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 1.81 by root, Sat May 5 17:09:19 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
6
7#include "ecb.h"//D
4 8
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 10#include "xthread.h"
7#include "schmorp.h" 11#include "schmorp.h"
8 12
26#if PREFER_1_1 30#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 32#endif
29 33
30#ifdef __APPLE__ 34#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
32#else 36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
33 #include <CL/opencl.h> 40#include CLHDR(opencl.h)
41
42#if 0
43#ifndef CL_VERSION_1_2
44 #include CLHDR(cl_d3d9.h)
45#endif
46#endif
47
48#if _WIN32
49 #include CLHDR(cl_d3d10.h)
50 #if CL_VERSION_1_2
51 #include CLHDR<cl_d3d11.h>
52 #endif
53 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 54#endif
35 55
36#ifndef CL_VERSION_1_2 56#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 57 #undef PREFER_1_1
38 #define PREFER_1_1 1 58 #define PREFER_1_1 1
39#endif 59#endif
40 60
61// make sure all constants we might use are actually defined
62#include "default.h"
63
41typedef cl_platform_id OpenCL__Platform; 64typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 65typedef cl_device_id OpenCL__Device;
66typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 67typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 68typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 69typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 70typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 71typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 72typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 73typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 74typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 75typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 76typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 77typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 78typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 79typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 80typedef cl_event OpenCL__UserEvent;
61 81
62typedef SV *FUTURE; 82typedef struct mapped * OpenCL__Mapped;
83
84static HV
85 *stash_platform,
86 *stash_device,
87 *stash_subdevice,
88 *stash_context,
89 *stash_queue,
90 *stash_program,
91 *stash_kernel,
92 *stash_sampler,
93 *stash_event,
94 *stash_userevent,
95 *stash_memory,
96 *stash_buffer,
97 *stash_bufferobj,
98 *stash_image,
99 *stash_image1d,
100 *stash_image1darray,
101 *stash_image1dbuffer,
102 *stash_image2d,
103 *stash_image2darray,
104 *stash_image3d,
105 *stash_mapped,
106 *stash_mappedbuffer,
107 *stash_mappedimage;
63 108
64/*****************************************************************************/ 109/*****************************************************************************/
65 110
66// name must include a leading underscore 111// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 112// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 152}
108 153
109/*****************************************************************************/ 154/*****************************************************************************/
110 155
111/* up to two temporary buffers */ 156/* up to two temporary buffers */
112static void * 157static void * ecb_noinline
113tmpbuf (size_t size) 158tmpbuf (size_t size)
114{ 159{
115 enum { buffers = 3 }; 160 enum { buffers = 4 };
116 static int idx; 161 static int idx;
117 static void *buf [buffers]; 162 static void *buf [buffers];
118 static size_t len [buffers]; 163 static size_t len [buffers];
119 164
120 idx = (idx + 1) % buffers; 165 idx = (idx + 1) % buffers;
127 } 172 }
128 173
129 return buf [idx]; 174 return buf [idx];
130} 175}
131 176
177static const char * ecb_noinline
178cv_get_name (CV *cv)
179{
180 static char fullname [256];
181
182 GV *gv = CvGV (cv); // gv better be nonzero
183
184 HV *stash = GvSTASH (gv);
185 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
186 const char *gvname = GvNAME (gv);
187
188 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
189 return fullname;
190}
191
132/*****************************************************************************/ 192/*****************************************************************************/
133 193
134typedef struct 194typedef struct
135{ 195{
136 IV iv; 196 IV iv;
137 const char *name; 197 const char *name;
138 #define const_iv(name) { (IV)CL_ ## name, # name }, 198 #define const_iv(name) { (IV)CL_ ## name, # name },
139} ivstr; 199} ivstr;
200
201typedef struct
202{
203 NV nv;
204 const char *name;
205 #define const_nv(name) { (NV)CL_ ## name, # name },
206} nvstr;
140 207
141static const char * 208static const char *
142iv2str (IV value, const ivstr *base, int count, const char *fallback) 209iv2str (IV value, const ivstr *base, int count, const char *fallback)
143{ 210{
144 int i; 211 int i;
193 if (res) \ 260 if (res) \
194 FAIL (name); 261 FAIL (name);
195 262
196/*****************************************************************************/ 263/*****************************************************************************/
197 264
265static SV *
266new_clobj (HV *stash, IV id)
267{
268 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
269}
270
271#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
272#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
273
274/* cl objects are either \$iv, or [$iv, ...] */
275/* they can be upgraded at runtime to the array form */
276static void * ecb_noinline
277SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
278{
279 // sv_derived_from is quite slow :(
280 if (SvROK (sv) && sv_derived_from (sv, pkg))
281 return (void *)SvIV (SvRV (sv));
282
283 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
284}
285
286// the "no-inherit" version of the above
287static void * ecb_noinline
288SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
289{
290 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
291 return (void *)SvIV (SvRV (sv));
292
293 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
294}
295
296/*****************************************************************************/
297
198static cl_context_properties * 298static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 299SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 300{
201 if (!sv || !SvOK (sv)) 301 if (!sv || !SvOK (sv))
202 if (extra) 302 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 303 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 304 else
210 int i, len = av_len (av) + 1; 310 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 311 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 312 cl_context_properties *l = p;
213 313
214 if (len & 1) 314 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 315 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 316
217 while (extracount--) 317 while (extracount--)
218 *l++ = *extra++; 318 *l++ = *extra++;
219 319
220 for (i = 0; i < len; i += 2) 320 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 323 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 324 cl_context_properties v = SvIV (p_sv); // code below can override
225 325
226 switch (t) 326 switch (t)
227 { 327 {
328 case CL_CONTEXT_PLATFORM:
329 if (SvROK (p_sv))
330 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
331 break;
332
228 case CL_GLX_DISPLAY_KHR: 333 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 334 if (!SvOK (p_sv))
230 { 335 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 336 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 337 if (func)
255 *l = 0; 360 *l = 0;
256 361
257 return p; 362 return p;
258 } 363 }
259 364
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 365 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 366}
262 367
263/*****************************************************************************/ 368// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 369// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 370static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 371object_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{ 372{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 373 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
275 return (void *)SvIV (SvRV (sv)); 374 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
375 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
276 376
277 croak ("%s: %s is not of type %s", func, svname, pkg); 377 AV *av = (AV *)SvRV (arg);
378 void **list = 0;
379 cl_uint count = av_len (av) + 1;
380
381 if (count)
382 {
383 list = tmpbuf (sizeof (*list) * count);
384 int i;
385 for (i = 0; i < count; ++i)
386 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
387 }
388
389 if (!count && !or_undef)
390 croak ("%s: '%s' must contain at least one %s object",
391 cv_get_name (cv), argname, klass);
392
393 *rcount = count;
394 return (void *)list;
278} 395}
279 396
280/*****************************************************************************/ 397/*****************************************************************************/
281/* callback stuff */ 398/* callback stuff */
282 399
304static void (*eq_signal_func)(void *signal_arg, int value); 421static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 422static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 423static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 424static eq_item *eq_head, *eq_tail;
308 425
309static void 426static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 427eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 428{
312 eq_item *item = malloc (sizeof (eq_item)); 429 eq_item *item = malloc (sizeof (eq_item));
313 430
314 item->next = 0; 431 item->next = 0;
381} 498}
382 499
383/*****************************************************************************/ 500/*****************************************************************************/
384/* context notify */ 501/* context notify */
385 502
386static void 503static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 504eq_context_push (void *data1, void *data2, void *data3)
388{ 505{
389 dSP; 506 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 507 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 508 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 530 { \
414 pfn_notify = eq_context_notify; \ 531 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 532 user_data = s_get_cv (notify); \
416 } 533 }
417 534
418static SV * 535static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 536new_clobj_context (cl_context ctx, void *user_data)
420{ 537{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 538 SV *sv = new_clobj (stash_context, (IV)ctx);
422 539
423 if (user_data) 540 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 541 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 542
426 return sv; 543 return sv;
433 550
434static void 551static void
435eq_program_push (void *data1, void *data2, void *data3) 552eq_program_push (void *data1, void *data2, void *data3)
436{ 553{
437 dSP; 554 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 555 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 556 PUTBACK;
440} 557}
441 558
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 559static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 560
444static void CL_CALLBACK 561static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 562eq_program_notify (cl_program program, void *user_data)
446{ 563{
564 clRetainProgram (program);
565
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 566 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
567}
568
569typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
570
571static program_callback ecb_noinline
572make_program_callback (SV *notify, void **ruser_data)
573{
574 if (SvOK (notify))
575 {
576 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
577 return eq_program_notify;
578 }
579 else
580 {
581 *ruser_data = 0;
582 return 0;
583 }
448} 584}
449 585
450struct build_args 586struct build_args
451{ 587{
452 cl_program program; 588 cl_program program;
466 else 602 else
467 clReleaseProgram (arg->program); 603 clReleaseProgram (arg->program);
468 604
469 free (arg->options); 605 free (arg->options);
470 free (arg); 606 free (arg);
607
608 return 0;
471} 609}
472 610
473static void 611static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 612build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 613{
484 xthread_t id; 622 xthread_t id;
485 thread_create (&id, build_program_thread, arg); 623 thread_create (&id, build_program_thread, arg);
486} 624}
487 625
488/*****************************************************************************/ 626/*****************************************************************************/
627/* mem object destructor notify */
628
629static void ecb_noinline
630eq_destructor_push (void *data1, void *data2, void *data3)
631{
632}
633
634static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
635
636static void CL_CALLBACK
637eq_destructor_notify (cl_mem memobj, void *user_data)
638{
639 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
640}
641
642/*****************************************************************************/
489/* event objects */ 643/* event objects */
490 644
491static void 645static void
492eq_event_push (void *data1, void *data2, void *data3) 646eq_event_push (void *data1, void *data2, void *data3)
493{ 647{
494 dSP; 648 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 649 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 650 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 651 PUTBACK;
498} 652}
499 653
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 654static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 659 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 660 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 661}
508 662
509/*****************************************************************************/ 663/*****************************************************************************/
664/* utilities for XS code */
510 665
511static size_t 666static size_t
512img_row_pitch (cl_mem img) 667img_row_pitch (cl_mem img)
513{ 668{
514 size_t res; 669 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 670 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 671 return res;
517} 672}
518 673
519static cl_event * 674static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 675event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 676{
522 cl_uint count = *rcount; 677 cl_uint count = *rcount;
523 678
524 if (!count) 679 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
680 *rcount = count = 0;
681
682 if (!count && !extra)
525 return 0; 683 return 0;
526 684
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 685 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 686 int i = 0;
529 687
530 do 688 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 689 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 690 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 691
536 while (count); 692 if (extra)
693 list [i++] = extra;
537 694
538 *rcount = i; 695 *rcount = i;
539 696
540 return i ? list : 0; 697 return i ? list : 0;
541} 698}
542 699
543#define EVENT_LIST(items,count) \ 700#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 701 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 702 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 703
547#define INFO(class) \ 704#define INFO(class) \
548{ \ 705{ \
549 size_t size; \ 706 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 707 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
554 SvCUR_set (sv, size); \ 711 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 712 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 713 XPUSHs (sv); \
557} 714}
558 715
716/*****************************************************************************/
717/* mapped_xxx */
718
719static OpenCL__Mapped
720SvMAPPED (SV *sv)
721{
722 // no typechecking atm., keep your fingers crossed
723 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
724}
725
726struct mapped
727{
728 cl_command_queue queue;
729 cl_mem memobj;
730 void *ptr;
731 size_t cb;
732 cl_event event;
733 size_t row_pitch;
734 size_t slice_pitch;
735
736 size_t element_size;
737 size_t width, height, depth;
738};
739
740static SV *
741mapped_new (
742 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
743 void *ptr, size_t cb, cl_event ev,
744 size_t row_pitch, size_t slice_pitch, size_t element_size,
745 size_t width, size_t height, size_t depth
746)
747{
748 SV *data = newSV (0);
749 SvUPGRADE (data, SVt_PVMG);
750
751 OpenCL__Mapped mapped;
752 New (0, mapped, 1, struct mapped);
753
754 clRetainCommandQueue (queue);
755
756 mapped->queue = queue;
757 mapped->memobj = memobj;
758 mapped->ptr = ptr;
759 mapped->cb = cb;
760 mapped->event = ev;
761 mapped->row_pitch = row_pitch;
762 mapped->slice_pitch = slice_pitch;
763
764 mapped->element_size = element_size;
765 mapped->width = width;
766 mapped->height = height;
767 mapped->depth = depth;
768
769 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
770
771 if (SvLEN (data))
772 Safefree (data);
773
774 SvPVX (data) = (char *)ptr;
775 SvCUR_set (data, cb);
776 SvLEN_set (data, 0);
777 SvPOK_only (data);
778
779 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
780
781 if (!(flags & CL_MAP_WRITE))
782 SvREADONLY_on (data);
783
784 return obj;
785}
786
787static void
788mapped_detach (SV *sv, OpenCL__Mapped mapped)
789{
790 SV *data = SvRV (sv);
791
792 // the next check checks both before AND after detach, where SvPVX should STILL be 0
793 if (SvPVX (data) != (char *)mapped->ptr)
794 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
795 else
796 {
797 SvREADONLY_off (data);
798 SvCUR_set (data, 0);
799 SvPVX (data) = 0;
800 SvOK_off (data);
801 }
802
803 mapped->ptr = 0;
804}
805
806static void
807mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
808{
809 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
810 cl_event ev;
811
812 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
813
814 clReleaseEvent (mapped->event);
815 mapped->event = ev;
816
817 mapped_detach (self, mapped);
818}
819
820static size_t
821mapped_element_size (OpenCL__Mapped self)
822{
823 if (!self->element_size)
824 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
825
826 return self->element_size;
827}
828
829/*****************************************************************************/
830
559MODULE = OpenCL PACKAGE = OpenCL 831MODULE = OpenCL PACKAGE = OpenCL
560 832
561PROTOTYPES: ENABLE 833PROTOTYPES: ENABLE
562 834
563void 835void
572 eq_signal_arg = (void*)arg; 844 eq_signal_arg = (void*)arg;
573 845
574BOOT: 846BOOT:
575{ 847{
576 HV *stash = gv_stashpv ("OpenCL", 1); 848 HV *stash = gv_stashpv ("OpenCL", 1);
849
577 static const ivstr *civ, const_iv[] = { 850 static const ivstr *civ, const_iv[] = {
578 { sizeof (cl_char ), "SIZEOF_CHAR" }, 851 { sizeof (cl_char ), "SIZEOF_CHAR" },
579 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 852 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
580 { sizeof (cl_short ), "SIZEOF_SHORT" }, 853 { sizeof (cl_short ), "SIZEOF_SHORT" },
581 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 854 { sizeof (cl_ushort), "SIZEOF_USHORT" },
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 857 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 858 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 859 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 860 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 861 { sizeof (cl_double), "SIZEOF_DOUBLE" },
862 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 863#include "constiv.h"
590 }; 864 };
591 865
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 866 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)); 867 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
868
869 static const nvstr *cnv, const_nv[] = {
870#include "constnv.h"
871 };
872
873 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
874 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
875
876 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
877
878 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
879 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
880 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
881 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
882 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
883 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
884 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
885 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
886 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
887 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
888 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
889 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
890 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
891 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
892 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
893 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
894 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
895 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
896 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
897 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
898 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
899 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
900 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 901
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 902 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 903}
597 904
598cl_int 905cl_int
601 RETVAL = res; 908 RETVAL = res;
602 OUTPUT: 909 OUTPUT:
603 RETVAL 910 RETVAL
604 911
605const char * 912const char *
606err2str (cl_int err) 913err2str (cl_int err = res)
607 914
608const char * 915const char *
609enum2str (cl_uint value) 916enum2str (cl_uint value)
610 917
611void 918void
619 list = tmpbuf (sizeof (*list) * count); 926 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 927 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 928
622 EXTEND (SP, count); 929 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 930 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 931 PUSH_CLOBJ (stash_platform, list [i]);
625 932
626void 933void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 934context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 935 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 936 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 937 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 938 XPUSH_CLOBJ_CONTEXT;
632 939
633void 940void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 941context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 942 PPCODE:
636 /* der Gipfel der Kunst */ 943 cl_uint device_count;
944 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
945
946 CONTEXT_NOTIFY_CALLBACK;
947 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
948 XPUSH_CLOBJ_CONTEXT;
637 949
638void 950void
639wait_for_events (...) 951wait_for_events (...)
640 CODE: 952 CODE:
641 EVENT_LIST (0, items); 953 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 954 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 955
644PROTOTYPES: DISABLE 956PROTOTYPES: DISABLE
645 957
646MODULE = OpenCL PACKAGE = OpenCL::Platform 958MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 1001 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 1002 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 1003
692 EXTEND (SP, count); 1004 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 1005 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1006 PUSH_CLOBJ (stash_device, list [i]);
695 1007
696void 1008void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 1009context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 1010 PPCODE:
1011 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1012 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1013
1014 cl_uint device_count;
1015 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1016
1017 CONTEXT_NOTIFY_CALLBACK;
1018 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1019 XPUSH_CLOBJ_CONTEXT;
1020
1021void
1022context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1023 PPCODE:
1024 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1025 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1026
1027 CONTEXT_NOTIFY_CALLBACK;
1028 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1029 XPUSH_CLOBJ_CONTEXT;
1030
1031MODULE = OpenCL PACKAGE = OpenCL::Device
1032
1033void
1034info (OpenCL::Device self, cl_device_info name)
1035 PPCODE:
1036 INFO (Device)
1037
1038#if CL_VERSION_1_2
1039
1040void
1041sub_devices (OpenCL::Device self, SV *properties)
1042 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1043 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1044 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 1045
702 AV *av = (AV *)SvRV (devices); 1046 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 1047
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1048 cl_uint count = av_len ((AV *)properties) + 1;
1049 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 1050
706 int i; 1051 int i;
707 for (i = num_devices; i--; ) 1052 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1053 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 1054
710 CONTEXT_NOTIFY_CALLBACK; 1055 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 1056
714void 1057 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) 1058 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 1059 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1060 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 1061
720 CONTEXT_NOTIFY_CALLBACK; 1062 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1063 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 1064 PUSH_CLOBJ (stash_subdevice, list [i]);
723 1065
724MODULE = OpenCL PACKAGE = OpenCL::Device 1066#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 1067
731#BEGIN:device 1068#BEGIN:device
732 1069
733void 1070void
734type (OpenCL::Device self) 1071type (OpenCL::Device self)
1072 ALIAS:
1073 type = CL_DEVICE_TYPE
1074 address_bits = CL_DEVICE_ADDRESS_BITS
1075 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1076 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1077 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1078 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1079 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1080 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1081 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1082 properties = CL_DEVICE_QUEUE_PROPERTIES
1083 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1084 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
735 PPCODE: 1085 PPCODE:
736 cl_device_type value [1]; 1086 cl_ulong value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
738 EXTEND (SP, 1); 1088 EXTEND (SP, 1);
739 const int i = 0; 1089 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1090 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1091
742void 1092void
743vendor_id (OpenCL::Device self) 1093vendor_id (OpenCL::Device self)
744 ALIAS: 1094 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1095 vendor_id = CL_DEVICE_VENDOR_ID
756 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1106 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
757 image_support = CL_DEVICE_IMAGE_SUPPORT 1107 image_support = CL_DEVICE_IMAGE_SUPPORT
758 max_samplers = CL_DEVICE_MAX_SAMPLERS 1108 max_samplers = CL_DEVICE_MAX_SAMPLERS
759 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1109 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
760 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1110 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1111 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
761 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1112 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
762 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1113 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1114 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
763 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1115 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
764 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1116 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
765 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1117 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
766 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1118 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
767 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1119 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
805 EXTEND (SP, n); 1157 EXTEND (SP, n);
806 for (i = 0; i < n; ++i) 1158 for (i = 0; i < n; ++i)
807 PUSHs (sv_2mortal (newSVuv (value [i]))); 1159 PUSHs (sv_2mortal (newSVuv (value [i])));
808 1160
809void 1161void
810address_bits (OpenCL::Device self)
811 PPCODE:
812 cl_bitfield value [1];
813 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
814 EXTEND (SP, 1);
815 const int i = 0;
816 PUSHs (sv_2mortal (newSVuv (value [i])));
817
818void
819max_mem_alloc_size (OpenCL::Device self)
820 ALIAS:
821 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
822 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
823 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
824 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
825 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
826 PPCODE:
827 cl_ulong value [1];
828 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
829 EXTEND (SP, 1);
830 const int i = 0;
831 PUSHs (sv_2mortal (newSVuv (value [i])));
832
833void
834single_fp_config (OpenCL::Device self)
835 ALIAS:
836 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
837 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
838 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
839 PPCODE:
840 cl_device_fp_config value [1];
841 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
842 EXTEND (SP, 1);
843 const int i = 0;
844 PUSHs (sv_2mortal (newSVuv (value [i])));
845
846void
847global_mem_cache_type (OpenCL::Device self)
848 PPCODE:
849 cl_device_mem_cache_type value [1];
850 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
851 EXTEND (SP, 1);
852 const int i = 0;
853 PUSHs (sv_2mortal (newSVuv (value [i])));
854
855void
856local_mem_type (OpenCL::Device self)
857 PPCODE:
858 cl_device_local_mem_type value [1];
859 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
860 EXTEND (SP, 1);
861 const int i = 0;
862 PUSHs (sv_2mortal (newSVuv (value [i])));
863
864void
865error_correction_support (OpenCL::Device self) 1162error_correction_support (OpenCL::Device self)
866 ALIAS: 1163 ALIAS:
867 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1164 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
868 endian_little = CL_DEVICE_ENDIAN_LITTLE 1165 endian_little = CL_DEVICE_ENDIAN_LITTLE
869 available = CL_DEVICE_AVAILABLE 1166 available = CL_DEVICE_AVAILABLE
870 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1167 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
871 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1168 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
872 PPCODE: 1169 PPCODE:
873 cl_bool value [1]; 1170 cl_uint value [1];
874 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1171 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
875 EXTEND (SP, 1); 1172 EXTEND (SP, 1);
876 const int i = 0; 1173 const int i = 0;
877 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1174 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
878
879void
880execution_capabilities (OpenCL::Device self)
881 PPCODE:
882 cl_device_exec_capabilities value [1];
883 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
884 EXTEND (SP, 1);
885 const int i = 0;
886 PUSHs (sv_2mortal (newSVuv (value [i])));
887
888void
889properties (OpenCL::Device self)
890 PPCODE:
891 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1);
894 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i])));
896 1175
897void 1176void
898platform (OpenCL::Device self) 1177platform (OpenCL::Device self)
899 PPCODE: 1178 PPCODE:
900 cl_platform_id value [1]; 1179 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1180 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1181 EXTEND (SP, 1);
903 const int i = 0; 1182 const int i = 0;
904 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1183 PUSH_CLOBJ (stash_platform, value [i]);
906 }
907 1184
908void 1185void
909name (OpenCL::Device self) 1186name (OpenCL::Device self)
910 ALIAS: 1187 ALIAS:
911 name = CL_DEVICE_NAME 1188 name = CL_DEVICE_NAME
928 PPCODE: 1205 PPCODE:
929 cl_device_id value [1]; 1206 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1207 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1208 EXTEND (SP, 1);
932 const int i = 0; 1209 const int i = 0;
933 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1210 PUSH_CLOBJ (stash_device, value [i]);
935 }
936 1211
937void 1212void
938partition_types_ext (OpenCL::Device self) 1213partition_types_ext (OpenCL::Device self)
939 ALIAS: 1214 ALIAS:
940 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1215 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
950 for (i = 0; i < n; ++i) 1225 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1226 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1227
953#END:device 1228#END:device
954 1229
1230MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1231
1232#if CL_VERSION_1_2
1233
1234void
1235DESTROY (OpenCL::SubDevice self)
1236 CODE:
1237 clReleaseDevice (self);
1238
1239#endif
1240
955MODULE = OpenCL PACKAGE = OpenCL::Context 1241MODULE = OpenCL PACKAGE = OpenCL::Context
956 1242
957void 1243void
958DESTROY (OpenCL::Context context) 1244DESTROY (OpenCL::Context self)
959 CODE: 1245 CODE:
960 clReleaseContext (context); 1246 clReleaseContext (self);
961 1247
962void 1248void
963info (OpenCL::Context self, cl_context_info name) 1249info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1250 PPCODE:
965 INFO (Context) 1251 INFO (Context)
966 1252
967void 1253void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1254queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1255 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1256 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1257 XPUSH_CLOBJ (stash_queue, queue);
972 1258
973void 1259void
974user_event (OpenCL::Context self) 1260user_event (OpenCL::Context self)
975 PPCODE: 1261 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1262 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1263 XPUSH_CLOBJ (stash_userevent, ev);
978 1264
979void 1265void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1266buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1267 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1268 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?"); 1269 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1270
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1272 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1273
988void 1274void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1275buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1276 PPCODE:
991 STRLEN len; 1277 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1278 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1279 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?"); 1280 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)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1282 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1283
998#if CL_VERSION_1_2 1284#if CL_VERSION_1_2
999 1285
1000void 1286void
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) 1287image (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: 1288 PPCODE:
1003 STRLEN len; 1289 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1290 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1291 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1292 const cl_image_desc desc = {
1007 type, 1293 type,
1008 width, height, depth, 1294 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1295 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1296 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1297 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1298 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1299 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1300 HV *stash = stash_image;
1015 switch (type) 1301 switch (type)
1016 { 1302 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1303 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1304 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1305 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1306 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1307 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1308 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1309 }
1024 XPUSH_CLOBJ (klass, mem); 1310 XPUSH_CLOBJ (stash, mem);
1025 1311
1026#endif 1312#endif
1027 1313
1028void 1314void
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) 1315image2d (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)); 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1322#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1323 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)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1325#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1326 XPUSH_CLOBJ (stash_image2d, mem);
1041 1327
1042void 1328void
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) 1329image3d (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: 1330 PPCODE:
1045 STRLEN len; 1331 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1336#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1337 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)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1339#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1340 XPUSH_CLOBJ (stash_image3d, mem);
1055 1341
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1342#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1343
1058void 1344void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1345gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1346 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1348 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1349
1064void 1350void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1351gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1352 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1353 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1354 XPUSH_CLOBJ (stash_image2d, mem);
1069 1355
1070#if CL_VERSION_1_2 1356#if CL_VERSION_1_2
1071 1357
1072void 1358void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1359gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1360 ALIAS:
1075 PPCODE: 1361 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1362 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1363 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1364 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1365 HV *stash = stash_memory;
1080 switch (type) 1366 switch (type)
1081 { 1367 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1368 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1369 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1370 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1371 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1372 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1373 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1374 }
1089 XPUSH_CLOBJ (klass, mem); 1375 XPUSH_CLOBJ (stash, mem);
1090 1376
1091#endif 1377#endif
1092 1378
1093void 1379void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1380gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1382#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1384#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1385 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1386#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1387 XPUSH_CLOBJ (stash_image2d, mem);
1102 1388
1103void 1389void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1390gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1391 PPCODE:
1106#if PREFER_1_1 1392#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1394#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1395 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1396#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1397 XPUSH_CLOBJ (stash_image3d, mem);
1112 1398
1113#endif 1399#endif
1114 1400
1115void 1401void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1402supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1422
1137void 1423void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1424sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1425 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1426 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1427 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1428
1143void 1429void
1144program_with_source (OpenCL::Context self, SV *program) 1430program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1431 PPCODE:
1146 STRLEN len; 1432 STRLEN len;
1147 size_t len2; 1433 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1434 const char *ptr = SvPVbyte (program, len);
1149 1435
1150 len2 = len; 1436 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1437 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1438 XPUSH_CLOBJ (stash_program, prog);
1439
1440void
1441program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1442 PPCODE:
1443 cl_uint device_count;
1444 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1445
1446 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1447 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1448
1449 binaries = SvRV (binaries);
1450
1451 if (device_count != av_len ((AV *)binaries) + 1)
1452 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1453
1454 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1455 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1456 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1457
1458 int i;
1459 for (i = 0; i < device_count; ++i)
1460 {
1461 STRLEN len;
1462 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1463 length_list [i] = len;
1464 }
1465
1466 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1467 length_list, binary_list,
1468 GIMME_V == G_ARRAY ? status_list : 0, &res));
1469
1470 EXTEND (SP, 2);
1471 PUSH_CLOBJ (stash_program, prog);
1472
1473 if (GIMME_V == G_ARRAY)
1474 {
1475 AV *av = newAV ();
1476 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1477
1478 for (i = device_count; i--; )
1479 av_store (av, i, newSViv (status_list [i]));
1480 }
1481
1482#if CL_VERSION_1_2
1483
1484void
1485program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1486 PPCODE:
1487 cl_uint device_count;
1488 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1489
1490 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1491
1492 XPUSH_CLOBJ (stash_program, prog);
1493
1494void
1495link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1496 CODE:
1497 cl_uint device_count = 0;
1498 cl_device_id *device_list = 0;
1499
1500 if (SvOK (devices))
1501 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1502
1503 cl_uint program_count;
1504 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1505
1506 void *user_data;
1507 program_callback pfn_notify = make_program_callback (notify, &user_data);
1508
1509 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1510 program_count, program_list, pfn_notify, user_data, &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514#endif
1153 1515
1154#BEGIN:context 1516#BEGIN:context
1155 1517
1156void 1518void
1157reference_count (OpenCL::Context self) 1519reference_count (OpenCL::Context self)
1173 cl_device_id *value = tmpbuf (size); 1535 cl_device_id *value = tmpbuf (size);
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1536 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1537 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1538 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1539 for (i = 0; i < n; ++i)
1178 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1540 PUSH_CLOBJ (stash_device, value [i]);
1180 }
1181 1541
1182void 1542void
1183properties (OpenCL::Context self) 1543properties (OpenCL::Context self)
1184 PPCODE: 1544 PPCODE:
1185 size_t size; 1545 size_t size;
1199DESTROY (OpenCL::Queue self) 1559DESTROY (OpenCL::Queue self)
1200 CODE: 1560 CODE:
1201 clReleaseCommandQueue (self); 1561 clReleaseCommandQueue (self);
1202 1562
1203void 1563void
1204enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1564read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1565 ALIAS:
1566 enqueue_read_buffer = 0
1205 PPCODE: 1567 PPCODE:
1206 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
1208 1569
1209 SvUPGRADE (data, SVt_PV); 1570 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1571 SvGROW (data, len);
1211 SvPOK_only (data); 1572 SvPOK_only (data);
1212 SvCUR_set (data, len); 1573 SvCUR_set (data, len);
1574
1575 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)); 1576 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1214 1577
1215 if (ev) 1578 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
1217 1580
1218void 1581void
1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1582write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1583 ALIAS:
1584 enqueue_write_buffer = 0
1220 PPCODE: 1585 PPCODE:
1221 cl_event ev = 0; 1586 EVENT_LIST (5);
1587
1222 STRLEN len; 1588 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1589 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5);
1225 1590
1591 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)); 1592 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1227 1593
1228 if (ev) 1594 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1595 XPUSH_CLOBJ (stash_event, ev);
1230 1596
1231#if CL_VERSION_1_2 1597#if CL_VERSION_1_2
1232 1598
1233void 1599void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1600fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1601 ALIAS:
1602 enqueue_fill_buffer = 0
1235 PPCODE: 1603 PPCODE:
1236 cl_event ev = 0; 1604 EVENT_LIST (5);
1605
1237 STRLEN len; 1606 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1607 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240 1608
1609 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)); 1610 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1611
1243 if (ev) 1612 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1613 XPUSH_CLOBJ (stash_event, ev);
1245 1614
1246void 1615void
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, ...) 1616fill_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, ...)
1617 ALIAS:
1618 enqueue_fill_image = 0
1248 PPCODE: 1619 PPCODE:
1249 cl_event ev = 0; 1620 EVENT_LIST (12);
1250 STRLEN len; 1621
1251 const size_t origin [3] = { x, y, z }; 1622 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1623 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254 1624
1255 const cl_float c_f [4] = { r, g, b, a }; 1625 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1626 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1627 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1628 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1631 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1632 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) 1633 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."); 1634 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1635
1636 cl_event ev = 0;
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1637 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)); 1638 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1639
1269 if (ev) 1640 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1641 XPUSH_CLOBJ (stash_event, ev);
1271 1642
1272#endif 1643#endif
1273 1644
1274void 1645void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1646copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1647 ALIAS:
1648 enqueue_copy_buffer = 0
1276 PPCODE: 1649 PPCODE:
1650 EVENT_LIST (6);
1651
1277 cl_event ev = 0; 1652 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)); 1653 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1654
1282 if (ev) 1655 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1656 XPUSH_CLOBJ (stash_event, ev);
1284 1657
1285void 1658void
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, ...) 1659read_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, ...)
1660 ALIAS:
1661 enqueue_read_buffer_rect = 0
1287 PPCODE: 1662 PPCODE:
1288 cl_event ev = 0; 1663 EVENT_LIST (17);
1664
1289 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1665 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 }; 1666 const size_t host_origin[3] = { host_x, host_y, host_z };
1291 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17);
1293 1668
1294 if (!buf_row_pitch) 1669 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1670 buf_row_pitch = region [0];
1296 1671
1297 if (!buf_slice_pitch) 1672 if (!buf_slice_pitch)
1307 1682
1308 SvUPGRADE (data, SVt_PV); 1683 SvUPGRADE (data, SVt_PV);
1309 SvGROW (data, len); 1684 SvGROW (data, len);
1310 SvPOK_only (data); 1685 SvPOK_only (data);
1311 SvCUR_set (data, len); 1686 SvCUR_set (data, len);
1687
1688 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)); 1689 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 1690
1314 if (ev) 1691 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1692 XPUSH_CLOBJ (stash_event, ev);
1316 1693
1317void 1694void
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, ...) 1695write_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, ...)
1696 ALIAS:
1697 enqueue_write_buffer_rect = 0
1319 PPCODE: 1698 PPCODE:
1320 cl_event ev = 0; 1699 EVENT_LIST (17);
1700
1321 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1701 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 }; 1702 const size_t host_origin[3] = { host_x, host_y, host_z };
1323 const size_t region[3] = { width, height, depth }; 1703 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1704 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1705 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17);
1327 1706
1328 if (!buf_row_pitch) 1707 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1708 buf_row_pitch = region [0];
1330 1709
1331 if (!buf_slice_pitch) 1710 if (!buf_slice_pitch)
1340 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1719 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1341 1720
1342 if (len < min_len) 1721 if (len < min_len)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1722 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1723
1724 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)); 1725 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 1726
1347 if (ev) 1727 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1728 XPUSH_CLOBJ (stash_event, ev);
1349 1729
1350void 1730void
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, ...) 1731copy_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, ...)
1732 ALIAS:
1733 enqueue_copy_buffer_rect = 0
1352 PPCODE: 1734 PPCODE:
1353 cl_event ev = 0; 1735 EVENT_LIST (16);
1736
1354 const size_t src_origin[3] = { src_x, src_y, src_z }; 1737 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 }; 1738 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1356 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16);
1358 1740
1741 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)); 1742 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 1743
1361 if (ev) 1744 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1745 XPUSH_CLOBJ (stash_event, ev);
1363 1746
1364void 1747void
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, ...) 1748read_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, ...)
1749 ALIAS:
1750 enqueue_read_image = 0
1366 PPCODE: 1751 PPCODE:
1367 cl_event ev = 0; 1752 EVENT_LIST (12);
1753
1368 const size_t src_origin[3] = { src_x, src_y, src_z }; 1754 const size_t src_origin[3] = { src_x, src_y, src_z };
1369 const size_t region[3] = { width, height, depth }; 1755 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12);
1371 1756
1372 if (!row_pitch) 1757 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1758 row_pitch = img_row_pitch (src);
1374 1759
1375 if (depth > 1 && !slice_pitch) 1760 if (depth > 1 && !slice_pitch)
1379 1764
1380 SvUPGRADE (data, SVt_PV); 1765 SvUPGRADE (data, SVt_PV);
1381 SvGROW (data, len); 1766 SvGROW (data, len);
1382 SvPOK_only (data); 1767 SvPOK_only (data);
1383 SvCUR_set (data, len); 1768 SvCUR_set (data, len);
1769
1770 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)); 1771 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 1772
1386 if (ev) 1773 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1774 XPUSH_CLOBJ (stash_event, ev);
1388 1775
1389void 1776void
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, ...) 1777write_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, ...)
1778 ALIAS:
1779 enqueue_write_image = 0
1391 PPCODE: 1780 PPCODE:
1392 cl_event ev = 0; 1781 EVENT_LIST (12);
1782
1393 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1783 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1394 const size_t region[3] = { width, height, depth }; 1784 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1785 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1786 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12);
1398 1787
1399 if (!row_pitch) 1788 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1789 row_pitch = img_row_pitch (dst);
1401 1790
1402 if (depth > 1 && !slice_pitch) 1791 if (depth > 1 && !slice_pitch)
1405 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1794 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1406 1795
1407 if (len < min_len) 1796 if (len < min_len)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1797 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1798
1799 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)); 1800 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 1801
1412 if (ev) 1802 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1803 XPUSH_CLOBJ (stash_event, ev);
1414 1804
1415void 1805void
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, ...) 1806copy_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, ...)
1807 ALIAS:
1808 enqueue_copy_image = 0
1417 PPCODE: 1809 PPCODE:
1418 cl_event ev = 0; 1810 EVENT_LIST (12);
1811
1419 const size_t src_origin[3] = { src_x, src_y, src_z }; 1812 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 }; 1813 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1421 const size_t region[3] = { width, height, depth }; 1814 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12);
1423 1815
1816 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)); 1817 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1818
1426 if (ev) 1819 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1820 XPUSH_CLOBJ (stash_event, ev);
1428 1821
1429void 1822void
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, ...) 1823copy_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, ...)
1824 ALIAS:
1825 enqueue_copy_image_to_buffer = 0
1431 PPCODE: 1826 PPCODE:
1827 EVENT_LIST (10);
1828
1829 const size_t src_origin[3] = { src_x, src_y, src_z };
1830 const size_t region [3] = { width, height, depth };
1831
1432 cl_event ev = 0; 1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839copy_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, ...)
1840 ALIAS:
1841 enqueue_copy_buffer_to_image = 0
1842 PPCODE:
1843 EVENT_LIST (10);
1844
1433 const size_t src_origin[3] = { src_x, src_y, src_z }; 1845 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1846 const size_t region [3] = { width, height, depth };
1847
1848 cl_event ev = 0;
1849 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1850
1851 if (ev)
1852 XPUSH_CLOBJ (stash_event, ev);
1853
1854void
1855map_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, ...)
1856 ALIAS:
1857 enqueue_map_buffer = 0
1858 PPCODE:
1859 EVENT_LIST (6);
1860
1861 size_t cb = SvIV (cb_);
1862
1863 if (!SvOK (cb_))
1864 {
1865 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1866 cb -= offset;
1867 }
1868
1869 cl_event ev;
1870 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1871 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1872
1873void
1874map_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, ...)
1875 ALIAS:
1876 enqueue_map_image = 0
1877 PPCODE:
1878 size_t width = SvIV (width_);
1879 if (!SvOK (width_))
1880 {
1881 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1882 width -= x;
1883 }
1884
1885 size_t height = SvIV (width_);
1886 if (!SvOK (height_))
1887 {
1888 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1889 height -= y;
1890
1891 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1892 if (!height)
1893 height = 1;
1894 }
1895
1896 size_t depth = SvIV (width_);
1897 if (!SvOK (depth_))
1898 {
1899 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1900 depth -= z;
1901
1902 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1903 if (!depth)
1904 depth = 1;
1905 }
1906
1907 const size_t origin[3] = { x, y, z };
1434 const size_t region[3] = { width, height, depth }; 1908 const size_t region[3] = { width, height, depth };
1909 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1910 EVENT_LIST (10);
1436 1911
1912 cl_event ev;
1913 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));
1914
1915 size_t cb = slice_pitch ? slice_pitch * region [2]
1916 : row_pitch ? row_pitch * region [1]
1917 : region [0];
1918
1919 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1920
1921void
1922unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1923 PPCODE:
1924 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1925 if (GIMME_V != G_VOID)
1926 {
1927 clRetainEvent (mapped->event);
1928 XPUSH_CLOBJ (stash_event, mapped->event);
1929 }
1930
1931void
1932task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1933 ALIAS:
1934 enqueue_task = 0
1935 PPCODE:
1936 EVENT_LIST (2);
1937
1938 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)); 1939 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1438 1940
1439 if (ev) 1941 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1942 XPUSH_CLOBJ (stash_event, ev);
1441 1943
1442void 1944void
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, ...) 1945nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1946 ALIAS:
1947 enqueue_nd_range_kernel = 0
1468 PPCODE: 1948 PPCODE:
1469 cl_event ev = 0; 1949 EVENT_LIST (5);
1950
1470 size_t *gwo = 0, *gws, *lws = 0; 1951 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1952 int gws_len;
1472 size_t *lists; 1953 size_t *lists;
1473 int i; 1954 int i;
1474 EVENT_LIST (5, items - 5);
1475 1955
1476 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1956 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1957 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1958
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1959 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1960
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1961 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1962
1483 gws = lists + gws_len * 0; 1963 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1964 for (i = 0; i < gws_len; ++i)
1965 {
1485 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1966 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1967 // at least nvidia crashes for 0-sized work group sizes, work around
1968 if (!gws [i])
1969 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1970 }
1486 1971
1487 if (SvOK (global_work_offset)) 1972 if (SvOK (global_work_offset))
1488 { 1973 {
1489 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1974 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"); 1975 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1498 } 1983 }
1499 1984
1500 if (SvOK (local_work_size)) 1985 if (SvOK (local_work_size))
1501 { 1986 {
1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1987 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"); 1988 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1504 1989
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1990 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"); 1991 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1507 1992
1508 lws = lists + gws_len * 2; 1993 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 1994 for (i = 0; i < gws_len; ++i)
1995 {
1510 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1996 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1997 // at least nvidia crashes for 0-sized work group sizes, work around
1998 if (!lws [i])
1999 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2000 }
1511 } 2001 }
1512 2002
2003 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)); 2004 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1514 2005
1515 if (ev) 2006 if (ev)
1516 XPUSH_CLOBJ ("OpenCL::Event", ev); 2007 XPUSH_CLOBJ (stash_event, ev);
2008
2009#if CL_VERSION_1_2
2010
2011void
2012migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2013 ALIAS:
2014 enqueue_migrate_mem_objects = 0
2015 PPCODE:
2016 EVENT_LIST (3);
2017
2018 cl_uint obj_count;
2019 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2020
2021 cl_event ev = 0;
2022 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2023
2024 if (ev)
2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#endif
1517 2028
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 2029#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 2030
1520void 2031void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2032acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 2033 ALIAS:
2034 release_gl_objects = 1
2035 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 2036 enqueue_release_gl_objects = 1
1524 PPCODE: 2037 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2038 EVENT_LIST (2);
1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2039
2040 cl_uint obj_count;
2041 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1527 2042
1528 cl_event ev = 0; 2043 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 2044
1538 if (ix) 2045 if (ix)
1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1540 else 2047 else
1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2048 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1542 2049
1543 if (ev) 2050 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 2051 XPUSH_CLOBJ (stash_event, ev);
1545 2052
1546#endif 2053#endif
1547 2054
1548void 2055void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 2056wait_for_events (OpenCL::Queue self, ...)
2057 ALIAS:
2058 enqueue_wait_for_events = 0
1550 CODE: 2059 CODE:
1551 EVENT_LIST (1, items - 1); 2060 EVENT_LIST (1);
1552#if PREFER_1_1 2061#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 2063#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2064 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 2065#endif
1557 2066
1558void 2067void
1559enqueue_marker (OpenCL::Queue self, ...) 2068marker (OpenCL::Queue self, ...)
2069 ALIAS:
2070 enqueue_marker = 0
1560 PPCODE: 2071 PPCODE:
2072 EVENT_LIST (1);
1561 cl_event ev = 0; 2073 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1 2074#if PREFER_1_1
1564 if (!event_list_count) 2075 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2076 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 2077 else
1567#if CL_VERSION_1_2 2078#if CL_VERSION_1_2
1574#endif 2085#endif
1575#else 2086#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2087 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif 2088#endif
1578 if (ev) 2089 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 2090 XPUSH_CLOBJ (stash_event, ev);
1580 2091
1581void 2092void
1582enqueue_barrier (OpenCL::Queue self, ...) 2093barrier (OpenCL::Queue self, ...)
2094 ALIAS:
2095 enqueue_barrier = 0
1583 PPCODE: 2096 PPCODE:
2097 EVENT_LIST (1);
1584 cl_event ev = 0; 2098 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1 2099#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 2100 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 2101 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 2102 else
1590#if CL_VERSION_1_2 2103#if CL_VERSION_1_2
1600#endif 2113#endif
1601#else 2114#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2115 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif 2116#endif
1604 if (ev) 2117 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 2118 XPUSH_CLOBJ (stash_event, ev);
1606 2119
1607void 2120void
1608flush (OpenCL::Queue self) 2121flush (OpenCL::Queue self)
1609 CODE: 2122 CODE:
1610 NEED_SUCCESS (Flush, (self)); 2123 NEED_SUCCESS (Flush, (self));
1626 PPCODE: 2139 PPCODE:
1627 cl_context value [1]; 2140 cl_context value [1];
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 2142 EXTEND (SP, 1);
1630 const int i = 0; 2143 const int i = 0;
1631 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 2144 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2145 PUSH_CLOBJ (stash_context, value [i]);
1634 }
1635 2146
1636void 2147void
1637device (OpenCL::Queue self) 2148device (OpenCL::Queue self)
1638 PPCODE: 2149 PPCODE:
1639 cl_device_id value [1]; 2150 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2151 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 2152 EXTEND (SP, 1);
1642 const int i = 0; 2153 const int i = 0;
1643 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2154 PUSH_CLOBJ (stash_device, value [i]);
1645 }
1646 2155
1647void 2156void
1648reference_count (OpenCL::Queue self) 2157reference_count (OpenCL::Queue self)
1649 PPCODE: 2158 PPCODE:
1650 cl_uint value [1]; 2159 cl_uint value [1];
1654 PUSHs (sv_2mortal (newSVuv (value [i]))); 2163 PUSHs (sv_2mortal (newSVuv (value [i])));
1655 2164
1656void 2165void
1657properties (OpenCL::Queue self) 2166properties (OpenCL::Queue self)
1658 PPCODE: 2167 PPCODE:
1659 cl_command_queue_properties value [1]; 2168 cl_ulong value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2169 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 2170 EXTEND (SP, 1);
1662 const int i = 0; 2171 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 2172 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 2173
1665#END:command_queue 2174#END:command_queue
1666 2175
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 2176MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 2177
1674void 2183void
1675info (OpenCL::Memory self, cl_mem_info name) 2184info (OpenCL::Memory self, cl_mem_info name)
1676 PPCODE: 2185 PPCODE:
1677 INFO (MemObject) 2186 INFO (MemObject)
1678 2187
2188void
2189destructor_callback (OpenCL::Memory self, SV *notify)
2190 PPCODE:
2191 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2192
1679#BEGIN:mem 2193#BEGIN:mem
1680 2194
1681void 2195void
1682type (OpenCL::Memory self) 2196type (OpenCL::Memory self)
2197 ALIAS:
2198 type = CL_MEM_TYPE
2199 map_count = CL_MEM_MAP_COUNT
2200 reference_count = CL_MEM_REFERENCE_COUNT
1683 PPCODE: 2201 PPCODE:
1684 cl_mem_object_type value [1]; 2202 cl_uint value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2203 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 2204 EXTEND (SP, 1);
1687 const int i = 0; 2205 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 2206 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 2207
1690void 2208void
1691flags (OpenCL::Memory self) 2209flags (OpenCL::Memory self)
1692 PPCODE: 2210 PPCODE:
1693 cl_mem_flags value [1]; 2211 cl_ulong value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2213 EXTEND (SP, 1);
1696 const int i = 0; 2214 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 2215 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 2216
1699void 2217void
1700size (OpenCL::Memory self) 2218size (OpenCL::Memory self)
1701 ALIAS: 2219 ALIAS:
1702 size = CL_MEM_SIZE 2220 size = CL_MEM_SIZE
1716 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1717 const int i = 0; 2235 const int i = 0;
1718 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2236 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1719 2237
1720void 2238void
1721map_count (OpenCL::Memory self)
1722 ALIAS:
1723 map_count = CL_MEM_MAP_COUNT
1724 reference_count = CL_MEM_REFERENCE_COUNT
1725 PPCODE:
1726 cl_uint value [1];
1727 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1728 EXTEND (SP, 1);
1729 const int i = 0;
1730 PUSHs (sv_2mortal (newSVuv (value [i])));
1731
1732void
1733context (OpenCL::Memory self) 2239context (OpenCL::Memory self)
1734 PPCODE: 2240 PPCODE:
1735 cl_context value [1]; 2241 cl_context value [1];
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2242 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 2243 EXTEND (SP, 1);
1738 const int i = 0; 2244 const int i = 0;
1739 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 2245 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2246 PUSH_CLOBJ (stash_context, value [i]);
1742 }
1743 2247
1744void 2248void
1745associated_memobject (OpenCL::Memory self) 2249associated_memobject (OpenCL::Memory self)
1746 PPCODE: 2250 PPCODE:
1747 cl_mem value [1]; 2251 cl_mem value [1];
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2253 EXTEND (SP, 1);
1750 const int i = 0; 2254 const int i = 0;
1751 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 2255 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2256 PUSH_CLOBJ (stash_memory, value [i]);
1754 }
1755 2257
1756#END:mem 2258#END:mem
1757 2259
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 2260#if cl_apple_gl_sharing || cl_khr_gl_sharing
1759 2261
1778 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2280 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1779 2281
1780 cl_buffer_region crdata = { origin, size }; 2282 cl_buffer_region crdata = { origin, size };
1781 2283
1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2284 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2285 XPUSH_CLOBJ (stash_buffer, mem);
1784 2286
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2287MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2288
1787void 2289void
1788image_info (OpenCL::Image self, cl_image_info name) 2290image_info (OpenCL::Image self, cl_image_info name)
1858 2360
1859#BEGIN:sampler 2361#BEGIN:sampler
1860 2362
1861void 2363void
1862reference_count (OpenCL::Sampler self) 2364reference_count (OpenCL::Sampler self)
2365 ALIAS:
2366 reference_count = CL_SAMPLER_REFERENCE_COUNT
2367 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2368 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1863 PPCODE: 2369 PPCODE:
1864 cl_uint value [1]; 2370 cl_uint value [1];
1865 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1866 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
1867 const int i = 0; 2373 const int i = 0;
1868 PUSHs (sv_2mortal (newSVuv (value [i]))); 2374 PUSHs (sv_2mortal (newSVuv (value [i])));
1869 2375
1870void 2376void
1872 PPCODE: 2378 PPCODE:
1873 cl_context value [1]; 2379 cl_context value [1];
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2380 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2381 EXTEND (SP, 1);
1876 const int i = 0; 2382 const int i = 0;
1877 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2384 PUSH_CLOBJ (stash_context, value [i]);
1880 }
1881
1882void
1883normalized_coords (OpenCL::Sampler self)
1884 PPCODE:
1885 cl_addressing_mode value [1];
1886 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1887 EXTEND (SP, 1);
1888 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i])));
1890
1891void
1892addressing_mode (OpenCL::Sampler self)
1893 PPCODE:
1894 cl_filter_mode value [1];
1895 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1896 EXTEND (SP, 1);
1897 const int i = 0;
1898 PUSHs (sv_2mortal (newSViv (value [i])));
1899 2385
1900void 2386void
1901filter_mode (OpenCL::Sampler self) 2387filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2388 PPCODE:
1903 cl_bool value [1]; 2389 cl_uint value [1];
1904 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1905 EXTEND (SP, 1); 2391 EXTEND (SP, 1);
1906 const int i = 0; 2392 const int i = 0;
1907 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2393 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1908 2394
1918void 2404void
1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2405build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS: 2406 ALIAS:
1921 build_async = 1 2407 build_async = 1
1922 CODE: 2408 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; 2409 cl_uint device_count = 0;
1926 cl_device_id *device_list = 0; 2410 cl_device_id *device_list = 0;
1927 2411
1928 if (SvOK (devices)) 2412 if (SvOK (devices))
1929 { 2413 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2414
2415 void *user_data;
2416 program_callback pfn_notify = make_program_callback (notify, &user_data);
2417
2418 if (ix)
2419 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2420 else
2421 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2422
2423#if CL_VERSION_1_2
2424
2425void
2426compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2427 CODE:
2428 cl_uint device_count = 0;
2429 cl_device_id *device_list = 0;
2430
2431 if (SvOK (devices))
2432 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2433
2434 cl_uint header_count = 0;
2435 cl_program *header_list = 0;
2436 const char **header_name = 0;
2437
2438 if (SvOK (headers))
2439 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2440 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2441 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1932 2442
1933 AV *av = (AV *)SvRV (devices); 2443 HV *hv = (HV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935 2444
1936 if (num_devices) 2445 header_count = hv_iterinit (hv);
2446 header_list = tmpbuf (sizeof (*header_list) * header_count);
2447 header_name = tmpbuf (sizeof (*header_name) * header_count);
2448
2449 HE *he;
2450 int i = 0;
2451 while (he = hv_iternext (hv))
1937 { 2452 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2453 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2454 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1939 int count; 2455 ++i;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 } 2456 }
1943 } 2457 }
1944 2458
1945 if (SvOK (notify)) 2459 void *user_data;
1946 { 2460 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 2461
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)); 2462 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2463 header_count, header_list, header_name, pfn_notify, user_data));
2464
2465#endif
1956 2466
1957void 2467void
1958build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2468build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1959 PPCODE: 2469 PPCODE:
1960 size_t size; 2470 size_t size;
1969#BEGIN:program_build 2479#BEGIN:program_build
1970 2480
1971void 2481void
1972build_status (OpenCL::Program self, OpenCL::Device device) 2482build_status (OpenCL::Program self, OpenCL::Device device)
1973 PPCODE: 2483 PPCODE:
1974 cl_build_status value [1]; 2484 cl_int value [1];
1975 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2485 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1976 EXTEND (SP, 1); 2486 EXTEND (SP, 1);
1977 const int i = 0; 2487 const int i = 0;
1978 PUSHs (sv_2mortal (newSViv (value [i]))); 2488 PUSHs (sv_2mortal (newSViv (value [i])));
1979 2489
1989 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2499 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1990 EXTEND (SP, 1); 2500 EXTEND (SP, 1);
1991 const int i = 0; 2501 const int i = 0;
1992 PUSHs (sv_2mortal (newSVpv (value, 0))); 2502 PUSHs (sv_2mortal (newSVpv (value, 0)));
1993 2503
2504void
2505binary_type (OpenCL::Program self, OpenCL::Device device)
2506 PPCODE:
2507 cl_uint value [1];
2508 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2509 EXTEND (SP, 1);
2510 const int i = 0;
2511 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2512
1994#END:program_build 2513#END:program_build
1995 2514
1996void 2515void
1997kernel (OpenCL::Program program, SV *function) 2516kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2517 PPCODE:
1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2518 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2519 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2520
2002void 2521void
2003kernels_in_program (OpenCL::Program program) 2522kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2523 PPCODE:
2005 cl_uint num_kernels; 2524 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2527 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2528
2010 int i; 2529 int i;
2011 EXTEND (SP, num_kernels); 2530 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2531 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2532 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2533
2015void 2534void
2016info (OpenCL::Program self, cl_program_info name) 2535info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2536 PPCODE:
2018 INFO (Program) 2537 INFO (Program)
2064 PPCODE: 2583 PPCODE:
2065 cl_context value [1]; 2584 cl_context value [1];
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2585 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2586 EXTEND (SP, 1);
2068 const int i = 0; 2587 const int i = 0;
2069 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2588 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2589 PUSH_CLOBJ (stash_context, value [i]);
2072 }
2073 2590
2074void 2591void
2075devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
2076 PPCODE: 2593 PPCODE:
2077 size_t size; 2594 size_t size;
2079 cl_device_id *value = tmpbuf (size); 2596 cl_device_id *value = tmpbuf (size);
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2599 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
2084 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2601 PUSH_CLOBJ (stash_device, value [i]);
2086 }
2087 2602
2088void 2603void
2089source (OpenCL::Program self) 2604source (OpenCL::Program self)
2090 PPCODE: 2605 PPCODE:
2091 size_t size; 2606 size_t size;
2116DESTROY (OpenCL::Kernel self) 2631DESTROY (OpenCL::Kernel self)
2117 CODE: 2632 CODE:
2118 clReleaseKernel (self); 2633 clReleaseKernel (self);
2119 2634
2120void 2635void
2636setf (OpenCL::Kernel self, const char *format, ...)
2637 CODE:
2638 int i;
2639 for (i = 2; ; ++i)
2640 {
2641 while (*format == ' ')
2642 ++format;
2643
2644 char type = *format++;
2645
2646 if (!type)
2647 break;
2648
2649 if (i >= items)
2650 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2651
2652 SV *sv = ST (i);
2653
2654 union
2655 {
2656 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2657 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2658 cl_half ch; cl_float cf; cl_double cd;
2659 cl_mem cm;
2660 cl_sampler ca;
2661 size_t cz;
2662 cl_event ce;
2663 } arg;
2664 size_t size;
2665 int nullarg = 0;
2666
2667 switch (type)
2668 {
2669 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2670 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2671 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2672 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2673 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2674 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2675 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2676 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2677
2678 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2679 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2680 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2681
2682 case 'z': nullarg = 1; size = SvIV (sv); break;
2683
2684 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2685 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2686 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2687
2688 default:
2689 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2690 }
2691
2692 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2693 if (res)
2694 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2695 }
2696
2697 if (i != items)
2698 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2699
2700void
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2701set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2702 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2703 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2704
2125void 2705void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2753 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2754
2175void 2755void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2756set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2757 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2758 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2759
2180void 2760void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2761set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2762 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2764
2185void 2765void
2186set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2766set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2187 ALIAS:
2188 set_image2d = 0
2189 set_image3d = 0
2190 CODE: 2767 CODE:
2191 clSetKernelArg (self, idx, sizeof (value), &value); 2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2192 2769
2193void 2770void
2194set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2771set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2195 CODE: 2772 CODE:
2196 clSetKernelArg (self, idx, sizeof (value), &value); 2773 clSetKernelArg (self, idx, sizeof (value), &value);
2240 PPCODE: 2817 PPCODE:
2241 cl_context value [1]; 2818 cl_context value [1];
2242 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2819 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2243 EXTEND (SP, 1); 2820 EXTEND (SP, 1);
2244 const int i = 0; 2821 const int i = 0;
2245 {
2246 NEED_SUCCESS (RetainContext, (value [i])); 2822 NEED_SUCCESS (RetainContext, (value [i]));
2247 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2823 PUSH_CLOBJ (stash_context, value [i]);
2248 }
2249 2824
2250void 2825void
2251program (OpenCL::Kernel self) 2826program (OpenCL::Kernel self)
2252 PPCODE: 2827 PPCODE:
2253 cl_program value [1]; 2828 cl_program value [1];
2254 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2829 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2255 EXTEND (SP, 1); 2830 EXTEND (SP, 1);
2256 const int i = 0; 2831 const int i = 0;
2257 {
2258 NEED_SUCCESS (RetainProgram, (value [i])); 2832 NEED_SUCCESS (RetainProgram, (value [i]));
2259 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2833 PUSH_CLOBJ (stash_program, value [i]);
2260 }
2261 2834
2262#END:kernel 2835#END:kernel
2263 2836
2264void 2837void
2265work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2838work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2311 const int i = 0; 2884 const int i = 0;
2312 PUSHs (sv_2mortal (newSVuv (value [i]))); 2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2313 2886
2314#END:kernel_work_group 2887#END:kernel_work_group
2315 2888
2889#if CL_VERSION_1_2
2890
2891void
2892arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2893 PPCODE:
2894 size_t size;
2895 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2896 SV *sv = sv_2mortal (newSV (size));
2897 SvUPGRADE (sv, SVt_PV);
2898 SvPOK_only (sv);
2899 SvCUR_set (sv, size);
2900 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2901 XPUSHs (sv);
2902
2903#BEGIN:kernel_arg
2904
2905void
2906arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2907 ALIAS:
2908 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2909 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2910 PPCODE:
2911 cl_uint value [1];
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2913 EXTEND (SP, 1);
2914 const int i = 0;
2915 PUSHs (sv_2mortal (newSVuv (value [i])));
2916
2917void
2918arg_type_name (OpenCL::Kernel self, cl_uint idx)
2919 ALIAS:
2920 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2921 arg_name = CL_KERNEL_ARG_NAME
2922 PPCODE:
2923 size_t size;
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2925 char *value = tmpbuf (size);
2926 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2927 EXTEND (SP, 1);
2928 const int i = 0;
2929 PUSHs (sv_2mortal (newSVpv (value, 0)));
2930
2931void
2932arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2933 PPCODE:
2934 cl_ulong value [1];
2935 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2936 EXTEND (SP, 1);
2937 const int i = 0;
2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2939
2940#END:kernel_arg
2941
2942#endif
2943
2316MODULE = OpenCL PACKAGE = OpenCL::Event 2944MODULE = OpenCL PACKAGE = OpenCL::Event
2317 2945
2318void 2946void
2319DESTROY (OpenCL::Event self) 2947DESTROY (OpenCL::Event self)
2320 CODE: 2948 CODE:
2342 PPCODE: 2970 PPCODE:
2343 cl_command_queue value [1]; 2971 cl_command_queue value [1];
2344 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2972 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2345 EXTEND (SP, 1); 2973 EXTEND (SP, 1);
2346 const int i = 0; 2974 const int i = 0;
2347 {
2348 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2975 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2349 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2976 PUSH_CLOBJ (stash_queue, value [i]);
2350 }
2351 2977
2352void 2978void
2353command_type (OpenCL::Event self) 2979command_type (OpenCL::Event self)
2354 PPCODE:
2355 cl_command_type value [1];
2356 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2357 EXTEND (SP, 1);
2358 const int i = 0;
2359 PUSHs (sv_2mortal (newSVuv (value [i])));
2360
2361void
2362reference_count (OpenCL::Event self)
2363 ALIAS: 2980 ALIAS:
2981 command_type = CL_EVENT_COMMAND_TYPE
2364 reference_count = CL_EVENT_REFERENCE_COUNT 2982 reference_count = CL_EVENT_REFERENCE_COUNT
2365 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2983 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2366 PPCODE: 2984 PPCODE:
2367 cl_uint value [1]; 2985 cl_uint value [1];
2368 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2986 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2375 PPCODE: 2993 PPCODE:
2376 cl_context value [1]; 2994 cl_context value [1];
2377 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2995 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2378 EXTEND (SP, 1); 2996 EXTEND (SP, 1);
2379 const int i = 0; 2997 const int i = 0;
2380 {
2381 NEED_SUCCESS (RetainContext, (value [i])); 2998 NEED_SUCCESS (RetainContext, (value [i]));
2382 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2999 PUSH_CLOBJ (stash_context, value [i]);
2383 }
2384 3000
2385#END:event 3001#END:event
2386 3002
2387void 3003void
2388profiling_info (OpenCL::Event self, cl_profiling_info name) 3004profiling_info (OpenCL::Event self, cl_profiling_info name)
2412void 3028void
2413set_status (OpenCL::UserEvent self, cl_int execution_status) 3029set_status (OpenCL::UserEvent self, cl_int execution_status)
2414 CODE: 3030 CODE:
2415 clSetUserEventStatus (self, execution_status); 3031 clSetUserEventStatus (self, execution_status);
2416 3032
3033MODULE = OpenCL PACKAGE = OpenCL::Mapped
3034
3035void
3036DESTROY (SV *self)
3037 CODE:
3038 OpenCL__Mapped mapped = SvMAPPED (self);
3039
3040 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3041 mapped_detach (self, mapped);
3042
3043 clReleaseCommandQueue (mapped->queue);
3044 clReleaseEvent (mapped->event);
3045 Safefree (mapped);
3046
3047void
3048unmap (OpenCL::Mapped self, ...)
3049 CODE:
3050 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3051
3052bool
3053mapped (OpenCL::Mapped self)
3054 CODE:
3055 RETVAL = !!self->ptr;
3056 OUTPUT:
3057 RETVAL
3058
3059void
3060wait (OpenCL::Mapped self)
3061 PPCODE:
3062 if (self->event)
3063 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3064
3065void
3066event (OpenCL::Mapped self)
3067 PPCODE:
3068 if (!self->event)
3069 XSRETURN_UNDEF;
3070
3071 clRetainEvent (self->event);
3072 XPUSH_CLOBJ (stash_event, self->event);
3073
3074#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3075#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3076#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3077#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3078#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3079#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3080
3081IV
3082size (OpenCL::Mapped self)
3083 ALIAS:
3084 size = MAPPED_OFFSET_CB
3085 row_pitch = MAPPED_OFFSET_ROW_PITCH
3086 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3087 width = MAPPED_OFFSET_WIDTH
3088 height = MAPPED_OFFSET_HEIGHT
3089 depth = MAPPED_OFFSET_DEPTH
3090 CODE:
3091 RETVAL = *(size_t *)((char *)self + ix);
3092 OUTPUT:
3093 RETVAL
3094
3095IV
3096ptr (OpenCL::Mapped self)
3097 CODE:
3098 RETVAL = PTR2IV (self->ptr);
3099 OUTPUT:
3100 RETVAL
3101
3102void
3103set (OpenCL::Mapped self, size_t offset, SV *data)
3104 CODE:
3105 STRLEN len;
3106 const char *ptr = SvPVbyte (data, len);
3107
3108 if (offset + len > self->cb)
3109 croak ("OpenCL::Mapped::set out of bound condition detected");
3110
3111 memcpy (offset + (char *)self->ptr, ptr, len);
3112
3113void
3114get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3115 PPCODE:
3116 if (!SvOK (ST (1)))
3117 count = self->width - x;
3118
3119 if (x + count > self->width)
3120 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3121
3122 if (y >= self->height)
3123 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3124
3125 if (z >= self->depth)
3126 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3127
3128 size_t element = mapped_element_size (self);
3129
3130 count *= element;
3131 x *= element;
3132
3133 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3134 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3135
3136void
3137set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3138 PPCODE:
3139 STRLEN count;
3140 char *dataptr = SvPVbyte (data, count);
3141 size_t element = mapped_element_size (self);
3142
3143 x *= element;
3144
3145 if (x + count > self->width * element)
3146 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3147
3148 if (y >= self->height)
3149 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3150
3151 if (z >= self->depth)
3152 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3153
3154 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3155 memcpy (ptr, dataptr, count);
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3158
3159MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3160
3161IV
3162element_size (OpenCL::Mapped self)
3163 CODE:
3164 RETVAL = mapped_element_size (self);
3165 OUTPUT:
3166 RETVAL
3167

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines