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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.58 by root, Sun Apr 29 18:39:31 2012 UTC vs.
Revision 1.76 by root, Sat May 5 14:49:52 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#include "ecb.h"//D
4 6
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 8#include "xthread.h"
7#include "schmorp.h" 9#include "schmorp.h"
8 10
26#if PREFER_1_1 28#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 30#endif
29 31
30#ifdef __APPLE__ 32#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
32#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
33 #include <CL/opencl.h> 38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 50#endif
35 51
36#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 53 #undef PREFER_1_1
38 #define PREFER_1_1 1 54 #define PREFER_1_1 1
39#endif 55#endif
40 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
41typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
61 77
62typedef SV *FUTURE; 78typedef struct mapped * OpenCL__Mapped;
79
80static HV
81 *stash_platform,
82 *stash_device,
83 *stash_subdevice,
84 *stash_context,
85 *stash_queue,
86 *stash_program,
87 *stash_kernel,
88 *stash_sampler,
89 *stash_event,
90 *stash_userevent,
91 *stash_memory,
92 *stash_buffer,
93 *stash_bufferobj,
94 *stash_image,
95 *stash_image1d,
96 *stash_image1darray,
97 *stash_image1dbuffer,
98 *stash_image2d,
99 *stash_image2darray,
100 *stash_image3d,
101 *stash_mapped,
102 *stash_mappedbuffer,
103 *stash_mappedimage;
63 104
64/*****************************************************************************/ 105/*****************************************************************************/
65 106
66// name must include a leading underscore 107// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 148}
108 149
109/*****************************************************************************/ 150/*****************************************************************************/
110 151
111/* up to two temporary buffers */ 152/* up to two temporary buffers */
112static void * 153static void * ecb_noinline
113tmpbuf (size_t size) 154tmpbuf (size_t size)
114{ 155{
115 enum { buffers = 3 }; 156 enum { buffers = 4 };
116 static int idx; 157 static int idx;
117 static void *buf [buffers]; 158 static void *buf [buffers];
118 static size_t len [buffers]; 159 static size_t len [buffers];
119 160
120 idx = (idx + 1) % buffers; 161 idx = (idx + 1) % buffers;
127 } 168 }
128 169
129 return buf [idx]; 170 return buf [idx];
130} 171}
131 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
132/*****************************************************************************/ 188/*****************************************************************************/
133 189
134typedef struct 190typedef struct
135{ 191{
136 IV iv; 192 IV iv;
137 const char *name; 193 const char *name;
138 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
139} ivstr; 195} ivstr;
196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
140 203
141static const char * 204static const char *
142iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
143{ 206{
144 int i; 207 int i;
193 if (res) \ 256 if (res) \
194 FAIL (name); 257 FAIL (name);
195 258
196/*****************************************************************************/ 259/*****************************************************************************/
197 260
261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
274{
275 // sv_derived_from is quite slow :(
276 if (SvROK (sv) && sv_derived_from (sv, pkg))
277 return (void *)SvIV (SvRV (sv));
278
279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
290}
291
292/*****************************************************************************/
293
198static cl_context_properties * 294static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 296{
201 if (!sv || !SvOK (sv)) 297 if (!sv || !SvOK (sv))
202 if (extra) 298 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 300 else
210 int i, len = av_len (av) + 1; 306 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 308 cl_context_properties *l = p;
213 309
214 if (len & 1) 310 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 312
217 while (extracount--) 313 while (extracount--)
218 *l++ = *extra++; 314 *l++ = *extra++;
219 315
220 for (i = 0; i < len; i += 2) 316 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 319 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 320 cl_context_properties v = SvIV (p_sv); // code below can override
225 321
226 switch (t) 322 switch (t)
227 { 323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
228 case CL_GLX_DISPLAY_KHR: 329 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 330 if (!SvOK (p_sv))
230 { 331 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 332 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 333 if (func)
255 *l = 0; 356 *l = 0;
256 357
257 return p; 358 return p;
258 } 359 }
259 360
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 362}
262 363
263/*****************************************************************************/ 364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 365// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 366static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 367object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{ 368{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
275 return (void *)SvIV (SvRV (sv)); 370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
276 372
277 croak ("%s: %s is not of type %s", func, svname, pkg); 373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
378 {
379 list = tmpbuf (sizeof (*list) * count);
380 int i;
381 for (i = 0; i < count; ++i)
382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
383 }
384
385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
278} 391}
279 392
280/*****************************************************************************/ 393/*****************************************************************************/
281/* callback stuff */ 394/* callback stuff */
282 395
304static void (*eq_signal_func)(void *signal_arg, int value); 417static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 418static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 419static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 420static eq_item *eq_head, *eq_tail;
308 421
309static void 422static void ecb_noinline
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{ 424{
312 eq_item *item = malloc (sizeof (eq_item)); 425 eq_item *item = malloc (sizeof (eq_item));
313 426
314 item->next = 0; 427 item->next = 0;
381} 494}
382 495
383/*****************************************************************************/ 496/*****************************************************************************/
384/* context notify */ 497/* context notify */
385 498
386static void 499static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 500eq_context_push (void *data1, void *data2, void *data3)
388{ 501{
389 dSP; 502 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 526 { \
414 pfn_notify = eq_context_notify; \ 527 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 528 user_data = s_get_cv (notify); \
416 } 529 }
417 530
418static SV * 531static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 532new_clobj_context (cl_context ctx, void *user_data)
420{ 533{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 534 SV *sv = new_clobj (stash_context, (IV)ctx);
422 535
423 if (user_data) 536 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 537 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 538
426 return sv; 539 return sv;
433 546
434static void 547static void
435eq_program_push (void *data1, void *data2, void *data3) 548eq_program_push (void *data1, void *data2, void *data3)
436{ 549{
437 dSP; 550 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 551 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 552 PUTBACK;
440} 553}
441 554
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 556
444static void CL_CALLBACK 557static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 558eq_program_notify (cl_program program, void *user_data)
446{ 559{
560 clRetainProgram (program);
561
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
448} 580}
449 581
450struct build_args 582struct build_args
451{ 583{
452 cl_program program; 584 cl_program program;
466 else 598 else
467 clReleaseProgram (arg->program); 599 clReleaseProgram (arg->program);
468 600
469 free (arg->options); 601 free (arg->options);
470 free (arg); 602 free (arg);
603
604 return 0;
471} 605}
472 606
473static void 607static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{ 609{
490 624
491static void 625static void
492eq_event_push (void *data1, void *data2, void *data3) 626eq_event_push (void *data1, void *data2, void *data3)
493{ 627{
494 dSP; 628 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 629 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 630 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 631 PUTBACK;
498} 632}
499 633
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 634static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 639 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 640 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 641}
508 642
509/*****************************************************************************/ 643/*****************************************************************************/
644/* utilities for XS code */
510 645
511static size_t 646static size_t
512img_row_pitch (cl_mem img) 647img_row_pitch (cl_mem img)
513{ 648{
514 size_t res; 649 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 650 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 651 return res;
517} 652}
518 653
519static cl_event * 654static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 655event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 656{
522 cl_uint count = *rcount; 657 cl_uint count = *rcount;
523 658
524 if (!count) 659 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
660 *rcount = count = 0;
661
662 if (!count && !extra)
525 return 0; 663 return 0;
526 664
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 665 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 666 int i = 0;
529 667
530 do 668 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 669 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 670 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 671
536 while (count); 672 if (extra)
673 list [i++] = extra;
537 674
538 *rcount = i; 675 *rcount = i;
539 676
540 return i ? list : 0; 677 return i ? list : 0;
541} 678}
542 679
543#define EVENT_LIST(items,count) \ 680#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 681 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 682 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 683
547#define INFO(class) \ 684#define INFO(class) \
548{ \ 685{ \
549 size_t size; \ 686 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 687 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
554 SvCUR_set (sv, size); \ 691 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 692 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 693 XPUSHs (sv); \
557} 694}
558 695
696/*****************************************************************************/
697/* mapped_xxx */
698
699static OpenCL__Mapped
700SvMAPPED (SV *sv)
701{
702 // no typechecking atm., keep your fingers crossed
703 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
704}
705
706struct mapped
707{
708 cl_command_queue queue;
709 cl_mem memobj;
710 void *ptr;
711 size_t cb;
712 cl_event event;
713 size_t row_pitch;
714 size_t slice_pitch;
715};
716
717static SV *
718mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
719{
720 SV *data = newSV (0);
721 SvUPGRADE (data, SVt_PVMG);
722
723 OpenCL__Mapped mapped;
724 New (0, mapped, 1, struct mapped);
725
726 clRetainCommandQueue (queue);
727
728 mapped->queue = queue;
729 mapped->memobj = memobj;
730 mapped->ptr = ptr;
731 mapped->cb = cb;
732 mapped->event = ev;
733 mapped->row_pitch = row_pitch;
734 mapped->slice_pitch = slice_pitch;
735
736 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
737
738 if (SvLEN (data))
739 Safefree (data);
740
741 SvPVX (data) = (char *)ptr;
742 SvCUR_set (data, cb);
743 SvLEN_set (data, 0);
744 SvPOK_only (data);
745
746 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
747
748 if (!(flags & CL_MAP_WRITE))
749 SvREADONLY_on (data);
750
751 return obj;
752}
753
754static void
755mapped_detach (SV *sv, OpenCL__Mapped mapped)
756{
757 SV *data = SvRV (sv);
758
759 // the next check checks both before AND after detach, where SvPVX should STILL be 0
760 if (SvPVX (data) != (char *)mapped->ptr)
761 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
762 else
763 {
764 SvREADONLY_off (data);
765 SvCUR_set (data, 0);
766 SvPVX (data) = 0;
767 SvOK_off (data);
768 }
769
770 mapped->ptr = 0;
771}
772
773static void
774mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
775{
776 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
777 cl_event ev;
778
779 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
780
781 clReleaseEvent (mapped->event);
782 mapped->event = ev;
783
784 mapped_detach (self, mapped);
785}
786
787/*****************************************************************************/
788
559MODULE = OpenCL PACKAGE = OpenCL 789MODULE = OpenCL PACKAGE = OpenCL
560 790
561PROTOTYPES: ENABLE 791PROTOTYPES: ENABLE
562 792
563void 793void
572 eq_signal_arg = (void*)arg; 802 eq_signal_arg = (void*)arg;
573 803
574BOOT: 804BOOT:
575{ 805{
576 HV *stash = gv_stashpv ("OpenCL", 1); 806 HV *stash = gv_stashpv ("OpenCL", 1);
807
577 static const ivstr *civ, const_iv[] = { 808 static const ivstr *civ, const_iv[] = {
578 { sizeof (cl_char ), "SIZEOF_CHAR" }, 809 { sizeof (cl_char ), "SIZEOF_CHAR" },
579 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 810 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
580 { sizeof (cl_short ), "SIZEOF_SHORT" }, 811 { sizeof (cl_short ), "SIZEOF_SHORT" },
581 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 812 { sizeof (cl_ushort), "SIZEOF_USHORT" },
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 815 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 816 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 817 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 818 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 819 { sizeof (cl_double), "SIZEOF_DOUBLE" },
820 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 821#include "constiv.h"
590 }; 822 };
591 823
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 824 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
593 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 825 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
826
827 static const nvstr *cnv, const_nv[] = {
828#include "constnv.h"
829 };
830
831 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
832 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
833
834 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
835
836 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
837 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
838 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
839 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
840 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
841 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
842 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
843 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
844 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
845 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
846 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
847 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
848 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
849 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
850 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
851 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
852 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
853 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
854 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
855 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
856 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
857 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
858 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 859
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 860 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 861}
597 862
598cl_int 863cl_int
619 list = tmpbuf (sizeof (*list) * count); 884 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 885 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 886
622 EXTEND (SP, count); 887 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 888 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 889 PUSH_CLOBJ (stash_platform, list [i]);
625 890
626void 891void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 892context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 893 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 894 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 895 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 896 XPUSH_CLOBJ_CONTEXT;
632 897
633void 898void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 899context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 900 PPCODE:
636 /* der Gipfel der Kunst */ 901 cl_uint device_count;
902 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
903
904 CONTEXT_NOTIFY_CALLBACK;
905 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
906 XPUSH_CLOBJ_CONTEXT;
637 907
638void 908void
639wait_for_events (...) 909wait_for_events (...)
640 CODE: 910 CODE:
641 EVENT_LIST (0, items); 911 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 912 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 913
644PROTOTYPES: DISABLE 914PROTOTYPES: DISABLE
645 915
646MODULE = OpenCL PACKAGE = OpenCL::Platform 916MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 959 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 960 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 961
692 EXTEND (SP, count); 962 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 963 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 964 PUSH_CLOBJ (stash_device, list [i]);
695 965
696void 966void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 967context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 968 PPCODE:
969 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
970 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
971
972 cl_uint device_count;
973 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
974
975 CONTEXT_NOTIFY_CALLBACK;
976 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
977 XPUSH_CLOBJ_CONTEXT;
978
979void
980context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
981 PPCODE:
982 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
983 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
984
985 CONTEXT_NOTIFY_CALLBACK;
986 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
987 XPUSH_CLOBJ_CONTEXT;
988
989MODULE = OpenCL PACKAGE = OpenCL::Device
990
991void
992info (OpenCL::Device self, cl_device_info name)
993 PPCODE:
994 INFO (Device)
995
996#if CL_VERSION_1_2
997
998void
999sub_devices (OpenCL::Device self, SV *properties)
1000 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1001 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1002 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 1003
702 AV *av = (AV *)SvRV (devices); 1004 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 1005
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1006 cl_uint count = av_len ((AV *)properties) + 1;
1007 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 1008
706 int i; 1009 int i;
707 for (i = num_devices; i--; ) 1010 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1011 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 1012
710 CONTEXT_NOTIFY_CALLBACK; 1013 props [count] = 0;
711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
712 XPUSH_CLOBJ_CONTEXT;
713 1014
714void 1015 cl_uint num_devices;
715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 1016 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 1017 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1018 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719 1019
720 CONTEXT_NOTIFY_CALLBACK; 1020 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1021 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 1022 PUSH_CLOBJ (stash_subdevice, list [i]);
723 1023
724MODULE = OpenCL PACKAGE = OpenCL::Device 1024#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 1025
731#BEGIN:device 1026#BEGIN:device
732 1027
733void 1028void
734type (OpenCL::Device self) 1029type (OpenCL::Device self)
735 PPCODE: 1030 PPCODE:
736 cl_device_type value [1]; 1031 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1032 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 1033 EXTEND (SP, 1);
739 const int i = 0; 1034 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1035 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1036
742void 1037void
743vendor_id (OpenCL::Device self) 1038vendor_id (OpenCL::Device self)
744 ALIAS: 1039 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1040 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1185 PPCODE:
891 cl_command_queue_properties value [1]; 1186 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
894 const int i = 0; 1189 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1190 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1191
897void 1192void
898platform (OpenCL::Device self) 1193platform (OpenCL::Device self)
899 PPCODE: 1194 PPCODE:
900 cl_platform_id value [1]; 1195 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1196 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1197 EXTEND (SP, 1);
903 const int i = 0; 1198 const int i = 0;
904 { 1199 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1200 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1201 }
907 1202
908void 1203void
909name (OpenCL::Device self) 1204name (OpenCL::Device self)
910 ALIAS: 1205 ALIAS:
929 cl_device_id value [1]; 1224 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1225 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1226 EXTEND (SP, 1);
932 const int i = 0; 1227 const int i = 0;
933 { 1228 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1229 PUSH_CLOBJ (stash_device, value [i]);
935 } 1230 }
936 1231
937void 1232void
938partition_types_ext (OpenCL::Device self) 1233partition_types_ext (OpenCL::Device self)
939 ALIAS: 1234 ALIAS:
950 for (i = 0; i < n; ++i) 1245 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1246 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1247
953#END:device 1248#END:device
954 1249
1250MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1251
1252#if CL_VERSION_1_2
1253
1254void
1255DESTROY (OpenCL::SubDevice self)
1256 CODE:
1257 clReleaseDevice (self);
1258
1259#endif
1260
955MODULE = OpenCL PACKAGE = OpenCL::Context 1261MODULE = OpenCL PACKAGE = OpenCL::Context
956 1262
957void 1263void
958DESTROY (OpenCL::Context context) 1264DESTROY (OpenCL::Context self)
959 CODE: 1265 CODE:
960 clReleaseContext (context); 1266 clReleaseContext (self);
961 1267
962void 1268void
963info (OpenCL::Context self, cl_context_info name) 1269info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1270 PPCODE:
965 INFO (Context) 1271 INFO (Context)
966 1272
967void 1273void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1274queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1275 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1276 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1277 XPUSH_CLOBJ (stash_queue, queue);
972 1278
973void 1279void
974user_event (OpenCL::Context self) 1280user_event (OpenCL::Context self)
975 PPCODE: 1281 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1282 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1283 XPUSH_CLOBJ (stash_userevent, ev);
978 1284
979void 1285void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1286buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1287 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1288 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
983 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1289 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1290
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1291 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1292 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1293
988void 1294void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1295buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1296 PPCODE:
991 STRLEN len; 1297 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1298 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1299 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
994 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1300 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
995 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1301 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1302 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1303
998#if CL_VERSION_1_2 1304#if CL_VERSION_1_2
999 1305
1000void 1306void
1001image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 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) 1307image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
1006 const cl_image_desc desc = { 1312 const cl_image_desc desc = {
1007 type, 1313 type,
1008 width, height, depth, 1314 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1315 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1316 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1317 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1318 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1320 HV *stash = stash_image;
1015 switch (type) 1321 switch (type)
1016 { 1322 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1323 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1324 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1325 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1326 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1327 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1328 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1329 }
1024 XPUSH_CLOBJ (klass, mem); 1330 XPUSH_CLOBJ (stash, mem);
1025 1331
1026#endif 1332#endif
1027 1333
1028void 1334void
1029image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1335image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
1035 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1342#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1343 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1038 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1344 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1345#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1346 XPUSH_CLOBJ (stash_image2d, mem);
1041 1347
1042void 1348void
1043image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1349image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1044 PPCODE: 1350 PPCODE:
1045 STRLEN len; 1351 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1355 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1356#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1357 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1052 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1358 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1359#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1360 XPUSH_CLOBJ (stash_image3d, mem);
1055 1361
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1362#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1363
1058void 1364void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1365gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1366 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1368 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1369
1064void 1370void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1371gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1372 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1373 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1374 XPUSH_CLOBJ (stash_image2d, mem);
1069 1375
1070#if CL_VERSION_1_2 1376#if CL_VERSION_1_2
1071 1377
1072void 1378void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1379gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1380 ALIAS:
1075 PPCODE: 1381 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1382 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1383 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1384 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1385 HV *stash = stash_memory;
1080 switch (type) 1386 switch (type)
1081 { 1387 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1388 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1389 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1390 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1391 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1392 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1393 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1394 }
1089 XPUSH_CLOBJ (klass, mem); 1395 XPUSH_CLOBJ (stash, mem);
1090 1396
1091#endif 1397#endif
1092 1398
1093void 1399void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1400gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1402#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1403 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1404#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1405 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1406#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1407 XPUSH_CLOBJ (stash_image2d, mem);
1102 1408
1103void 1409void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1410gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1411 PPCODE:
1106#if PREFER_1_1 1412#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1413 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1414#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1415 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1416#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1417 XPUSH_CLOBJ (stash_image3d, mem);
1112 1418
1113#endif 1419#endif
1114 1420
1115void 1421void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1422supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1442
1137void 1443void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1444sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1445 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1446 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1447 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1448
1143void 1449void
1144program_with_source (OpenCL::Context self, SV *program) 1450program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1451 PPCODE:
1146 STRLEN len; 1452 STRLEN len;
1147 size_t len2; 1453 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1454 const char *ptr = SvPVbyte (program, len);
1149 1455
1150 len2 = len; 1456 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1457 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1458 XPUSH_CLOBJ (stash_program, prog);
1459
1460void
1461program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1462 PPCODE:
1463 cl_uint device_count;
1464 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1465
1466 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1467 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1468
1469 binaries = SvRV (binaries);
1470
1471 if (device_count != av_len ((AV *)binaries) + 1)
1472 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1473
1474 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1475 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1476 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1477
1478 int i;
1479 for (i = 0; i < device_count; ++i)
1480 {
1481 STRLEN len;
1482 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1483 length_list [i] = len;
1484 }
1485
1486 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1487 length_list, binary_list,
1488 GIMME_V == G_ARRAY ? status_list : 0, &res));
1489
1490 EXTEND (SP, 2);
1491 PUSH_CLOBJ (stash_program, prog);
1492
1493 if (GIMME_V == G_ARRAY)
1494 {
1495 AV *av = newAV ();
1496 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1497
1498 for (i = device_count; i--; )
1499 av_store (av, i, newSViv (status_list [i]));
1500 }
1501
1502#if CL_VERSION_1_2
1503
1504void
1505program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1506 PPCODE:
1507 cl_uint device_count;
1508 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1509
1510 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514void
1515link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1516 CODE:
1517 cl_uint device_count = 0;
1518 cl_device_id *device_list = 0;
1519
1520 if (SvOK (devices))
1521 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1522
1523 cl_uint program_count;
1524 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1525
1526 void *user_data;
1527 program_callback pfn_notify = make_program_callback (notify, &user_data);
1528
1529 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1530 program_count, program_list, pfn_notify, user_data, &res));
1531
1532 XPUSH_CLOBJ (stash_program, prog);
1533
1534#endif
1153 1535
1154#BEGIN:context 1536#BEGIN:context
1155 1537
1156void 1538void
1157reference_count (OpenCL::Context self) 1539reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1556 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1557 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1558 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1559 for (i = 0; i < n; ++i)
1178 { 1560 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1561 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1562 }
1181 1563
1182void 1564void
1183properties (OpenCL::Context self) 1565properties (OpenCL::Context self)
1184 PPCODE: 1566 PPCODE:
1203void 1585void
1204read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1586read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1205 ALIAS: 1587 ALIAS:
1206 enqueue_read_buffer = 0 1588 enqueue_read_buffer = 0
1207 PPCODE: 1589 PPCODE:
1208 cl_event ev = 0;
1209 EVENT_LIST (6, items - 6); 1590 EVENT_LIST (6);
1210 1591
1211 SvUPGRADE (data, SVt_PV); 1592 SvUPGRADE (data, SVt_PV);
1212 SvGROW (data, len); 1593 SvGROW (data, len);
1213 SvPOK_only (data); 1594 SvPOK_only (data);
1214 SvCUR_set (data, len); 1595 SvCUR_set (data, len);
1596
1597 cl_event ev = 0;
1215 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1598 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1216 1599
1217 if (ev) 1600 if (ev)
1218 XPUSH_CLOBJ ("OpenCL::Event", ev); 1601 XPUSH_CLOBJ (stash_event, ev);
1219 1602
1220void 1603void
1221write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1604write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1222 ALIAS: 1605 ALIAS:
1223 enqueue_write_buffer = 0 1606 enqueue_write_buffer = 0
1224 PPCODE: 1607 PPCODE:
1225 cl_event ev = 0; 1608 EVENT_LIST (5);
1609
1226 STRLEN len; 1610 STRLEN len;
1227 char *ptr = SvPVbyte (data, len); 1611 char *ptr = SvPVbyte (data, len);
1228 EVENT_LIST (5, items - 5);
1229 1612
1613 cl_event ev = 0;
1230 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1231 1615
1232 if (ev) 1616 if (ev)
1233 XPUSH_CLOBJ ("OpenCL::Event", ev); 1617 XPUSH_CLOBJ (stash_event, ev);
1234 1618
1235#if CL_VERSION_1_2 1619#if CL_VERSION_1_2
1236 1620
1237void 1621void
1238fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1622fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1239 ALIAS: 1623 ALIAS:
1240 enqueue_fill_buffer = 0 1624 enqueue_fill_buffer = 0
1241 PPCODE: 1625 PPCODE:
1242 cl_event ev = 0; 1626 EVENT_LIST (5);
1627
1243 STRLEN len; 1628 STRLEN len;
1244 char *ptr = SvPVbyte (data, len); 1629 char *ptr = SvPVbyte (data, len);
1245 EVENT_LIST (5, items - 5);
1246 1630
1631 cl_event ev = 0;
1247 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1632 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1248 1633
1249 if (ev) 1634 if (ev)
1250 XPUSH_CLOBJ ("OpenCL::Event", ev); 1635 XPUSH_CLOBJ (stash_event, ev);
1251 1636
1252void 1637void
1253fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 1638fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1254 ALIAS: 1639 ALIAS:
1255 enqueue_fill_image = 0 1640 enqueue_fill_image = 0
1256 PPCODE: 1641 PPCODE:
1257 cl_event ev = 0; 1642 EVENT_LIST (12);
1258 STRLEN len; 1643
1259 const size_t origin [3] = { x, y, z }; 1644 const size_t origin [3] = { x, y, z };
1260 const size_t region [3] = { width, height, depth }; 1645 const size_t region [3] = { width, height, depth };
1261 EVENT_LIST (12, items - 12);
1262 1646
1263 const cl_float c_f [4] = { r, g, b, a }; 1647 const cl_float c_f [4] = { r, g, b, a };
1264 const cl_uint c_u [4] = { r, g, b, a }; 1648 const cl_uint c_u [4] = { r, g, b, a };
1265 const cl_int c_s [4] = { r, g, b, a }; 1649 const cl_int c_s [4] = { r, g, b, a };
1266 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1650 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1269 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1653 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1270 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1654 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1271 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1655 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1272 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1656 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1273 1657
1658 cl_event ev = 0;
1274 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1659 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1275 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1660 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1276 1661
1277 if (ev) 1662 if (ev)
1278 XPUSH_CLOBJ ("OpenCL::Event", ev); 1663 XPUSH_CLOBJ (stash_event, ev);
1279 1664
1280#endif 1665#endif
1281 1666
1282void 1667void
1283copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1668copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1284 ALIAS: 1669 ALIAS:
1285 enqueue_copy_buffer = 0 1670 enqueue_copy_buffer = 0
1286 PPCODE: 1671 PPCODE:
1672 EVENT_LIST (6);
1673
1287 cl_event ev = 0; 1674 cl_event ev = 0;
1288 EVENT_LIST (6, items - 6);
1289
1290 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1675 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1291 1676
1292 if (ev) 1677 if (ev)
1293 XPUSH_CLOBJ ("OpenCL::Event", ev); 1678 XPUSH_CLOBJ (stash_event, ev);
1294 1679
1295void 1680void
1296read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1681read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1297 ALIAS: 1682 ALIAS:
1298 enqueue_read_buffer_rect = 0 1683 enqueue_read_buffer_rect = 0
1299 PPCODE: 1684 PPCODE:
1300 cl_event ev = 0; 1685 EVENT_LIST (17);
1686
1301 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1687 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1302 const size_t host_origin[3] = { host_x, host_y, host_z }; 1688 const size_t host_origin[3] = { host_x, host_y, host_z };
1303 const size_t region[3] = { width, height, depth }; 1689 const size_t region[3] = { width, height, depth };
1304 EVENT_LIST (17, items - 17);
1305 1690
1306 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
1307 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
1308 1693
1309 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
1319 1704
1320 SvUPGRADE (data, SVt_PV); 1705 SvUPGRADE (data, SVt_PV);
1321 SvGROW (data, len); 1706 SvGROW (data, len);
1322 SvPOK_only (data); 1707 SvPOK_only (data);
1323 SvCUR_set (data, len); 1708 SvCUR_set (data, len);
1709
1710 cl_event ev = 0;
1324 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1711 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1325 1712
1326 if (ev) 1713 if (ev)
1327 XPUSH_CLOBJ ("OpenCL::Event", ev); 1714 XPUSH_CLOBJ (stash_event, ev);
1328 1715
1329void 1716void
1330write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1717write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1331 ALIAS: 1718 ALIAS:
1332 enqueue_write_buffer_rect = 0 1719 enqueue_write_buffer_rect = 0
1333 PPCODE: 1720 PPCODE:
1334 cl_event ev = 0; 1721 EVENT_LIST (17);
1722
1335 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1723 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1336 const size_t host_origin[3] = { host_x, host_y, host_z }; 1724 const size_t host_origin[3] = { host_x, host_y, host_z };
1337 const size_t region[3] = { width, height, depth }; 1725 const size_t region[3] = { width, height, depth };
1338 STRLEN len; 1726 STRLEN len;
1339 char *ptr = SvPVbyte (data, len); 1727 char *ptr = SvPVbyte (data, len);
1340 EVENT_LIST (17, items - 17);
1341 1728
1342 if (!buf_row_pitch) 1729 if (!buf_row_pitch)
1343 buf_row_pitch = region [0]; 1730 buf_row_pitch = region [0];
1344 1731
1345 if (!buf_slice_pitch) 1732 if (!buf_slice_pitch)
1354 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1741 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1355 1742
1356 if (len < min_len) 1743 if (len < min_len)
1357 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1744 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1358 1745
1746 cl_event ev = 0;
1359 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1747 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1360 1748
1361 if (ev) 1749 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1363 1751
1364void 1752void
1365copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1753copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1366 ALIAS: 1754 ALIAS:
1367 enqueue_copy_buffer_rect = 0 1755 enqueue_copy_buffer_rect = 0
1368 PPCODE: 1756 PPCODE:
1369 cl_event ev = 0; 1757 EVENT_LIST (16);
1758
1370 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1760 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth }; 1761 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16);
1374 1762
1763 cl_event ev = 0;
1375 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1764 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1376 1765
1377 if (ev) 1766 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev); 1767 XPUSH_CLOBJ (stash_event, ev);
1379 1768
1380void 1769void
1381read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1770read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1382 ALIAS: 1771 ALIAS:
1383 enqueue_read_image = 0 1772 enqueue_read_image = 0
1384 PPCODE: 1773 PPCODE:
1385 cl_event ev = 0; 1774 EVENT_LIST (12);
1775
1386 const size_t src_origin[3] = { src_x, src_y, src_z }; 1776 const size_t src_origin[3] = { src_x, src_y, src_z };
1387 const size_t region[3] = { width, height, depth }; 1777 const size_t region[3] = { width, height, depth };
1388 EVENT_LIST (12, items - 12);
1389 1778
1390 if (!row_pitch) 1779 if (!row_pitch)
1391 row_pitch = img_row_pitch (src); 1780 row_pitch = img_row_pitch (src);
1392 1781
1393 if (depth > 1 && !slice_pitch) 1782 if (depth > 1 && !slice_pitch)
1397 1786
1398 SvUPGRADE (data, SVt_PV); 1787 SvUPGRADE (data, SVt_PV);
1399 SvGROW (data, len); 1788 SvGROW (data, len);
1400 SvPOK_only (data); 1789 SvPOK_only (data);
1401 SvCUR_set (data, len); 1790 SvCUR_set (data, len);
1791
1792 cl_event ev = 0;
1402 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1793 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1403 1794
1404 if (ev) 1795 if (ev)
1405 XPUSH_CLOBJ ("OpenCL::Event", ev); 1796 XPUSH_CLOBJ (stash_event, ev);
1406 1797
1407void 1798void
1408write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1799write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1409 ALIAS: 1800 ALIAS:
1410 enqueue_write_image = 0 1801 enqueue_write_image = 0
1411 PPCODE: 1802 PPCODE:
1412 cl_event ev = 0; 1803 EVENT_LIST (12);
1804
1413 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1805 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1414 const size_t region[3] = { width, height, depth }; 1806 const size_t region[3] = { width, height, depth };
1415 STRLEN len; 1807 STRLEN len;
1416 char *ptr = SvPVbyte (data, len); 1808 char *ptr = SvPVbyte (data, len);
1417 EVENT_LIST (12, items - 12);
1418 1809
1419 if (!row_pitch) 1810 if (!row_pitch)
1420 row_pitch = img_row_pitch (dst); 1811 row_pitch = img_row_pitch (dst);
1421 1812
1422 if (depth > 1 && !slice_pitch) 1813 if (depth > 1 && !slice_pitch)
1425 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1816 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1426 1817
1427 if (len < min_len) 1818 if (len < min_len)
1428 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1819 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1429 1820
1821 cl_event ev = 0;
1430 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1822 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1431 1823
1432 if (ev) 1824 if (ev)
1433 XPUSH_CLOBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1434 1826
1435void 1827void
1436copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1828copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1437 ALIAS: 1829 ALIAS:
1438 enqueue_copy_image = 0 1830 enqueue_copy_image = 0
1439 PPCODE: 1831 PPCODE:
1440 cl_event ev = 0; 1832 EVENT_LIST (12);
1833
1441 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1442 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1835 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1443 const size_t region[3] = { width, height, depth }; 1836 const size_t region[3] = { width, height, depth };
1444 EVENT_LIST (12, items - 12);
1445 1837
1838 cl_event ev = 0;
1446 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1839 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1447 1840
1448 if (ev) 1841 if (ev)
1449 XPUSH_CLOBJ ("OpenCL::Event", ev); 1842 XPUSH_CLOBJ (stash_event, ev);
1450 1843
1451void 1844void
1452copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1845copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1453 ALIAS: 1846 ALIAS:
1454 enqueue_copy_image_to_buffer = 0 1847 enqueue_copy_image_to_buffer = 0
1455 PPCODE: 1848 PPCODE:
1849 EVENT_LIST (10);
1850
1851 const size_t src_origin[3] = { src_x, src_y, src_z };
1852 const size_t region [3] = { width, height, depth };
1853
1456 cl_event ev = 0; 1854 cl_event ev = 0;
1457 const size_t src_origin[3] = { src_x, src_y, src_z };
1458 const size_t region[3] = { width, height, depth };
1459 EVENT_LIST (10, items - 10);
1460
1461 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1855 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1462 1856
1463 if (ev) 1857 if (ev)
1464 XPUSH_CLOBJ ("OpenCL::Event", ev); 1858 XPUSH_CLOBJ (stash_event, ev);
1465 1859
1466void 1860void
1467copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1861copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1468 ALIAS: 1862 ALIAS:
1469 enqueue_copy_buffer_to_image = 0 1863 enqueue_copy_buffer_to_image = 0
1470 PPCODE: 1864 PPCODE:
1865 EVENT_LIST (10);
1866
1867 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1868 const size_t region [3] = { width, height, depth };
1869
1471 cl_event ev = 0; 1870 cl_event ev = 0;
1871 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1872
1873 if (ev)
1874 XPUSH_CLOBJ (stash_event, ev);
1875
1876void
1877map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1878 ALIAS:
1879 enqueue_map_buffer = 0
1880 PPCODE:
1881 EVENT_LIST (6);
1882
1883 size_t cb = SvIV (cb_);
1884
1885 if (!SvOK (cb_))
1886 {
1887 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1888 cb -= offset;
1889 }
1890
1891 cl_event ev;
1892 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1893 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1894
1895void
1896map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1897 ALIAS:
1898 enqueue_map_image = 0
1899 PPCODE:
1900 size_t width = SvIV (width_);
1901 if (!SvOK (width_))
1902 {
1903 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1904 width -= x;
1905 }
1906
1907 size_t height = SvIV (width_);
1908 if (!SvOK (height_))
1909 {
1910 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1911 height -= y;
1912 }
1913
1914 size_t depth = SvIV (width_);
1915 if (!SvOK (depth_))
1916 {
1917 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1918 depth -= z;
1919
1920 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1921 if (!depth)
1922 depth = 1;
1923 }
1924
1472 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1925 const size_t origin[3] = { x, y, z };
1473 const size_t region[3] = { width, height, depth }; 1926 const size_t region[3] = { width, height, depth };
1927 size_t row_pitch, slice_pitch;
1474 EVENT_LIST (10, items - 10); 1928 EVENT_LIST (10);
1475 1929
1476 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1930 cl_event ev;
1931 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1477 1932
1478 if (ev) 1933 size_t cb = slice_pitch ? slice_pitch * region [2]
1479 XPUSH_CLOBJ ("OpenCL::Event", ev); 1934 : row_pitch ? row_pitch * region [1]
1935 : region [0];
1936
1937 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1938
1939void
1940unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1941 PPCODE:
1942 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1943 if (GIMME_V != G_VOID)
1944 {
1945 clRetainEvent (mapped->event);
1946 XPUSH_CLOBJ (stash_event, mapped->event);
1947 }
1480 1948
1481void 1949void
1482task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1950task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1483 ALIAS: 1951 ALIAS:
1484 enqueue_task = 0 1952 enqueue_task = 0
1485 PPCODE: 1953 PPCODE:
1954 EVENT_LIST (2);
1955
1486 cl_event ev = 0; 1956 cl_event ev = 0;
1487 EVENT_LIST (2, items - 2);
1488
1489 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1957 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1490 1958
1491 if (ev) 1959 if (ev)
1492 XPUSH_CLOBJ ("OpenCL::Event", ev); 1960 XPUSH_CLOBJ (stash_event, ev);
1493 1961
1494void 1962void
1495nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1963nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1496 ALIAS: 1964 ALIAS:
1497 enqueue_nd_range_kernel = 0 1965 enqueue_nd_range_kernel = 0
1498 PPCODE: 1966 PPCODE:
1499 cl_event ev = 0; 1967 EVENT_LIST (5);
1968
1500 size_t *gwo = 0, *gws, *lws = 0; 1969 size_t *gwo = 0, *gws, *lws = 0;
1501 int gws_len; 1970 int gws_len;
1502 size_t *lists; 1971 size_t *lists;
1503 int i; 1972 int i;
1504 EVENT_LIST (5, items - 5);
1505 1973
1506 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1974 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1507 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1508 1976
1509 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1977 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1548 if (!lws [i]) 2016 if (!lws [i])
1549 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 2017 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1550 } 2018 }
1551 } 2019 }
1552 2020
2021 cl_event ev = 0;
1553 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2022 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1554 2023
1555 if (ev) 2024 if (ev)
1556 XPUSH_CLOBJ ("OpenCL::Event", ev); 2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#if CL_VERSION_1_2
2028
2029void
2030migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2031 ALIAS:
2032 enqueue_migrate_mem_objects = 0
2033 PPCODE:
2034 EVENT_LIST (3);
2035
2036 cl_uint obj_count;
2037 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2038
2039 cl_event ev = 0;
2040 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2041
2042 if (ev)
2043 XPUSH_CLOBJ (stash_event, ev);
2044
2045#endif
1557 2046
1558#if cl_apple_gl_sharing || cl_khr_gl_sharing 2047#if cl_apple_gl_sharing || cl_khr_gl_sharing
1559 2048
1560void 2049void
1561acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2050acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1562 ALIAS: 2051 ALIAS:
2052 release_gl_objects = 1
1563 enqueue_acquire_gl_objects = 0 2053 enqueue_acquire_gl_objects = 0
1564 ALIAS:
1565 enqueue_release_gl_objects = 1 2054 enqueue_release_gl_objects = 1
1566 PPCODE: 2055 PPCODE:
1567 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2056 EVENT_LIST (2);
1568 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2057
2058 cl_uint obj_count;
2059 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1569 2060
1570 cl_event ev = 0; 2061 cl_event ev = 0;
1571 EVENT_LIST (2, items - 2);
1572 AV *av = (AV *)SvRV (objects);
1573 cl_uint num_objects = av_len (av) + 1;
1574 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1575 int i;
1576
1577 for (i = num_objects; i--; )
1578 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1579 2062
1580 if (ix) 2063 if (ix)
1581 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2064 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1582 else 2065 else
1583 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2066 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1584 2067
1585 if (ev) 2068 if (ev)
1586 XPUSH_CLOBJ ("OpenCL::Event", ev); 2069 XPUSH_CLOBJ (stash_event, ev);
1587 2070
1588#endif 2071#endif
1589 2072
1590void 2073void
1591wait_for_events (OpenCL::Queue self, ...) 2074wait_for_events (OpenCL::Queue self, ...)
1592 ALIAS: 2075 ALIAS:
1593 enqueue_wait_for_events = 0 2076 enqueue_wait_for_events = 0
1594 CODE: 2077 CODE:
1595 EVENT_LIST (1, items - 1); 2078 EVENT_LIST (1);
1596#if PREFER_1_1 2079#if PREFER_1_1
1597 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1598#else 2081#else
1599 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2082 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1600#endif 2083#endif
1602void 2085void
1603marker (OpenCL::Queue self, ...) 2086marker (OpenCL::Queue self, ...)
1604 ALIAS: 2087 ALIAS:
1605 enqueue_marker = 0 2088 enqueue_marker = 0
1606 PPCODE: 2089 PPCODE:
2090 EVENT_LIST (1);
1607 cl_event ev = 0; 2091 cl_event ev = 0;
1608 EVENT_LIST (1, items - 1);
1609#if PREFER_1_1 2092#if PREFER_1_1
1610 if (!event_list_count) 2093 if (!event_list_count)
1611 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2094 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1612 else 2095 else
1613#if CL_VERSION_1_2 2096#if CL_VERSION_1_2
1620#endif 2103#endif
1621#else 2104#else
1622 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2105 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1623#endif 2106#endif
1624 if (ev) 2107 if (ev)
1625 XPUSH_CLOBJ ("OpenCL::Event", ev); 2108 XPUSH_CLOBJ (stash_event, ev);
1626 2109
1627void 2110void
1628barrier (OpenCL::Queue self, ...) 2111barrier (OpenCL::Queue self, ...)
1629 ALIAS: 2112 ALIAS:
1630 enqueue_barrier = 0 2113 enqueue_barrier = 0
1631 PPCODE: 2114 PPCODE:
2115 EVENT_LIST (1);
1632 cl_event ev = 0; 2116 cl_event ev = 0;
1633 EVENT_LIST (1, items - 1);
1634#if PREFER_1_1 2117#if PREFER_1_1
1635 if (!event_list_count && GIMME_V == G_VOID) 2118 if (!event_list_count && GIMME_V == G_VOID)
1636 NEED_SUCCESS (EnqueueBarrier, (self)); 2119 NEED_SUCCESS (EnqueueBarrier, (self));
1637 else 2120 else
1638#if CL_VERSION_1_2 2121#if CL_VERSION_1_2
1648#endif 2131#endif
1649#else 2132#else
1650 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2133 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1651#endif 2134#endif
1652 if (ev) 2135 if (ev)
1653 XPUSH_CLOBJ ("OpenCL::Event", ev); 2136 XPUSH_CLOBJ (stash_event, ev);
1654 2137
1655void 2138void
1656flush (OpenCL::Queue self) 2139flush (OpenCL::Queue self)
1657 CODE: 2140 CODE:
1658 NEED_SUCCESS (Flush, (self)); 2141 NEED_SUCCESS (Flush, (self));
1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2159 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1677 EXTEND (SP, 1); 2160 EXTEND (SP, 1);
1678 const int i = 0; 2161 const int i = 0;
1679 { 2162 {
1680 NEED_SUCCESS (RetainContext, (value [i])); 2163 NEED_SUCCESS (RetainContext, (value [i]));
1681 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2164 PUSH_CLOBJ (stash_context, value [i]);
1682 } 2165 }
1683 2166
1684void 2167void
1685device (OpenCL::Queue self) 2168device (OpenCL::Queue self)
1686 PPCODE: 2169 PPCODE:
1687 cl_device_id value [1]; 2170 cl_device_id value [1];
1688 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2171 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1689 EXTEND (SP, 1); 2172 EXTEND (SP, 1);
1690 const int i = 0; 2173 const int i = 0;
1691 { 2174 {
1692 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2175 PUSH_CLOBJ (stash_device, value [i]);
1693 } 2176 }
1694 2177
1695void 2178void
1696reference_count (OpenCL::Queue self) 2179reference_count (OpenCL::Queue self)
1697 PPCODE: 2180 PPCODE:
1706 PPCODE: 2189 PPCODE:
1707 cl_command_queue_properties value [1]; 2190 cl_command_queue_properties value [1];
1708 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2191 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1709 EXTEND (SP, 1); 2192 EXTEND (SP, 1);
1710 const int i = 0; 2193 const int i = 0;
1711 PUSHs (sv_2mortal (newSViv (value [i]))); 2194 PUSHs (sv_2mortal (newSVuv (value [i])));
1712 2195
1713#END:command_queue 2196#END:command_queue
1714 2197
1715MODULE = OpenCL PACKAGE = OpenCL::Memory 2198MODULE = OpenCL PACKAGE = OpenCL::Memory
1716 2199
1731 PPCODE: 2214 PPCODE:
1732 cl_mem_object_type value [1]; 2215 cl_mem_object_type value [1];
1733 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2216 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1734 EXTEND (SP, 1); 2217 EXTEND (SP, 1);
1735 const int i = 0; 2218 const int i = 0;
1736 PUSHs (sv_2mortal (newSViv (value [i]))); 2219 PUSHs (sv_2mortal (newSVuv (value [i])));
1737 2220
1738void 2221void
1739flags (OpenCL::Memory self) 2222flags (OpenCL::Memory self)
1740 PPCODE: 2223 PPCODE:
1741 cl_mem_flags value [1]; 2224 cl_mem_flags value [1];
1742 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2225 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1743 EXTEND (SP, 1); 2226 EXTEND (SP, 1);
1744 const int i = 0; 2227 const int i = 0;
1745 PUSHs (sv_2mortal (newSViv (value [i]))); 2228 PUSHs (sv_2mortal (newSVuv (value [i])));
1746 2229
1747void 2230void
1748size (OpenCL::Memory self) 2231size (OpenCL::Memory self)
1749 ALIAS: 2232 ALIAS:
1750 size = CL_MEM_SIZE 2233 size = CL_MEM_SIZE
1784 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2267 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1785 EXTEND (SP, 1); 2268 EXTEND (SP, 1);
1786 const int i = 0; 2269 const int i = 0;
1787 { 2270 {
1788 NEED_SUCCESS (RetainContext, (value [i])); 2271 NEED_SUCCESS (RetainContext, (value [i]));
1789 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2272 PUSH_CLOBJ (stash_context, value [i]);
1790 } 2273 }
1791 2274
1792void 2275void
1793associated_memobject (OpenCL::Memory self) 2276associated_memobject (OpenCL::Memory self)
1794 PPCODE: 2277 PPCODE:
1796 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2279 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1797 EXTEND (SP, 1); 2280 EXTEND (SP, 1);
1798 const int i = 0; 2281 const int i = 0;
1799 { 2282 {
1800 NEED_SUCCESS (RetainMemObject, (value [i])); 2283 NEED_SUCCESS (RetainMemObject, (value [i]));
1801 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2284 PUSH_CLOBJ (stash_memory, value [i]);
1802 } 2285 }
1803 2286
1804#END:mem 2287#END:mem
1805 2288
1806#if cl_apple_gl_sharing || cl_khr_gl_sharing 2289#if cl_apple_gl_sharing || cl_khr_gl_sharing
1826 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2309 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1827 2310
1828 cl_buffer_region crdata = { origin, size }; 2311 cl_buffer_region crdata = { origin, size };
1829 2312
1830 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2313 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1831 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2314 XPUSH_CLOBJ (stash_buffer, mem);
1832 2315
1833MODULE = OpenCL PACKAGE = OpenCL::Image 2316MODULE = OpenCL PACKAGE = OpenCL::Image
1834 2317
1835void 2318void
1836image_info (OpenCL::Image self, cl_image_info name) 2319image_info (OpenCL::Image self, cl_image_info name)
1922 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2405 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1923 EXTEND (SP, 1); 2406 EXTEND (SP, 1);
1924 const int i = 0; 2407 const int i = 0;
1925 { 2408 {
1926 NEED_SUCCESS (RetainContext, (value [i])); 2409 NEED_SUCCESS (RetainContext, (value [i]));
1927 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2410 PUSH_CLOBJ (stash_context, value [i]);
1928 } 2411 }
1929 2412
1930void 2413void
1931normalized_coords (OpenCL::Sampler self) 2414normalized_coords (OpenCL::Sampler self)
1932 PPCODE: 2415 PPCODE:
1933 cl_addressing_mode value [1]; 2416 cl_addressing_mode value [1];
1934 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2417 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1935 EXTEND (SP, 1); 2418 EXTEND (SP, 1);
1936 const int i = 0; 2419 const int i = 0;
1937 PUSHs (sv_2mortal (newSViv (value [i]))); 2420 PUSHs (sv_2mortal (newSVuv (value [i])));
1938 2421
1939void 2422void
1940addressing_mode (OpenCL::Sampler self) 2423addressing_mode (OpenCL::Sampler self)
1941 PPCODE: 2424 PPCODE:
1942 cl_filter_mode value [1]; 2425 cl_filter_mode value [1];
1943 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2426 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1944 EXTEND (SP, 1); 2427 EXTEND (SP, 1);
1945 const int i = 0; 2428 const int i = 0;
1946 PUSHs (sv_2mortal (newSViv (value [i]))); 2429 PUSHs (sv_2mortal (newSVuv (value [i])));
1947 2430
1948void 2431void
1949filter_mode (OpenCL::Sampler self) 2432filter_mode (OpenCL::Sampler self)
1950 PPCODE: 2433 PPCODE:
1951 cl_bool value [1]; 2434 cl_bool value [1];
1966void 2449void
1967build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2450build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1968 ALIAS: 2451 ALIAS:
1969 build_async = 1 2452 build_async = 1
1970 CODE: 2453 CODE:
1971 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1972 void *user_data = 0;
1973 cl_uint num_devices = 0; 2454 cl_uint device_count = 0;
1974 cl_device_id *device_list = 0; 2455 cl_device_id *device_list = 0;
1975 2456
1976 if (SvOK (devices)) 2457 if (SvOK (devices))
1977 { 2458 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2459
2460 void *user_data;
2461 program_callback pfn_notify = make_program_callback (notify, &user_data);
2462
2463 if (ix)
2464 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2465 else
2466 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2467
2468#if CL_VERSION_1_2
2469
2470void
2471compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2472 CODE:
2473 cl_uint device_count = 0;
2474 cl_device_id *device_list = 0;
2475
2476 if (SvOK (devices))
2477 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2478
2479 cl_uint header_count = 0;
2480 cl_program *header_list = 0;
2481 const char **header_name = 0;
2482
2483 if (SvOK (headers))
2484 {
1978 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2485 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1979 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2486 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1980 2487
1981 AV *av = (AV *)SvRV (devices); 2488 HV *hv = (HV *)SvRV (devices);
1982 num_devices = av_len (av) + 1;
1983 2489
1984 if (num_devices) 2490 header_count = hv_iterinit (hv);
2491 header_list = tmpbuf (sizeof (*header_list) * header_count);
2492 header_name = tmpbuf (sizeof (*header_name) * header_count);
2493
2494 HE *he;
2495 int i = 0;
2496 while (he = hv_iternext (hv))
1985 { 2497 {
1986 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2498 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2499 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1987 int count; 2500 ++i;
1988 for (count = 0; count < num_devices; ++count)
1989 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1990 } 2501 }
1991 } 2502 }
1992 2503
1993 if (SvOK (notify)) 2504 void *user_data;
1994 { 2505 program_callback pfn_notify = make_program_callback (notify, &user_data);
1995 NEED_SUCCESS (RetainProgram, (self));
1996 pfn_notify = eq_program_notify;
1997 user_data = SvREFCNT_inc (s_get_cv (notify));
1998 }
1999 2506
2000 if (ix)
2001 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2002 else
2003 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2507 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2508 header_count, header_list, header_name, pfn_notify, user_data));
2509
2510#endif
2004 2511
2005void 2512void
2006build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2513build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2007 PPCODE: 2514 PPCODE:
2008 size_t size; 2515 size_t size;
2037 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2544 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2038 EXTEND (SP, 1); 2545 EXTEND (SP, 1);
2039 const int i = 0; 2546 const int i = 0;
2040 PUSHs (sv_2mortal (newSVpv (value, 0))); 2547 PUSHs (sv_2mortal (newSVpv (value, 0)));
2041 2548
2549void
2550binary_type (OpenCL::Program self, OpenCL::Device device)
2551 PPCODE:
2552 cl_program_binary_type value [1];
2553 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2554 EXTEND (SP, 1);
2555 const int i = 0;
2556 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2557
2042#END:program_build 2558#END:program_build
2043 2559
2044void 2560void
2045kernel (OpenCL::Program program, SV *function) 2561kernel (OpenCL::Program program, SV *function)
2046 PPCODE: 2562 PPCODE:
2047 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2563 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2048 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2564 XPUSH_CLOBJ (stash_kernel, kernel);
2049 2565
2050void 2566void
2051kernels_in_program (OpenCL::Program program) 2567kernels_in_program (OpenCL::Program program)
2052 PPCODE: 2568 PPCODE:
2053 cl_uint num_kernels; 2569 cl_uint num_kernels;
2056 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2572 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2057 2573
2058 int i; 2574 int i;
2059 EXTEND (SP, num_kernels); 2575 EXTEND (SP, num_kernels);
2060 for (i = 0; i < num_kernels; ++i) 2576 for (i = 0; i < num_kernels; ++i)
2061 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2577 PUSH_CLOBJ (stash_kernel, kernels [i]);
2062 2578
2063void 2579void
2064info (OpenCL::Program self, cl_program_info name) 2580info (OpenCL::Program self, cl_program_info name)
2065 PPCODE: 2581 PPCODE:
2066 INFO (Program) 2582 INFO (Program)
2114 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2630 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2115 EXTEND (SP, 1); 2631 EXTEND (SP, 1);
2116 const int i = 0; 2632 const int i = 0;
2117 { 2633 {
2118 NEED_SUCCESS (RetainContext, (value [i])); 2634 NEED_SUCCESS (RetainContext, (value [i]));
2119 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2635 PUSH_CLOBJ (stash_context, value [i]);
2120 } 2636 }
2121 2637
2122void 2638void
2123devices (OpenCL::Program self) 2639devices (OpenCL::Program self)
2124 PPCODE: 2640 PPCODE:
2128 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2644 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2129 int i, n = size / sizeof (*value); 2645 int i, n = size / sizeof (*value);
2130 EXTEND (SP, n); 2646 EXTEND (SP, n);
2131 for (i = 0; i < n; ++i) 2647 for (i = 0; i < n; ++i)
2132 { 2648 {
2133 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2649 PUSH_CLOBJ (stash_device, value [i]);
2134 } 2650 }
2135 2651
2136void 2652void
2137source (OpenCL::Program self) 2653source (OpenCL::Program self)
2138 PPCODE: 2654 PPCODE:
2193 cl_sampler ca; 2709 cl_sampler ca;
2194 size_t cz; 2710 size_t cz;
2195 cl_event ce; 2711 cl_event ce;
2196 } arg; 2712 } arg;
2197 size_t size; 2713 size_t size;
2714 int nullarg = 0;
2198 2715
2199 switch (type) 2716 switch (type)
2200 { 2717 {
2201 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break; 2718 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2202 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break; 2719 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2208 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break; 2725 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2209 2726
2210 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break; 2727 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2211 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2728 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2212 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2729 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2213 case 'z': arg.cz = SvUV (sv); size = sizeof (arg.cz); break;
2214 2730
2731 case 'z': nullarg = 1; size = SvIV (sv); break;
2732
2215 case 'm': arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2733 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2216 case 'a': arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2734 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2217 case 'e': arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2735 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2218 2736
2219 default: 2737 default:
2220 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2738 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2221 } 2739 }
2222 2740
2223 clSetKernelArg (self, i - 2, size, &arg); 2741 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2742 if (res)
2743 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2224 } 2744 }
2225 2745
2226 if (i != items) 2746 if (i != items)
2227 croak ("OpenCL::Kernel::setf format string too short (too many arguments)"); 2747 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2228 2748
2282 clSetKernelArg (self, idx, sizeof (value), &value); 2802 clSetKernelArg (self, idx, sizeof (value), &value);
2283 2803
2284void 2804void
2285set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2805set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2286 CODE: 2806 CODE:
2287 clSetKernelArg (self, idx, sizeof (value), &value); 2807 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2288 2808
2289void 2809void
2290set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2810set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2291 CODE: 2811 CODE:
2292 clSetKernelArg (self, idx, sizeof (value), &value); 2812 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2293 2813
2294void 2814void
2295set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2815set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2296 ALIAS:
2297 set_image2d = 0
2298 set_image3d = 0
2299 CODE: 2816 CODE:
2300 clSetKernelArg (self, idx, sizeof (value), &value); 2817 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2301 2818
2302void 2819void
2303set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2820set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2304 CODE: 2821 CODE:
2305 clSetKernelArg (self, idx, sizeof (value), &value); 2822 clSetKernelArg (self, idx, sizeof (value), &value);
2351 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2868 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2352 EXTEND (SP, 1); 2869 EXTEND (SP, 1);
2353 const int i = 0; 2870 const int i = 0;
2354 { 2871 {
2355 NEED_SUCCESS (RetainContext, (value [i])); 2872 NEED_SUCCESS (RetainContext, (value [i]));
2356 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2873 PUSH_CLOBJ (stash_context, value [i]);
2357 } 2874 }
2358 2875
2359void 2876void
2360program (OpenCL::Kernel self) 2877program (OpenCL::Kernel self)
2361 PPCODE: 2878 PPCODE:
2363 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2880 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2364 EXTEND (SP, 1); 2881 EXTEND (SP, 1);
2365 const int i = 0; 2882 const int i = 0;
2366 { 2883 {
2367 NEED_SUCCESS (RetainProgram, (value [i])); 2884 NEED_SUCCESS (RetainProgram, (value [i]));
2368 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2885 PUSH_CLOBJ (stash_program, value [i]);
2369 } 2886 }
2370 2887
2371#END:kernel 2888#END:kernel
2372 2889
2373void 2890void
2420 const int i = 0; 2937 const int i = 0;
2421 PUSHs (sv_2mortal (newSVuv (value [i]))); 2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2422 2939
2423#END:kernel_work_group 2940#END:kernel_work_group
2424 2941
2942#if CL_VERSION_1_2
2943
2944void
2945arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2946 PPCODE:
2947 size_t size;
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2949 SV *sv = sv_2mortal (newSV (size));
2950 SvUPGRADE (sv, SVt_PV);
2951 SvPOK_only (sv);
2952 SvCUR_set (sv, size);
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2954 XPUSHs (sv);
2955
2956#BEGIN:kernel_arg
2957
2958void
2959arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2960 PPCODE:
2961 cl_kernel_arg_address_qualifier value [1];
2962 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2963 EXTEND (SP, 1);
2964 const int i = 0;
2965 PUSHs (sv_2mortal (newSVuv (value [i])));
2966
2967void
2968arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2969 PPCODE:
2970 cl_kernel_arg_access_qualifier value [1];
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVuv (value [i])));
2975
2976void
2977arg_type_name (OpenCL::Kernel self, cl_uint idx)
2978 ALIAS:
2979 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2980 arg_name = CL_KERNEL_ARG_NAME
2981 PPCODE:
2982 size_t size;
2983 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2984 char *value = tmpbuf (size);
2985 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2986 EXTEND (SP, 1);
2987 const int i = 0;
2988 PUSHs (sv_2mortal (newSVpv (value, 0)));
2989
2990void
2991arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2992 PPCODE:
2993 cl_kernel_arg_type_qualifier value [1];
2994 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2995 EXTEND (SP, 1);
2996 const int i = 0;
2997 PUSHs (sv_2mortal (newSVuv (value [i])));
2998
2999#END:kernel_arg
3000
3001#endif
3002
2425MODULE = OpenCL PACKAGE = OpenCL::Event 3003MODULE = OpenCL PACKAGE = OpenCL::Event
2426 3004
2427void 3005void
2428DESTROY (OpenCL::Event self) 3006DESTROY (OpenCL::Event self)
2429 CODE: 3007 CODE:
2453 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3031 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2454 EXTEND (SP, 1); 3032 EXTEND (SP, 1);
2455 const int i = 0; 3033 const int i = 0;
2456 { 3034 {
2457 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3035 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2458 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 3036 PUSH_CLOBJ (stash_queue, value [i]);
2459 } 3037 }
2460 3038
2461void 3039void
2462command_type (OpenCL::Event self) 3040command_type (OpenCL::Event self)
2463 PPCODE: 3041 PPCODE:
2486 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3064 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2487 EXTEND (SP, 1); 3065 EXTEND (SP, 1);
2488 const int i = 0; 3066 const int i = 0;
2489 { 3067 {
2490 NEED_SUCCESS (RetainContext, (value [i])); 3068 NEED_SUCCESS (RetainContext, (value [i]));
2491 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3069 PUSH_CLOBJ (stash_context, value [i]);
2492 } 3070 }
2493 3071
2494#END:event 3072#END:event
2495 3073
2496void 3074void
2521void 3099void
2522set_status (OpenCL::UserEvent self, cl_int execution_status) 3100set_status (OpenCL::UserEvent self, cl_int execution_status)
2523 CODE: 3101 CODE:
2524 clSetUserEventStatus (self, execution_status); 3102 clSetUserEventStatus (self, execution_status);
2525 3103
3104MODULE = OpenCL PACKAGE = OpenCL::Mapped
3105
3106void
3107DESTROY (SV *self)
3108 CODE:
3109 OpenCL__Mapped mapped = SvMAPPED (self);
3110
3111 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3112 mapped_detach (self, mapped);
3113
3114 clReleaseCommandQueue (mapped->queue);
3115 clReleaseEvent (mapped->event);
3116 Safefree (mapped);
3117
3118void
3119unmap (OpenCL::Mapped self, ...)
3120 CODE:
3121 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3122
3123bool
3124mapped (OpenCL::Mapped self)
3125 CODE:
3126 RETVAL = !!self->ptr;
3127 OUTPUT:
3128 RETVAL
3129
3130void
3131wait (OpenCL::Mapped self)
3132 PPCODE:
3133 if (self->event)
3134 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3135
3136void
3137event (OpenCL::Mapped self)
3138 PPCODE:
3139 if (!self->event)
3140 XSRETURN_UNDEF;
3141
3142 clRetainEvent (self->event);
3143 XPUSH_CLOBJ (stash_event, self->event);
3144
3145size_t
3146size (OpenCL::Mapped self)
3147 CODE:
3148 RETVAL = self->cb;
3149 OUTPUT:
3150 RETVAL
3151
3152IV
3153ptr (OpenCL::Mapped self)
3154 CODE:
3155 RETVAL = PTR2IV (self->ptr);
3156 OUTPUT:
3157 RETVAL
3158
3159void
3160set (OpenCL::Mapped self, size_t offset, SV *data)
3161 CODE:
3162 STRLEN len;
3163 const char *ptr = SvPVbyte (data, len);
3164
3165 if (offset + len > self->cb)
3166 croak ("OpenCL::Mapped::set out of bound condition detected");
3167
3168 memcpy (offset + (char *)self->ptr, ptr, len);
3169
3170MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3171
3172MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3173
3174IV
3175row_pitch (OpenCL::Mapped self)
3176 ALIAS:
3177 slice_pitch = 1
3178 CODE:
3179 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3180 OUTPUT:
3181 RETVAL
3182
3183
3184

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines