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 1.74 by root, Sat May 5 13:56:00 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#include "ecb.h"//D
4 6
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 8#include "xthread.h"
7#include "schmorp.h" 9#include "schmorp.h"
8 10
26#if PREFER_1_1 28#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 30#endif
29 31
30#ifdef __APPLE__ 32#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
32#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
33 #include <CL/opencl.h> 38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 50#endif
35 51
36#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 53 #undef PREFER_1_1
38 #define PREFER_1_1 1 54 #define PREFER_1_1 1
39#endif 55#endif
40 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
41typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
61 77
62typedef struct mapped * OpenCL__Mapped; 78typedef struct mapped * OpenCL__Mapped;
63 79
64typedef SV *FUTURE;
65
66static HV 80static HV
67 *stash_platform, 81 *stash_platform,
68 *stash_device, 82 *stash_device,
83 *stash_subdevice,
69 *stash_context, 84 *stash_context,
70 *stash_queue, 85 *stash_queue,
71 *stash_program, 86 *stash_program,
72 *stash_kernel, 87 *stash_kernel,
73 *stash_sampler, 88 *stash_sampler,
133} 148}
134 149
135/*****************************************************************************/ 150/*****************************************************************************/
136 151
137/* up to two temporary buffers */ 152/* up to two temporary buffers */
138static void * 153static void * ecb_noinline
139tmpbuf (size_t size) 154tmpbuf (size_t size)
140{ 155{
141 enum { buffers = 3 }; 156 enum { buffers = 4 };
142 static int idx; 157 static int idx;
143 static void *buf [buffers]; 158 static void *buf [buffers];
144 static size_t len [buffers]; 159 static size_t len [buffers];
145 160
146 idx = (idx + 1) % buffers; 161 idx = (idx + 1) % buffers;
151 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
152 buf [idx] = malloc (len [idx]); 167 buf [idx] = malloc (len [idx]);
153 } 168 }
154 169
155 return buf [idx]; 170 return buf [idx];
171}
172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
156} 186}
157 187
158/*****************************************************************************/ 188/*****************************************************************************/
159 189
160typedef struct 190typedef struct
219 if (res) \ 249 if (res) \
220 FAIL (name); 250 FAIL (name);
221 251
222/*****************************************************************************/ 252/*****************************************************************************/
223 253
254static SV *
255new_clobj (HV *stash, IV id)
256{
257 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
258}
259
260#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
261#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
262
263/* cl objects are either \$iv, or [$iv, ...] */
264/* they can be upgraded at runtime to the array form */
265static void * ecb_noinline
266SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
267{
268 // sv_derived_from is quite slow :(
269 if (SvROK (sv) && sv_derived_from (sv, pkg))
270 return (void *)SvIV (SvRV (sv));
271
272 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
273}
274
275// the "no-inherit" version of the above
276static void * ecb_noinline
277SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
278{
279 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
280 return (void *)SvIV (SvRV (sv));
281
282 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
283}
284
285/*****************************************************************************/
286
224static cl_context_properties * 287static cl_context_properties * ecb_noinline
225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 288SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
226{ 289{
227 if (!sv || !SvOK (sv)) 290 if (!sv || !SvOK (sv))
228 if (extra) 291 if (extra)
229 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 292 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
230 else 293 else
236 int i, len = av_len (av) + 1; 299 int i, len = av_len (av) + 1;
237 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 300 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
238 cl_context_properties *l = p; 301 cl_context_properties *l = p;
239 302
240 if (len & 1) 303 if (len & 1)
241 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 304 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
242 305
243 while (extracount--) 306 while (extracount--)
244 *l++ = *extra++; 307 *l++ = *extra++;
245 308
246 for (i = 0; i < len; i += 2) 309 for (i = 0; i < len; i += 2)
249 SV *p_sv = *av_fetch (av, i + 1, 0); 312 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override 313 cl_context_properties v = SvIV (p_sv); // code below can override
251 314
252 switch (t) 315 switch (t)
253 { 316 {
317 case CL_CONTEXT_PLATFORM:
318 if (SvROK (p_sv))
319 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
320 break;
321
254 case CL_GLX_DISPLAY_KHR: 322 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv)) 323 if (!SvOK (p_sv))
256 { 324 {
257 void *func = glsym ("_glXGetCurrentDisplay"); 325 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func) 326 if (func)
281 *l = 0; 349 *l = 0;
282 350
283 return p; 351 return p;
284 } 352 }
285 353
286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 354 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
287} 355}
288 356
289/*****************************************************************************/ 357// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
290 358// is a pointer (and only on well-behaved systems).
291static SV * 359static void * ecb_noinline
292new_clobj (HV *stash, IV id) 360object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
293{ 361{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash)); 362 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
295} 363 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
364 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
296 365
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id))) 366 AV *av = (AV *)SvRV (arg);
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id))) 367 void **list = 0;
368 cl_uint count = av_len (av) + 1;
299 369
300/* cl objects are either \$iv, or [$iv, ...] */ 370 if (count)
301/* they can be upgraded at runtime to the array form */ 371 {
302static void * 372 list = tmpbuf (sizeof (*list) * count);
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 373 int i;
304{ 374 for (i = 0; i < count; ++i)
305 if (SvROK (sv) && sv_derived_from (sv, pkg)) 375 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
306 return (void *)SvIV (SvRV (sv)); 376 }
307 377
308 croak ("%s: %s is not of type %s", func, svname, pkg); 378 if (!count && !or_undef)
379 croak ("%s: '%s' must contain at least one %s object",
380 cv_get_name (cv), argname, klass);
381
382 *rcount = count;
383 return (void *)list;
309} 384}
310 385
311/*****************************************************************************/ 386/*****************************************************************************/
312/* callback stuff */ 387/* callback stuff */
313 388
335static void (*eq_signal_func)(void *signal_arg, int value); 410static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg; 411static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT; 412static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail; 413static eq_item *eq_head, *eq_tail;
339 414
340static void 415static void ecb_noinline
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 416eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{ 417{
343 eq_item *item = malloc (sizeof (eq_item)); 418 eq_item *item = malloc (sizeof (eq_item));
344 419
345 item->next = 0; 420 item->next = 0;
412} 487}
413 488
414/*****************************************************************************/ 489/*****************************************************************************/
415/* context notify */ 490/* context notify */
416 491
417static void 492static void ecb_noinline
418eq_context_push (void *data1, void *data2, void *data3) 493eq_context_push (void *data1, void *data2, void *data3)
419{ 494{
420 dSP; 495 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0))); 496 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 497 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
444 { \ 519 { \
445 pfn_notify = eq_context_notify; \ 520 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \ 521 user_data = s_get_cv (notify); \
447 } 522 }
448 523
449static SV * 524static SV * ecb_noinline
450new_clobj_context (cl_context ctx, void *user_data) 525new_clobj_context (cl_context ctx, void *user_data)
451{ 526{
452 SV *sv = new_clobj (stash_context, (IV)ctx); 527 SV *sv = new_clobj (stash_context, (IV)ctx);
453 528
454 if (user_data) 529 if (user_data)
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 548static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474 549
475static void CL_CALLBACK 550static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data) 551eq_program_notify (cl_program program, void *user_data)
477{ 552{
553 clRetainProgram (program);
554
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 555 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
556}
557
558typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
559
560static program_callback ecb_noinline
561make_program_callback (SV *notify, void **ruser_data)
562{
563 if (SvOK (notify))
564 {
565 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
566 return eq_program_notify;
567 }
568 else
569 {
570 *ruser_data = 0;
571 return 0;
572 }
479} 573}
480 574
481struct build_args 575struct build_args
482{ 576{
483 cl_program program; 577 cl_program program;
497 else 591 else
498 clReleaseProgram (arg->program); 592 clReleaseProgram (arg->program);
499 593
500 free (arg->options); 594 free (arg->options);
501 free (arg); 595 free (arg);
596
597 return 0;
502} 598}
503 599
504static void 600static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 601build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{ 602{
536 clRetainEvent (event); 632 clRetainEvent (event);
537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 633 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
538} 634}
539 635
540/*****************************************************************************/ 636/*****************************************************************************/
541/* mapped_xxx */ 637/* 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/*****************************************************************************/
615 638
616static size_t 639static size_t
617img_row_pitch (cl_mem img) 640img_row_pitch (cl_mem img)
618{ 641{
619 size_t res; 642 size_t res;
620 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 643 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
621 return res; 644 return res;
622} 645}
623 646
624static cl_event * 647static cl_event * ecb_noinline
625event_list (SV **items, cl_uint *rcount, cl_event extra) 648event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
626{ 649{
627 cl_uint count = *rcount; 650 cl_uint count = *rcount;
651
652 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
653 *rcount = count = 0;
628 654
629 if (!count && !extra) 655 if (!count && !extra)
630 return 0; 656 return 0;
631 657
632 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 658 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
633 int i = 0; 659 int i = 0;
634 660
635 while (count) 661 while (count--)
636 {
637 --count;
638 if (SvOK (items [count])) 662 if (SvOK (items [count]))
639 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 663 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
640 }
641 664
642 if (extra) 665 if (extra)
643 list [i++] = extra; 666 list [i++] = extra;
644 667
645 *rcount = i; 668 *rcount = i;
646 669
647 return i ? list : 0; 670 return i ? list : 0;
648} 671}
649 672
650#define EVENT_LIST(skip) \ 673#define EVENT_LIST(skip) \
651 cl_uint event_list_count = items - skip; \ 674 cl_uint event_list_count = items - (skip); \
652 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 675 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
653 676
654#define INFO(class) \ 677#define INFO(class) \
655{ \ 678{ \
656 size_t size; \ 679 size_t size; \
657 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 680 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
660 SvPOK_only (sv); \ 683 SvPOK_only (sv); \
661 SvCUR_set (sv, size); \ 684 SvCUR_set (sv, size); \
662 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 685 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
663 XPUSHs (sv); \ 686 XPUSHs (sv); \
664} 687}
688
689/*****************************************************************************/
690/* mapped_xxx */
691
692static OpenCL__Mapped
693SvMAPPED (SV *sv)
694{
695 // no typechecking atm., keep your fingers crossed
696 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
697}
698
699struct mapped
700{
701 cl_command_queue queue;
702 cl_mem memobj;
703 void *ptr;
704 size_t cb;
705 cl_event event;
706 size_t row_pitch;
707 size_t slice_pitch;
708};
709
710static SV *
711mapped_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)
712{
713 SV *data = newSV (0);
714 SvUPGRADE (data, SVt_PVMG);
715
716 OpenCL__Mapped mapped;
717 New (0, mapped, 1, struct mapped);
718
719 clRetainCommandQueue (queue);
720
721 mapped->queue = queue;
722 mapped->memobj = memobj;
723 mapped->ptr = ptr;
724 mapped->cb = cb;
725 mapped->event = ev;
726 mapped->row_pitch = row_pitch;
727 mapped->slice_pitch = slice_pitch;
728
729 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
730
731 if (SvLEN (data))
732 Safefree (data);
733
734 SvPVX (data) = (char *)ptr;
735 SvCUR_set (data, cb);
736 SvLEN_set (data, 0);
737 SvPOK_only (data);
738
739 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
740
741 if (!(flags & CL_MAP_WRITE))
742 SvREADONLY_on (data);
743
744 return obj;
745}
746
747static void
748mapped_detach (SV *sv, OpenCL__Mapped mapped)
749{
750 SV *data = SvRV (sv);
751
752 // the next check checks both before AND after detach, where SvPVX should STILL be 0
753 if (SvPVX (data) != (char *)mapped->ptr)
754 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
755 else
756 {
757 SvREADONLY_off (data);
758 SvCUR_set (data, 0);
759 SvPVX (data) = 0;
760 SvOK_off (data);
761 }
762
763 mapped->ptr = 0;
764}
765
766static void
767mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
768{
769 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
770 cl_event ev;
771
772 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
773
774 clReleaseEvent (mapped->event);
775 mapped->event = ev;
776
777 mapped_detach (self, mapped);
778}
779
780/*****************************************************************************/
665 781
666MODULE = OpenCL PACKAGE = OpenCL 782MODULE = OpenCL PACKAGE = OpenCL
667 783
668PROTOTYPES: ENABLE 784PROTOTYPES: ENABLE
669 785
691 { sizeof (cl_long ), "SIZEOF_LONG" }, 807 { sizeof (cl_long ), "SIZEOF_LONG" },
692 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 808 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
693 { sizeof (cl_half ), "SIZEOF_HALF" }, 809 { sizeof (cl_half ), "SIZEOF_HALF" },
694 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 810 { sizeof (cl_float ), "SIZEOF_FLOAT" },
695 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 811 { sizeof (cl_double), "SIZEOF_DOUBLE" },
812 { PREFER_1_1 , "PREFER_1_1" },
696#include "constiv.h" 813#include "constiv.h"
697 }; 814 };
698 815
699 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 816 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)); 817 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
701 818
702 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 819 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
703 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 820 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
821 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
704 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 822 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
705 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 823 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
706 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 824 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
707 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 825 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
708 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 826 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
749 list = tmpbuf (sizeof (*list) * count); 867 list = tmpbuf (sizeof (*list) * count);
750 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 868 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
751 869
752 EXTEND (SP, count); 870 EXTEND (SP, count);
753 for (i = 0; i < count; ++i) 871 for (i = 0; i < count; ++i)
754 PUSH_CLOBJ (stash_platform, (IV)list [i]); 872 PUSH_CLOBJ (stash_platform, list [i]);
755 873
756void 874void
757context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 875context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
758 PPCODE: 876 PPCODE:
759 CONTEXT_NOTIFY_CALLBACK; 877 CONTEXT_NOTIFY_CALLBACK;
760 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 878 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
761 XPUSH_CLOBJ_CONTEXT; 879 XPUSH_CLOBJ_CONTEXT;
762 880
763void 881void
764context (FUTURE properties, FUTURE devices, FUTURE notify) 882context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
765 PPCODE: 883 PPCODE:
766 /* der Gipfel der Kunst */ 884 cl_uint device_count;
885 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
886
887 CONTEXT_NOTIFY_CALLBACK;
888 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
889 XPUSH_CLOBJ_CONTEXT;
767 890
768void 891void
769wait_for_events (...) 892wait_for_events (...)
770 CODE: 893 CODE:
771 EVENT_LIST (0); 894 EVENT_LIST (0);
822 EXTEND (SP, count); 945 EXTEND (SP, count);
823 for (i = 0; i < count; ++i) 946 for (i = 0; i < count; ++i)
824 PUSH_CLOBJ (stash_device, list [i]); 947 PUSH_CLOBJ (stash_device, list [i]);
825 948
826void 949void
827context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 950context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
828 PPCODE: 951 PPCODE:
952 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
953 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
954
955 cl_uint device_count;
956 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
957
958 CONTEXT_NOTIFY_CALLBACK;
959 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
960 XPUSH_CLOBJ_CONTEXT;
961
962void
963context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
964 PPCODE:
965 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
966 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
967
968 CONTEXT_NOTIFY_CALLBACK;
969 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
970 XPUSH_CLOBJ_CONTEXT;
971
972MODULE = OpenCL PACKAGE = OpenCL::Device
973
974void
975info (OpenCL::Device self, cl_device_info name)
976 PPCODE:
977 INFO (Device)
978
979#if CL_VERSION_1_2
980
981void
982sub_devices (OpenCL::Device self, SV *properties)
983 PPCODE:
829 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 984 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
830 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 985 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
831 986
832 AV *av = (AV *)SvRV (devices); 987 properties = SvRV (properties);
833 cl_uint num_devices = av_len (av) + 1; 988
834 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 989 cl_uint count = av_len ((AV *)properties) + 1;
990 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
835 991
836 int i; 992 int i;
837 for (i = num_devices; i--; ) 993 for (i = 0; i < count; ++i)
838 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 994 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
839 995
840 CONTEXT_NOTIFY_CALLBACK; 996 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;
843 997
844void 998 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) 999 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
846 PPCODE: 1000 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
847 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1001 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 1002
850 CONTEXT_NOTIFY_CALLBACK; 1003 EXTEND (SP, num_devices);
851 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1004 for (i = 0; i < count; ++i)
852 XPUSH_CLOBJ_CONTEXT; 1005 PUSH_CLOBJ (stash_subdevice, list [i]);
853 1006
854MODULE = OpenCL PACKAGE = OpenCL::Device 1007#endif
855
856void
857info (OpenCL::Device self, cl_device_info name)
858 PPCODE:
859 INFO (Device)
860 1008
861#BEGIN:device 1009#BEGIN:device
862 1010
863void 1011void
864type (OpenCL::Device self) 1012type (OpenCL::Device self)
1080 for (i = 0; i < n; ++i) 1228 for (i = 0; i < n; ++i)
1081 PUSHs (sv_2mortal (newSVuv (value [i]))); 1229 PUSHs (sv_2mortal (newSVuv (value [i])));
1082 1230
1083#END:device 1231#END:device
1084 1232
1233MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1234
1235#if CL_VERSION_1_2
1236
1237void
1238DESTROY (OpenCL::SubDevice self)
1239 CODE:
1240 clReleaseDevice (self);
1241
1242#endif
1243
1085MODULE = OpenCL PACKAGE = OpenCL::Context 1244MODULE = OpenCL PACKAGE = OpenCL::Context
1086 1245
1087void 1246void
1088DESTROY (OpenCL::Context context) 1247DESTROY (OpenCL::Context self)
1089 CODE: 1248 CODE:
1090 clReleaseContext (context); 1249 clReleaseContext (self);
1091 1250
1092void 1251void
1093info (OpenCL::Context self, cl_context_info name) 1252info (OpenCL::Context self, cl_context_info name)
1094 PPCODE: 1253 PPCODE:
1095 INFO (Context) 1254 INFO (Context)
1136 const cl_image_desc desc = { 1295 const cl_image_desc desc = {
1137 type, 1296 type,
1138 width, height, depth, 1297 width, height, depth,
1139 array_size, row_pitch, slice_pitch, 1298 array_size, row_pitch, slice_pitch,
1140 num_mip_level, num_samples, 1299 num_mip_level, num_samples,
1141 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1300 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1142 }; 1301 };
1143 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1302 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1144 HV *stash = stash_image; 1303 HV *stash = stash_image;
1145 switch (type) 1304 switch (type)
1146 { 1305 {
1279 1438
1280 len2 = len; 1439 len2 = len;
1281 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1440 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1282 XPUSH_CLOBJ (stash_program, prog); 1441 XPUSH_CLOBJ (stash_program, prog);
1283 1442
1443void
1444program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1445 PPCODE:
1446 cl_uint device_count;
1447 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1448
1449 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1450 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1451
1452 binaries = SvRV (binaries);
1453
1454 if (device_count != av_len ((AV *)binaries) + 1)
1455 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1456
1457 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1458 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1459 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1460
1461 int i;
1462 for (i = 0; i < device_count; ++i)
1463 {
1464 STRLEN len;
1465 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1466 length_list [i] = len;
1467 }
1468
1469 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1470 length_list, binary_list,
1471 GIMME_V == G_ARRAY ? status_list : 0, &res));
1472
1473 EXTEND (SP, 2);
1474 PUSH_CLOBJ (stash_program, prog);
1475
1476 if (GIMME_V == G_ARRAY)
1477 {
1478 AV *av = newAV ();
1479 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1480
1481 for (i = device_count; i--; )
1482 av_store (av, i, newSViv (status_list [i]));
1483 }
1484
1485#if CL_VERSION_1_2
1486
1487void
1488program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1489 PPCODE:
1490 cl_uint device_count;
1491 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1492
1493 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1494
1495 XPUSH_CLOBJ (stash_program, prog);
1496
1497void
1498link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1499 CODE:
1500 cl_uint device_count = 0;
1501 cl_device_id *device_list = 0;
1502
1503 if (SvOK (devices))
1504 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1505
1506 cl_uint program_count;
1507 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1508
1509 void *user_data;
1510 program_callback pfn_notify = make_program_callback (notify, &user_data);
1511
1512 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1513 program_count, program_list, pfn_notify, user_data, &res));
1514
1515 XPUSH_CLOBJ (stash_program, prog);
1516
1517#endif
1518
1284#BEGIN:context 1519#BEGIN:context
1285 1520
1286void 1521void
1287reference_count (OpenCL::Context self) 1522reference_count (OpenCL::Context self)
1288 ALIAS: 1523 ALIAS:
1333void 1568void
1334read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1569read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1335 ALIAS: 1570 ALIAS:
1336 enqueue_read_buffer = 0 1571 enqueue_read_buffer = 0
1337 PPCODE: 1572 PPCODE:
1338 cl_event ev = 0;
1339 EVENT_LIST (6); 1573 EVENT_LIST (6);
1340 1574
1341 SvUPGRADE (data, SVt_PV); 1575 SvUPGRADE (data, SVt_PV);
1342 SvGROW (data, len); 1576 SvGROW (data, len);
1343 SvPOK_only (data); 1577 SvPOK_only (data);
1344 SvCUR_set (data, len); 1578 SvCUR_set (data, len);
1579
1580 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)); 1581 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1346 1582
1347 if (ev) 1583 if (ev)
1348 XPUSH_CLOBJ (stash_event, ev); 1584 XPUSH_CLOBJ (stash_event, ev);
1349 1585
1350void 1586void
1351write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1587write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1352 ALIAS: 1588 ALIAS:
1353 enqueue_write_buffer = 0 1589 enqueue_write_buffer = 0
1354 PPCODE: 1590 PPCODE:
1355 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1356 STRLEN len; 1593 STRLEN len;
1357 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1358 EVENT_LIST (5);
1359 1595
1596 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)); 1597 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1361 1598
1362 if (ev) 1599 if (ev)
1363 XPUSH_CLOBJ (stash_event, ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1364 1601
1367void 1604void
1368fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1605fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1369 ALIAS: 1606 ALIAS:
1370 enqueue_fill_buffer = 0 1607 enqueue_fill_buffer = 0
1371 PPCODE: 1608 PPCODE:
1372 cl_event ev = 0; 1609 EVENT_LIST (5);
1610
1373 STRLEN len; 1611 STRLEN len;
1374 char *ptr = SvPVbyte (data, len); 1612 char *ptr = SvPVbyte (data, len);
1375 EVENT_LIST (5);
1376 1613
1614 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)); 1615 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1378 1616
1379 if (ev) 1617 if (ev)
1380 XPUSH_CLOBJ (stash_event, ev); 1618 XPUSH_CLOBJ (stash_event, ev);
1381 1619
1382void 1620void
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, ...) 1621fill_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: 1622 ALIAS:
1385 enqueue_fill_image = 0 1623 enqueue_fill_image = 0
1386 PPCODE: 1624 PPCODE:
1387 cl_event ev = 0; 1625 EVENT_LIST (12);
1388 STRLEN len; 1626
1389 const size_t origin [3] = { x, y, z }; 1627 const size_t origin [3] = { x, y, z };
1390 const size_t region [3] = { width, height, depth }; 1628 const size_t region [3] = { width, height, depth };
1391 EVENT_LIST (12);
1392 1629
1393 const cl_float c_f [4] = { r, g, b, a }; 1630 const cl_float c_f [4] = { r, g, b, a };
1394 const cl_uint c_u [4] = { r, g, b, a }; 1631 const cl_uint c_u [4] = { r, g, b, a };
1395 const cl_int c_s [4] = { r, g, b, a }; 1632 const cl_int c_s [4] = { r, g, b, a };
1396 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1633 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1399 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1636 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1400 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1637 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) 1638 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."); 1639 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1403 1640
1641 cl_event ev = 0;
1404 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1642 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)); 1643 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1406 1644
1407 if (ev) 1645 if (ev)
1408 XPUSH_CLOBJ (stash_event, ev); 1646 XPUSH_CLOBJ (stash_event, ev);
1412void 1650void
1413copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1651copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1414 ALIAS: 1652 ALIAS:
1415 enqueue_copy_buffer = 0 1653 enqueue_copy_buffer = 0
1416 PPCODE: 1654 PPCODE:
1655 EVENT_LIST (6);
1656
1417 cl_event ev = 0; 1657 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)); 1658 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1421 1659
1422 if (ev) 1660 if (ev)
1423 XPUSH_CLOBJ (stash_event, ev); 1661 XPUSH_CLOBJ (stash_event, ev);
1424 1662
1425void 1663void
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, ...) 1664read_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: 1665 ALIAS:
1428 enqueue_read_buffer_rect = 0 1666 enqueue_read_buffer_rect = 0
1429 PPCODE: 1667 PPCODE:
1430 cl_event ev = 0; 1668 EVENT_LIST (17);
1669
1431 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1670 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 }; 1671 const size_t host_origin[3] = { host_x, host_y, host_z };
1433 const size_t region[3] = { width, height, depth }; 1672 const size_t region[3] = { width, height, depth };
1434 EVENT_LIST (17);
1435 1673
1436 if (!buf_row_pitch) 1674 if (!buf_row_pitch)
1437 buf_row_pitch = region [0]; 1675 buf_row_pitch = region [0];
1438 1676
1439 if (!buf_slice_pitch) 1677 if (!buf_slice_pitch)
1449 1687
1450 SvUPGRADE (data, SVt_PV); 1688 SvUPGRADE (data, SVt_PV);
1451 SvGROW (data, len); 1689 SvGROW (data, len);
1452 SvPOK_only (data); 1690 SvPOK_only (data);
1453 SvCUR_set (data, len); 1691 SvCUR_set (data, len);
1692
1693 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)); 1694 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 1695
1456 if (ev) 1696 if (ev)
1457 XPUSH_CLOBJ (stash_event, ev); 1697 XPUSH_CLOBJ (stash_event, ev);
1458 1698
1459void 1699void
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, ...) 1700write_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: 1701 ALIAS:
1462 enqueue_write_buffer_rect = 0 1702 enqueue_write_buffer_rect = 0
1463 PPCODE: 1703 PPCODE:
1464 cl_event ev = 0; 1704 EVENT_LIST (17);
1705
1465 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1706 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 }; 1707 const size_t host_origin[3] = { host_x, host_y, host_z };
1467 const size_t region[3] = { width, height, depth }; 1708 const size_t region[3] = { width, height, depth };
1468 STRLEN len; 1709 STRLEN len;
1469 char *ptr = SvPVbyte (data, len); 1710 char *ptr = SvPVbyte (data, len);
1470 EVENT_LIST (17);
1471 1711
1472 if (!buf_row_pitch) 1712 if (!buf_row_pitch)
1473 buf_row_pitch = region [0]; 1713 buf_row_pitch = region [0];
1474 1714
1475 if (!buf_slice_pitch) 1715 if (!buf_slice_pitch)
1484 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1724 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1485 1725
1486 if (len < min_len) 1726 if (len < min_len)
1487 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1727 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1488 1728
1729 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)); 1730 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 1731
1491 if (ev) 1732 if (ev)
1492 XPUSH_CLOBJ (stash_event, ev); 1733 XPUSH_CLOBJ (stash_event, ev);
1493 1734
1494void 1735void
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, ...) 1736copy_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: 1737 ALIAS:
1497 enqueue_copy_buffer_rect = 0 1738 enqueue_copy_buffer_rect = 0
1498 PPCODE: 1739 PPCODE:
1499 cl_event ev = 0; 1740 EVENT_LIST (16);
1741
1500 const size_t src_origin[3] = { src_x, src_y, src_z }; 1742 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 }; 1743 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1502 const size_t region[3] = { width, height, depth }; 1744 const size_t region[3] = { width, height, depth };
1503 EVENT_LIST (16);
1504 1745
1746 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)); 1747 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 1748
1507 if (ev) 1749 if (ev)
1508 XPUSH_CLOBJ (stash_event, ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1509 1751
1510void 1752void
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, ...) 1753read_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: 1754 ALIAS:
1513 enqueue_read_image = 0 1755 enqueue_read_image = 0
1514 PPCODE: 1756 PPCODE:
1515 cl_event ev = 0; 1757 EVENT_LIST (12);
1758
1516 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1517 const size_t region[3] = { width, height, depth }; 1760 const size_t region[3] = { width, height, depth };
1518 EVENT_LIST (12);
1519 1761
1520 if (!row_pitch) 1762 if (!row_pitch)
1521 row_pitch = img_row_pitch (src); 1763 row_pitch = img_row_pitch (src);
1522 1764
1523 if (depth > 1 && !slice_pitch) 1765 if (depth > 1 && !slice_pitch)
1527 1769
1528 SvUPGRADE (data, SVt_PV); 1770 SvUPGRADE (data, SVt_PV);
1529 SvGROW (data, len); 1771 SvGROW (data, len);
1530 SvPOK_only (data); 1772 SvPOK_only (data);
1531 SvCUR_set (data, len); 1773 SvCUR_set (data, len);
1774
1775 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)); 1776 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 1777
1534 if (ev) 1778 if (ev)
1535 XPUSH_CLOBJ (stash_event, ev); 1779 XPUSH_CLOBJ (stash_event, ev);
1536 1780
1537void 1781void
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, ...) 1782write_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: 1783 ALIAS:
1540 enqueue_write_image = 0 1784 enqueue_write_image = 0
1541 PPCODE: 1785 PPCODE:
1542 cl_event ev = 0; 1786 EVENT_LIST (12);
1787
1543 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1788 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1544 const size_t region[3] = { width, height, depth }; 1789 const size_t region[3] = { width, height, depth };
1545 STRLEN len; 1790 STRLEN len;
1546 char *ptr = SvPVbyte (data, len); 1791 char *ptr = SvPVbyte (data, len);
1547 EVENT_LIST (12);
1548 1792
1549 if (!row_pitch) 1793 if (!row_pitch)
1550 row_pitch = img_row_pitch (dst); 1794 row_pitch = img_row_pitch (dst);
1551 1795
1552 if (depth > 1 && !slice_pitch) 1796 if (depth > 1 && !slice_pitch)
1555 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1799 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1556 1800
1557 if (len < min_len) 1801 if (len < min_len)
1558 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1802 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1559 1803
1804 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)); 1805 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 1806
1562 if (ev) 1807 if (ev)
1563 XPUSH_CLOBJ (stash_event, ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1564 1809
1565void 1810void
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, ...) 1811copy_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: 1812 ALIAS:
1568 enqueue_copy_image = 0 1813 enqueue_copy_image = 0
1569 PPCODE: 1814 PPCODE:
1570 cl_event ev = 0; 1815 EVENT_LIST (12);
1816
1571 const size_t src_origin[3] = { src_x, src_y, src_z }; 1817 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 }; 1818 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1573 const size_t region[3] = { width, height, depth }; 1819 const size_t region[3] = { width, height, depth };
1574 EVENT_LIST (12);
1575 1820
1821 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)); 1822 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577 1823
1578 if (ev) 1824 if (ev)
1579 XPUSH_CLOBJ (stash_event, ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1580 1826
1581void 1827void
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, ...) 1828copy_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: 1829 ALIAS:
1584 enqueue_copy_image_to_buffer = 0 1830 enqueue_copy_image_to_buffer = 0
1585 PPCODE: 1831 PPCODE:
1586 cl_event ev = 0; 1832 EVENT_LIST (10);
1833
1587 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1588 const size_t region [3] = { width, height, depth }; 1835 const size_t region [3] = { width, height, depth };
1589 EVENT_LIST (10);
1590 1836
1837 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)); 1838 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1592 1839
1593 if (ev) 1840 if (ev)
1594 XPUSH_CLOBJ (stash_event, ev); 1841 XPUSH_CLOBJ (stash_event, ev);
1595 1842
1596void 1843void
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, ...) 1844copy_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: 1845 ALIAS:
1599 enqueue_copy_buffer_to_image = 0 1846 enqueue_copy_buffer_to_image = 0
1600 PPCODE: 1847 PPCODE:
1601 cl_event ev = 0; 1848 EVENT_LIST (10);
1849
1602 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1850 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1603 const size_t region [3] = { width, height, depth }; 1851 const size_t region [3] = { width, height, depth };
1604 EVENT_LIST (10);
1605 1852
1853 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)); 1854 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1607 1855
1608 if (ev) 1856 if (ev)
1609 XPUSH_CLOBJ (stash_event, ev); 1857 XPUSH_CLOBJ (stash_event, ev);
1610 1858
1611void 1859void
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, ...) 1860map_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: 1861 ALIAS:
1614 enqueue_map_buffer = 0 1862 enqueue_map_buffer = 0
1615 PPCODE: 1863 PPCODE:
1864 EVENT_LIST (6);
1865
1866 size_t cb = SvIV (cb_);
1867
1868 if (!SvOK (cb_))
1869 {
1870 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1871 cb -= offset;
1872 }
1873
1616 cl_event ev; 1874 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)); 1875 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)); 1876 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1621 1877
1622void 1878void
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, ...) 1879map_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: 1880 ALIAS:
1625 enqueue_map_image = 0 1881 enqueue_map_image = 0
1626 PPCODE: 1882 PPCODE:
1627 cl_event ev; 1883 size_t width = SvIV (width_);
1884 if (!SvOK (width_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1887 width -= x;
1888 }
1889
1890 size_t height = SvIV (width_);
1891 if (!SvOK (height_))
1892 {
1893 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1894 height -= y;
1895 }
1896
1897 size_t depth = SvIV (width_);
1898 if (!SvOK (depth_))
1899 {
1900 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1901 depth -= z;
1902
1903 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1904 if (!depth)
1905 depth = 1;
1906 }
1907
1628 const size_t origin[3] = { x, y, z }; 1908 const size_t origin[3] = { x, y, z };
1629 const size_t region[3] = { width, height, depth }; 1909 const size_t region[3] = { width, height, depth };
1630 size_t row_pitch, slice_pitch; 1910 size_t row_pitch, slice_pitch;
1631 EVENT_LIST (11); 1911 EVENT_LIST (10);
1632 1912
1913 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)); 1914 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 1915
1635 size_t cb = slice_pitch ? slice_pitch * region [2] 1916 size_t cb = slice_pitch ? slice_pitch * region [2]
1636 : row_pitch ? row_pitch * region [1] 1917 : row_pitch ? row_pitch * region [1]
1637 : region [0]; 1918 : region [0];
1639 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch)); 1920 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1640 1921
1641void 1922void
1642unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1923unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1643 PPCODE: 1924 PPCODE:
1644 cl_uint event_list_count = items - 2; 1925 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) 1926 if (GIMME_V != G_VOID)
1927 {
1928 clRetainEvent (mapped->event);
1651 XPUSH_CLOBJ (stash_event, mapped->event); 1929 XPUSH_CLOBJ (stash_event, mapped->event);
1930 }
1652 1931
1653void 1932void
1654task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1933task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1655 ALIAS: 1934 ALIAS:
1656 enqueue_task = 0 1935 enqueue_task = 0
1657 PPCODE: 1936 PPCODE:
1937 EVENT_LIST (2);
1938
1658 cl_event ev = 0; 1939 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)); 1940 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1662 1941
1663 if (ev) 1942 if (ev)
1664 XPUSH_CLOBJ (stash_event, ev); 1943 XPUSH_CLOBJ (stash_event, ev);
1665 1944
1666void 1945void
1667nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1946nd_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: 1947 ALIAS:
1669 enqueue_nd_range_kernel = 0 1948 enqueue_nd_range_kernel = 0
1670 PPCODE: 1949 PPCODE:
1671 cl_event ev = 0; 1950 EVENT_LIST (5);
1951
1672 size_t *gwo = 0, *gws, *lws = 0; 1952 size_t *gwo = 0, *gws, *lws = 0;
1673 int gws_len; 1953 int gws_len;
1674 size_t *lists; 1954 size_t *lists;
1675 int i; 1955 int i;
1676 EVENT_LIST (5);
1677 1956
1678 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1957 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1679 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1958 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1680 1959
1681 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1960 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1720 if (!lws [i]) 1999 if (!lws [i])
1721 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 2000 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1722 } 2001 }
1723 } 2002 }
1724 2003
2004 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)); 2005 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1726 2006
1727 if (ev) 2007 if (ev)
1728 XPUSH_CLOBJ (stash_event, ev); 2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#if CL_VERSION_1_2
2011
2012void
2013migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2014 ALIAS:
2015 enqueue_migrate_mem_objects = 0
2016 PPCODE:
2017 EVENT_LIST (3);
2018
2019 cl_uint obj_count;
2020 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2021
2022 cl_event ev = 0;
2023 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2024
2025 if (ev)
2026 XPUSH_CLOBJ (stash_event, ev);
2027
2028#endif
1729 2029
1730#if cl_apple_gl_sharing || cl_khr_gl_sharing 2030#if cl_apple_gl_sharing || cl_khr_gl_sharing
1731 2031
1732void 2032void
1733acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2033acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1734 ALIAS: 2034 ALIAS:
1735 release_gl_objects = 1 2035 release_gl_objects = 1
1736 enqueue_acquire_gl_objects = 0 2036 enqueue_acquire_gl_objects = 0
1737 enqueue_release_gl_objects = 1 2037 enqueue_release_gl_objects = 1
1738 PPCODE: 2038 PPCODE:
1739 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2039 EVENT_LIST (2);
1740 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2040
2041 cl_uint obj_count;
2042 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1741 2043
1742 cl_event ev = 0; 2044 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 2045
1752 if (ix) 2046 if (ix)
1753 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2047 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1754 else 2048 else
1755 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2049 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1756 2050
1757 if (ev) 2051 if (ev)
1758 XPUSH_CLOBJ (stash_event, ev); 2052 XPUSH_CLOBJ (stash_event, ev);
1759 2053
1760#endif 2054#endif
1774void 2068void
1775marker (OpenCL::Queue self, ...) 2069marker (OpenCL::Queue self, ...)
1776 ALIAS: 2070 ALIAS:
1777 enqueue_marker = 0 2071 enqueue_marker = 0
1778 PPCODE: 2072 PPCODE:
2073 EVENT_LIST (1);
1779 cl_event ev = 0; 2074 cl_event ev = 0;
1780 EVENT_LIST (1);
1781#if PREFER_1_1 2075#if PREFER_1_1
1782 if (!event_list_count) 2076 if (!event_list_count)
1783 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2077 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1784 else 2078 else
1785#if CL_VERSION_1_2 2079#if CL_VERSION_1_2
1799void 2093void
1800barrier (OpenCL::Queue self, ...) 2094barrier (OpenCL::Queue self, ...)
1801 ALIAS: 2095 ALIAS:
1802 enqueue_barrier = 0 2096 enqueue_barrier = 0
1803 PPCODE: 2097 PPCODE:
2098 EVENT_LIST (1);
1804 cl_event ev = 0; 2099 cl_event ev = 0;
1805 EVENT_LIST (1);
1806#if PREFER_1_1 2100#if PREFER_1_1
1807 if (!event_list_count && GIMME_V == G_VOID) 2101 if (!event_list_count && GIMME_V == G_VOID)
1808 NEED_SUCCESS (EnqueueBarrier, (self)); 2102 NEED_SUCCESS (EnqueueBarrier, (self));
1809 else 2103 else
1810#if CL_VERSION_1_2 2104#if CL_VERSION_1_2
2138void 2432void
2139build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2433build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2140 ALIAS: 2434 ALIAS:
2141 build_async = 1 2435 build_async = 1
2142 CODE: 2436 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; 2437 cl_uint device_count = 0;
2146 cl_device_id *device_list = 0; 2438 cl_device_id *device_list = 0;
2147 2439
2148 if (SvOK (devices)) 2440 if (SvOK (devices))
2149 { 2441 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2442
2443 void *user_data;
2444 program_callback pfn_notify = make_program_callback (notify, &user_data);
2445
2446 if (ix)
2447 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2448 else
2449 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2450
2451#if CL_VERSION_1_2
2452
2453void
2454compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2455 CODE:
2456 cl_uint device_count = 0;
2457 cl_device_id *device_list = 0;
2458
2459 if (SvOK (devices))
2460 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2461
2462 cl_uint header_count = 0;
2463 cl_program *header_list = 0;
2464 const char **header_name = 0;
2465
2466 if (SvOK (headers))
2467 {
2150 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2468 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2151 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2469 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2152 2470
2153 AV *av = (AV *)SvRV (devices); 2471 HV *hv = (HV *)SvRV (devices);
2154 num_devices = av_len (av) + 1;
2155 2472
2156 if (num_devices) 2473 header_count = hv_iterinit (hv);
2474 header_list = tmpbuf (sizeof (*header_list) * header_count);
2475 header_name = tmpbuf (sizeof (*header_name) * header_count);
2476
2477 HE *he;
2478 int i = 0;
2479 while (he = hv_iternext (hv))
2157 { 2480 {
2158 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2481 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2482 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2159 int count; 2483 ++i;
2160 for (count = 0; count < num_devices; ++count)
2161 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2162 } 2484 }
2163 } 2485 }
2164 2486
2165 if (SvOK (notify)) 2487 void *user_data;
2166 { 2488 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 2489
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)); 2490 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2491 header_count, header_list, header_name, pfn_notify, user_data));
2492
2493#endif
2176 2494
2177void 2495void
2178build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2496build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2179 PPCODE: 2497 PPCODE:
2180 size_t size; 2498 size_t size;
2208 char *value = tmpbuf (size); 2526 char *value = tmpbuf (size);
2209 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2527 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2210 EXTEND (SP, 1); 2528 EXTEND (SP, 1);
2211 const int i = 0; 2529 const int i = 0;
2212 PUSHs (sv_2mortal (newSVpv (value, 0))); 2530 PUSHs (sv_2mortal (newSVpv (value, 0)));
2531
2532void
2533binary_type (OpenCL::Program self, OpenCL::Device device)
2534 PPCODE:
2535 cl_program_binary_type value [1];
2536 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2537 EXTEND (SP, 1);
2538 const int i = 0;
2539 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2213 2540
2214#END:program_build 2541#END:program_build
2215 2542
2216void 2543void
2217kernel (OpenCL::Program program, SV *function) 2544kernel (OpenCL::Program program, SV *function)
2384 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2711 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2385 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2712 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2386 2713
2387 case 'z': nullarg = 1; size = SvIV (sv); break; 2714 case 'z': nullarg = 1; size = SvIV (sv); break;
2388 2715
2389 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2716 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; 2717 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; 2718 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2392 2719
2393 default: 2720 default:
2394 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2721 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2395 } 2722 }
2396 2723
2596 const int i = 0; 2923 const int i = 0;
2597 PUSHs (sv_2mortal (newSVuv (value [i]))); 2924 PUSHs (sv_2mortal (newSVuv (value [i])));
2598 2925
2599#END:kernel_work_group 2926#END:kernel_work_group
2600 2927
2928#if CL_VERSION_1_2
2929
2930void
2931arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2932 PPCODE:
2933 size_t size;
2934 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2935 SV *sv = sv_2mortal (newSV (size));
2936 SvUPGRADE (sv, SVt_PV);
2937 SvPOK_only (sv);
2938 SvCUR_set (sv, size);
2939 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2940 XPUSHs (sv);
2941
2942#BEGIN:kernel_arg
2943
2944void
2945arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2946 PPCODE:
2947 cl_kernel_arg_address_qualifier value [1];
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2949 EXTEND (SP, 1);
2950 const int i = 0;
2951 PUSHs (sv_2mortal (newSVuv (value [i])));
2952
2953void
2954arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2955 PPCODE:
2956 cl_kernel_arg_access_qualifier value [1];
2957 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2958 EXTEND (SP, 1);
2959 const int i = 0;
2960 PUSHs (sv_2mortal (newSVuv (value [i])));
2961
2962void
2963arg_type_name (OpenCL::Kernel self, cl_uint idx)
2964 ALIAS:
2965 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2966 arg_name = CL_KERNEL_ARG_NAME
2967 PPCODE:
2968 size_t size;
2969 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2970 char *value = tmpbuf (size);
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVpv (value, 0)));
2975
2976void
2977arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2978 PPCODE:
2979 cl_kernel_arg_type_qualifier value [1];
2980 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2981 EXTEND (SP, 1);
2982 const int i = 0;
2983 PUSHs (sv_2mortal (newSVuv (value [i])));
2984
2985#END:kernel_arg
2986
2987#endif
2988
2601MODULE = OpenCL PACKAGE = OpenCL::Event 2989MODULE = OpenCL PACKAGE = OpenCL::Event
2602 2990
2603void 2991void
2604DESTROY (OpenCL::Event self) 2992DESTROY (OpenCL::Event self)
2605 CODE: 2993 CODE:
2708 3096
2709 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0); 3097 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2710 mapped_detach (self, mapped); 3098 mapped_detach (self, mapped);
2711 3099
2712 clReleaseCommandQueue (mapped->queue); 3100 clReleaseCommandQueue (mapped->queue);
2713
2714 if (mapped->event)
2715 clReleaseEvent (mapped->event); 3101 clReleaseEvent (mapped->event);
2716
2717 Safefree (mapped); 3102 Safefree (mapped);
3103
3104void
3105unmap (OpenCL::Mapped self, ...)
3106 CODE:
3107 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2718 3108
2719bool 3109bool
2720mapped (OpenCL::Mapped self) 3110mapped (OpenCL::Mapped self)
2721 CODE: 3111 CODE:
2722 RETVAL = !!self->ptr; 3112 RETVAL = !!self->ptr;
2749ptr (OpenCL::Mapped self) 3139ptr (OpenCL::Mapped self)
2750 CODE: 3140 CODE:
2751 RETVAL = PTR2IV (self->ptr); 3141 RETVAL = PTR2IV (self->ptr);
2752 OUTPUT: 3142 OUTPUT:
2753 RETVAL 3143 RETVAL
3144
3145void
3146set (OpenCL::Mapped self, size_t offset, SV *data)
3147 CODE:
3148 STRLEN len;
3149 const char *ptr = SvPVbyte (data, len);
3150
3151 if (offset + len > self->cb)
3152 croak ("OpenCL::Mapped::set out of bound condition detected");
3153
3154 memcpy (offset + (char *)self->ptr, ptr, len);
2754 3155
2755MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer 3156MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2756 3157
2757MODULE = OpenCL PACKAGE = OpenCL::MappedImage 3158MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2758 3159

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines