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 1.80 by root, Sat May 5 16:44:02 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#ifndef CL_VERSION_1_2
43 #include CLHDR(cl_d3d9.h)
44#endif
45
46#if _WIN32
47 #include CLHDR(cl_d3d10.h)
48 #if CL_VERSION_1_2
49 #include CLHDR<cl_d3d11.h>
50 #endif
51 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 52#endif
35 53
36#ifndef CL_VERSION_1_2 54#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 55 #undef PREFER_1_1
38 #define PREFER_1_1 1 56 #define PREFER_1_1 1
39#endif 57#endif
40 58
59// make sure all constants we might use are actually defined
60#include "default.h"
61
41typedef cl_platform_id OpenCL__Platform; 62typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 63typedef cl_device_id OpenCL__Device;
64typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 65typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 66typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 67typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 68typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 69typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 70typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 71typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 72typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 73typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 74typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 75typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 76typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 77typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 78typedef cl_event OpenCL__UserEvent;
61 79
62typedef SV *FUTURE; 80typedef struct mapped * OpenCL__Mapped;
81
82static HV
83 *stash_platform,
84 *stash_device,
85 *stash_subdevice,
86 *stash_context,
87 *stash_queue,
88 *stash_program,
89 *stash_kernel,
90 *stash_sampler,
91 *stash_event,
92 *stash_userevent,
93 *stash_memory,
94 *stash_buffer,
95 *stash_bufferobj,
96 *stash_image,
97 *stash_image1d,
98 *stash_image1darray,
99 *stash_image1dbuffer,
100 *stash_image2d,
101 *stash_image2darray,
102 *stash_image3d,
103 *stash_mapped,
104 *stash_mappedbuffer,
105 *stash_mappedimage;
63 106
64/*****************************************************************************/ 107/*****************************************************************************/
65 108
66// name must include a leading underscore 109// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 110// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 150}
108 151
109/*****************************************************************************/ 152/*****************************************************************************/
110 153
111/* up to two temporary buffers */ 154/* up to two temporary buffers */
112static void * 155static void * ecb_noinline
113tmpbuf (size_t size) 156tmpbuf (size_t size)
114{ 157{
115 enum { buffers = 3 }; 158 enum { buffers = 4 };
116 static int idx; 159 static int idx;
117 static void *buf [buffers]; 160 static void *buf [buffers];
118 static size_t len [buffers]; 161 static size_t len [buffers];
119 162
120 idx = (idx + 1) % buffers; 163 idx = (idx + 1) % buffers;
127 } 170 }
128 171
129 return buf [idx]; 172 return buf [idx];
130} 173}
131 174
175static const char * ecb_noinline
176cv_get_name (CV *cv)
177{
178 static char fullname [256];
179
180 GV *gv = CvGV (cv); // gv better be nonzero
181
182 HV *stash = GvSTASH (gv);
183 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
184 const char *gvname = GvNAME (gv);
185
186 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
187 return fullname;
188}
189
132/*****************************************************************************/ 190/*****************************************************************************/
133 191
134typedef struct 192typedef struct
135{ 193{
136 IV iv; 194 IV iv;
137 const char *name; 195 const char *name;
138 #define const_iv(name) { (IV)CL_ ## name, # name }, 196 #define const_iv(name) { (IV)CL_ ## name, # name },
139} ivstr; 197} ivstr;
198
199typedef struct
200{
201 NV nv;
202 const char *name;
203 #define const_nv(name) { (NV)CL_ ## name, # name },
204} nvstr;
140 205
141static const char * 206static const char *
142iv2str (IV value, const ivstr *base, int count, const char *fallback) 207iv2str (IV value, const ivstr *base, int count, const char *fallback)
143{ 208{
144 int i; 209 int i;
193 if (res) \ 258 if (res) \
194 FAIL (name); 259 FAIL (name);
195 260
196/*****************************************************************************/ 261/*****************************************************************************/
197 262
263static SV *
264new_clobj (HV *stash, IV id)
265{
266 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
267}
268
269#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
270#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
271
272/* cl objects are either \$iv, or [$iv, ...] */
273/* they can be upgraded at runtime to the array form */
274static void * ecb_noinline
275SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
276{
277 // sv_derived_from is quite slow :(
278 if (SvROK (sv) && sv_derived_from (sv, pkg))
279 return (void *)SvIV (SvRV (sv));
280
281 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
282}
283
284// the "no-inherit" version of the above
285static void * ecb_noinline
286SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
287{
288 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
289 return (void *)SvIV (SvRV (sv));
290
291 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
292}
293
294/*****************************************************************************/
295
198static cl_context_properties * 296static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 297SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 298{
201 if (!sv || !SvOK (sv)) 299 if (!sv || !SvOK (sv))
202 if (extra) 300 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 301 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 302 else
210 int i, len = av_len (av) + 1; 308 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 309 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 310 cl_context_properties *l = p;
213 311
214 if (len & 1) 312 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 313 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 314
217 while (extracount--) 315 while (extracount--)
218 *l++ = *extra++; 316 *l++ = *extra++;
219 317
220 for (i = 0; i < len; i += 2) 318 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 321 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 322 cl_context_properties v = SvIV (p_sv); // code below can override
225 323
226 switch (t) 324 switch (t)
227 { 325 {
326 case CL_CONTEXT_PLATFORM:
327 if (SvROK (p_sv))
328 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
329 break;
330
228 case CL_GLX_DISPLAY_KHR: 331 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 332 if (!SvOK (p_sv))
230 { 333 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 334 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 335 if (func)
255 *l = 0; 358 *l = 0;
256 359
257 return p; 360 return p;
258 } 361 }
259 362
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 363 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 364}
262 365
263/*****************************************************************************/ 366// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 367// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 368static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 369object_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{ 370{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 371 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
275 return (void *)SvIV (SvRV (sv)); 372 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
373 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
276 374
277 croak ("%s: %s is not of type %s", func, svname, pkg); 375 AV *av = (AV *)SvRV (arg);
376 void **list = 0;
377 cl_uint count = av_len (av) + 1;
378
379 if (count)
380 {
381 list = tmpbuf (sizeof (*list) * count);
382 int i;
383 for (i = 0; i < count; ++i)
384 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
385 }
386
387 if (!count && !or_undef)
388 croak ("%s: '%s' must contain at least one %s object",
389 cv_get_name (cv), argname, klass);
390
391 *rcount = count;
392 return (void *)list;
278} 393}
279 394
280/*****************************************************************************/ 395/*****************************************************************************/
281/* callback stuff */ 396/* callback stuff */
282 397
304static void (*eq_signal_func)(void *signal_arg, int value); 419static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 420static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 421static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 422static eq_item *eq_head, *eq_tail;
308 423
309static void 424static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 425eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 426{
312 eq_item *item = malloc (sizeof (eq_item)); 427 eq_item *item = malloc (sizeof (eq_item));
313 428
314 item->next = 0; 429 item->next = 0;
381} 496}
382 497
383/*****************************************************************************/ 498/*****************************************************************************/
384/* context notify */ 499/* context notify */
385 500
386static void 501static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 502eq_context_push (void *data1, void *data2, void *data3)
388{ 503{
389 dSP; 504 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 505 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 506 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 528 { \
414 pfn_notify = eq_context_notify; \ 529 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 530 user_data = s_get_cv (notify); \
416 } 531 }
417 532
418static SV * 533static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 534new_clobj_context (cl_context ctx, void *user_data)
420{ 535{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 536 SV *sv = new_clobj (stash_context, (IV)ctx);
422 537
423 if (user_data) 538 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 539 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 540
426 return sv; 541 return sv;
433 548
434static void 549static void
435eq_program_push (void *data1, void *data2, void *data3) 550eq_program_push (void *data1, void *data2, void *data3)
436{ 551{
437 dSP; 552 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 553 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 554 PUTBACK;
440} 555}
441 556
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 557static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 558
444static void CL_CALLBACK 559static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 560eq_program_notify (cl_program program, void *user_data)
446{ 561{
562 clRetainProgram (program);
563
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 564 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
565}
566
567typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
568
569static program_callback ecb_noinline
570make_program_callback (SV *notify, void **ruser_data)
571{
572 if (SvOK (notify))
573 {
574 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
575 return eq_program_notify;
576 }
577 else
578 {
579 *ruser_data = 0;
580 return 0;
581 }
448} 582}
449 583
450struct build_args 584struct build_args
451{ 585{
452 cl_program program; 586 cl_program program;
466 else 600 else
467 clReleaseProgram (arg->program); 601 clReleaseProgram (arg->program);
468 602
469 free (arg->options); 603 free (arg->options);
470 free (arg); 604 free (arg);
605
606 return 0;
471} 607}
472 608
473static void 609static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 610build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 611{
484 xthread_t id; 620 xthread_t id;
485 thread_create (&id, build_program_thread, arg); 621 thread_create (&id, build_program_thread, arg);
486} 622}
487 623
488/*****************************************************************************/ 624/*****************************************************************************/
625/* mem object destructor notify */
626
627static void ecb_noinline
628eq_destructor_push (void *data1, void *data2, void *data3)
629{
630}
631
632static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
633
634static void CL_CALLBACK
635eq_destructor_notify (cl_mem memobj, void *user_data)
636{
637 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
638}
639
640/*****************************************************************************/
489/* event objects */ 641/* event objects */
490 642
491static void 643static void
492eq_event_push (void *data1, void *data2, void *data3) 644eq_event_push (void *data1, void *data2, void *data3)
493{ 645{
494 dSP; 646 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 647 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 648 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 649 PUTBACK;
498} 650}
499 651
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 652static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 657 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 658 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 659}
508 660
509/*****************************************************************************/ 661/*****************************************************************************/
662/* utilities for XS code */
510 663
511static size_t 664static size_t
512img_row_pitch (cl_mem img) 665img_row_pitch (cl_mem img)
513{ 666{
514 size_t res; 667 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 668 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 669 return res;
517} 670}
518 671
519static cl_event * 672static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 673event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 674{
522 cl_uint count = *rcount; 675 cl_uint count = *rcount;
523 676
524 if (!count) 677 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
678 *rcount = count = 0;
679
680 if (!count && !extra)
525 return 0; 681 return 0;
526 682
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 683 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 684 int i = 0;
529 685
530 do 686 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 687 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 688 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 689
536 while (count); 690 if (extra)
691 list [i++] = extra;
537 692
538 *rcount = i; 693 *rcount = i;
539 694
540 return i ? list : 0; 695 return i ? list : 0;
541} 696}
542 697
543#define EVENT_LIST(items,count) \ 698#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 699 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 700 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 701
547#define INFO(class) \ 702#define INFO(class) \
548{ \ 703{ \
549 size_t size; \ 704 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 705 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
554 SvCUR_set (sv, size); \ 709 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 710 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 711 XPUSHs (sv); \
557} 712}
558 713
714/*****************************************************************************/
715/* mapped_xxx */
716
717static OpenCL__Mapped
718SvMAPPED (SV *sv)
719{
720 // no typechecking atm., keep your fingers crossed
721 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
722}
723
724struct mapped
725{
726 cl_command_queue queue;
727 cl_mem memobj;
728 void *ptr;
729 size_t cb;
730 cl_event event;
731 size_t row_pitch;
732 size_t slice_pitch;
733
734 size_t element_size;
735 size_t width, height, depth;
736};
737
738static SV *
739mapped_new (
740 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
741 void *ptr, size_t cb, cl_event ev,
742 size_t row_pitch, size_t slice_pitch, size_t element_size,
743 size_t width, size_t height, size_t depth
744)
745{
746 SV *data = newSV (0);
747 SvUPGRADE (data, SVt_PVMG);
748
749 OpenCL__Mapped mapped;
750 New (0, mapped, 1, struct mapped);
751
752 clRetainCommandQueue (queue);
753
754 mapped->queue = queue;
755 mapped->memobj = memobj;
756 mapped->ptr = ptr;
757 mapped->cb = cb;
758 mapped->event = ev;
759 mapped->row_pitch = row_pitch;
760 mapped->slice_pitch = slice_pitch;
761
762 mapped->element_size = element_size;
763 mapped->width = width;
764 mapped->height = height;
765 mapped->depth = depth;
766
767 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
768
769 if (SvLEN (data))
770 Safefree (data);
771
772 SvPVX (data) = (char *)ptr;
773 SvCUR_set (data, cb);
774 SvLEN_set (data, 0);
775 SvPOK_only (data);
776
777 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
778
779 if (!(flags & CL_MAP_WRITE))
780 SvREADONLY_on (data);
781
782 return obj;
783}
784
785static void
786mapped_detach (SV *sv, OpenCL__Mapped mapped)
787{
788 SV *data = SvRV (sv);
789
790 // the next check checks both before AND after detach, where SvPVX should STILL be 0
791 if (SvPVX (data) != (char *)mapped->ptr)
792 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
793 else
794 {
795 SvREADONLY_off (data);
796 SvCUR_set (data, 0);
797 SvPVX (data) = 0;
798 SvOK_off (data);
799 }
800
801 mapped->ptr = 0;
802}
803
804static void
805mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
806{
807 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
808 cl_event ev;
809
810 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
811
812 clReleaseEvent (mapped->event);
813 mapped->event = ev;
814
815 mapped_detach (self, mapped);
816}
817
818static size_t
819mapped_element_size (OpenCL__Mapped self)
820{
821 if (!self->element_size)
822 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
823
824 return self->element_size;
825}
826
827/*****************************************************************************/
828
559MODULE = OpenCL PACKAGE = OpenCL 829MODULE = OpenCL PACKAGE = OpenCL
560 830
561PROTOTYPES: ENABLE 831PROTOTYPES: ENABLE
562 832
563void 833void
572 eq_signal_arg = (void*)arg; 842 eq_signal_arg = (void*)arg;
573 843
574BOOT: 844BOOT:
575{ 845{
576 HV *stash = gv_stashpv ("OpenCL", 1); 846 HV *stash = gv_stashpv ("OpenCL", 1);
847
577 static const ivstr *civ, const_iv[] = { 848 static const ivstr *civ, const_iv[] = {
578 { sizeof (cl_char ), "SIZEOF_CHAR" }, 849 { sizeof (cl_char ), "SIZEOF_CHAR" },
579 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 850 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
580 { sizeof (cl_short ), "SIZEOF_SHORT" }, 851 { sizeof (cl_short ), "SIZEOF_SHORT" },
581 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 852 { sizeof (cl_ushort), "SIZEOF_USHORT" },
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 855 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 856 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 857 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 858 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 859 { sizeof (cl_double), "SIZEOF_DOUBLE" },
860 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 861#include "constiv.h"
590 }; 862 };
591 863
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 864 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)); 865 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
866
867 static const nvstr *cnv, const_nv[] = {
868#include "constnv.h"
869 };
870
871 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
872 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
873
874 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
875
876 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
877 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
878 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
879 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
880 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
881 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
882 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
883 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
884 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
885 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
886 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
887 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
888 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
889 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
890 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
891 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
892 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
893 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
894 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
895 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
896 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
897 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
898 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 899
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 900 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 901}
597 902
598cl_int 903cl_int
601 RETVAL = res; 906 RETVAL = res;
602 OUTPUT: 907 OUTPUT:
603 RETVAL 908 RETVAL
604 909
605const char * 910const char *
606err2str (cl_int err) 911err2str (cl_int err = res)
607 912
608const char * 913const char *
609enum2str (cl_uint value) 914enum2str (cl_uint value)
610 915
611void 916void
619 list = tmpbuf (sizeof (*list) * count); 924 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 925 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 926
622 EXTEND (SP, count); 927 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 929 PUSH_CLOBJ (stash_platform, list [i]);
625 930
626void 931void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 932context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 933 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 934 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 935 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 936 XPUSH_CLOBJ_CONTEXT;
632 937
633void 938void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 939context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 940 PPCODE:
636 /* der Gipfel der Kunst */ 941 cl_uint device_count;
942 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
943
944 CONTEXT_NOTIFY_CALLBACK;
945 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
946 XPUSH_CLOBJ_CONTEXT;
637 947
638void 948void
639wait_for_events (...) 949wait_for_events (...)
640 CODE: 950 CODE:
641 EVENT_LIST (0, items); 951 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 952 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 953
644PROTOTYPES: DISABLE 954PROTOTYPES: DISABLE
645 955
646MODULE = OpenCL PACKAGE = OpenCL::Platform 956MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 999 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 1000 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 1001
692 EXTEND (SP, count); 1002 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 1003 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1004 PUSH_CLOBJ (stash_device, list [i]);
695 1005
696void 1006void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 1007context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 1008 PPCODE:
1009 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1010 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1011
1012 cl_uint device_count;
1013 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1014
1015 CONTEXT_NOTIFY_CALLBACK;
1016 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1017 XPUSH_CLOBJ_CONTEXT;
1018
1019void
1020context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1021 PPCODE:
1022 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1023 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1024
1025 CONTEXT_NOTIFY_CALLBACK;
1026 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1027 XPUSH_CLOBJ_CONTEXT;
1028
1029MODULE = OpenCL PACKAGE = OpenCL::Device
1030
1031void
1032info (OpenCL::Device self, cl_device_info name)
1033 PPCODE:
1034 INFO (Device)
1035
1036#if CL_VERSION_1_2
1037
1038void
1039sub_devices (OpenCL::Device self, SV *properties)
1040 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1041 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1042 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 1043
702 AV *av = (AV *)SvRV (devices); 1044 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 1045
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1046 cl_uint count = av_len ((AV *)properties) + 1;
1047 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 1048
706 int i; 1049 int i;
707 for (i = num_devices; i--; ) 1050 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1051 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 1052
710 CONTEXT_NOTIFY_CALLBACK; 1053 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 1054
714void 1055 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) 1056 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 1057 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1058 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 1059
720 CONTEXT_NOTIFY_CALLBACK; 1060 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1061 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 1062 PUSH_CLOBJ (stash_subdevice, list [i]);
723 1063
724MODULE = OpenCL PACKAGE = OpenCL::Device 1064#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 1065
731#BEGIN:device 1066#BEGIN:device
732 1067
733void 1068void
734type (OpenCL::Device self) 1069type (OpenCL::Device self)
1070 ALIAS:
1071 type = CL_DEVICE_TYPE
1072 address_bits = CL_DEVICE_ADDRESS_BITS
1073 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1074 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1075 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1076 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1077 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1078 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1079 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1080 properties = CL_DEVICE_QUEUE_PROPERTIES
1081 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1082 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
735 PPCODE: 1083 PPCODE:
736 cl_device_type value [1]; 1084 cl_ulong value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1085 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
738 EXTEND (SP, 1); 1086 EXTEND (SP, 1);
739 const int i = 0; 1087 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1088 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1089
742void 1090void
743vendor_id (OpenCL::Device self) 1091vendor_id (OpenCL::Device self)
744 ALIAS: 1092 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1093 vendor_id = CL_DEVICE_VENDOR_ID
756 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1104 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
757 image_support = CL_DEVICE_IMAGE_SUPPORT 1105 image_support = CL_DEVICE_IMAGE_SUPPORT
758 max_samplers = CL_DEVICE_MAX_SAMPLERS 1106 max_samplers = CL_DEVICE_MAX_SAMPLERS
759 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1107 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
760 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1108 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1109 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
761 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1110 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
762 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1111 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1112 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
763 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1113 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
764 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1114 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
765 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1115 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
766 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1116 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
767 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1117 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
805 EXTEND (SP, n); 1155 EXTEND (SP, n);
806 for (i = 0; i < n; ++i) 1156 for (i = 0; i < n; ++i)
807 PUSHs (sv_2mortal (newSVuv (value [i]))); 1157 PUSHs (sv_2mortal (newSVuv (value [i])));
808 1158
809void 1159void
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) 1160error_correction_support (OpenCL::Device self)
866 ALIAS: 1161 ALIAS:
867 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1162 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
868 endian_little = CL_DEVICE_ENDIAN_LITTLE 1163 endian_little = CL_DEVICE_ENDIAN_LITTLE
869 available = CL_DEVICE_AVAILABLE 1164 available = CL_DEVICE_AVAILABLE
870 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1165 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
871 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1166 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
872 PPCODE: 1167 PPCODE:
873 cl_bool value [1]; 1168 cl_uint value [1];
874 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1169 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
875 EXTEND (SP, 1); 1170 EXTEND (SP, 1);
876 const int i = 0; 1171 const int i = 0;
877 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1172 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 1173
897void 1174void
898platform (OpenCL::Device self) 1175platform (OpenCL::Device self)
899 PPCODE: 1176 PPCODE:
900 cl_platform_id value [1]; 1177 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1178 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1179 EXTEND (SP, 1);
903 const int i = 0; 1180 const int i = 0;
904 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1181 PUSH_CLOBJ (stash_platform, value [i]);
906 }
907 1182
908void 1183void
909name (OpenCL::Device self) 1184name (OpenCL::Device self)
910 ALIAS: 1185 ALIAS:
911 name = CL_DEVICE_NAME 1186 name = CL_DEVICE_NAME
928 PPCODE: 1203 PPCODE:
929 cl_device_id value [1]; 1204 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1205 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1206 EXTEND (SP, 1);
932 const int i = 0; 1207 const int i = 0;
933 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1208 PUSH_CLOBJ (stash_device, value [i]);
935 }
936 1209
937void 1210void
938partition_types_ext (OpenCL::Device self) 1211partition_types_ext (OpenCL::Device self)
939 ALIAS: 1212 ALIAS:
940 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1213 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
950 for (i = 0; i < n; ++i) 1223 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1224 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1225
953#END:device 1226#END:device
954 1227
1228MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1229
1230#if CL_VERSION_1_2
1231
1232void
1233DESTROY (OpenCL::SubDevice self)
1234 CODE:
1235 clReleaseDevice (self);
1236
1237#endif
1238
955MODULE = OpenCL PACKAGE = OpenCL::Context 1239MODULE = OpenCL PACKAGE = OpenCL::Context
956 1240
957void 1241void
958DESTROY (OpenCL::Context context) 1242DESTROY (OpenCL::Context self)
959 CODE: 1243 CODE:
960 clReleaseContext (context); 1244 clReleaseContext (self);
961 1245
962void 1246void
963info (OpenCL::Context self, cl_context_info name) 1247info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1248 PPCODE:
965 INFO (Context) 1249 INFO (Context)
966 1250
967void 1251void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1252queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1253 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1254 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1255 XPUSH_CLOBJ (stash_queue, queue);
972 1256
973void 1257void
974user_event (OpenCL::Context self) 1258user_event (OpenCL::Context self)
975 PPCODE: 1259 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1260 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1261 XPUSH_CLOBJ (stash_userevent, ev);
978 1262
979void 1263void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1264buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1265 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1266 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?"); 1267 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1268
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1269 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1270 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1271
988void 1272void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1273buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1274 PPCODE:
991 STRLEN len; 1275 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1276 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1277 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?"); 1278 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)); 1279 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1280 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1281
998#if CL_VERSION_1_2 1282#if CL_VERSION_1_2
999 1283
1000void 1284void
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) 1285image (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: 1286 PPCODE:
1003 STRLEN len; 1287 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1288 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1289 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1290 const cl_image_desc desc = {
1007 type, 1291 type,
1008 width, height, depth, 1292 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1293 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1294 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1295 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1296 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1297 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1298 HV *stash = stash_image;
1015 switch (type) 1299 switch (type)
1016 { 1300 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1301 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1302 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1303 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1304 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1305 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1306 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1307 }
1024 XPUSH_CLOBJ (klass, mem); 1308 XPUSH_CLOBJ (stash, mem);
1025 1309
1026#endif 1310#endif
1027 1311
1028void 1312void
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) 1313image2d (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)); 1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1320#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1321 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)); 1322 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1323#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1324 XPUSH_CLOBJ (stash_image2d, mem);
1041 1325
1042void 1326void
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) 1327image3d (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: 1328 PPCODE:
1045 STRLEN len; 1329 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1333 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1334#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1335 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)); 1336 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1337#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1338 XPUSH_CLOBJ (stash_image3d, mem);
1055 1339
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1340#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1341
1058void 1342void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1343gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1344 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1345 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1346 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1347
1064void 1348void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1349gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1350 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1351 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1352 XPUSH_CLOBJ (stash_image2d, mem);
1069 1353
1070#if CL_VERSION_1_2 1354#if CL_VERSION_1_2
1071 1355
1072void 1356void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1357gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1358 ALIAS:
1075 PPCODE: 1359 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1360 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1361 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1362 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1363 HV *stash = stash_memory;
1080 switch (type) 1364 switch (type)
1081 { 1365 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1366 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1367 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1368 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1369 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1370 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1371 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1372 }
1089 XPUSH_CLOBJ (klass, mem); 1373 XPUSH_CLOBJ (stash, mem);
1090 1374
1091#endif 1375#endif
1092 1376
1093void 1377void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1378gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1380#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1381 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1382#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1384#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1385 XPUSH_CLOBJ (stash_image2d, mem);
1102 1386
1103void 1387void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1388gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1389 PPCODE:
1106#if PREFER_1_1 1390#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1391 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1392#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1394#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1395 XPUSH_CLOBJ (stash_image3d, mem);
1112 1396
1113#endif 1397#endif
1114 1398
1115void 1399void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1400supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1420
1137void 1421void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1422sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1423 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1424 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1425 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1426
1143void 1427void
1144program_with_source (OpenCL::Context self, SV *program) 1428program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1429 PPCODE:
1146 STRLEN len; 1430 STRLEN len;
1147 size_t len2; 1431 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1432 const char *ptr = SvPVbyte (program, len);
1149 1433
1150 len2 = len; 1434 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1435 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1436 XPUSH_CLOBJ (stash_program, prog);
1437
1438void
1439program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1440 PPCODE:
1441 cl_uint device_count;
1442 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1443
1444 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1445 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1446
1447 binaries = SvRV (binaries);
1448
1449 if (device_count != av_len ((AV *)binaries) + 1)
1450 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1451
1452 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1453 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1454 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1455
1456 int i;
1457 for (i = 0; i < device_count; ++i)
1458 {
1459 STRLEN len;
1460 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1461 length_list [i] = len;
1462 }
1463
1464 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1465 length_list, binary_list,
1466 GIMME_V == G_ARRAY ? status_list : 0, &res));
1467
1468 EXTEND (SP, 2);
1469 PUSH_CLOBJ (stash_program, prog);
1470
1471 if (GIMME_V == G_ARRAY)
1472 {
1473 AV *av = newAV ();
1474 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1475
1476 for (i = device_count; i--; )
1477 av_store (av, i, newSViv (status_list [i]));
1478 }
1479
1480#if CL_VERSION_1_2
1481
1482void
1483program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1484 PPCODE:
1485 cl_uint device_count;
1486 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1489
1490 XPUSH_CLOBJ (stash_program, prog);
1491
1492void
1493link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1494 CODE:
1495 cl_uint device_count = 0;
1496 cl_device_id *device_list = 0;
1497
1498 if (SvOK (devices))
1499 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1500
1501 cl_uint program_count;
1502 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1503
1504 void *user_data;
1505 program_callback pfn_notify = make_program_callback (notify, &user_data);
1506
1507 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1508 program_count, program_list, pfn_notify, user_data, &res));
1509
1510 XPUSH_CLOBJ (stash_program, prog);
1511
1512#endif
1153 1513
1154#BEGIN:context 1514#BEGIN:context
1155 1515
1156void 1516void
1157reference_count (OpenCL::Context self) 1517reference_count (OpenCL::Context self)
1173 cl_device_id *value = tmpbuf (size); 1533 cl_device_id *value = tmpbuf (size);
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1534 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1535 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1536 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1537 for (i = 0; i < n; ++i)
1178 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1538 PUSH_CLOBJ (stash_device, value [i]);
1180 }
1181 1539
1182void 1540void
1183properties (OpenCL::Context self) 1541properties (OpenCL::Context self)
1184 PPCODE: 1542 PPCODE:
1185 size_t size; 1543 size_t size;
1199DESTROY (OpenCL::Queue self) 1557DESTROY (OpenCL::Queue self)
1200 CODE: 1558 CODE:
1201 clReleaseCommandQueue (self); 1559 clReleaseCommandQueue (self);
1202 1560
1203void 1561void
1204enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1562read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1563 ALIAS:
1564 enqueue_read_buffer = 0
1205 PPCODE: 1565 PPCODE:
1206 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1566 EVENT_LIST (6);
1208 1567
1209 SvUPGRADE (data, SVt_PV); 1568 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1569 SvGROW (data, len);
1211 SvPOK_only (data); 1570 SvPOK_only (data);
1212 SvCUR_set (data, len); 1571 SvCUR_set (data, len);
1572
1573 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)); 1574 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1214 1575
1215 if (ev) 1576 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1577 XPUSH_CLOBJ (stash_event, ev);
1217 1578
1218void 1579void
1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1580write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1581 ALIAS:
1582 enqueue_write_buffer = 0
1220 PPCODE: 1583 PPCODE:
1221 cl_event ev = 0; 1584 EVENT_LIST (5);
1585
1222 STRLEN len; 1586 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1587 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5);
1225 1588
1589 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)); 1590 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1227 1591
1228 if (ev) 1592 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1593 XPUSH_CLOBJ (stash_event, ev);
1230 1594
1231#if CL_VERSION_1_2 1595#if CL_VERSION_1_2
1232 1596
1233void 1597void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1598fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1599 ALIAS:
1600 enqueue_fill_buffer = 0
1235 PPCODE: 1601 PPCODE:
1236 cl_event ev = 0; 1602 EVENT_LIST (5);
1603
1237 STRLEN len; 1604 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1605 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240 1606
1607 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)); 1608 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1609
1243 if (ev) 1610 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1611 XPUSH_CLOBJ (stash_event, ev);
1245 1612
1246void 1613void
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, ...) 1614fill_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, ...)
1615 ALIAS:
1616 enqueue_fill_image = 0
1248 PPCODE: 1617 PPCODE:
1249 cl_event ev = 0; 1618 EVENT_LIST (12);
1250 STRLEN len; 1619
1251 const size_t origin [3] = { x, y, z }; 1620 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1621 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254 1622
1255 const cl_float c_f [4] = { r, g, b, a }; 1623 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1624 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1625 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1626 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1629 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1630 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) 1631 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."); 1632 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1633
1634 cl_event ev = 0;
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1635 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)); 1636 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1637
1269 if (ev) 1638 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1639 XPUSH_CLOBJ (stash_event, ev);
1271 1640
1272#endif 1641#endif
1273 1642
1274void 1643void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1644copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1645 ALIAS:
1646 enqueue_copy_buffer = 0
1276 PPCODE: 1647 PPCODE:
1648 EVENT_LIST (6);
1649
1277 cl_event ev = 0; 1650 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)); 1651 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1652
1282 if (ev) 1653 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1654 XPUSH_CLOBJ (stash_event, ev);
1284 1655
1285void 1656void
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, ...) 1657read_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, ...)
1658 ALIAS:
1659 enqueue_read_buffer_rect = 0
1287 PPCODE: 1660 PPCODE:
1288 cl_event ev = 0; 1661 EVENT_LIST (17);
1662
1289 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1663 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 }; 1664 const size_t host_origin[3] = { host_x, host_y, host_z };
1291 const size_t region[3] = { width, height, depth }; 1665 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17);
1293 1666
1294 if (!buf_row_pitch) 1667 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1668 buf_row_pitch = region [0];
1296 1669
1297 if (!buf_slice_pitch) 1670 if (!buf_slice_pitch)
1307 1680
1308 SvUPGRADE (data, SVt_PV); 1681 SvUPGRADE (data, SVt_PV);
1309 SvGROW (data, len); 1682 SvGROW (data, len);
1310 SvPOK_only (data); 1683 SvPOK_only (data);
1311 SvCUR_set (data, len); 1684 SvCUR_set (data, len);
1685
1686 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)); 1687 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 1688
1314 if (ev) 1689 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1690 XPUSH_CLOBJ (stash_event, ev);
1316 1691
1317void 1692void
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, ...) 1693write_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, ...)
1694 ALIAS:
1695 enqueue_write_buffer_rect = 0
1319 PPCODE: 1696 PPCODE:
1320 cl_event ev = 0; 1697 EVENT_LIST (17);
1698
1321 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1699 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 }; 1700 const size_t host_origin[3] = { host_x, host_y, host_z };
1323 const size_t region[3] = { width, height, depth }; 1701 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1702 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1703 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17);
1327 1704
1328 if (!buf_row_pitch) 1705 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1706 buf_row_pitch = region [0];
1330 1707
1331 if (!buf_slice_pitch) 1708 if (!buf_slice_pitch)
1340 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1717 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1341 1718
1342 if (len < min_len) 1719 if (len < min_len)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1720 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1721
1722 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)); 1723 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 1724
1347 if (ev) 1725 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1726 XPUSH_CLOBJ (stash_event, ev);
1349 1727
1350void 1728void
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, ...) 1729copy_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, ...)
1730 ALIAS:
1731 enqueue_copy_buffer_rect = 0
1352 PPCODE: 1732 PPCODE:
1353 cl_event ev = 0; 1733 EVENT_LIST (16);
1734
1354 const size_t src_origin[3] = { src_x, src_y, src_z }; 1735 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 }; 1736 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1356 const size_t region[3] = { width, height, depth }; 1737 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16);
1358 1738
1739 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)); 1740 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 1741
1361 if (ev) 1742 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1743 XPUSH_CLOBJ (stash_event, ev);
1363 1744
1364void 1745void
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, ...) 1746read_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, ...)
1747 ALIAS:
1748 enqueue_read_image = 0
1366 PPCODE: 1749 PPCODE:
1367 cl_event ev = 0; 1750 EVENT_LIST (12);
1751
1368 const size_t src_origin[3] = { src_x, src_y, src_z }; 1752 const size_t src_origin[3] = { src_x, src_y, src_z };
1369 const size_t region[3] = { width, height, depth }; 1753 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12);
1371 1754
1372 if (!row_pitch) 1755 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1756 row_pitch = img_row_pitch (src);
1374 1757
1375 if (depth > 1 && !slice_pitch) 1758 if (depth > 1 && !slice_pitch)
1379 1762
1380 SvUPGRADE (data, SVt_PV); 1763 SvUPGRADE (data, SVt_PV);
1381 SvGROW (data, len); 1764 SvGROW (data, len);
1382 SvPOK_only (data); 1765 SvPOK_only (data);
1383 SvCUR_set (data, len); 1766 SvCUR_set (data, len);
1767
1768 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)); 1769 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 1770
1386 if (ev) 1771 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1772 XPUSH_CLOBJ (stash_event, ev);
1388 1773
1389void 1774void
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, ...) 1775write_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, ...)
1776 ALIAS:
1777 enqueue_write_image = 0
1391 PPCODE: 1778 PPCODE:
1392 cl_event ev = 0; 1779 EVENT_LIST (12);
1780
1393 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1781 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1394 const size_t region[3] = { width, height, depth }; 1782 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1783 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1784 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12);
1398 1785
1399 if (!row_pitch) 1786 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1787 row_pitch = img_row_pitch (dst);
1401 1788
1402 if (depth > 1 && !slice_pitch) 1789 if (depth > 1 && !slice_pitch)
1405 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1792 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1406 1793
1407 if (len < min_len) 1794 if (len < min_len)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1795 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1796
1797 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)); 1798 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 1799
1412 if (ev) 1800 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1801 XPUSH_CLOBJ (stash_event, ev);
1414 1802
1415void 1803void
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, ...) 1804copy_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, ...)
1805 ALIAS:
1806 enqueue_copy_image = 0
1417 PPCODE: 1807 PPCODE:
1418 cl_event ev = 0; 1808 EVENT_LIST (12);
1809
1419 const size_t src_origin[3] = { src_x, src_y, src_z }; 1810 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 }; 1811 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1421 const size_t region[3] = { width, height, depth }; 1812 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12);
1423 1813
1814 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)); 1815 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1816
1426 if (ev) 1817 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1818 XPUSH_CLOBJ (stash_event, ev);
1428 1819
1429void 1820void
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, ...) 1821copy_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, ...)
1822 ALIAS:
1823 enqueue_copy_image_to_buffer = 0
1431 PPCODE: 1824 PPCODE:
1825 EVENT_LIST (10);
1826
1827 const size_t src_origin[3] = { src_x, src_y, src_z };
1828 const size_t region [3] = { width, height, depth };
1829
1432 cl_event ev = 0; 1830 cl_event ev = 0;
1831 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1832
1833 if (ev)
1834 XPUSH_CLOBJ (stash_event, ev);
1835
1836void
1837copy_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, ...)
1838 ALIAS:
1839 enqueue_copy_buffer_to_image = 0
1840 PPCODE:
1841 EVENT_LIST (10);
1842
1433 const size_t src_origin[3] = { src_x, src_y, src_z }; 1843 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1844 const size_t region [3] = { width, height, depth };
1845
1846 cl_event ev = 0;
1847 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1848
1849 if (ev)
1850 XPUSH_CLOBJ (stash_event, ev);
1851
1852void
1853map_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, ...)
1854 ALIAS:
1855 enqueue_map_buffer = 0
1856 PPCODE:
1857 EVENT_LIST (6);
1858
1859 size_t cb = SvIV (cb_);
1860
1861 if (!SvOK (cb_))
1862 {
1863 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1864 cb -= offset;
1865 }
1866
1867 cl_event ev;
1868 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1869 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1870
1871void
1872map_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, ...)
1873 ALIAS:
1874 enqueue_map_image = 0
1875 PPCODE:
1876 size_t width = SvIV (width_);
1877 if (!SvOK (width_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1880 width -= x;
1881 }
1882
1883 size_t height = SvIV (width_);
1884 if (!SvOK (height_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1887 height -= y;
1888
1889 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1890 if (!height)
1891 height = 1;
1892 }
1893
1894 size_t depth = SvIV (width_);
1895 if (!SvOK (depth_))
1896 {
1897 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1898 depth -= z;
1899
1900 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1901 if (!depth)
1902 depth = 1;
1903 }
1904
1905 const size_t origin[3] = { x, y, z };
1434 const size_t region[3] = { width, height, depth }; 1906 const size_t region[3] = { width, height, depth };
1907 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1908 EVENT_LIST (10);
1436 1909
1910 cl_event ev;
1911 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));
1912
1913 size_t cb = slice_pitch ? slice_pitch * region [2]
1914 : row_pitch ? row_pitch * region [1]
1915 : region [0];
1916
1917 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1918
1919void
1920unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1921 PPCODE:
1922 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1923 if (GIMME_V != G_VOID)
1924 {
1925 clRetainEvent (mapped->event);
1926 XPUSH_CLOBJ (stash_event, mapped->event);
1927 }
1928
1929void
1930task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1931 ALIAS:
1932 enqueue_task = 0
1933 PPCODE:
1934 EVENT_LIST (2);
1935
1936 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)); 1937 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1438 1938
1439 if (ev) 1939 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1940 XPUSH_CLOBJ (stash_event, ev);
1441 1941
1442void 1942void
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, ...) 1943nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1944 ALIAS:
1945 enqueue_nd_range_kernel = 0
1468 PPCODE: 1946 PPCODE:
1469 cl_event ev = 0; 1947 EVENT_LIST (5);
1948
1470 size_t *gwo = 0, *gws, *lws = 0; 1949 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1950 int gws_len;
1472 size_t *lists; 1951 size_t *lists;
1473 int i; 1952 int i;
1474 EVENT_LIST (5, items - 5);
1475 1953
1476 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1954 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1956
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1957 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1958
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1959 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1960
1483 gws = lists + gws_len * 0; 1961 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1962 for (i = 0; i < gws_len; ++i)
1963 {
1485 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1964 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1965 // at least nvidia crashes for 0-sized work group sizes, work around
1966 if (!gws [i])
1967 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1968 }
1486 1969
1487 if (SvOK (global_work_offset)) 1970 if (SvOK (global_work_offset))
1488 { 1971 {
1489 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1972 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"); 1973 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1498 } 1981 }
1499 1982
1500 if (SvOK (local_work_size)) 1983 if (SvOK (local_work_size))
1501 { 1984 {
1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1985 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"); 1986 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1504 1987
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1988 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"); 1989 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1507 1990
1508 lws = lists + gws_len * 2; 1991 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 1992 for (i = 0; i < gws_len; ++i)
1993 {
1510 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1994 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1995 // at least nvidia crashes for 0-sized work group sizes, work around
1996 if (!lws [i])
1997 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1998 }
1511 } 1999 }
1512 2000
2001 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)); 2002 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1514 2003
1515 if (ev) 2004 if (ev)
1516 XPUSH_CLOBJ ("OpenCL::Event", ev); 2005 XPUSH_CLOBJ (stash_event, ev);
2006
2007#if CL_VERSION_1_2
2008
2009void
2010migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2011 ALIAS:
2012 enqueue_migrate_mem_objects = 0
2013 PPCODE:
2014 EVENT_LIST (3);
2015
2016 cl_uint obj_count;
2017 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2018
2019 cl_event ev = 0;
2020 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2021
2022 if (ev)
2023 XPUSH_CLOBJ (stash_event, ev);
2024
2025#endif
1517 2026
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 2027#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 2028
1520void 2029void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2030acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 2031 ALIAS:
2032 release_gl_objects = 1
2033 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 2034 enqueue_release_gl_objects = 1
1524 PPCODE: 2035 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2036 EVENT_LIST (2);
1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2037
2038 cl_uint obj_count;
2039 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1527 2040
1528 cl_event ev = 0; 2041 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 2042
1538 if (ix) 2043 if (ix)
1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2044 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1540 else 2045 else
1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1542 2047
1543 if (ev) 2048 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 2049 XPUSH_CLOBJ (stash_event, ev);
1545 2050
1546#endif 2051#endif
1547 2052
1548void 2053void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 2054wait_for_events (OpenCL::Queue self, ...)
2055 ALIAS:
2056 enqueue_wait_for_events = 0
1550 CODE: 2057 CODE:
1551 EVENT_LIST (1, items - 1); 2058 EVENT_LIST (1);
1552#if PREFER_1_1 2059#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2060 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 2061#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2062 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 2063#endif
1557 2064
1558void 2065void
1559enqueue_marker (OpenCL::Queue self, ...) 2066marker (OpenCL::Queue self, ...)
2067 ALIAS:
2068 enqueue_marker = 0
1560 PPCODE: 2069 PPCODE:
2070 EVENT_LIST (1);
1561 cl_event ev = 0; 2071 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1 2072#if PREFER_1_1
1564 if (!event_list_count) 2073 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2074 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 2075 else
1567#if CL_VERSION_1_2 2076#if CL_VERSION_1_2
1574#endif 2083#endif
1575#else 2084#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2085 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif 2086#endif
1578 if (ev) 2087 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 2088 XPUSH_CLOBJ (stash_event, ev);
1580 2089
1581void 2090void
1582enqueue_barrier (OpenCL::Queue self, ...) 2091barrier (OpenCL::Queue self, ...)
2092 ALIAS:
2093 enqueue_barrier = 0
1583 PPCODE: 2094 PPCODE:
2095 EVENT_LIST (1);
1584 cl_event ev = 0; 2096 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1 2097#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 2098 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 2099 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 2100 else
1590#if CL_VERSION_1_2 2101#if CL_VERSION_1_2
1600#endif 2111#endif
1601#else 2112#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2113 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif 2114#endif
1604 if (ev) 2115 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 2116 XPUSH_CLOBJ (stash_event, ev);
1606 2117
1607void 2118void
1608flush (OpenCL::Queue self) 2119flush (OpenCL::Queue self)
1609 CODE: 2120 CODE:
1610 NEED_SUCCESS (Flush, (self)); 2121 NEED_SUCCESS (Flush, (self));
1626 PPCODE: 2137 PPCODE:
1627 cl_context value [1]; 2138 cl_context value [1];
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2139 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 2140 EXTEND (SP, 1);
1630 const int i = 0; 2141 const int i = 0;
1631 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 2142 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2143 PUSH_CLOBJ (stash_context, value [i]);
1634 }
1635 2144
1636void 2145void
1637device (OpenCL::Queue self) 2146device (OpenCL::Queue self)
1638 PPCODE: 2147 PPCODE:
1639 cl_device_id value [1]; 2148 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2149 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 2150 EXTEND (SP, 1);
1642 const int i = 0; 2151 const int i = 0;
1643 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2152 PUSH_CLOBJ (stash_device, value [i]);
1645 }
1646 2153
1647void 2154void
1648reference_count (OpenCL::Queue self) 2155reference_count (OpenCL::Queue self)
1649 PPCODE: 2156 PPCODE:
1650 cl_uint value [1]; 2157 cl_uint value [1];
1654 PUSHs (sv_2mortal (newSVuv (value [i]))); 2161 PUSHs (sv_2mortal (newSVuv (value [i])));
1655 2162
1656void 2163void
1657properties (OpenCL::Queue self) 2164properties (OpenCL::Queue self)
1658 PPCODE: 2165 PPCODE:
1659 cl_command_queue_properties value [1]; 2166 cl_ulong value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2167 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 2168 EXTEND (SP, 1);
1662 const int i = 0; 2169 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 2170 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 2171
1665#END:command_queue 2172#END:command_queue
1666 2173
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 2174MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 2175
1674void 2181void
1675info (OpenCL::Memory self, cl_mem_info name) 2182info (OpenCL::Memory self, cl_mem_info name)
1676 PPCODE: 2183 PPCODE:
1677 INFO (MemObject) 2184 INFO (MemObject)
1678 2185
2186void
2187destructor_callback (OpenCL::Memory self, SV *notify)
2188 PPCODE:
2189 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2190
1679#BEGIN:mem 2191#BEGIN:mem
1680 2192
1681void 2193void
1682type (OpenCL::Memory self) 2194type (OpenCL::Memory self)
2195 ALIAS:
2196 type = CL_MEM_TYPE
2197 map_count = CL_MEM_MAP_COUNT
2198 reference_count = CL_MEM_REFERENCE_COUNT
1683 PPCODE: 2199 PPCODE:
1684 cl_mem_object_type value [1]; 2200 cl_uint value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2201 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 2202 EXTEND (SP, 1);
1687 const int i = 0; 2203 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 2204 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 2205
1690void 2206void
1691flags (OpenCL::Memory self) 2207flags (OpenCL::Memory self)
1692 PPCODE: 2208 PPCODE:
1693 cl_mem_flags value [1]; 2209 cl_ulong value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2210 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2211 EXTEND (SP, 1);
1696 const int i = 0; 2212 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 2213 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 2214
1699void 2215void
1700size (OpenCL::Memory self) 2216size (OpenCL::Memory self)
1701 ALIAS: 2217 ALIAS:
1702 size = CL_MEM_SIZE 2218 size = CL_MEM_SIZE
1716 EXTEND (SP, 1); 2232 EXTEND (SP, 1);
1717 const int i = 0; 2233 const int i = 0;
1718 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2234 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1719 2235
1720void 2236void
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) 2237context (OpenCL::Memory self)
1734 PPCODE: 2238 PPCODE:
1735 cl_context value [1]; 2239 cl_context value [1];
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 2241 EXTEND (SP, 1);
1738 const int i = 0; 2242 const int i = 0;
1739 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 2243 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2244 PUSH_CLOBJ (stash_context, value [i]);
1742 }
1743 2245
1744void 2246void
1745associated_memobject (OpenCL::Memory self) 2247associated_memobject (OpenCL::Memory self)
1746 PPCODE: 2248 PPCODE:
1747 cl_mem value [1]; 2249 cl_mem value [1];
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1750 const int i = 0; 2252 const int i = 0;
1751 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 2253 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2254 PUSH_CLOBJ (stash_memory, value [i]);
1754 }
1755 2255
1756#END:mem 2256#END:mem
1757 2257
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 2258#if cl_apple_gl_sharing || cl_khr_gl_sharing
1759 2259
1778 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2278 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1779 2279
1780 cl_buffer_region crdata = { origin, size }; 2280 cl_buffer_region crdata = { origin, size };
1781 2281
1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2282 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2283 XPUSH_CLOBJ (stash_buffer, mem);
1784 2284
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2285MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2286
1787void 2287void
1788image_info (OpenCL::Image self, cl_image_info name) 2288image_info (OpenCL::Image self, cl_image_info name)
1858 2358
1859#BEGIN:sampler 2359#BEGIN:sampler
1860 2360
1861void 2361void
1862reference_count (OpenCL::Sampler self) 2362reference_count (OpenCL::Sampler self)
2363 ALIAS:
2364 reference_count = CL_SAMPLER_REFERENCE_COUNT
2365 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2366 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1863 PPCODE: 2367 PPCODE:
1864 cl_uint value [1]; 2368 cl_uint value [1];
1865 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2369 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1866 EXTEND (SP, 1); 2370 EXTEND (SP, 1);
1867 const int i = 0; 2371 const int i = 0;
1868 PUSHs (sv_2mortal (newSVuv (value [i]))); 2372 PUSHs (sv_2mortal (newSVuv (value [i])));
1869 2373
1870void 2374void
1872 PPCODE: 2376 PPCODE:
1873 cl_context value [1]; 2377 cl_context value [1];
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2378 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2379 EXTEND (SP, 1);
1876 const int i = 0; 2380 const int i = 0;
1877 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2381 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2382 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 2383
1900void 2384void
1901filter_mode (OpenCL::Sampler self) 2385filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2386 PPCODE:
1903 cl_bool value [1]; 2387 cl_uint value [1];
1904 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1905 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1906 const int i = 0; 2390 const int i = 0;
1907 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2391 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1908 2392
1918void 2402void
1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2403build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS: 2404 ALIAS:
1921 build_async = 1 2405 build_async = 1
1922 CODE: 2406 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; 2407 cl_uint device_count = 0;
1926 cl_device_id *device_list = 0; 2408 cl_device_id *device_list = 0;
1927 2409
1928 if (SvOK (devices)) 2410 if (SvOK (devices))
1929 { 2411 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2412
2413 void *user_data;
2414 program_callback pfn_notify = make_program_callback (notify, &user_data);
2415
2416 if (ix)
2417 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2418 else
2419 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2420
2421#if CL_VERSION_1_2
2422
2423void
2424compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2425 CODE:
2426 cl_uint device_count = 0;
2427 cl_device_id *device_list = 0;
2428
2429 if (SvOK (devices))
2430 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2431
2432 cl_uint header_count = 0;
2433 cl_program *header_list = 0;
2434 const char **header_name = 0;
2435
2436 if (SvOK (headers))
2437 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2439 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1932 2440
1933 AV *av = (AV *)SvRV (devices); 2441 HV *hv = (HV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935 2442
1936 if (num_devices) 2443 header_count = hv_iterinit (hv);
2444 header_list = tmpbuf (sizeof (*header_list) * header_count);
2445 header_name = tmpbuf (sizeof (*header_name) * header_count);
2446
2447 HE *he;
2448 int i = 0;
2449 while (he = hv_iternext (hv))
1937 { 2450 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2451 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2452 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1939 int count; 2453 ++i;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 } 2454 }
1943 } 2455 }
1944 2456
1945 if (SvOK (notify)) 2457 void *user_data;
1946 { 2458 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 2459
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)); 2460 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2461 header_count, header_list, header_name, pfn_notify, user_data));
2462
2463#endif
1956 2464
1957void 2465void
1958build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2466build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1959 PPCODE: 2467 PPCODE:
1960 size_t size; 2468 size_t size;
1969#BEGIN:program_build 2477#BEGIN:program_build
1970 2478
1971void 2479void
1972build_status (OpenCL::Program self, OpenCL::Device device) 2480build_status (OpenCL::Program self, OpenCL::Device device)
1973 PPCODE: 2481 PPCODE:
1974 cl_build_status value [1]; 2482 cl_int value [1];
1975 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2483 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1976 EXTEND (SP, 1); 2484 EXTEND (SP, 1);
1977 const int i = 0; 2485 const int i = 0;
1978 PUSHs (sv_2mortal (newSViv (value [i]))); 2486 PUSHs (sv_2mortal (newSViv (value [i])));
1979 2487
1989 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2497 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1990 EXTEND (SP, 1); 2498 EXTEND (SP, 1);
1991 const int i = 0; 2499 const int i = 0;
1992 PUSHs (sv_2mortal (newSVpv (value, 0))); 2500 PUSHs (sv_2mortal (newSVpv (value, 0)));
1993 2501
2502void
2503binary_type (OpenCL::Program self, OpenCL::Device device)
2504 PPCODE:
2505 cl_uint value [1];
2506 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2507 EXTEND (SP, 1);
2508 const int i = 0;
2509 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2510
1994#END:program_build 2511#END:program_build
1995 2512
1996void 2513void
1997kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2515 PPCODE:
1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2518
2002void 2519void
2003kernels_in_program (OpenCL::Program program) 2520kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2521 PPCODE:
2005 cl_uint num_kernels; 2522 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2526
2010 int i; 2527 int i;
2011 EXTEND (SP, num_kernels); 2528 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2529 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2531
2015void 2532void
2016info (OpenCL::Program self, cl_program_info name) 2533info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2534 PPCODE:
2018 INFO (Program) 2535 INFO (Program)
2064 PPCODE: 2581 PPCODE:
2065 cl_context value [1]; 2582 cl_context value [1];
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
2068 const int i = 0; 2585 const int i = 0;
2069 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2586 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2587 PUSH_CLOBJ (stash_context, value [i]);
2072 }
2073 2588
2074void 2589void
2075devices (OpenCL::Program self) 2590devices (OpenCL::Program self)
2076 PPCODE: 2591 PPCODE:
2077 size_t size; 2592 size_t size;
2079 cl_device_id *value = tmpbuf (size); 2594 cl_device_id *value = tmpbuf (size);
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2595 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2596 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2597 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2598 for (i = 0; i < n; ++i)
2084 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2599 PUSH_CLOBJ (stash_device, value [i]);
2086 }
2087 2600
2088void 2601void
2089source (OpenCL::Program self) 2602source (OpenCL::Program self)
2090 PPCODE: 2603 PPCODE:
2091 size_t size; 2604 size_t size;
2116DESTROY (OpenCL::Kernel self) 2629DESTROY (OpenCL::Kernel self)
2117 CODE: 2630 CODE:
2118 clReleaseKernel (self); 2631 clReleaseKernel (self);
2119 2632
2120void 2633void
2634setf (OpenCL::Kernel self, const char *format, ...)
2635 CODE:
2636 int i;
2637 for (i = 2; ; ++i)
2638 {
2639 while (*format == ' ')
2640 ++format;
2641
2642 char type = *format++;
2643
2644 if (!type)
2645 break;
2646
2647 if (i >= items)
2648 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2649
2650 SV *sv = ST (i);
2651
2652 union
2653 {
2654 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2655 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2656 cl_half ch; cl_float cf; cl_double cd;
2657 cl_mem cm;
2658 cl_sampler ca;
2659 size_t cz;
2660 cl_event ce;
2661 } arg;
2662 size_t size;
2663 int nullarg = 0;
2664
2665 switch (type)
2666 {
2667 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2668 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2669 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2670 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2671 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2672 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2673 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2674 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2675
2676 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2677 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2678 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2679
2680 case 'z': nullarg = 1; size = SvIV (sv); break;
2681
2682 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2683 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2684 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2685
2686 default:
2687 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2688 }
2689
2690 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2691 if (res)
2692 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2693 }
2694
2695 if (i != items)
2696 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2697
2698void
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2699set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2700 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2701 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2702
2125void 2703void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2751 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2752
2175void 2753void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2754set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2755 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2756 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2757
2180void 2758void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2759set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2760 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2761 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2762
2185void 2763void
2186set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2764set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2187 ALIAS:
2188 set_image2d = 0
2189 set_image3d = 0
2190 CODE: 2765 CODE:
2191 clSetKernelArg (self, idx, sizeof (value), &value); 2766 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2192 2767
2193void 2768void
2194set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2769set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2195 CODE: 2770 CODE:
2196 clSetKernelArg (self, idx, sizeof (value), &value); 2771 clSetKernelArg (self, idx, sizeof (value), &value);
2240 PPCODE: 2815 PPCODE:
2241 cl_context value [1]; 2816 cl_context value [1];
2242 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2817 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2243 EXTEND (SP, 1); 2818 EXTEND (SP, 1);
2244 const int i = 0; 2819 const int i = 0;
2245 {
2246 NEED_SUCCESS (RetainContext, (value [i])); 2820 NEED_SUCCESS (RetainContext, (value [i]));
2247 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2821 PUSH_CLOBJ (stash_context, value [i]);
2248 }
2249 2822
2250void 2823void
2251program (OpenCL::Kernel self) 2824program (OpenCL::Kernel self)
2252 PPCODE: 2825 PPCODE:
2253 cl_program value [1]; 2826 cl_program value [1];
2254 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2827 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2255 EXTEND (SP, 1); 2828 EXTEND (SP, 1);
2256 const int i = 0; 2829 const int i = 0;
2257 {
2258 NEED_SUCCESS (RetainProgram, (value [i])); 2830 NEED_SUCCESS (RetainProgram, (value [i]));
2259 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2831 PUSH_CLOBJ (stash_program, value [i]);
2260 }
2261 2832
2262#END:kernel 2833#END:kernel
2263 2834
2264void 2835void
2265work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2836work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2311 const int i = 0; 2882 const int i = 0;
2312 PUSHs (sv_2mortal (newSVuv (value [i]))); 2883 PUSHs (sv_2mortal (newSVuv (value [i])));
2313 2884
2314#END:kernel_work_group 2885#END:kernel_work_group
2315 2886
2887#if CL_VERSION_1_2
2888
2889void
2890arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2891 PPCODE:
2892 size_t size;
2893 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2894 SV *sv = sv_2mortal (newSV (size));
2895 SvUPGRADE (sv, SVt_PV);
2896 SvPOK_only (sv);
2897 SvCUR_set (sv, size);
2898 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2899 XPUSHs (sv);
2900
2901#BEGIN:kernel_arg
2902
2903void
2904arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2905 ALIAS:
2906 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2907 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2908 PPCODE:
2909 cl_uint value [1];
2910 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2911 EXTEND (SP, 1);
2912 const int i = 0;
2913 PUSHs (sv_2mortal (newSVuv (value [i])));
2914
2915void
2916arg_type_name (OpenCL::Kernel self, cl_uint idx)
2917 ALIAS:
2918 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2919 arg_name = CL_KERNEL_ARG_NAME
2920 PPCODE:
2921 size_t size;
2922 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2923 char *value = tmpbuf (size);
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2925 EXTEND (SP, 1);
2926 const int i = 0;
2927 PUSHs (sv_2mortal (newSVpv (value, 0)));
2928
2929void
2930arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2931 PPCODE:
2932 cl_ulong value [1];
2933 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2934 EXTEND (SP, 1);
2935 const int i = 0;
2936 PUSHs (sv_2mortal (newSVuv (value [i])));
2937
2938#END:kernel_arg
2939
2940#endif
2941
2316MODULE = OpenCL PACKAGE = OpenCL::Event 2942MODULE = OpenCL PACKAGE = OpenCL::Event
2317 2943
2318void 2944void
2319DESTROY (OpenCL::Event self) 2945DESTROY (OpenCL::Event self)
2320 CODE: 2946 CODE:
2342 PPCODE: 2968 PPCODE:
2343 cl_command_queue value [1]; 2969 cl_command_queue value [1];
2344 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2970 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2345 EXTEND (SP, 1); 2971 EXTEND (SP, 1);
2346 const int i = 0; 2972 const int i = 0;
2347 {
2348 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2973 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2349 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2974 PUSH_CLOBJ (stash_queue, value [i]);
2350 }
2351 2975
2352void 2976void
2353command_type (OpenCL::Event self) 2977command_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: 2978 ALIAS:
2979 command_type = CL_EVENT_COMMAND_TYPE
2364 reference_count = CL_EVENT_REFERENCE_COUNT 2980 reference_count = CL_EVENT_REFERENCE_COUNT
2365 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2981 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2366 PPCODE: 2982 PPCODE:
2367 cl_uint value [1]; 2983 cl_uint value [1];
2368 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2984 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2375 PPCODE: 2991 PPCODE:
2376 cl_context value [1]; 2992 cl_context value [1];
2377 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2993 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2378 EXTEND (SP, 1); 2994 EXTEND (SP, 1);
2379 const int i = 0; 2995 const int i = 0;
2380 {
2381 NEED_SUCCESS (RetainContext, (value [i])); 2996 NEED_SUCCESS (RetainContext, (value [i]));
2382 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2997 PUSH_CLOBJ (stash_context, value [i]);
2383 }
2384 2998
2385#END:event 2999#END:event
2386 3000
2387void 3001void
2388profiling_info (OpenCL::Event self, cl_profiling_info name) 3002profiling_info (OpenCL::Event self, cl_profiling_info name)
2412void 3026void
2413set_status (OpenCL::UserEvent self, cl_int execution_status) 3027set_status (OpenCL::UserEvent self, cl_int execution_status)
2414 CODE: 3028 CODE:
2415 clSetUserEventStatus (self, execution_status); 3029 clSetUserEventStatus (self, execution_status);
2416 3030
3031MODULE = OpenCL PACKAGE = OpenCL::Mapped
3032
3033void
3034DESTROY (SV *self)
3035 CODE:
3036 OpenCL__Mapped mapped = SvMAPPED (self);
3037
3038 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3039 mapped_detach (self, mapped);
3040
3041 clReleaseCommandQueue (mapped->queue);
3042 clReleaseEvent (mapped->event);
3043 Safefree (mapped);
3044
3045void
3046unmap (OpenCL::Mapped self, ...)
3047 CODE:
3048 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3049
3050bool
3051mapped (OpenCL::Mapped self)
3052 CODE:
3053 RETVAL = !!self->ptr;
3054 OUTPUT:
3055 RETVAL
3056
3057void
3058wait (OpenCL::Mapped self)
3059 PPCODE:
3060 if (self->event)
3061 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3062
3063void
3064event (OpenCL::Mapped self)
3065 PPCODE:
3066 if (!self->event)
3067 XSRETURN_UNDEF;
3068
3069 clRetainEvent (self->event);
3070 XPUSH_CLOBJ (stash_event, self->event);
3071
3072#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3073#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3074#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3075#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3076#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3077#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3078
3079IV
3080size (OpenCL::Mapped self)
3081 ALIAS:
3082 size = MAPPED_OFFSET_CB
3083 row_pitch = MAPPED_OFFSET_ROW_PITCH
3084 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3085 width = MAPPED_OFFSET_WIDTH
3086 height = MAPPED_OFFSET_HEIGHT
3087 depth = MAPPED_OFFSET_DEPTH
3088 CODE:
3089 RETVAL = *(size_t *)((char *)self + ix);
3090 OUTPUT:
3091 RETVAL
3092
3093IV
3094ptr (OpenCL::Mapped self)
3095 CODE:
3096 RETVAL = PTR2IV (self->ptr);
3097 OUTPUT:
3098 RETVAL
3099
3100void
3101set (OpenCL::Mapped self, size_t offset, SV *data)
3102 CODE:
3103 STRLEN len;
3104 const char *ptr = SvPVbyte (data, len);
3105
3106 if (offset + len > self->cb)
3107 croak ("OpenCL::Mapped::set out of bound condition detected");
3108
3109 memcpy (offset + (char *)self->ptr, ptr, len);
3110
3111void
3112get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3113 PPCODE:
3114 if (!SvOK (ST (1)))
3115 count = self->width - x;
3116
3117 if (x + count > self->width)
3118 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3119
3120 if (y >= self->height)
3121 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3122
3123 if (z >= self->depth)
3124 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3125
3126 size_t element = mapped_element_size (self);
3127
3128 count *= element;
3129 x *= element;
3130
3131 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3132 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3133
3134void
3135set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3136 PPCODE:
3137 STRLEN count;
3138 char *dataptr = SvPVbyte (data, count);
3139 size_t element = mapped_element_size (self);
3140
3141 x *= element;
3142
3143 if (x + count > self->width * element)
3144 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3145
3146 if (y >= self->height)
3147 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3148
3149 if (z >= self->depth)
3150 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3151
3152 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3153 memcpy (ptr, dataptr, count);
3154
3155MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3158
3159IV
3160element_size (OpenCL::Mapped self)
3161 CODE:
3162 RETVAL = mapped_element_size (self);
3163 OUTPUT:
3164 RETVAL
3165

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines