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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.61 by root, Tue May 1 16:37:23 2012 UTC vs.
Revision 1.80 by root, Sat May 5 16:44:02 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
6
7#include "ecb.h"//D
4 8
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 10#include "xthread.h"
7#include "schmorp.h" 11#include "schmorp.h"
8 12
26#if PREFER_1_1 30#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 32#endif
29 33
30#ifdef __APPLE__ 34#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
32#else 36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
33 #include <CL/opencl.h> 40#include CLHDR(opencl.h)
41
42#ifndef CL_VERSION_1_2
43 #include CLHDR(cl_d3d9.h)
44#endif
45
46#if _WIN32
47 #include CLHDR(cl_d3d10.h)
48 #if CL_VERSION_1_2
49 #include CLHDR<cl_d3d11.h>
50 #endif
51 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 52#endif
35 53
36#ifndef CL_VERSION_1_2 54#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 55 #undef PREFER_1_1
38 #define PREFER_1_1 1 56 #define PREFER_1_1 1
39#endif 57#endif
40 58
59// make sure all constants we might use are actually defined
60#include "default.h"
61
41typedef cl_platform_id OpenCL__Platform; 62typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 63typedef cl_device_id OpenCL__Device;
64typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 65typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 66typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 67typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 68typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 69typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 70typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 71typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 72typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 73typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 74typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 75typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 76typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 77typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 78typedef cl_event OpenCL__UserEvent;
61 79
62typedef struct mapped * OpenCL__Mapped; 80typedef struct mapped * OpenCL__Mapped;
63 81
64typedef SV *FUTURE;
65
66static HV 82static HV
67 *stash_platform, 83 *stash_platform,
68 *stash_device, 84 *stash_device,
85 *stash_subdevice,
69 *stash_context, 86 *stash_context,
70 *stash_queue, 87 *stash_queue,
71 *stash_program, 88 *stash_program,
72 *stash_kernel, 89 *stash_kernel,
73 *stash_sampler, 90 *stash_sampler,
133} 150}
134 151
135/*****************************************************************************/ 152/*****************************************************************************/
136 153
137/* up to two temporary buffers */ 154/* up to two temporary buffers */
138static void * 155static void * ecb_noinline
139tmpbuf (size_t size) 156tmpbuf (size_t size)
140{ 157{
141 enum { buffers = 3 }; 158 enum { buffers = 4 };
142 static int idx; 159 static int idx;
143 static void *buf [buffers]; 160 static void *buf [buffers];
144 static size_t len [buffers]; 161 static size_t len [buffers];
145 162
146 idx = (idx + 1) % buffers; 163 idx = (idx + 1) % buffers;
153 } 170 }
154 171
155 return buf [idx]; 172 return buf [idx];
156} 173}
157 174
175static const char * ecb_noinline
176cv_get_name (CV *cv)
177{
178 static char fullname [256];
179
180 GV *gv = CvGV (cv); // gv better be nonzero
181
182 HV *stash = GvSTASH (gv);
183 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
184 const char *gvname = GvNAME (gv);
185
186 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
187 return fullname;
188}
189
158/*****************************************************************************/ 190/*****************************************************************************/
159 191
160typedef struct 192typedef struct
161{ 193{
162 IV iv; 194 IV iv;
163 const char *name; 195 const char *name;
164 #define const_iv(name) { (IV)CL_ ## name, # name }, 196 #define const_iv(name) { (IV)CL_ ## name, # name },
165} ivstr; 197} ivstr;
198
199typedef struct
200{
201 NV nv;
202 const char *name;
203 #define const_nv(name) { (NV)CL_ ## name, # name },
204} nvstr;
166 205
167static const char * 206static const char *
168iv2str (IV value, const ivstr *base, int count, const char *fallback) 207iv2str (IV value, const ivstr *base, int count, const char *fallback)
169{ 208{
170 int i; 209 int i;
219 if (res) \ 258 if (res) \
220 FAIL (name); 259 FAIL (name);
221 260
222/*****************************************************************************/ 261/*****************************************************************************/
223 262
263static SV *
264new_clobj (HV *stash, IV id)
265{
266 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
267}
268
269#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
270#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
271
272/* cl objects are either \$iv, or [$iv, ...] */
273/* they can be upgraded at runtime to the array form */
274static void * ecb_noinline
275SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
276{
277 // sv_derived_from is quite slow :(
278 if (SvROK (sv) && sv_derived_from (sv, pkg))
279 return (void *)SvIV (SvRV (sv));
280
281 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
282}
283
284// the "no-inherit" version of the above
285static void * ecb_noinline
286SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
287{
288 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
289 return (void *)SvIV (SvRV (sv));
290
291 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
292}
293
294/*****************************************************************************/
295
224static cl_context_properties * 296static cl_context_properties * ecb_noinline
225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 297SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
226{ 298{
227 if (!sv || !SvOK (sv)) 299 if (!sv || !SvOK (sv))
228 if (extra) 300 if (extra)
229 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 301 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
230 else 302 else
236 int i, len = av_len (av) + 1; 308 int i, len = av_len (av) + 1;
237 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 309 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
238 cl_context_properties *l = p; 310 cl_context_properties *l = p;
239 311
240 if (len & 1) 312 if (len & 1)
241 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 313 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
242 314
243 while (extracount--) 315 while (extracount--)
244 *l++ = *extra++; 316 *l++ = *extra++;
245 317
246 for (i = 0; i < len; i += 2) 318 for (i = 0; i < len; i += 2)
249 SV *p_sv = *av_fetch (av, i + 1, 0); 321 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override 322 cl_context_properties v = SvIV (p_sv); // code below can override
251 323
252 switch (t) 324 switch (t)
253 { 325 {
326 case CL_CONTEXT_PLATFORM:
327 if (SvROK (p_sv))
328 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
329 break;
330
254 case CL_GLX_DISPLAY_KHR: 331 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv)) 332 if (!SvOK (p_sv))
256 { 333 {
257 void *func = glsym ("_glXGetCurrentDisplay"); 334 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func) 335 if (func)
281 *l = 0; 358 *l = 0;
282 359
283 return p; 360 return p;
284 } 361 }
285 362
286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 363 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
287} 364}
288 365
289/*****************************************************************************/ 366// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
290 367// is a pointer (and only on well-behaved systems).
291static SV * 368static void * ecb_noinline
292new_clobj (HV *stash, IV id) 369object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
293{ 370{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash)); 371 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
295} 372 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
373 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
296 374
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id))) 375 AV *av = (AV *)SvRV (arg);
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id))) 376 void **list = 0;
377 cl_uint count = av_len (av) + 1;
299 378
300/* cl objects are either \$iv, or [$iv, ...] */ 379 if (count)
301/* they can be upgraded at runtime to the array form */ 380 {
302static void * 381 list = tmpbuf (sizeof (*list) * count);
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 382 int i;
304{ 383 for (i = 0; i < count; ++i)
305 if (SvROK (sv) && sv_derived_from (sv, pkg)) 384 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
306 return (void *)SvIV (SvRV (sv)); 385 }
307 386
308 croak ("%s: %s is not of type %s", func, svname, pkg); 387 if (!count && !or_undef)
388 croak ("%s: '%s' must contain at least one %s object",
389 cv_get_name (cv), argname, klass);
390
391 *rcount = count;
392 return (void *)list;
309} 393}
310 394
311/*****************************************************************************/ 395/*****************************************************************************/
312/* callback stuff */ 396/* callback stuff */
313 397
335static void (*eq_signal_func)(void *signal_arg, int value); 419static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg; 420static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT; 421static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail; 422static eq_item *eq_head, *eq_tail;
339 423
340static void 424static void ecb_noinline
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 425eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{ 426{
343 eq_item *item = malloc (sizeof (eq_item)); 427 eq_item *item = malloc (sizeof (eq_item));
344 428
345 item->next = 0; 429 item->next = 0;
412} 496}
413 497
414/*****************************************************************************/ 498/*****************************************************************************/
415/* context notify */ 499/* context notify */
416 500
417static void 501static void ecb_noinline
418eq_context_push (void *data1, void *data2, void *data3) 502eq_context_push (void *data1, void *data2, void *data3)
419{ 503{
420 dSP; 504 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0))); 505 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 506 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
444 { \ 528 { \
445 pfn_notify = eq_context_notify; \ 529 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \ 530 user_data = s_get_cv (notify); \
447 } 531 }
448 532
449static SV * 533static SV * ecb_noinline
450new_clobj_context (cl_context ctx, void *user_data) 534new_clobj_context (cl_context ctx, void *user_data)
451{ 535{
452 SV *sv = new_clobj (stash_context, (IV)ctx); 536 SV *sv = new_clobj (stash_context, (IV)ctx);
453 537
454 if (user_data) 538 if (user_data)
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 557static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474 558
475static void CL_CALLBACK 559static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data) 560eq_program_notify (cl_program program, void *user_data)
477{ 561{
562 clRetainProgram (program);
563
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 564 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
565}
566
567typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
568
569static program_callback ecb_noinline
570make_program_callback (SV *notify, void **ruser_data)
571{
572 if (SvOK (notify))
573 {
574 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
575 return eq_program_notify;
576 }
577 else
578 {
579 *ruser_data = 0;
580 return 0;
581 }
479} 582}
480 583
481struct build_args 584struct build_args
482{ 585{
483 cl_program program; 586 cl_program program;
497 else 600 else
498 clReleaseProgram (arg->program); 601 clReleaseProgram (arg->program);
499 602
500 free (arg->options); 603 free (arg->options);
501 free (arg); 604 free (arg);
605
606 return 0;
502} 607}
503 608
504static void 609static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 610build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{ 611{
515 xthread_t id; 620 xthread_t id;
516 thread_create (&id, build_program_thread, arg); 621 thread_create (&id, build_program_thread, arg);
517} 622}
518 623
519/*****************************************************************************/ 624/*****************************************************************************/
625/* mem object destructor notify */
626
627static void ecb_noinline
628eq_destructor_push (void *data1, void *data2, void *data3)
629{
630}
631
632static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
633
634static void CL_CALLBACK
635eq_destructor_notify (cl_mem memobj, void *user_data)
636{
637 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
638}
639
640/*****************************************************************************/
520/* event objects */ 641/* event objects */
521 642
522static void 643static void
523eq_event_push (void *data1, void *data2, void *data3) 644eq_event_push (void *data1, void *data2, void *data3)
524{ 645{
536 clRetainEvent (event); 657 clRetainEvent (event);
537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 658 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
538} 659}
539 660
540/*****************************************************************************/ 661/*****************************************************************************/
541/* mapped_xxx */ 662/* utilities for XS code */
542
543static OpenCL__Mapped
544SvMAPPED (SV *sv)
545{
546 // no typechecking atm., keep your fingers crossed
547 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
548}
549
550struct mapped
551{
552 cl_command_queue queue;
553 cl_mem memobj;
554 void *ptr;
555 size_t cb;
556 cl_event event;
557 size_t row_pitch;
558 size_t slice_pitch;
559};
560
561static SV *
562mapped_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)
563{
564 SV *data = newSV (0);
565 SvUPGRADE (data, SVt_PVMG);
566
567 OpenCL__Mapped mapped;
568 New (0, mapped, 1, struct mapped);
569
570 clRetainCommandQueue (queue);
571
572 mapped->queue = queue;
573 mapped->memobj = memobj;
574 mapped->ptr = ptr;
575 mapped->cb = cb;
576 mapped->event = ev;
577 mapped->row_pitch = row_pitch;
578 mapped->slice_pitch = slice_pitch;
579
580 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
581
582 if (!(flags & CL_MAP_WRITE))
583 SvREADONLY_on (data);
584
585 if (SvLEN (data))
586 Safefree (data);
587
588 SvPVX (data) = (char *)ptr;
589 SvCUR_set (data, cb);
590 SvLEN_set (data, 0);
591 SvPOK_only (data);
592
593 return sv_2mortal (sv_bless (newRV_noinc (data), stash));
594}
595
596static void
597mapped_detach (SV *sv, OpenCL__Mapped mapped)
598{
599 SV *data = SvRV (sv);
600
601 if (SvPVX (data) != (char *)mapped->ptr)
602 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
603 else
604 {
605 SvREADONLY_off (data);
606 SvCUR_set (data, 0);
607 SvPVX (data) = 0;
608 SvOK_off (data);
609 }
610
611 mapped->ptr = 0;
612}
613
614/*****************************************************************************/
615 663
616static size_t 664static size_t
617img_row_pitch (cl_mem img) 665img_row_pitch (cl_mem img)
618{ 666{
619 size_t res; 667 size_t res;
620 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 668 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
621 return res; 669 return res;
622} 670}
623 671
624static cl_event * 672static cl_event * ecb_noinline
625event_list (SV **items, cl_uint *rcount, cl_event extra) 673event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
626{ 674{
627 cl_uint count = *rcount; 675 cl_uint count = *rcount;
676
677 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
678 *rcount = count = 0;
628 679
629 if (!count && !extra) 680 if (!count && !extra)
630 return 0; 681 return 0;
631 682
632 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 683 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
633 int i = 0; 684 int i = 0;
634 685
635 while (count) 686 while (count--)
636 {
637 --count;
638 if (SvOK (items [count])) 687 if (SvOK (items [count]))
639 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 688 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
640 }
641 689
642 if (extra) 690 if (extra)
643 list [i++] = extra; 691 list [i++] = extra;
644 692
645 *rcount = i; 693 *rcount = i;
646 694
647 return i ? list : 0; 695 return i ? list : 0;
648} 696}
649 697
650#define EVENT_LIST(skip) \ 698#define EVENT_LIST(skip) \
651 cl_uint event_list_count = items - skip; \ 699 cl_uint event_list_count = items - (skip); \
652 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 700 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
653 701
654#define INFO(class) \ 702#define INFO(class) \
655{ \ 703{ \
656 size_t size; \ 704 size_t size; \
657 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 705 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
661 SvCUR_set (sv, size); \ 709 SvCUR_set (sv, size); \
662 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 710 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
663 XPUSHs (sv); \ 711 XPUSHs (sv); \
664} 712}
665 713
714/*****************************************************************************/
715/* mapped_xxx */
716
717static OpenCL__Mapped
718SvMAPPED (SV *sv)
719{
720 // no typechecking atm., keep your fingers crossed
721 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
722}
723
724struct mapped
725{
726 cl_command_queue queue;
727 cl_mem memobj;
728 void *ptr;
729 size_t cb;
730 cl_event event;
731 size_t row_pitch;
732 size_t slice_pitch;
733
734 size_t element_size;
735 size_t width, height, depth;
736};
737
738static SV *
739mapped_new (
740 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
741 void *ptr, size_t cb, cl_event ev,
742 size_t row_pitch, size_t slice_pitch, size_t element_size,
743 size_t width, size_t height, size_t depth
744)
745{
746 SV *data = newSV (0);
747 SvUPGRADE (data, SVt_PVMG);
748
749 OpenCL__Mapped mapped;
750 New (0, mapped, 1, struct mapped);
751
752 clRetainCommandQueue (queue);
753
754 mapped->queue = queue;
755 mapped->memobj = memobj;
756 mapped->ptr = ptr;
757 mapped->cb = cb;
758 mapped->event = ev;
759 mapped->row_pitch = row_pitch;
760 mapped->slice_pitch = slice_pitch;
761
762 mapped->element_size = element_size;
763 mapped->width = width;
764 mapped->height = height;
765 mapped->depth = depth;
766
767 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
768
769 if (SvLEN (data))
770 Safefree (data);
771
772 SvPVX (data) = (char *)ptr;
773 SvCUR_set (data, cb);
774 SvLEN_set (data, 0);
775 SvPOK_only (data);
776
777 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
778
779 if (!(flags & CL_MAP_WRITE))
780 SvREADONLY_on (data);
781
782 return obj;
783}
784
785static void
786mapped_detach (SV *sv, OpenCL__Mapped mapped)
787{
788 SV *data = SvRV (sv);
789
790 // the next check checks both before AND after detach, where SvPVX should STILL be 0
791 if (SvPVX (data) != (char *)mapped->ptr)
792 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
793 else
794 {
795 SvREADONLY_off (data);
796 SvCUR_set (data, 0);
797 SvPVX (data) = 0;
798 SvOK_off (data);
799 }
800
801 mapped->ptr = 0;
802}
803
804static void
805mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
806{
807 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
808 cl_event ev;
809
810 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
811
812 clReleaseEvent (mapped->event);
813 mapped->event = ev;
814
815 mapped_detach (self, mapped);
816}
817
818static size_t
819mapped_element_size (OpenCL__Mapped self)
820{
821 if (!self->element_size)
822 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
823
824 return self->element_size;
825}
826
827/*****************************************************************************/
828
666MODULE = OpenCL PACKAGE = OpenCL 829MODULE = OpenCL PACKAGE = OpenCL
667 830
668PROTOTYPES: ENABLE 831PROTOTYPES: ENABLE
669 832
670void 833void
679 eq_signal_arg = (void*)arg; 842 eq_signal_arg = (void*)arg;
680 843
681BOOT: 844BOOT:
682{ 845{
683 HV *stash = gv_stashpv ("OpenCL", 1); 846 HV *stash = gv_stashpv ("OpenCL", 1);
847
684 static const ivstr *civ, const_iv[] = { 848 static const ivstr *civ, const_iv[] = {
685 { sizeof (cl_char ), "SIZEOF_CHAR" }, 849 { sizeof (cl_char ), "SIZEOF_CHAR" },
686 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 850 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
687 { sizeof (cl_short ), "SIZEOF_SHORT" }, 851 { sizeof (cl_short ), "SIZEOF_SHORT" },
688 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 852 { sizeof (cl_ushort), "SIZEOF_USHORT" },
691 { sizeof (cl_long ), "SIZEOF_LONG" }, 855 { sizeof (cl_long ), "SIZEOF_LONG" },
692 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 856 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
693 { sizeof (cl_half ), "SIZEOF_HALF" }, 857 { sizeof (cl_half ), "SIZEOF_HALF" },
694 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 858 { sizeof (cl_float ), "SIZEOF_FLOAT" },
695 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 859 { sizeof (cl_double), "SIZEOF_DOUBLE" },
860 { PREFER_1_1 , "PREFER_1_1" },
696#include "constiv.h" 861#include "constiv.h"
697 }; 862 };
698 863
699 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 864 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
700 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 865 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
701 866
867 static const nvstr *cnv, const_nv[] = {
868#include "constnv.h"
869 };
870
871 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
872 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
873
874 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
875
702 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 876 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
703 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 877 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
878 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
704 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 879 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
705 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 880 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
706 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 881 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
707 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 882 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
708 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 883 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
749 list = tmpbuf (sizeof (*list) * count); 924 list = tmpbuf (sizeof (*list) * count);
750 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 925 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
751 926
752 EXTEND (SP, count); 927 EXTEND (SP, count);
753 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
754 PUSH_CLOBJ (stash_platform, (IV)list [i]); 929 PUSH_CLOBJ (stash_platform, list [i]);
755 930
756void 931void
757context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 932context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
758 PPCODE: 933 PPCODE:
759 CONTEXT_NOTIFY_CALLBACK; 934 CONTEXT_NOTIFY_CALLBACK;
760 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 935 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
761 XPUSH_CLOBJ_CONTEXT; 936 XPUSH_CLOBJ_CONTEXT;
762 937
763void 938void
764context (FUTURE properties, FUTURE devices, FUTURE notify) 939context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
765 PPCODE: 940 PPCODE:
766 /* der Gipfel der Kunst */ 941 cl_uint device_count;
942 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
943
944 CONTEXT_NOTIFY_CALLBACK;
945 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
946 XPUSH_CLOBJ_CONTEXT;
767 947
768void 948void
769wait_for_events (...) 949wait_for_events (...)
770 CODE: 950 CODE:
771 EVENT_LIST (0); 951 EVENT_LIST (0);
822 EXTEND (SP, count); 1002 EXTEND (SP, count);
823 for (i = 0; i < count; ++i) 1003 for (i = 0; i < count; ++i)
824 PUSH_CLOBJ (stash_device, list [i]); 1004 PUSH_CLOBJ (stash_device, list [i]);
825 1005
826void 1006void
827context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 1007context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
828 PPCODE: 1008 PPCODE:
1009 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1010 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1011
1012 cl_uint device_count;
1013 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1014
1015 CONTEXT_NOTIFY_CALLBACK;
1016 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1017 XPUSH_CLOBJ_CONTEXT;
1018
1019void
1020context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1021 PPCODE:
1022 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1023 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1024
1025 CONTEXT_NOTIFY_CALLBACK;
1026 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1027 XPUSH_CLOBJ_CONTEXT;
1028
1029MODULE = OpenCL PACKAGE = OpenCL::Device
1030
1031void
1032info (OpenCL::Device self, cl_device_info name)
1033 PPCODE:
1034 INFO (Device)
1035
1036#if CL_VERSION_1_2
1037
1038void
1039sub_devices (OpenCL::Device self, SV *properties)
1040 PPCODE:
829 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1041 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
830 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1042 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
831 1043
832 AV *av = (AV *)SvRV (devices); 1044 properties = SvRV (properties);
833 cl_uint num_devices = av_len (av) + 1; 1045
834 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1046 cl_uint count = av_len ((AV *)properties) + 1;
1047 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
835 1048
836 int i; 1049 int i;
837 for (i = num_devices; i--; ) 1050 for (i = 0; i < count; ++i)
838 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1051 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
839 1052
840 CONTEXT_NOTIFY_CALLBACK; 1053 props [count] = 0;
841 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
842 XPUSH_CLOBJ_CONTEXT;
843 1054
844void 1055 cl_uint num_devices;
845context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 1056 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
846 PPCODE: 1057 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
847 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1058 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
848 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
849 1059
850 CONTEXT_NOTIFY_CALLBACK; 1060 EXTEND (SP, num_devices);
851 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1061 for (i = 0; i < count; ++i)
852 XPUSH_CLOBJ_CONTEXT; 1062 PUSH_CLOBJ (stash_subdevice, list [i]);
853 1063
854MODULE = OpenCL PACKAGE = OpenCL::Device 1064#endif
855
856void
857info (OpenCL::Device self, cl_device_info name)
858 PPCODE:
859 INFO (Device)
860 1065
861#BEGIN:device 1066#BEGIN:device
862 1067
863void 1068void
864type (OpenCL::Device self) 1069type (OpenCL::Device self)
1070 ALIAS:
1071 type = CL_DEVICE_TYPE
1072 address_bits = CL_DEVICE_ADDRESS_BITS
1073 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1074 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1075 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1076 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1077 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1078 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1079 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1080 properties = CL_DEVICE_QUEUE_PROPERTIES
1081 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1082 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
865 PPCODE: 1083 PPCODE:
866 cl_device_type value [1]; 1084 cl_ulong value [1];
867 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1085 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
868 EXTEND (SP, 1); 1086 EXTEND (SP, 1);
869 const int i = 0; 1087 const int i = 0;
870 PUSHs (sv_2mortal (newSVuv (value [i]))); 1088 PUSHs (sv_2mortal (newSVuv (value [i])));
871 1089
872void 1090void
886 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1104 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
887 image_support = CL_DEVICE_IMAGE_SUPPORT 1105 image_support = CL_DEVICE_IMAGE_SUPPORT
888 max_samplers = CL_DEVICE_MAX_SAMPLERS 1106 max_samplers = CL_DEVICE_MAX_SAMPLERS
889 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1107 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
890 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1108 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1109 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
891 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1110 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
892 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1111 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1112 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
893 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1113 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
894 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1114 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
895 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1115 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
896 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1116 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
897 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1117 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
935 EXTEND (SP, n); 1155 EXTEND (SP, n);
936 for (i = 0; i < n; ++i) 1156 for (i = 0; i < n; ++i)
937 PUSHs (sv_2mortal (newSVuv (value [i]))); 1157 PUSHs (sv_2mortal (newSVuv (value [i])));
938 1158
939void 1159void
940address_bits (OpenCL::Device self)
941 PPCODE:
942 cl_bitfield value [1];
943 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
944 EXTEND (SP, 1);
945 const int i = 0;
946 PUSHs (sv_2mortal (newSVuv (value [i])));
947
948void
949max_mem_alloc_size (OpenCL::Device self)
950 ALIAS:
951 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
952 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
953 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
954 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
955 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
956 PPCODE:
957 cl_ulong value [1];
958 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
959 EXTEND (SP, 1);
960 const int i = 0;
961 PUSHs (sv_2mortal (newSVuv (value [i])));
962
963void
964single_fp_config (OpenCL::Device self)
965 ALIAS:
966 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
967 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
968 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
969 PPCODE:
970 cl_device_fp_config value [1];
971 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
972 EXTEND (SP, 1);
973 const int i = 0;
974 PUSHs (sv_2mortal (newSVuv (value [i])));
975
976void
977global_mem_cache_type (OpenCL::Device self)
978 PPCODE:
979 cl_device_mem_cache_type value [1];
980 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
981 EXTEND (SP, 1);
982 const int i = 0;
983 PUSHs (sv_2mortal (newSVuv (value [i])));
984
985void
986local_mem_type (OpenCL::Device self)
987 PPCODE:
988 cl_device_local_mem_type value [1];
989 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
990 EXTEND (SP, 1);
991 const int i = 0;
992 PUSHs (sv_2mortal (newSVuv (value [i])));
993
994void
995error_correction_support (OpenCL::Device self) 1160error_correction_support (OpenCL::Device self)
996 ALIAS: 1161 ALIAS:
997 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1162 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
998 endian_little = CL_DEVICE_ENDIAN_LITTLE 1163 endian_little = CL_DEVICE_ENDIAN_LITTLE
999 available = CL_DEVICE_AVAILABLE 1164 available = CL_DEVICE_AVAILABLE
1000 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1165 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1001 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1166 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1002 PPCODE: 1167 PPCODE:
1003 cl_bool value [1]; 1168 cl_uint value [1];
1004 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1169 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1005 EXTEND (SP, 1); 1170 EXTEND (SP, 1);
1006 const int i = 0; 1171 const int i = 0;
1007 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1172 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1008
1009void
1010execution_capabilities (OpenCL::Device self)
1011 PPCODE:
1012 cl_device_exec_capabilities value [1];
1013 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1014 EXTEND (SP, 1);
1015 const int i = 0;
1016 PUSHs (sv_2mortal (newSVuv (value [i])));
1017
1018void
1019properties (OpenCL::Device self)
1020 PPCODE:
1021 cl_command_queue_properties value [1];
1022 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1023 EXTEND (SP, 1);
1024 const int i = 0;
1025 PUSHs (sv_2mortal (newSVuv (value [i])));
1026 1173
1027void 1174void
1028platform (OpenCL::Device self) 1175platform (OpenCL::Device self)
1029 PPCODE: 1176 PPCODE:
1030 cl_platform_id value [1]; 1177 cl_platform_id value [1];
1031 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1178 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1032 EXTEND (SP, 1); 1179 EXTEND (SP, 1);
1033 const int i = 0; 1180 const int i = 0;
1034 {
1035 PUSH_CLOBJ (stash_platform, value [i]); 1181 PUSH_CLOBJ (stash_platform, value [i]);
1036 }
1037 1182
1038void 1183void
1039name (OpenCL::Device self) 1184name (OpenCL::Device self)
1040 ALIAS: 1185 ALIAS:
1041 name = CL_DEVICE_NAME 1186 name = CL_DEVICE_NAME
1058 PPCODE: 1203 PPCODE:
1059 cl_device_id value [1]; 1204 cl_device_id value [1];
1060 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1205 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1061 EXTEND (SP, 1); 1206 EXTEND (SP, 1);
1062 const int i = 0; 1207 const int i = 0;
1063 {
1064 PUSH_CLOBJ (stash_device, value [i]); 1208 PUSH_CLOBJ (stash_device, value [i]);
1065 }
1066 1209
1067void 1210void
1068partition_types_ext (OpenCL::Device self) 1211partition_types_ext (OpenCL::Device self)
1069 ALIAS: 1212 ALIAS:
1070 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1213 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1080 for (i = 0; i < n; ++i) 1223 for (i = 0; i < n; ++i)
1081 PUSHs (sv_2mortal (newSVuv (value [i]))); 1224 PUSHs (sv_2mortal (newSVuv (value [i])));
1082 1225
1083#END:device 1226#END:device
1084 1227
1228MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1229
1230#if CL_VERSION_1_2
1231
1232void
1233DESTROY (OpenCL::SubDevice self)
1234 CODE:
1235 clReleaseDevice (self);
1236
1237#endif
1238
1085MODULE = OpenCL PACKAGE = OpenCL::Context 1239MODULE = OpenCL PACKAGE = OpenCL::Context
1086 1240
1087void 1241void
1088DESTROY (OpenCL::Context context) 1242DESTROY (OpenCL::Context self)
1089 CODE: 1243 CODE:
1090 clReleaseContext (context); 1244 clReleaseContext (self);
1091 1245
1092void 1246void
1093info (OpenCL::Context self, cl_context_info name) 1247info (OpenCL::Context self, cl_context_info name)
1094 PPCODE: 1248 PPCODE:
1095 INFO (Context) 1249 INFO (Context)
1136 const cl_image_desc desc = { 1290 const cl_image_desc desc = {
1137 type, 1291 type,
1138 width, height, depth, 1292 width, height, depth,
1139 array_size, row_pitch, slice_pitch, 1293 array_size, row_pitch, slice_pitch,
1140 num_mip_level, num_samples, 1294 num_mip_level, num_samples,
1141 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1295 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1142 }; 1296 };
1143 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1297 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1144 HV *stash = stash_image; 1298 HV *stash = stash_image;
1145 switch (type) 1299 switch (type)
1146 { 1300 {
1279 1433
1280 len2 = len; 1434 len2 = len;
1281 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1435 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1282 XPUSH_CLOBJ (stash_program, prog); 1436 XPUSH_CLOBJ (stash_program, prog);
1283 1437
1438void
1439program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1440 PPCODE:
1441 cl_uint device_count;
1442 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1443
1444 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1445 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1446
1447 binaries = SvRV (binaries);
1448
1449 if (device_count != av_len ((AV *)binaries) + 1)
1450 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1451
1452 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1453 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1454 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1455
1456 int i;
1457 for (i = 0; i < device_count; ++i)
1458 {
1459 STRLEN len;
1460 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1461 length_list [i] = len;
1462 }
1463
1464 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1465 length_list, binary_list,
1466 GIMME_V == G_ARRAY ? status_list : 0, &res));
1467
1468 EXTEND (SP, 2);
1469 PUSH_CLOBJ (stash_program, prog);
1470
1471 if (GIMME_V == G_ARRAY)
1472 {
1473 AV *av = newAV ();
1474 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1475
1476 for (i = device_count; i--; )
1477 av_store (av, i, newSViv (status_list [i]));
1478 }
1479
1480#if CL_VERSION_1_2
1481
1482void
1483program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1484 PPCODE:
1485 cl_uint device_count;
1486 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1489
1490 XPUSH_CLOBJ (stash_program, prog);
1491
1492void
1493link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1494 CODE:
1495 cl_uint device_count = 0;
1496 cl_device_id *device_list = 0;
1497
1498 if (SvOK (devices))
1499 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1500
1501 cl_uint program_count;
1502 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1503
1504 void *user_data;
1505 program_callback pfn_notify = make_program_callback (notify, &user_data);
1506
1507 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1508 program_count, program_list, pfn_notify, user_data, &res));
1509
1510 XPUSH_CLOBJ (stash_program, prog);
1511
1512#endif
1513
1284#BEGIN:context 1514#BEGIN:context
1285 1515
1286void 1516void
1287reference_count (OpenCL::Context self) 1517reference_count (OpenCL::Context self)
1288 ALIAS: 1518 ALIAS:
1303 cl_device_id *value = tmpbuf (size); 1533 cl_device_id *value = tmpbuf (size);
1304 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1534 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1305 int i, n = size / sizeof (*value); 1535 int i, n = size / sizeof (*value);
1306 EXTEND (SP, n); 1536 EXTEND (SP, n);
1307 for (i = 0; i < n; ++i) 1537 for (i = 0; i < n; ++i)
1308 {
1309 PUSH_CLOBJ (stash_device, value [i]); 1538 PUSH_CLOBJ (stash_device, value [i]);
1310 }
1311 1539
1312void 1540void
1313properties (OpenCL::Context self) 1541properties (OpenCL::Context self)
1314 PPCODE: 1542 PPCODE:
1315 size_t size; 1543 size_t size;
1333void 1561void
1334read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1562read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1335 ALIAS: 1563 ALIAS:
1336 enqueue_read_buffer = 0 1564 enqueue_read_buffer = 0
1337 PPCODE: 1565 PPCODE:
1338 cl_event ev = 0;
1339 EVENT_LIST (6); 1566 EVENT_LIST (6);
1340 1567
1341 SvUPGRADE (data, SVt_PV); 1568 SvUPGRADE (data, SVt_PV);
1342 SvGROW (data, len); 1569 SvGROW (data, len);
1343 SvPOK_only (data); 1570 SvPOK_only (data);
1344 SvCUR_set (data, len); 1571 SvCUR_set (data, len);
1572
1573 cl_event ev = 0;
1345 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1574 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1346 1575
1347 if (ev) 1576 if (ev)
1348 XPUSH_CLOBJ (stash_event, ev); 1577 XPUSH_CLOBJ (stash_event, ev);
1349 1578
1350void 1579void
1351write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1580write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1352 ALIAS: 1581 ALIAS:
1353 enqueue_write_buffer = 0 1582 enqueue_write_buffer = 0
1354 PPCODE: 1583 PPCODE:
1355 cl_event ev = 0; 1584 EVENT_LIST (5);
1585
1356 STRLEN len; 1586 STRLEN len;
1357 char *ptr = SvPVbyte (data, len); 1587 char *ptr = SvPVbyte (data, len);
1358 EVENT_LIST (5);
1359 1588
1589 cl_event ev = 0;
1360 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1590 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1361 1591
1362 if (ev) 1592 if (ev)
1363 XPUSH_CLOBJ (stash_event, ev); 1593 XPUSH_CLOBJ (stash_event, ev);
1364 1594
1367void 1597void
1368fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1598fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1369 ALIAS: 1599 ALIAS:
1370 enqueue_fill_buffer = 0 1600 enqueue_fill_buffer = 0
1371 PPCODE: 1601 PPCODE:
1372 cl_event ev = 0; 1602 EVENT_LIST (5);
1603
1373 STRLEN len; 1604 STRLEN len;
1374 char *ptr = SvPVbyte (data, len); 1605 char *ptr = SvPVbyte (data, len);
1375 EVENT_LIST (5);
1376 1606
1607 cl_event ev = 0;
1377 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1608 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1378 1609
1379 if (ev) 1610 if (ev)
1380 XPUSH_CLOBJ (stash_event, ev); 1611 XPUSH_CLOBJ (stash_event, ev);
1381 1612
1382void 1613void
1383fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 1614fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1384 ALIAS: 1615 ALIAS:
1385 enqueue_fill_image = 0 1616 enqueue_fill_image = 0
1386 PPCODE: 1617 PPCODE:
1387 cl_event ev = 0; 1618 EVENT_LIST (12);
1388 STRLEN len; 1619
1389 const size_t origin [3] = { x, y, z }; 1620 const size_t origin [3] = { x, y, z };
1390 const size_t region [3] = { width, height, depth }; 1621 const size_t region [3] = { width, height, depth };
1391 EVENT_LIST (12);
1392 1622
1393 const cl_float c_f [4] = { r, g, b, a }; 1623 const cl_float c_f [4] = { r, g, b, a };
1394 const cl_uint c_u [4] = { r, g, b, a }; 1624 const cl_uint c_u [4] = { r, g, b, a };
1395 const cl_int c_s [4] = { r, g, b, a }; 1625 const cl_int c_s [4] = { r, g, b, a };
1396 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1626 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1399 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1629 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1400 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1630 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1401 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1631 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1402 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1632 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1403 1633
1634 cl_event ev = 0;
1404 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1635 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1405 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1636 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1406 1637
1407 if (ev) 1638 if (ev)
1408 XPUSH_CLOBJ (stash_event, ev); 1639 XPUSH_CLOBJ (stash_event, ev);
1412void 1643void
1413copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1644copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1414 ALIAS: 1645 ALIAS:
1415 enqueue_copy_buffer = 0 1646 enqueue_copy_buffer = 0
1416 PPCODE: 1647 PPCODE:
1648 EVENT_LIST (6);
1649
1417 cl_event ev = 0; 1650 cl_event ev = 0;
1418 EVENT_LIST (6);
1419
1420 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1651 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1421 1652
1422 if (ev) 1653 if (ev)
1423 XPUSH_CLOBJ (stash_event, ev); 1654 XPUSH_CLOBJ (stash_event, ev);
1424 1655
1425void 1656void
1426read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1657read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1427 ALIAS: 1658 ALIAS:
1428 enqueue_read_buffer_rect = 0 1659 enqueue_read_buffer_rect = 0
1429 PPCODE: 1660 PPCODE:
1430 cl_event ev = 0; 1661 EVENT_LIST (17);
1662
1431 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1663 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1432 const size_t host_origin[3] = { host_x, host_y, host_z }; 1664 const size_t host_origin[3] = { host_x, host_y, host_z };
1433 const size_t region[3] = { width, height, depth }; 1665 const size_t region[3] = { width, height, depth };
1434 EVENT_LIST (17);
1435 1666
1436 if (!buf_row_pitch) 1667 if (!buf_row_pitch)
1437 buf_row_pitch = region [0]; 1668 buf_row_pitch = region [0];
1438 1669
1439 if (!buf_slice_pitch) 1670 if (!buf_slice_pitch)
1449 1680
1450 SvUPGRADE (data, SVt_PV); 1681 SvUPGRADE (data, SVt_PV);
1451 SvGROW (data, len); 1682 SvGROW (data, len);
1452 SvPOK_only (data); 1683 SvPOK_only (data);
1453 SvCUR_set (data, len); 1684 SvCUR_set (data, len);
1685
1686 cl_event ev = 0;
1454 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1687 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1455 1688
1456 if (ev) 1689 if (ev)
1457 XPUSH_CLOBJ (stash_event, ev); 1690 XPUSH_CLOBJ (stash_event, ev);
1458 1691
1459void 1692void
1460write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1693write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1461 ALIAS: 1694 ALIAS:
1462 enqueue_write_buffer_rect = 0 1695 enqueue_write_buffer_rect = 0
1463 PPCODE: 1696 PPCODE:
1464 cl_event ev = 0; 1697 EVENT_LIST (17);
1698
1465 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1699 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1466 const size_t host_origin[3] = { host_x, host_y, host_z }; 1700 const size_t host_origin[3] = { host_x, host_y, host_z };
1467 const size_t region[3] = { width, height, depth }; 1701 const size_t region[3] = { width, height, depth };
1468 STRLEN len; 1702 STRLEN len;
1469 char *ptr = SvPVbyte (data, len); 1703 char *ptr = SvPVbyte (data, len);
1470 EVENT_LIST (17);
1471 1704
1472 if (!buf_row_pitch) 1705 if (!buf_row_pitch)
1473 buf_row_pitch = region [0]; 1706 buf_row_pitch = region [0];
1474 1707
1475 if (!buf_slice_pitch) 1708 if (!buf_slice_pitch)
1484 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1717 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1485 1718
1486 if (len < min_len) 1719 if (len < min_len)
1487 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1720 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1488 1721
1722 cl_event ev = 0;
1489 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1723 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1490 1724
1491 if (ev) 1725 if (ev)
1492 XPUSH_CLOBJ (stash_event, ev); 1726 XPUSH_CLOBJ (stash_event, ev);
1493 1727
1494void 1728void
1495copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1729copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1496 ALIAS: 1730 ALIAS:
1497 enqueue_copy_buffer_rect = 0 1731 enqueue_copy_buffer_rect = 0
1498 PPCODE: 1732 PPCODE:
1499 cl_event ev = 0; 1733 EVENT_LIST (16);
1734
1500 const size_t src_origin[3] = { src_x, src_y, src_z }; 1735 const size_t src_origin[3] = { src_x, src_y, src_z };
1501 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1736 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1502 const size_t region[3] = { width, height, depth }; 1737 const size_t region[3] = { width, height, depth };
1503 EVENT_LIST (16);
1504 1738
1739 cl_event ev = 0;
1505 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1740 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1506 1741
1507 if (ev) 1742 if (ev)
1508 XPUSH_CLOBJ (stash_event, ev); 1743 XPUSH_CLOBJ (stash_event, ev);
1509 1744
1510void 1745void
1511read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1746read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1512 ALIAS: 1747 ALIAS:
1513 enqueue_read_image = 0 1748 enqueue_read_image = 0
1514 PPCODE: 1749 PPCODE:
1515 cl_event ev = 0; 1750 EVENT_LIST (12);
1751
1516 const size_t src_origin[3] = { src_x, src_y, src_z }; 1752 const size_t src_origin[3] = { src_x, src_y, src_z };
1517 const size_t region[3] = { width, height, depth }; 1753 const size_t region[3] = { width, height, depth };
1518 EVENT_LIST (12);
1519 1754
1520 if (!row_pitch) 1755 if (!row_pitch)
1521 row_pitch = img_row_pitch (src); 1756 row_pitch = img_row_pitch (src);
1522 1757
1523 if (depth > 1 && !slice_pitch) 1758 if (depth > 1 && !slice_pitch)
1527 1762
1528 SvUPGRADE (data, SVt_PV); 1763 SvUPGRADE (data, SVt_PV);
1529 SvGROW (data, len); 1764 SvGROW (data, len);
1530 SvPOK_only (data); 1765 SvPOK_only (data);
1531 SvCUR_set (data, len); 1766 SvCUR_set (data, len);
1767
1768 cl_event ev = 0;
1532 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1769 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1533 1770
1534 if (ev) 1771 if (ev)
1535 XPUSH_CLOBJ (stash_event, ev); 1772 XPUSH_CLOBJ (stash_event, ev);
1536 1773
1537void 1774void
1538write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1775write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1539 ALIAS: 1776 ALIAS:
1540 enqueue_write_image = 0 1777 enqueue_write_image = 0
1541 PPCODE: 1778 PPCODE:
1542 cl_event ev = 0; 1779 EVENT_LIST (12);
1780
1543 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1781 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1544 const size_t region[3] = { width, height, depth }; 1782 const size_t region[3] = { width, height, depth };
1545 STRLEN len; 1783 STRLEN len;
1546 char *ptr = SvPVbyte (data, len); 1784 char *ptr = SvPVbyte (data, len);
1547 EVENT_LIST (12);
1548 1785
1549 if (!row_pitch) 1786 if (!row_pitch)
1550 row_pitch = img_row_pitch (dst); 1787 row_pitch = img_row_pitch (dst);
1551 1788
1552 if (depth > 1 && !slice_pitch) 1789 if (depth > 1 && !slice_pitch)
1555 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1792 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1556 1793
1557 if (len < min_len) 1794 if (len < min_len)
1558 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1795 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1559 1796
1797 cl_event ev = 0;
1560 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1798 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1561 1799
1562 if (ev) 1800 if (ev)
1563 XPUSH_CLOBJ (stash_event, ev); 1801 XPUSH_CLOBJ (stash_event, ev);
1564 1802
1565void 1803void
1566copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1804copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1567 ALIAS: 1805 ALIAS:
1568 enqueue_copy_image = 0 1806 enqueue_copy_image = 0
1569 PPCODE: 1807 PPCODE:
1570 cl_event ev = 0; 1808 EVENT_LIST (12);
1809
1571 const size_t src_origin[3] = { src_x, src_y, src_z }; 1810 const size_t src_origin[3] = { src_x, src_y, src_z };
1572 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1811 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1573 const size_t region[3] = { width, height, depth }; 1812 const size_t region[3] = { width, height, depth };
1574 EVENT_LIST (12);
1575 1813
1814 cl_event ev = 0;
1576 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1815 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577 1816
1578 if (ev) 1817 if (ev)
1579 XPUSH_CLOBJ (stash_event, ev); 1818 XPUSH_CLOBJ (stash_event, ev);
1580 1819
1581void 1820void
1582copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1821copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1583 ALIAS: 1822 ALIAS:
1584 enqueue_copy_image_to_buffer = 0 1823 enqueue_copy_image_to_buffer = 0
1585 PPCODE: 1824 PPCODE:
1586 cl_event ev = 0; 1825 EVENT_LIST (10);
1826
1587 const size_t src_origin[3] = { src_x, src_y, src_z }; 1827 const size_t src_origin[3] = { src_x, src_y, src_z };
1588 const size_t region [3] = { width, height, depth }; 1828 const size_t region [3] = { width, height, depth };
1589 EVENT_LIST (10);
1590 1829
1830 cl_event ev = 0;
1591 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1831 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1592 1832
1593 if (ev) 1833 if (ev)
1594 XPUSH_CLOBJ (stash_event, ev); 1834 XPUSH_CLOBJ (stash_event, ev);
1595 1835
1596void 1836void
1597copy_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, ...) 1837copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1598 ALIAS: 1838 ALIAS:
1599 enqueue_copy_buffer_to_image = 0 1839 enqueue_copy_buffer_to_image = 0
1600 PPCODE: 1840 PPCODE:
1601 cl_event ev = 0; 1841 EVENT_LIST (10);
1842
1602 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1843 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1603 const size_t region [3] = { width, height, depth }; 1844 const size_t region [3] = { width, height, depth };
1604 EVENT_LIST (10);
1605 1845
1846 cl_event ev = 0;
1606 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1847 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1607 1848
1608 if (ev) 1849 if (ev)
1609 XPUSH_CLOBJ (stash_event, ev); 1850 XPUSH_CLOBJ (stash_event, ev);
1610 1851
1611void 1852void
1612map_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, size_t cb = 0, ...) 1853map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1613 ALIAS: 1854 ALIAS:
1614 enqueue_map_buffer = 0 1855 enqueue_map_buffer = 0
1615 PPCODE: 1856 PPCODE:
1857 EVENT_LIST (6);
1858
1859 size_t cb = SvIV (cb_);
1860
1861 if (!SvOK (cb_))
1862 {
1863 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1864 cb -= offset;
1865 }
1866
1616 cl_event ev; 1867 cl_event ev;
1617 EVENT_LIST (7);
1618
1619 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res)); 1868 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1620 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0)); 1869 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1621 1870
1622void 1871void
1623map_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, size_t width = 0, size_t height = 0, size_t depth = 0, ...) 1872map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1624 ALIAS: 1873 ALIAS:
1625 enqueue_map_image = 0 1874 enqueue_map_image = 0
1626 PPCODE: 1875 PPCODE:
1627 cl_event ev; 1876 size_t width = SvIV (width_);
1877 if (!SvOK (width_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1880 width -= x;
1881 }
1882
1883 size_t height = SvIV (width_);
1884 if (!SvOK (height_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1887 height -= y;
1888
1889 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1890 if (!height)
1891 height = 1;
1892 }
1893
1894 size_t depth = SvIV (width_);
1895 if (!SvOK (depth_))
1896 {
1897 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1898 depth -= z;
1899
1900 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1901 if (!depth)
1902 depth = 1;
1903 }
1904
1628 const size_t origin[3] = { x, y, z }; 1905 const size_t origin[3] = { x, y, z };
1629 const size_t region[3] = { width, height, depth }; 1906 const size_t region[3] = { width, height, depth };
1630 size_t row_pitch, slice_pitch; 1907 size_t row_pitch, slice_pitch;
1631 EVENT_LIST (11); 1908 EVENT_LIST (10);
1632 1909
1910 cl_event ev;
1633 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)); 1911 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1634 1912
1635 size_t cb = slice_pitch ? slice_pitch * region [2] 1913 size_t cb = slice_pitch ? slice_pitch * region [2]
1636 : row_pitch ? row_pitch * region [1] 1914 : row_pitch ? row_pitch * region [1]
1637 : region [0]; 1915 : region [0];
1638 1916
1639 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch)); 1917 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1640 1918
1641void 1919void
1642unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1920unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1643 PPCODE: 1921 PPCODE:
1644 cl_uint event_list_count = items - 2; 1922 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1645 cl_event *event_list_ptr = event_list (&ST (2), &event_list_count, mapped->event);
1646
1647 NEED_SUCCESS (EnqueueUnmapMemObject, (self, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &mapped->event));
1648 mapped_detach (ST (1), mapped);
1649
1650 if (GIMME_V != G_VOID) 1923 if (GIMME_V != G_VOID)
1924 {
1925 clRetainEvent (mapped->event);
1651 XPUSH_CLOBJ (stash_event, mapped->event); 1926 XPUSH_CLOBJ (stash_event, mapped->event);
1927 }
1652 1928
1653void 1929void
1654task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1930task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1655 ALIAS: 1931 ALIAS:
1656 enqueue_task = 0 1932 enqueue_task = 0
1657 PPCODE: 1933 PPCODE:
1934 EVENT_LIST (2);
1935
1658 cl_event ev = 0; 1936 cl_event ev = 0;
1659 EVENT_LIST (2);
1660
1661 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1937 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1662 1938
1663 if (ev) 1939 if (ev)
1664 XPUSH_CLOBJ (stash_event, ev); 1940 XPUSH_CLOBJ (stash_event, ev);
1665 1941
1666void 1942void
1667nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1943nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1668 ALIAS: 1944 ALIAS:
1669 enqueue_nd_range_kernel = 0 1945 enqueue_nd_range_kernel = 0
1670 PPCODE: 1946 PPCODE:
1671 cl_event ev = 0; 1947 EVENT_LIST (5);
1948
1672 size_t *gwo = 0, *gws, *lws = 0; 1949 size_t *gwo = 0, *gws, *lws = 0;
1673 int gws_len; 1950 int gws_len;
1674 size_t *lists; 1951 size_t *lists;
1675 int i; 1952 int i;
1676 EVENT_LIST (5);
1677 1953
1678 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1954 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1679 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1680 1956
1681 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1957 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1720 if (!lws [i]) 1996 if (!lws [i])
1721 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 1997 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1722 } 1998 }
1723 } 1999 }
1724 2000
2001 cl_event ev = 0;
1725 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2002 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1726 2003
1727 if (ev) 2004 if (ev)
1728 XPUSH_CLOBJ (stash_event, ev); 2005 XPUSH_CLOBJ (stash_event, ev);
2006
2007#if CL_VERSION_1_2
2008
2009void
2010migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2011 ALIAS:
2012 enqueue_migrate_mem_objects = 0
2013 PPCODE:
2014 EVENT_LIST (3);
2015
2016 cl_uint obj_count;
2017 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2018
2019 cl_event ev = 0;
2020 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2021
2022 if (ev)
2023 XPUSH_CLOBJ (stash_event, ev);
2024
2025#endif
1729 2026
1730#if cl_apple_gl_sharing || cl_khr_gl_sharing 2027#if cl_apple_gl_sharing || cl_khr_gl_sharing
1731 2028
1732void 2029void
1733acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2030acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1734 ALIAS: 2031 ALIAS:
1735 release_gl_objects = 1 2032 release_gl_objects = 1
1736 enqueue_acquire_gl_objects = 0 2033 enqueue_acquire_gl_objects = 0
1737 enqueue_release_gl_objects = 1 2034 enqueue_release_gl_objects = 1
1738 PPCODE: 2035 PPCODE:
1739 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2036 EVENT_LIST (2);
1740 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2037
2038 cl_uint obj_count;
2039 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1741 2040
1742 cl_event ev = 0; 2041 cl_event ev = 0;
1743 EVENT_LIST (2);
1744 AV *av = (AV *)SvRV (objects);
1745 cl_uint num_objects = av_len (av) + 1;
1746 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1747 int i;
1748
1749 for (i = num_objects; i--; )
1750 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1751 2042
1752 if (ix) 2043 if (ix)
1753 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2044 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1754 else 2045 else
1755 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1756 2047
1757 if (ev) 2048 if (ev)
1758 XPUSH_CLOBJ (stash_event, ev); 2049 XPUSH_CLOBJ (stash_event, ev);
1759 2050
1760#endif 2051#endif
1774void 2065void
1775marker (OpenCL::Queue self, ...) 2066marker (OpenCL::Queue self, ...)
1776 ALIAS: 2067 ALIAS:
1777 enqueue_marker = 0 2068 enqueue_marker = 0
1778 PPCODE: 2069 PPCODE:
2070 EVENT_LIST (1);
1779 cl_event ev = 0; 2071 cl_event ev = 0;
1780 EVENT_LIST (1);
1781#if PREFER_1_1 2072#if PREFER_1_1
1782 if (!event_list_count) 2073 if (!event_list_count)
1783 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2074 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1784 else 2075 else
1785#if CL_VERSION_1_2 2076#if CL_VERSION_1_2
1799void 2090void
1800barrier (OpenCL::Queue self, ...) 2091barrier (OpenCL::Queue self, ...)
1801 ALIAS: 2092 ALIAS:
1802 enqueue_barrier = 0 2093 enqueue_barrier = 0
1803 PPCODE: 2094 PPCODE:
2095 EVENT_LIST (1);
1804 cl_event ev = 0; 2096 cl_event ev = 0;
1805 EVENT_LIST (1);
1806#if PREFER_1_1 2097#if PREFER_1_1
1807 if (!event_list_count && GIMME_V == G_VOID) 2098 if (!event_list_count && GIMME_V == G_VOID)
1808 NEED_SUCCESS (EnqueueBarrier, (self)); 2099 NEED_SUCCESS (EnqueueBarrier, (self));
1809 else 2100 else
1810#if CL_VERSION_1_2 2101#if CL_VERSION_1_2
1846 PPCODE: 2137 PPCODE:
1847 cl_context value [1]; 2138 cl_context value [1];
1848 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2139 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1849 EXTEND (SP, 1); 2140 EXTEND (SP, 1);
1850 const int i = 0; 2141 const int i = 0;
1851 {
1852 NEED_SUCCESS (RetainContext, (value [i])); 2142 NEED_SUCCESS (RetainContext, (value [i]));
1853 PUSH_CLOBJ (stash_context, value [i]); 2143 PUSH_CLOBJ (stash_context, value [i]);
1854 }
1855 2144
1856void 2145void
1857device (OpenCL::Queue self) 2146device (OpenCL::Queue self)
1858 PPCODE: 2147 PPCODE:
1859 cl_device_id value [1]; 2148 cl_device_id value [1];
1860 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2149 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1861 EXTEND (SP, 1); 2150 EXTEND (SP, 1);
1862 const int i = 0; 2151 const int i = 0;
1863 {
1864 PUSH_CLOBJ (stash_device, value [i]); 2152 PUSH_CLOBJ (stash_device, value [i]);
1865 }
1866 2153
1867void 2154void
1868reference_count (OpenCL::Queue self) 2155reference_count (OpenCL::Queue self)
1869 PPCODE: 2156 PPCODE:
1870 cl_uint value [1]; 2157 cl_uint value [1];
1874 PUSHs (sv_2mortal (newSVuv (value [i]))); 2161 PUSHs (sv_2mortal (newSVuv (value [i])));
1875 2162
1876void 2163void
1877properties (OpenCL::Queue self) 2164properties (OpenCL::Queue self)
1878 PPCODE: 2165 PPCODE:
1879 cl_command_queue_properties value [1]; 2166 cl_ulong value [1];
1880 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2167 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1881 EXTEND (SP, 1); 2168 EXTEND (SP, 1);
1882 const int i = 0; 2169 const int i = 0;
1883 PUSHs (sv_2mortal (newSVuv (value [i]))); 2170 PUSHs (sv_2mortal (newSVuv (value [i])));
1884 2171
1894void 2181void
1895info (OpenCL::Memory self, cl_mem_info name) 2182info (OpenCL::Memory self, cl_mem_info name)
1896 PPCODE: 2183 PPCODE:
1897 INFO (MemObject) 2184 INFO (MemObject)
1898 2185
2186void
2187destructor_callback (OpenCL::Memory self, SV *notify)
2188 PPCODE:
2189 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2190
1899#BEGIN:mem 2191#BEGIN:mem
1900 2192
1901void 2193void
1902type (OpenCL::Memory self) 2194type (OpenCL::Memory self)
2195 ALIAS:
2196 type = CL_MEM_TYPE
2197 map_count = CL_MEM_MAP_COUNT
2198 reference_count = CL_MEM_REFERENCE_COUNT
1903 PPCODE: 2199 PPCODE:
1904 cl_mem_object_type value [1]; 2200 cl_uint value [1];
1905 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2201 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1906 EXTEND (SP, 1); 2202 EXTEND (SP, 1);
1907 const int i = 0; 2203 const int i = 0;
1908 PUSHs (sv_2mortal (newSVuv (value [i]))); 2204 PUSHs (sv_2mortal (newSVuv (value [i])));
1909 2205
1910void 2206void
1911flags (OpenCL::Memory self) 2207flags (OpenCL::Memory self)
1912 PPCODE: 2208 PPCODE:
1913 cl_mem_flags value [1]; 2209 cl_ulong value [1];
1914 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2210 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1915 EXTEND (SP, 1); 2211 EXTEND (SP, 1);
1916 const int i = 0; 2212 const int i = 0;
1917 PUSHs (sv_2mortal (newSVuv (value [i]))); 2213 PUSHs (sv_2mortal (newSVuv (value [i])));
1918 2214
1936 EXTEND (SP, 1); 2232 EXTEND (SP, 1);
1937 const int i = 0; 2233 const int i = 0;
1938 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2234 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1939 2235
1940void 2236void
1941map_count (OpenCL::Memory self)
1942 ALIAS:
1943 map_count = CL_MEM_MAP_COUNT
1944 reference_count = CL_MEM_REFERENCE_COUNT
1945 PPCODE:
1946 cl_uint value [1];
1947 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1948 EXTEND (SP, 1);
1949 const int i = 0;
1950 PUSHs (sv_2mortal (newSVuv (value [i])));
1951
1952void
1953context (OpenCL::Memory self) 2237context (OpenCL::Memory self)
1954 PPCODE: 2238 PPCODE:
1955 cl_context value [1]; 2239 cl_context value [1];
1956 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1957 EXTEND (SP, 1); 2241 EXTEND (SP, 1);
1958 const int i = 0; 2242 const int i = 0;
1959 {
1960 NEED_SUCCESS (RetainContext, (value [i])); 2243 NEED_SUCCESS (RetainContext, (value [i]));
1961 PUSH_CLOBJ (stash_context, value [i]); 2244 PUSH_CLOBJ (stash_context, value [i]);
1962 }
1963 2245
1964void 2246void
1965associated_memobject (OpenCL::Memory self) 2247associated_memobject (OpenCL::Memory self)
1966 PPCODE: 2248 PPCODE:
1967 cl_mem value [1]; 2249 cl_mem value [1];
1968 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1969 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1970 const int i = 0; 2252 const int i = 0;
1971 {
1972 NEED_SUCCESS (RetainMemObject, (value [i])); 2253 NEED_SUCCESS (RetainMemObject, (value [i]));
1973 PUSH_CLOBJ (stash_memory, value [i]); 2254 PUSH_CLOBJ (stash_memory, value [i]);
1974 }
1975 2255
1976#END:mem 2256#END:mem
1977 2257
1978#if cl_apple_gl_sharing || cl_khr_gl_sharing 2258#if cl_apple_gl_sharing || cl_khr_gl_sharing
1979 2259
2078 2358
2079#BEGIN:sampler 2359#BEGIN:sampler
2080 2360
2081void 2361void
2082reference_count (OpenCL::Sampler self) 2362reference_count (OpenCL::Sampler self)
2363 ALIAS:
2364 reference_count = CL_SAMPLER_REFERENCE_COUNT
2365 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2366 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
2083 PPCODE: 2367 PPCODE:
2084 cl_uint value [1]; 2368 cl_uint value [1];
2085 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2369 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
2086 EXTEND (SP, 1); 2370 EXTEND (SP, 1);
2087 const int i = 0; 2371 const int i = 0;
2088 PUSHs (sv_2mortal (newSVuv (value [i]))); 2372 PUSHs (sv_2mortal (newSVuv (value [i])));
2089 2373
2090void 2374void
2092 PPCODE: 2376 PPCODE:
2093 cl_context value [1]; 2377 cl_context value [1];
2094 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2378 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2095 EXTEND (SP, 1); 2379 EXTEND (SP, 1);
2096 const int i = 0; 2380 const int i = 0;
2097 {
2098 NEED_SUCCESS (RetainContext, (value [i])); 2381 NEED_SUCCESS (RetainContext, (value [i]));
2099 PUSH_CLOBJ (stash_context, value [i]); 2382 PUSH_CLOBJ (stash_context, value [i]);
2100 }
2101
2102void
2103normalized_coords (OpenCL::Sampler self)
2104 PPCODE:
2105 cl_addressing_mode value [1];
2106 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2107 EXTEND (SP, 1);
2108 const int i = 0;
2109 PUSHs (sv_2mortal (newSVuv (value [i])));
2110
2111void
2112addressing_mode (OpenCL::Sampler self)
2113 PPCODE:
2114 cl_filter_mode value [1];
2115 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2116 EXTEND (SP, 1);
2117 const int i = 0;
2118 PUSHs (sv_2mortal (newSVuv (value [i])));
2119 2383
2120void 2384void
2121filter_mode (OpenCL::Sampler self) 2385filter_mode (OpenCL::Sampler self)
2122 PPCODE: 2386 PPCODE:
2123 cl_bool value [1]; 2387 cl_uint value [1];
2124 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2125 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
2126 const int i = 0; 2390 const int i = 0;
2127 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2391 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2128 2392
2138void 2402void
2139build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2403build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2140 ALIAS: 2404 ALIAS:
2141 build_async = 1 2405 build_async = 1
2142 CODE: 2406 CODE:
2143 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2144 void *user_data = 0;
2145 cl_uint num_devices = 0; 2407 cl_uint device_count = 0;
2146 cl_device_id *device_list = 0; 2408 cl_device_id *device_list = 0;
2147 2409
2148 if (SvOK (devices)) 2410 if (SvOK (devices))
2149 { 2411 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2412
2413 void *user_data;
2414 program_callback pfn_notify = make_program_callback (notify, &user_data);
2415
2416 if (ix)
2417 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2418 else
2419 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2420
2421#if CL_VERSION_1_2
2422
2423void
2424compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2425 CODE:
2426 cl_uint device_count = 0;
2427 cl_device_id *device_list = 0;
2428
2429 if (SvOK (devices))
2430 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2431
2432 cl_uint header_count = 0;
2433 cl_program *header_list = 0;
2434 const char **header_name = 0;
2435
2436 if (SvOK (headers))
2437 {
2150 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2151 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2439 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2152 2440
2153 AV *av = (AV *)SvRV (devices); 2441 HV *hv = (HV *)SvRV (devices);
2154 num_devices = av_len (av) + 1;
2155 2442
2156 if (num_devices) 2443 header_count = hv_iterinit (hv);
2444 header_list = tmpbuf (sizeof (*header_list) * header_count);
2445 header_name = tmpbuf (sizeof (*header_name) * header_count);
2446
2447 HE *he;
2448 int i = 0;
2449 while (he = hv_iternext (hv))
2157 { 2450 {
2158 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2451 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2452 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2159 int count; 2453 ++i;
2160 for (count = 0; count < num_devices; ++count)
2161 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2162 } 2454 }
2163 } 2455 }
2164 2456
2165 if (SvOK (notify)) 2457 void *user_data;
2166 { 2458 program_callback pfn_notify = make_program_callback (notify, &user_data);
2167 NEED_SUCCESS (RetainProgram, (self));
2168 pfn_notify = eq_program_notify;
2169 user_data = SvREFCNT_inc (s_get_cv (notify));
2170 }
2171 2459
2172 if (ix)
2173 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2174 else
2175 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2460 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2461 header_count, header_list, header_name, pfn_notify, user_data));
2462
2463#endif
2176 2464
2177void 2465void
2178build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2466build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2179 PPCODE: 2467 PPCODE:
2180 size_t size; 2468 size_t size;
2189#BEGIN:program_build 2477#BEGIN:program_build
2190 2478
2191void 2479void
2192build_status (OpenCL::Program self, OpenCL::Device device) 2480build_status (OpenCL::Program self, OpenCL::Device device)
2193 PPCODE: 2481 PPCODE:
2194 cl_build_status value [1]; 2482 cl_int value [1];
2195 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2483 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2196 EXTEND (SP, 1); 2484 EXTEND (SP, 1);
2197 const int i = 0; 2485 const int i = 0;
2198 PUSHs (sv_2mortal (newSViv (value [i]))); 2486 PUSHs (sv_2mortal (newSViv (value [i])));
2199 2487
2208 char *value = tmpbuf (size); 2496 char *value = tmpbuf (size);
2209 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2497 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2210 EXTEND (SP, 1); 2498 EXTEND (SP, 1);
2211 const int i = 0; 2499 const int i = 0;
2212 PUSHs (sv_2mortal (newSVpv (value, 0))); 2500 PUSHs (sv_2mortal (newSVpv (value, 0)));
2501
2502void
2503binary_type (OpenCL::Program self, OpenCL::Device device)
2504 PPCODE:
2505 cl_uint value [1];
2506 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2507 EXTEND (SP, 1);
2508 const int i = 0;
2509 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2213 2510
2214#END:program_build 2511#END:program_build
2215 2512
2216void 2513void
2217kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
2284 PPCODE: 2581 PPCODE:
2285 cl_context value [1]; 2582 cl_context value [1];
2286 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2287 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
2288 const int i = 0; 2585 const int i = 0;
2289 {
2290 NEED_SUCCESS (RetainContext, (value [i])); 2586 NEED_SUCCESS (RetainContext, (value [i]));
2291 PUSH_CLOBJ (stash_context, value [i]); 2587 PUSH_CLOBJ (stash_context, value [i]);
2292 }
2293 2588
2294void 2589void
2295devices (OpenCL::Program self) 2590devices (OpenCL::Program self)
2296 PPCODE: 2591 PPCODE:
2297 size_t size; 2592 size_t size;
2299 cl_device_id *value = tmpbuf (size); 2594 cl_device_id *value = tmpbuf (size);
2300 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2595 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2301 int i, n = size / sizeof (*value); 2596 int i, n = size / sizeof (*value);
2302 EXTEND (SP, n); 2597 EXTEND (SP, n);
2303 for (i = 0; i < n; ++i) 2598 for (i = 0; i < n; ++i)
2304 {
2305 PUSH_CLOBJ (stash_device, value [i]); 2599 PUSH_CLOBJ (stash_device, value [i]);
2306 }
2307 2600
2308void 2601void
2309source (OpenCL::Program self) 2602source (OpenCL::Program self)
2310 PPCODE: 2603 PPCODE:
2311 size_t size; 2604 size_t size;
2384 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2677 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2385 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2678 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2386 2679
2387 case 'z': nullarg = 1; size = SvIV (sv); break; 2680 case 'z': nullarg = 1; size = SvIV (sv); break;
2388 2681
2389 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2682 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2390 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2683 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2391 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2684 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2392 2685
2393 default: 2686 default:
2394 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2687 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2395 } 2688 }
2396 2689
2467 CODE: 2760 CODE:
2468 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2761 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2469 2762
2470void 2763void
2471set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2764set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2472 ALIAS:
2473 set_image2d = 0
2474 set_image3d = 0
2475 CODE: 2765 CODE:
2476 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2766 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2477 2767
2478void 2768void
2479set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2769set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2525 PPCODE: 2815 PPCODE:
2526 cl_context value [1]; 2816 cl_context value [1];
2527 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2817 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2528 EXTEND (SP, 1); 2818 EXTEND (SP, 1);
2529 const int i = 0; 2819 const int i = 0;
2530 {
2531 NEED_SUCCESS (RetainContext, (value [i])); 2820 NEED_SUCCESS (RetainContext, (value [i]));
2532 PUSH_CLOBJ (stash_context, value [i]); 2821 PUSH_CLOBJ (stash_context, value [i]);
2533 }
2534 2822
2535void 2823void
2536program (OpenCL::Kernel self) 2824program (OpenCL::Kernel self)
2537 PPCODE: 2825 PPCODE:
2538 cl_program value [1]; 2826 cl_program value [1];
2539 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2827 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2540 EXTEND (SP, 1); 2828 EXTEND (SP, 1);
2541 const int i = 0; 2829 const int i = 0;
2542 {
2543 NEED_SUCCESS (RetainProgram, (value [i])); 2830 NEED_SUCCESS (RetainProgram, (value [i]));
2544 PUSH_CLOBJ (stash_program, value [i]); 2831 PUSH_CLOBJ (stash_program, value [i]);
2545 }
2546 2832
2547#END:kernel 2833#END:kernel
2548 2834
2549void 2835void
2550work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2836work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2596 const int i = 0; 2882 const int i = 0;
2597 PUSHs (sv_2mortal (newSVuv (value [i]))); 2883 PUSHs (sv_2mortal (newSVuv (value [i])));
2598 2884
2599#END:kernel_work_group 2885#END:kernel_work_group
2600 2886
2887#if CL_VERSION_1_2
2888
2889void
2890arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2891 PPCODE:
2892 size_t size;
2893 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2894 SV *sv = sv_2mortal (newSV (size));
2895 SvUPGRADE (sv, SVt_PV);
2896 SvPOK_only (sv);
2897 SvCUR_set (sv, size);
2898 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2899 XPUSHs (sv);
2900
2901#BEGIN:kernel_arg
2902
2903void
2904arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2905 ALIAS:
2906 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2907 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2908 PPCODE:
2909 cl_uint value [1];
2910 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2911 EXTEND (SP, 1);
2912 const int i = 0;
2913 PUSHs (sv_2mortal (newSVuv (value [i])));
2914
2915void
2916arg_type_name (OpenCL::Kernel self, cl_uint idx)
2917 ALIAS:
2918 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2919 arg_name = CL_KERNEL_ARG_NAME
2920 PPCODE:
2921 size_t size;
2922 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2923 char *value = tmpbuf (size);
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2925 EXTEND (SP, 1);
2926 const int i = 0;
2927 PUSHs (sv_2mortal (newSVpv (value, 0)));
2928
2929void
2930arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2931 PPCODE:
2932 cl_ulong value [1];
2933 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2934 EXTEND (SP, 1);
2935 const int i = 0;
2936 PUSHs (sv_2mortal (newSVuv (value [i])));
2937
2938#END:kernel_arg
2939
2940#endif
2941
2601MODULE = OpenCL PACKAGE = OpenCL::Event 2942MODULE = OpenCL PACKAGE = OpenCL::Event
2602 2943
2603void 2944void
2604DESTROY (OpenCL::Event self) 2945DESTROY (OpenCL::Event self)
2605 CODE: 2946 CODE:
2627 PPCODE: 2968 PPCODE:
2628 cl_command_queue value [1]; 2969 cl_command_queue value [1];
2629 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2970 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2630 EXTEND (SP, 1); 2971 EXTEND (SP, 1);
2631 const int i = 0; 2972 const int i = 0;
2632 {
2633 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2973 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2634 PUSH_CLOBJ (stash_queue, value [i]); 2974 PUSH_CLOBJ (stash_queue, value [i]);
2635 }
2636 2975
2637void 2976void
2638command_type (OpenCL::Event self) 2977command_type (OpenCL::Event self)
2639 PPCODE:
2640 cl_command_type value [1];
2641 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2642 EXTEND (SP, 1);
2643 const int i = 0;
2644 PUSHs (sv_2mortal (newSVuv (value [i])));
2645
2646void
2647reference_count (OpenCL::Event self)
2648 ALIAS: 2978 ALIAS:
2979 command_type = CL_EVENT_COMMAND_TYPE
2649 reference_count = CL_EVENT_REFERENCE_COUNT 2980 reference_count = CL_EVENT_REFERENCE_COUNT
2650 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2981 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2651 PPCODE: 2982 PPCODE:
2652 cl_uint value [1]; 2983 cl_uint value [1];
2653 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2984 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2660 PPCODE: 2991 PPCODE:
2661 cl_context value [1]; 2992 cl_context value [1];
2662 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2993 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2663 EXTEND (SP, 1); 2994 EXTEND (SP, 1);
2664 const int i = 0; 2995 const int i = 0;
2665 {
2666 NEED_SUCCESS (RetainContext, (value [i])); 2996 NEED_SUCCESS (RetainContext, (value [i]));
2667 PUSH_CLOBJ (stash_context, value [i]); 2997 PUSH_CLOBJ (stash_context, value [i]);
2668 }
2669 2998
2670#END:event 2999#END:event
2671 3000
2672void 3001void
2673profiling_info (OpenCL::Event self, cl_profiling_info name) 3002profiling_info (OpenCL::Event self, cl_profiling_info name)
2708 3037
2709 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0); 3038 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2710 mapped_detach (self, mapped); 3039 mapped_detach (self, mapped);
2711 3040
2712 clReleaseCommandQueue (mapped->queue); 3041 clReleaseCommandQueue (mapped->queue);
2713
2714 if (mapped->event)
2715 clReleaseEvent (mapped->event); 3042 clReleaseEvent (mapped->event);
2716
2717 Safefree (mapped); 3043 Safefree (mapped);
3044
3045void
3046unmap (OpenCL::Mapped self, ...)
3047 CODE:
3048 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2718 3049
2719bool 3050bool
2720mapped (OpenCL::Mapped self) 3051mapped (OpenCL::Mapped self)
2721 CODE: 3052 CODE:
2722 RETVAL = !!self->ptr; 3053 RETVAL = !!self->ptr;
2736 XSRETURN_UNDEF; 3067 XSRETURN_UNDEF;
2737 3068
2738 clRetainEvent (self->event); 3069 clRetainEvent (self->event);
2739 XPUSH_CLOBJ (stash_event, self->event); 3070 XPUSH_CLOBJ (stash_event, self->event);
2740 3071
2741size_t 3072#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3073#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3074#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3075#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3076#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3077#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3078
3079IV
2742size (OpenCL::Mapped self) 3080size (OpenCL::Mapped self)
3081 ALIAS:
3082 size = MAPPED_OFFSET_CB
3083 row_pitch = MAPPED_OFFSET_ROW_PITCH
3084 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3085 width = MAPPED_OFFSET_WIDTH
3086 height = MAPPED_OFFSET_HEIGHT
3087 depth = MAPPED_OFFSET_DEPTH
2743 CODE: 3088 CODE:
2744 RETVAL = self->cb; 3089 RETVAL = *(size_t *)((char *)self + ix);
2745 OUTPUT: 3090 OUTPUT:
2746 RETVAL 3091 RETVAL
2747 3092
2748IV 3093IV
2749ptr (OpenCL::Mapped self) 3094ptr (OpenCL::Mapped self)
2750 CODE: 3095 CODE:
2751 RETVAL = PTR2IV (self->ptr); 3096 RETVAL = PTR2IV (self->ptr);
2752 OUTPUT: 3097 OUTPUT:
2753 RETVAL 3098 RETVAL
2754 3099
3100void
3101set (OpenCL::Mapped self, size_t offset, SV *data)
3102 CODE:
3103 STRLEN len;
3104 const char *ptr = SvPVbyte (data, len);
3105
3106 if (offset + len > self->cb)
3107 croak ("OpenCL::Mapped::set out of bound condition detected");
3108
3109 memcpy (offset + (char *)self->ptr, ptr, len);
3110
3111void
3112get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3113 PPCODE:
3114 if (!SvOK (ST (1)))
3115 count = self->width - x;
3116
3117 if (x + count > self->width)
3118 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3119
3120 if (y >= self->height)
3121 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3122
3123 if (z >= self->depth)
3124 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3125
3126 size_t element = mapped_element_size (self);
3127
3128 count *= element;
3129 x *= element;
3130
3131 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3132 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3133
3134void
3135set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3136 PPCODE:
3137 STRLEN count;
3138 char *dataptr = SvPVbyte (data, count);
3139 size_t element = mapped_element_size (self);
3140
3141 x *= element;
3142
3143 if (x + count > self->width * element)
3144 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3145
3146 if (y >= self->height)
3147 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3148
3149 if (z >= self->depth)
3150 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3151
3152 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3153 memcpy (ptr, dataptr, count);
3154
2755MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer 3155MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2756 3156
2757MODULE = OpenCL PACKAGE = OpenCL::MappedImage 3157MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2758 3158
2759IV 3159IV
2760row_pitch (OpenCL::Mapped self) 3160element_size (OpenCL::Mapped self)
2761 ALIAS:
2762 slice_pitch = 1
2763 CODE: 3161 CODE:
2764 RETVAL = ix ? self->slice_pitch : self->row_pitch; 3162 RETVAL = mapped_element_size (self);
2765 OUTPUT: 3163 OUTPUT:
2766 RETVAL 3164 RETVAL
2767 3165
2768
2769

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines