2
1253
1254void
1255DESTROY (OpenCL::SubDevice self)
1256 CODE:
1257 clReleaseDevice (self);
1258
1259#endif
1260
955MODULE = OpenCL PACKAGE = OpenCL::Context 1261MODULE = OpenCL PACKAGE = OpenCL::Context
956 1262
957void 1263void
958DESTROY (OpenCL::Context context) 1264DESTROY (OpenCL::Context self)
959 CODE: 1265 CODE:
960 clReleaseContext (context); 1266 clReleaseContext (self);
961 1267
962void 1268void
963info (OpenCL::Context self, cl_context_info name) 1269info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1270 PPCODE:
965 INFO (Context) 1271 INFO (Context)
966 1272
967void 1273void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1274queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1275 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1276 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1277 XPUSH_CLOBJ (stash_queue, queue);
972 1278
973void 1279void
974user_event (OpenCL::Context self) 1280user_event (OpenCL::Context self)
975 PPCODE: 1281 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1282 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1283 XPUSH_CLOBJ (stash_userevent, ev);
978 1284
979void 1285void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1286buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1287 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1288 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?"); 1289 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1290
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1291 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1292 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1293
988void 1294void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1295buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1296 PPCODE:
991 STRLEN len; 1297 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1298 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1299 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?"); 1300 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)); 1301 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1302 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1303
998#if CL_VERSION_1_2 1304#if CL_VERSION_1_2
999 1305
1000void 1306void
1001image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1307image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
1002 PPCODE: 1308 PPCODE:
1003 STRLEN len; 1309 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1310 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1311 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1312 const cl_image_desc desc = {
1007 type, 1313 type,
1008 width, height, depth, 1314 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1315 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1316 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1317 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1318 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1320 HV *stash = stash_image;
1015 switch (type) 1321 switch (type)
1016 { 1322 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1323 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1324 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1325 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1326 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1327 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1328 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1329 }
1024 XPUSH_CLOBJ (klass, mem); 1330 XPUSH_CLOBJ (stash, mem);
1025 1331
1026#endif 1332#endif
1027 1333
1028void 1334void
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) 1335image2d (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)); 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1342#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1343 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)); 1344 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1345#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1346 XPUSH_CLOBJ (stash_image2d, mem);
1041 1347
1042void 1348void
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) 1349image3d (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: 1350 PPCODE:
1045 STRLEN len; 1351 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1355 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1356#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1357 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)); 1358 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1359#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1360 XPUSH_CLOBJ (stash_image3d, mem);
1055 1361
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1362#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1363
1058void 1364void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1365gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1366 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1368 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1369
1064void 1370void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1371gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1372 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1373 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1374 XPUSH_CLOBJ (stash_image2d, mem);
1069 1375
1070#if CL_VERSION_1_2 1376#if CL_VERSION_1_2
1071 1377
1072void 1378void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1379gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1380 ALIAS:
1075 PPCODE: 1381 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1382 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1383 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1384 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1385 HV *stash = stash_memory;
1080 switch (type) 1386 switch (type)
1081 { 1387 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1388 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1389 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1390 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1391 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1392 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1393 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1394 }
1089 XPUSH_CLOBJ (klass, mem); 1395 XPUSH_CLOBJ (stash, mem);
1090 1396
1091#endif 1397#endif
1092 1398
1093void 1399void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1400gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1402#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1403 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1404#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1405 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1406#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1407 XPUSH_CLOBJ (stash_image2d, mem);
1102 1408
1103void 1409void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1410gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1411 PPCODE:
1106#if PREFER_1_1 1412#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1413 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1414#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1415 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1416#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1417 XPUSH_CLOBJ (stash_image3d, mem);
1112 1418
1113#endif 1419#endif
1114 1420
1115void 1421void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1422supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1442
1137void 1443void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1444sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1445 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1446 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1447 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1448
1143void 1449void
1144program_with_source (OpenCL::Context self, SV *program) 1450program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1451 PPCODE:
1146 STRLEN len; 1452 STRLEN len;
1147 size_t len2; 1453 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1454 const char *ptr = SvPVbyte (program, len);
1149 1455
1150 len2 = len; 1456 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1457 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1458 XPUSH_CLOBJ (stash_program, prog);
1459
1460void
1461program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1462 PPCODE:
1463 cl_uint device_count;
1464 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1465
1466 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1467 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1468
1469 binaries = SvRV (binaries);
1470
1471 if (device_count != av_len ((AV *)binaries) + 1)
1472 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1473
1474 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1475 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1476 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1477
1478 int i;
1479 for (i = 0; i < device_count; ++i)
1480 {
1481 STRLEN len;
1482 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1483 length_list [i] = len;
1484 }
1485
1486 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1487 length_list, binary_list,
1488 GIMME_V == G_ARRAY ? status_list : 0, &res));
1489
1490 EXTEND (SP, 2);
1491 PUSH_CLOBJ (stash_program, prog);
1492
1493 if (GIMME_V == G_ARRAY)
1494 {
1495 AV *av = newAV ();
1496 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1497
1498 for (i = device_count; i--; )
1499 av_store (av, i, newSViv (status_list [i]));
1500 }
1501
1502#if CL_VERSION_1_2
1503
1504void
1505program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1506 PPCODE:
1507 cl_uint device_count;
1508 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1509
1510 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514void
1515link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1516 CODE:
1517 cl_uint device_count = 0;
1518 cl_device_id *device_list = 0;
1519
1520 if (SvOK (devices))
1521 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1522
1523 cl_uint program_count;
1524 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1525
1526 void *user_data;
1527 program_callback pfn_notify = make_program_callback (notify, &user_data);
1528
1529 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1530 program_count, program_list, pfn_notify, user_data, &res));
1531
1532 XPUSH_CLOBJ (stash_program, prog);
1533
1534#endif
1153 1535
1154#BEGIN:context 1536#BEGIN:context
1155 1537
1156void 1538void
1157reference_count (OpenCL::Context self) 1539reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1556 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1557 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1558 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1559 for (i = 0; i < n; ++i)
1178 { 1560 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1561 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1562 }
1181 1563
1182void 1564void
1183properties (OpenCL::Context self) 1565properties (OpenCL::Context self)
1184 PPCODE: 1566 PPCODE:
1199DESTROY (OpenCL::Queue self) 1581DESTROY (OpenCL::Queue self)
1200 CODE: 1582 CODE:
1201 clReleaseCommandQueue (self); 1583 clReleaseCommandQueue (self);
1202 1584
1203void 1585void
1204enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1586read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1587 ALIAS:
1588 enqueue_read_buffer = 0
1205 PPCODE: 1589 PPCODE:
1206 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1590 EVENT_LIST (6);
1208 1591
1209 SvUPGRADE (data, SVt_PV); 1592 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1593 SvGROW (data, len);
1211 SvPOK_only (data); 1594 SvPOK_only (data);
1212 SvCUR_set (data, len); 1595 SvCUR_set (data, len);
1596
1597 cl_event ev = 0;
1213 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1598 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1214 1599
1215 if (ev) 1600 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1601 XPUSH_CLOBJ (stash_event, ev);
1217 1602
1218void 1603void
1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1604write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1605 ALIAS:
1606 enqueue_write_buffer = 0
1220 PPCODE: 1607 PPCODE:
1221 cl_event ev = 0; 1608 EVENT_LIST (5);
1609
1222 STRLEN len; 1610 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1611 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5);
1225 1612
1613 cl_event ev = 0;
1226 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1227 1615
1228 if (ev) 1616 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1617 XPUSH_CLOBJ (stash_event, ev);
1230 1618
1231#if CL_VERSION_1_2 1619#if CL_VERSION_1_2
1232 1620
1233void 1621void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1622fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1623 ALIAS:
1624 enqueue_fill_buffer = 0
1235 PPCODE: 1625 PPCODE:
1236 cl_event ev = 0; 1626 EVENT_LIST (5);
1627
1237 STRLEN len; 1628 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1629 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240 1630
1631 cl_event ev = 0;
1241 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1632 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1633
1243 if (ev) 1634 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1635 XPUSH_CLOBJ (stash_event, ev);
1245 1636
1246void 1637void
1247enqueue_fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 1638fill_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, ...)
1639 ALIAS:
1640 enqueue_fill_image = 0
1248 PPCODE: 1641 PPCODE:
1249 cl_event ev = 0; 1642 EVENT_LIST (12);
1250 STRLEN len; 1643
1251 const size_t origin [3] = { x, y, z }; 1644 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1645 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254 1646
1255 const cl_float c_f [4] = { r, g, b, a }; 1647 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1648 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1649 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1650 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1653 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1654 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1263 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1655 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1264 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1656 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1657
1658 cl_event ev = 0;
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1659 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1267 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1660 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1661
1269 if (ev) 1662 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1663 XPUSH_CLOBJ (stash_event, ev);
1271 1664
1272#endif 1665#endif
1273 1666
1274void 1667void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1668copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1669 ALIAS:
1670 enqueue_copy_buffer = 0
1276 PPCODE: 1671 PPCODE:
1672 EVENT_LIST (6);
1673
1277 cl_event ev = 0; 1674 cl_event ev = 0;
1278 EVENT_LIST (6, items - 6);
1279
1280 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1675 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1676
1282 if (ev) 1677 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1678 XPUSH_CLOBJ (stash_event, ev);
1284 1679
1285void 1680void
1286enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1681read_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, ...)
1682 ALIAS:
1683 enqueue_read_buffer_rect = 0
1287 PPCODE: 1684 PPCODE:
1288 cl_event ev = 0; 1685 EVENT_LIST (17);
1686
1289 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1687 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1290 const size_t host_origin[3] = { host_x, host_y, host_z }; 1688 const size_t host_origin[3] = { host_x, host_y, host_z };
1291 const size_t region[3] = { width, height, depth }; 1689 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17);
1293 1690
1294 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
1296 1693
1297 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
1307 1704
1308 SvUPGRADE (data, SVt_PV); 1705 SvUPGRADE (data, SVt_PV);
1309 SvGROW (data, len); 1706 SvGROW (data, len);
1310 SvPOK_only (data); 1707 SvPOK_only (data);
1311 SvCUR_set (data, len); 1708 SvCUR_set (data, len);
1709
1710 cl_event ev = 0;
1312 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1711 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1313 1712
1314 if (ev) 1713 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1714 XPUSH_CLOBJ (stash_event, ev);
1316 1715
1317void 1716void
1318enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1717write_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, ...)
1718 ALIAS:
1719 enqueue_write_buffer_rect = 0
1319 PPCODE: 1720 PPCODE:
1320 cl_event ev = 0; 1721 EVENT_LIST (17);
1722
1321 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1723 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1322 const size_t host_origin[3] = { host_x, host_y, host_z }; 1724 const size_t host_origin[3] = { host_x, host_y, host_z };
1323 const size_t region[3] = { width, height, depth }; 1725 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1726 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1727 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17);
1327 1728
1328 if (!buf_row_pitch) 1729 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1730 buf_row_pitch = region [0];
1330 1731
1331 if (!buf_slice_pitch) 1732 if (!buf_slice_pitch)
1340 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1741 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1341 1742
1342 if (len < min_len) 1743 if (len < min_len)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1744 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1745
1746 cl_event ev = 0;
1345 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1747 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1346 1748
1347 if (ev) 1749 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1349 1751
1350void 1752void
1351enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1753copy_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, ...)
1754 ALIAS:
1755 enqueue_copy_buffer_rect = 0
1352 PPCODE: 1756 PPCODE:
1353 cl_event ev = 0; 1757 EVENT_LIST (16);
1758
1354 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 };
1355 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1760 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1356 const size_t region[3] = { width, height, depth }; 1761 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16);
1358 1762
1763 cl_event ev = 0;
1359 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1764 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1360 1765
1361 if (ev) 1766 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1767 XPUSH_CLOBJ (stash_event, ev);
1363 1768
1364void 1769void
1365enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1770read_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, ...)
1771 ALIAS:
1772 enqueue_read_image = 0
1366 PPCODE: 1773 PPCODE:
1367 cl_event ev = 0; 1774 EVENT_LIST (12);
1775
1368 const size_t src_origin[3] = { src_x, src_y, src_z }; 1776 const size_t src_origin[3] = { src_x, src_y, src_z };
1369 const size_t region[3] = { width, height, depth }; 1777 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12);
1371 1778
1372 if (!row_pitch) 1779 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1780 row_pitch = img_row_pitch (src);
1374 1781
1375 if (depth > 1 && !slice_pitch) 1782 if (depth > 1 && !slice_pitch)
1379 1786
1380 SvUPGRADE (data, SVt_PV); 1787 SvUPGRADE (data, SVt_PV);
1381 SvGROW (data, len); 1788 SvGROW (data, len);
1382 SvPOK_only (data); 1789 SvPOK_only (data);
1383 SvCUR_set (data, len); 1790 SvCUR_set (data, len);
1791
1792 cl_event ev = 0;
1384 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1793 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1385 1794
1386 if (ev) 1795 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1796 XPUSH_CLOBJ (stash_event, ev);
1388 1797
1389void 1798void
1390enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1799write_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, ...)
1800 ALIAS:
1801 enqueue_write_image = 0
1391 PPCODE: 1802 PPCODE:
1392 cl_event ev = 0; 1803 EVENT_LIST (12);
1804
1393 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1805 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1394 const size_t region[3] = { width, height, depth }; 1806 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1807 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1808 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12);
1398 1809
1399 if (!row_pitch) 1810 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1811 row_pitch = img_row_pitch (dst);
1401 1812
1402 if (depth > 1 && !slice_pitch) 1813 if (depth > 1 && !slice_pitch)
1405 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1816 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1406 1817
1407 if (len < min_len) 1818 if (len < min_len)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1819 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1820
1821 cl_event ev = 0;
1410 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1822 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1411 1823
1412 if (ev) 1824 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1414 1826
1415void 1827void
1416enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1828copy_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, ...)
1829 ALIAS:
1830 enqueue_copy_image = 0
1417 PPCODE: 1831 PPCODE:
1418 cl_event ev = 0; 1832 EVENT_LIST (12);
1833
1419 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1420 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1835 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1421 const size_t region[3] = { width, height, depth }; 1836 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12);
1423 1837
1838 cl_event ev = 0;
1424 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1839 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1840
1426 if (ev) 1841 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1842 XPUSH_CLOBJ (stash_event, ev);
1428 1843
1429void 1844void
1430enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1845copy_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, ...)
1846 ALIAS:
1847 enqueue_copy_image_to_buffer = 0
1431 PPCODE: 1848 PPCODE:
1849 EVENT_LIST (10);
1850
1851 const size_t src_origin[3] = { src_x, src_y, src_z };
1852 const size_t region [3] = { width, height, depth };
1853
1432 cl_event ev = 0; 1854 cl_event ev = 0;
1855 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1856
1857 if (ev)
1858 XPUSH_CLOBJ (stash_event, ev);
1859
1860void
1861copy_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, ...)
1862 ALIAS:
1863 enqueue_copy_buffer_to_image = 0
1864 PPCODE:
1865 EVENT_LIST (10);
1866
1433 const size_t src_origin[3] = { src_x, src_y, src_z }; 1867 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1868 const size_t region [3] = { width, height, depth };
1869
1870 cl_event ev = 0;
1871 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1872
1873 if (ev)
1874 XPUSH_CLOBJ (stash_event, ev);
1875
1876void
1877map_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, ...)
1878 ALIAS:
1879 enqueue_map_buffer = 0
1880 PPCODE:
1881 EVENT_LIST (6);
1882
1883 size_t cb = SvIV (cb_);
1884
1885 if (!SvOK (cb_))
1886 {
1887 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1888 cb -= offset;
1889 }
1890
1891 cl_event ev;
1892 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1893 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1894
1895void
1896map_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, ...)
1897 ALIAS:
1898 enqueue_map_image = 0
1899 PPCODE:
1900 size_t width = SvIV (width_);
1901 if (!SvOK (width_))
1902 {
1903 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1904 width -= x;
1905 }
1906
1907 size_t height = SvIV (width_);
1908 if (!SvOK (height_))
1909 {
1910 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1911 height -= y;
1912 }
1913
1914 size_t depth = SvIV (width_);
1915 if (!SvOK (depth_))
1916 {
1917 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1918 depth -= z;
1919
1920 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1921 if (!depth)
1922 depth = 1;
1923 }
1924
1925 const size_t origin[3] = { x, y, z };
1434 const size_t region[3] = { width, height, depth }; 1926 const size_t region[3] = { width, height, depth };
1927 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1928 EVENT_LIST (10);
1436 1929
1930 cl_event ev;
1931 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));
1932
1933 size_t cb = slice_pitch ? slice_pitch * region [2]
1934 : row_pitch ? row_pitch * region [1]
1935 : region [0];
1936
1937 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1938
1939void
1940unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1941 PPCODE:
1942 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1943 if (GIMME_V != G_VOID)
1944 {
1945 clRetainEvent (mapped->event);
1946 XPUSH_CLOBJ (stash_event, mapped->event);
1947 }
1948
1949void
1950task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1951 ALIAS:
1952 enqueue_task = 0
1953 PPCODE:
1954 EVENT_LIST (2);
1955
1956 cl_event ev = 0;
1437 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1957 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1438 1958
1439 if (ev) 1959 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1960 XPUSH_CLOBJ (stash_event, ev);
1441 1961
1442void 1962void
1443enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1444 PPCODE:
1445 cl_event ev = 0;
1446 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1447 const size_t region[3] = { width, height, depth };
1448 EVENT_LIST (10, items - 10);
1449
1450 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1451
1452 if (ev)
1453 XPUSH_CLOBJ ("OpenCL::Event", ev);
1454
1455void
1456enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1457 PPCODE:
1458 cl_event ev = 0;
1459 EVENT_LIST (2, items - 2);
1460
1461 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1462
1463 if (ev)
1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
1465
1466void
1467enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1963nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1964 ALIAS:
1965 enqueue_nd_range_kernel = 0
1468 PPCODE: 1966 PPCODE:
1469 cl_event ev = 0; 1967 EVENT_LIST (5);
1968
1470 size_t *gwo = 0, *gws, *lws = 0; 1969 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1970 int gws_len;
1472 size_t *lists; 1971 size_t *lists;
1473 int i; 1972 int i;
1474 EVENT_LIST (5, items - 5);
1475 1973
1476 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1974 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1976
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1977 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1978
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1979 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1980
1483 gws = lists + gws_len * 0; 1981 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1982 for (i = 0; i < gws_len; ++i)
1983 {
1485 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1984 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1985 // at least nvidia crashes for 0-sized work group sizes, work around
1986 if (!gws [i])
1987 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1988 }
1486 1989
1487 if (SvOK (global_work_offset)) 1990 if (SvOK (global_work_offset))
1488 { 1991 {
1489 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1992 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1490 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1993 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1498 } 2001 }
1499 2002
1500 if (SvOK (local_work_size)) 2003 if (SvOK (local_work_size))
1501 { 2004 {
1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 2005 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1503 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 2006 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1504 2007
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 2008 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1506 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 2009 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1507 2010
1508 lws = lists + gws_len * 2; 2011 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 2012 for (i = 0; i < gws_len; ++i)
2013 {
1510 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 2014 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
2015 // at least nvidia crashes for 0-sized work group sizes, work around
2016 if (!lws [i])
2017 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2018 }
1511 } 2019 }
1512 2020
2021 cl_event ev = 0;
1513 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2022 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1514 2023
1515 if (ev) 2024 if (ev)
1516 XPUSH_CLOBJ ("OpenCL::Event", ev); 2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#if CL_VERSION_1_2
2028
2029void
2030migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2031 ALIAS:
2032 enqueue_migrate_mem_objects = 0
2033 PPCODE:
2034 EVENT_LIST (3);
2035
2036 cl_uint obj_count;
2037 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2038
2039 cl_event ev = 0;
2040 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2041
2042 if (ev)
2043 XPUSH_CLOBJ (stash_event, ev);
2044
2045#endif
1517 2046
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 2047#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 2048
1520void 2049void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2050acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 2051 ALIAS:
2052 release_gl_objects = 1
2053 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 2054 enqueue_release_gl_objects = 1
1524 PPCODE: 2055 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2056 EVENT_LIST (2);
1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2057
2058 cl_uint obj_count;
2059 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1527 2060
1528 cl_event ev = 0; 2061 cl_event ev = 0;
1529 EVENT_LIST (2, items - 2);
1530 AV *av = (AV *)SvRV (objects);
1531 cl_uint num_objects = av_len (av) + 1;
1532 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1533 int i;
1534
1535 for (i = num_objects; i--; )
1536 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1537 2062
1538 if (ix) 2063 if (ix)
1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2064 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1540 else 2065 else
1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2066 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1542 2067
1543 if (ev) 2068 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 2069 XPUSH_CLOBJ (stash_event, ev);
1545 2070
1546#endif 2071#endif
1547 2072
1548void 2073void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 2074wait_for_events (OpenCL::Queue self, ...)
2075 ALIAS:
2076 enqueue_wait_for_events = 0
1550 CODE: 2077 CODE:
1551 EVENT_LIST (1, items - 1); 2078 EVENT_LIST (1);
1552#if PREFER_1_1 2079#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 2081#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2082 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 2083#endif
1557 2084
1558void 2085void
1559enqueue_marker (OpenCL::Queue self, ...) 2086marker (OpenCL::Queue self, ...)
2087 ALIAS:
2088 enqueue_marker = 0
1560 PPCODE: 2089 PPCODE:
2090 EVENT_LIST (1);
1561 cl_event ev = 0; 2091 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1 2092#if PREFER_1_1
1564 if (!event_list_count) 2093 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2094 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 2095 else
1567#if CL_VERSION_1_2 2096#if CL_VERSION_1_2
1574#endif 2103#endif
1575#else 2104#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2105 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif 2106#endif
1578 if (ev) 2107 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 2108 XPUSH_CLOBJ (stash_event, ev);
1580 2109
1581void 2110void
1582enqueue_barrier (OpenCL::Queue self, ...) 2111barrier (OpenCL::Queue self, ...)
2112 ALIAS:
2113 enqueue_barrier = 0
1583 PPCODE: 2114 PPCODE:
2115 EVENT_LIST (1);
1584 cl_event ev = 0; 2116 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1 2117#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 2118 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 2119 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 2120 else
1590#if CL_VERSION_1_2 2121#if CL_VERSION_1_2
1600#endif 2131#endif
1601#else 2132#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2133 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif 2134#endif
1604 if (ev) 2135 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 2136 XPUSH_CLOBJ (stash_event, ev);
1606 2137
1607void 2138void
1608flush (OpenCL::Queue self) 2139flush (OpenCL::Queue self)
1609 CODE: 2140 CODE:
1610 NEED_SUCCESS (Flush, (self)); 2141 NEED_SUCCESS (Flush, (self));
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2159 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 2160 EXTEND (SP, 1);
1630 const int i = 0; 2161 const int i = 0;
1631 { 2162 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 2163 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2164 PUSH_CLOBJ (stash_context, value [i]);
1634 } 2165 }
1635 2166
1636void 2167void
1637device (OpenCL::Queue self) 2168device (OpenCL::Queue self)
1638 PPCODE: 2169 PPCODE:
1639 cl_device_id value [1]; 2170 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2171 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 2172 EXTEND (SP, 1);
1642 const int i = 0; 2173 const int i = 0;
1643 { 2174 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2175 PUSH_CLOBJ (stash_device, value [i]);
1645 } 2176 }
1646 2177
1647void 2178void
1648reference_count (OpenCL::Queue self) 2179reference_count (OpenCL::Queue self)
1649 PPCODE: 2180 PPCODE:
1658 PPCODE: 2189 PPCODE:
1659 cl_command_queue_properties value [1]; 2190 cl_command_queue_properties value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2191 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 2192 EXTEND (SP, 1);
1662 const int i = 0; 2193 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 2194 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 2195
1665#END:command_queue 2196#END:command_queue
1666 2197
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 2198MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 2199
1673 2204
1674void 2205void
1675info (OpenCL::Memory self, cl_mem_info name) 2206info (OpenCL::Memory self, cl_mem_info name)
1676 PPCODE: 2207 PPCODE:
1677 INFO (MemObject) 2208 INFO (MemObject)
2209
2210void
2211destructor_callback (OpenCL::Memory self, SV *notify)
2212 PPCODE:
1678 2213
1679#BEGIN:mem 2214#BEGIN:mem
1680 2215
1681void 2216void
1682type (OpenCL::Memory self) 2217type (OpenCL::Memory self)
1683 PPCODE: 2218 PPCODE:
1684 cl_mem_object_type value [1]; 2219 cl_mem_object_type value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2220 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 2221 EXTEND (SP, 1);
1687 const int i = 0; 2222 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 2223 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 2224
1690void 2225void
1691flags (OpenCL::Memory self) 2226flags (OpenCL::Memory self)
1692 PPCODE: 2227 PPCODE:
1693 cl_mem_flags value [1]; 2228 cl_mem_flags value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2230 EXTEND (SP, 1);
1696 const int i = 0; 2231 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 2232 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 2233
1699void 2234void
1700size (OpenCL::Memory self) 2235size (OpenCL::Memory self)
1701 ALIAS: 2236 ALIAS:
1702 size = CL_MEM_SIZE 2237 size = CL_MEM_SIZE
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2271 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 2272 EXTEND (SP, 1);
1738 const int i = 0; 2273 const int i = 0;
1739 { 2274 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 2275 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2276 PUSH_CLOBJ (stash_context, value [i]);
1742 } 2277 }
1743 2278
1744void 2279void
1745associated_memobject (OpenCL::Memory self) 2280associated_memobject (OpenCL::Memory self)
1746 PPCODE: 2281 PPCODE:
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2283 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2284 EXTEND (SP, 1);
1750 const int i = 0; 2285 const int i = 0;
1751 { 2286 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 2287 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2288 PUSH_CLOBJ (stash_memory, value [i]);
1754 } 2289 }
1755 2290
1756#END:mem 2291#END:mem
1757 2292
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 2293#if cl_apple_gl_sharing || cl_khr_gl_sharing
1778 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2313 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1779 2314
1780 cl_buffer_region crdata = { origin, size }; 2315 cl_buffer_region crdata = { origin, size };
1781 2316
1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2317 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2318 XPUSH_CLOBJ (stash_buffer, mem);
1784 2319
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2320MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2321
1787void 2322void
1788image_info (OpenCL::Image self, cl_image_info name) 2323image_info (OpenCL::Image self, cl_image_info name)
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2409 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2410 EXTEND (SP, 1);
1876 const int i = 0; 2411 const int i = 0;
1877 { 2412 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2413 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2414 PUSH_CLOBJ (stash_context, value [i]);
1880 } 2415 }
1881 2416
1882void 2417void
1883normalized_coords (OpenCL::Sampler self) 2418normalized_coords (OpenCL::Sampler self)
1884 PPCODE: 2419 PPCODE:
1885 cl_addressing_mode value [1]; 2420 cl_addressing_mode value [1];
1886 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2421 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1887 EXTEND (SP, 1); 2422 EXTEND (SP, 1);
1888 const int i = 0; 2423 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i]))); 2424 PUSHs (sv_2mortal (newSVuv (value [i])));
1890 2425
1891void 2426void
1892addressing_mode (OpenCL::Sampler self) 2427addressing_mode (OpenCL::Sampler self)
1893 PPCODE: 2428 PPCODE:
1894 cl_filter_mode value [1]; 2429 cl_filter_mode value [1];
1895 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2430 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1896 EXTEND (SP, 1); 2431 EXTEND (SP, 1);
1897 const int i = 0; 2432 const int i = 0;
1898 PUSHs (sv_2mortal (newSViv (value [i]))); 2433 PUSHs (sv_2mortal (newSVuv (value [i])));
1899 2434
1900void 2435void
1901filter_mode (OpenCL::Sampler self) 2436filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2437 PPCODE:
1903 cl_bool value [1]; 2438 cl_bool value [1];
1918void 2453void
1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2454build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS: 2455 ALIAS:
1921 build_async = 1 2456 build_async = 1
1922 CODE: 2457 CODE:
1923 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1924 void *user_data = 0;
1925 cl_uint num_devices = 0; 2458 cl_uint device_count = 0;
1926 cl_device_id *device_list = 0; 2459 cl_device_id *device_list = 0;
1927 2460
1928 if (SvOK (devices)) 2461 if (SvOK (devices))
1929 { 2462 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2463
2464 void *user_data;
2465 program_callback pfn_notify = make_program_callback (notify, &user_data);
2466
2467 if (ix)
2468 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2469 else
2470 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2471
2472#if CL_VERSION_1_2
2473
2474void
2475compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2476 CODE:
2477 cl_uint device_count = 0;
2478 cl_device_id *device_list = 0;
2479
2480 if (SvOK (devices))
2481 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2482
2483 cl_uint header_count = 0;
2484 cl_program *header_list = 0;
2485 const char **header_name = 0;
2486
2487 if (SvOK (headers))
2488 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2489 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2490 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1932 2491
1933 AV *av = (AV *)SvRV (devices); 2492 HV *hv = (HV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935 2493
1936 if (num_devices) 2494 header_count = hv_iterinit (hv);
2495 header_list = tmpbuf (sizeof (*header_list) * header_count);
2496 header_name = tmpbuf (sizeof (*header_name) * header_count);
2497
2498 HE *he;
2499 int i = 0;
2500 while (he = hv_iternext (hv))
1937 { 2501 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2502 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2503 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1939 int count; 2504 ++i;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 } 2505 }
1943 } 2506 }
1944 2507
1945 if (SvOK (notify)) 2508 void *user_data;
1946 { 2509 program_callback pfn_notify = make_program_callback (notify, &user_data);
1947 NEED_SUCCESS (RetainProgram, (self));
1948 pfn_notify = eq_program_notify;
1949 user_data = SvREFCNT_inc (s_get_cv (notify));
1950 }
1951 2510
1952 if (ix)
1953 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1954 else
1955 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2511 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2512 header_count, header_list, header_name, pfn_notify, user_data));
2513
2514#endif
1956 2515
1957void 2516void
1958build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2517build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1959 PPCODE: 2518 PPCODE:
1960 size_t size; 2519 size_t size;
1989 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2548 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1990 EXTEND (SP, 1); 2549 EXTEND (SP, 1);
1991 const int i = 0; 2550 const int i = 0;
1992 PUSHs (sv_2mortal (newSVpv (value, 0))); 2551 PUSHs (sv_2mortal (newSVpv (value, 0)));
1993 2552
2553void
2554binary_type (OpenCL::Program self, OpenCL::Device device)
2555 PPCODE:
2556 cl_program_binary_type value [1];
2557 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2558 EXTEND (SP, 1);
2559 const int i = 0;
2560 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2561
1994#END:program_build 2562#END:program_build
1995 2563
1996void 2564void
1997kernel (OpenCL::Program program, SV *function) 2565kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2566 PPCODE:
1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2567 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2568 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2569
2002void 2570void
2003kernels_in_program (OpenCL::Program program) 2571kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2572 PPCODE:
2005 cl_uint num_kernels; 2573 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2576 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2577
2010 int i; 2578 int i;
2011 EXTEND (SP, num_kernels); 2579 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2580 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2581 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2582
2015void 2583void
2016info (OpenCL::Program self, cl_program_info name) 2584info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2585 PPCODE:
2018 INFO (Program) 2586 INFO (Program)
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2634 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2635 EXTEND (SP, 1);
2068 const int i = 0; 2636 const int i = 0;
2069 { 2637 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2638 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2639 PUSH_CLOBJ (stash_context, value [i]);
2072 } 2640 }
2073 2641
2074void 2642void
2075devices (OpenCL::Program self) 2643devices (OpenCL::Program self)
2076 PPCODE: 2644 PPCODE:
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2648 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2649 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2650 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2651 for (i = 0; i < n; ++i)
2084 { 2652 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2653 PUSH_CLOBJ (stash_device, value [i]);
2086 } 2654 }
2087 2655
2088void 2656void
2089source (OpenCL::Program self) 2657source (OpenCL::Program self)
2090 PPCODE: 2658 PPCODE:
2116DESTROY (OpenCL::Kernel self) 2684DESTROY (OpenCL::Kernel self)
2117 CODE: 2685 CODE:
2118 clReleaseKernel (self); 2686 clReleaseKernel (self);
2119 2687
2120void 2688void
2689setf (OpenCL::Kernel self, const char *format, ...)
2690 CODE:
2691 int i;
2692 for (i = 2; ; ++i)
2693 {
2694 while (*format == ' ')
2695 ++format;
2696
2697 char type = *format++;
2698
2699 if (!type)
2700 break;
2701
2702 if (i >= items)
2703 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2704
2705 SV *sv = ST (i);
2706
2707 union
2708 {
2709 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2710 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2711 cl_half ch; cl_float cf; cl_double cd;
2712 cl_mem cm;
2713 cl_sampler ca;
2714 size_t cz;
2715 cl_event ce;
2716 } arg;
2717 size_t size;
2718 int nullarg = 0;
2719
2720 switch (type)
2721 {
2722 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2723 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2724 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2725 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2726 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2727 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2728 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2729 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2730
2731 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2732 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2733 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2734
2735 case 'z': nullarg = 1; size = SvIV (sv); break;
2736
2737 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2738 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2739 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2740
2741 default:
2742 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2743 }
2744
2745 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2746 if (res)
2747 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2748 }
2749
2750 if (i != items)
2751 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2752
2753void
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2754set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2755 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2756 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2757
2125void 2758void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2806 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2807
2175void 2808void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2809set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2810 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2811 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2812
2180void 2813void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2814set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2815 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2816 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2817
2185void 2818void
2186set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2819set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2187 ALIAS:
2188 set_image2d = 0
2189 set_image3d = 0
2190 CODE: 2820 CODE:
2191 clSetKernelArg (self, idx, sizeof (value), &value); 2821 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2192 2822
2193void 2823void
2194set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2824set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2195 CODE: 2825 CODE:
2196 clSetKernelArg (self, idx, sizeof (value), &value); 2826 clSetKernelArg (self, idx, sizeof (value), &value);
2242 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2872 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2243 EXTEND (SP, 1); 2873 EXTEND (SP, 1);
2244 const int i = 0; 2874 const int i = 0;
2245 { 2875 {
2246 NEED_SUCCESS (RetainContext, (value [i])); 2876 NEED_SUCCESS (RetainContext, (value [i]));
2247 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2877 PUSH_CLOBJ (stash_context, value [i]);
2248 } 2878 }
2249 2879
2250void 2880void
2251program (OpenCL::Kernel self) 2881program (OpenCL::Kernel self)
2252 PPCODE: 2882 PPCODE:
2254 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2884 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2255 EXTEND (SP, 1); 2885 EXTEND (SP, 1);
2256 const int i = 0; 2886 const int i = 0;
2257 { 2887 {
2258 NEED_SUCCESS (RetainProgram, (value [i])); 2888 NEED_SUCCESS (RetainProgram, (value [i]));
2259 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2889 PUSH_CLOBJ (stash_program, value [i]);
2260 } 2890 }
2261 2891
2262#END:kernel 2892#END:kernel
2263 2893
2264void 2894void
2311 const int i = 0; 2941 const int i = 0;
2312 PUSHs (sv_2mortal (newSVuv (value [i]))); 2942 PUSHs (sv_2mortal (newSVuv (value [i])));
2313 2943
2314#END:kernel_work_group 2944#END:kernel_work_group
2315 2945
2946#if CL_VERSION_1_2
2947
2948void
2949arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2950 PPCODE:
2951 size_t size;
2952 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2953 SV *sv = sv_2mortal (newSV (size));
2954 SvUPGRADE (sv, SVt_PV);
2955 SvPOK_only (sv);
2956 SvCUR_set (sv, size);
2957 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2958 XPUSHs (sv);
2959
2960#BEGIN:kernel_arg
2961
2962void
2963arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2964 PPCODE:
2965 cl_kernel_arg_address_qualifier value [1];
2966 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2967 EXTEND (SP, 1);
2968 const int i = 0;
2969 PUSHs (sv_2mortal (newSVuv (value [i])));
2970
2971void
2972arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2973 PPCODE:
2974 cl_kernel_arg_access_qualifier value [1];
2975 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2976 EXTEND (SP, 1);
2977 const int i = 0;
2978 PUSHs (sv_2mortal (newSVuv (value [i])));
2979
2980void
2981arg_type_name (OpenCL::Kernel self, cl_uint idx)
2982 ALIAS:
2983 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2984 arg_name = CL_KERNEL_ARG_NAME
2985 PPCODE:
2986 size_t size;
2987 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2988 char *value = tmpbuf (size);
2989 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2990 EXTEND (SP, 1);
2991 const int i = 0;
2992 PUSHs (sv_2mortal (newSVpv (value, 0)));
2993
2994void
2995arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2996 PPCODE:
2997 cl_kernel_arg_type_qualifier value [1];
2998 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2999 EXTEND (SP, 1);
3000 const int i = 0;
3001 PUSHs (sv_2mortal (newSVuv (value [i])));
3002
3003#END:kernel_arg
3004
3005#endif
3006
2316MODULE = OpenCL PACKAGE = OpenCL::Event 3007MODULE = OpenCL PACKAGE = OpenCL::Event
2317 3008
2318void 3009void
2319DESTROY (OpenCL::Event self) 3010DESTROY (OpenCL::Event self)
2320 CODE: 3011 CODE:
2344 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3035 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2345 EXTEND (SP, 1); 3036 EXTEND (SP, 1);
2346 const int i = 0; 3037 const int i = 0;
2347 { 3038 {
2348 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3039 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2349 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 3040 PUSH_CLOBJ (stash_queue, value [i]);
2350 } 3041 }
2351 3042
2352void 3043void
2353command_type (OpenCL::Event self) 3044command_type (OpenCL::Event self)
2354 PPCODE: 3045 PPCODE:
2377 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3068 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2378 EXTEND (SP, 1); 3069 EXTEND (SP, 1);
2379 const int i = 0; 3070 const int i = 0;
2380 { 3071 {
2381 NEED_SUCCESS (RetainContext, (value [i])); 3072 NEED_SUCCESS (RetainContext, (value [i]));
2382 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3073 PUSH_CLOBJ (stash_context, value [i]);
2383 } 3074 }
2384 3075
2385#END:event 3076#END:event
2386 3077
2387void 3078void
2412void 3103void
2413set_status (OpenCL::UserEvent self, cl_int execution_status) 3104set_status (OpenCL::UserEvent self, cl_int execution_status)
2414 CODE: 3105 CODE:
2415 clSetUserEventStatus (self, execution_status); 3106 clSetUserEventStatus (self, execution_status);
2416 3107
3108MODULE = OpenCL PACKAGE = OpenCL::Mapped
3109
3110void
3111DESTROY (SV *self)
3112 CODE:
3113 OpenCL__Mapped mapped = SvMAPPED (self);
3114
3115 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3116 mapped_detach (self, mapped);
3117
3118 clReleaseCommandQueue (mapped->queue);
3119 clReleaseEvent (mapped->event);
3120 Safefree (mapped);
3121
3122void
3123unmap (OpenCL::Mapped self, ...)
3124 CODE:
3125 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3126
3127bool
3128mapped (OpenCL::Mapped self)
3129 CODE:
3130 RETVAL = !!self->ptr;
3131 OUTPUT:
3132 RETVAL
3133
3134void
3135wait (OpenCL::Mapped self)
3136 PPCODE:
3137 if (self->event)
3138 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3139
3140void
3141event (OpenCL::Mapped self)
3142 PPCODE:
3143 if (!self->event)
3144 XSRETURN_UNDEF;
3145
3146 clRetainEvent (self->event);
3147 XPUSH_CLOBJ (stash_event, self->event);
3148
3149size_t
3150size (OpenCL::Mapped self)
3151 CODE:
3152 RETVAL = self->cb;
3153 OUTPUT:
3154 RETVAL
3155
3156IV
3157ptr (OpenCL::Mapped self)
3158 CODE:
3159 RETVAL = PTR2IV (self->ptr);
3160 OUTPUT:
3161 RETVAL
3162
3163void
3164set (OpenCL::Mapped self, size_t offset, SV *data)
3165 CODE:
3166 STRLEN len;
3167 const char *ptr = SvPVbyte (data, len);
3168
3169 if (offset + len > self->cb)
3170 croak ("OpenCL::Mapped::set out of bound condition detected");
3171
3172 memcpy (offset + (char *)self->ptr, ptr, len);
3173
3174MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3175
3176MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3177
3178IV
3179row_pitch (OpenCL::Mapped self)
3180 ALIAS:
3181 slice_pitch = 1
3182 CODE:
3183 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3184 OUTPUT:
3185 RETVAL
3186
3187
3188

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines