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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.47 by root, Tue Apr 24 13:30:49 2012 UTC vs.
Revision 1.75 by root, Sat May 5 14:18:35 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
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
4 10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
7#endif 13#endif
8 14
22#if PREFER_1_1 28#if PREFER_1_1
23 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
24#endif 30#endif
25 31
26#ifdef __APPLE__ 32#ifdef __APPLE__
27 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
28#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
29 #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>
30#endif 50#endif
31 51
32#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
33 #undef PREFER_1_1 53 #undef PREFER_1_1
34 #define PREFER_1_1 1 54 #define PREFER_1_1 1
35#endif 55#endif
36 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
37typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
38typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
39typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
40typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
41typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
42typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
43typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
44typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
45typedef cl_mem OpenCL__Image2D;
46typedef cl_mem OpenCL__Image3D;
47typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
48typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
49typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
50typedef cl_mem OpenCL__Image2D_ornull;
51typedef cl_mem OpenCL__Image3D_ornull;
52typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
53typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
54typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
55typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
56typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
57 77
58typedef SV *FUTURE; 78typedef struct mapped * OpenCL__Mapped;
79
80static HV
81 *stash_platform,
82 *stash_device,
83 *stash_subdevice,
84 *stash_context,
85 *stash_queue,
86 *stash_program,
87 *stash_kernel,
88 *stash_sampler,
89 *stash_event,
90 *stash_userevent,
91 *stash_memory,
92 *stash_buffer,
93 *stash_bufferobj,
94 *stash_image,
95 *stash_image1d,
96 *stash_image1darray,
97 *stash_image1dbuffer,
98 *stash_image2d,
99 *stash_image2darray,
100 *stash_image3d,
101 *stash_mapped,
102 *stash_mappedbuffer,
103 *stash_mappedimage;
59 104
60/*****************************************************************************/ 105/*****************************************************************************/
61 106
62// name must include a leading underscore 107// name must include a leading underscore
63// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
103} 148}
104 149
105/*****************************************************************************/ 150/*****************************************************************************/
106 151
107/* up to two temporary buffers */ 152/* up to two temporary buffers */
108static void * 153static void * ecb_noinline
109tmpbuf (size_t size) 154tmpbuf (size_t size)
110{ 155{
111 enum { buffers = 3 }; 156 enum { buffers = 4 };
112 static int idx; 157 static int idx;
113 static void *buf [buffers]; 158 static void *buf [buffers];
114 static size_t len [buffers]; 159 static size_t len [buffers];
115 160
116 idx = (idx + 1) % buffers; 161 idx = (idx + 1) % buffers;
121 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
122 buf [idx] = malloc (len [idx]); 167 buf [idx] = malloc (len [idx]);
123 } 168 }
124 169
125 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;
126} 186}
127 187
128/*****************************************************************************/ 188/*****************************************************************************/
129 189
130typedef struct 190typedef struct
189 if (res) \ 249 if (res) \
190 FAIL (name); 250 FAIL (name);
191 251
192/*****************************************************************************/ 252/*****************************************************************************/
193 253
194#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
195#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
196
197static void * 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
198SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 266SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
199{ 267{
268 // sv_derived_from is quite slow :(
200 if (SvROK (sv) && sv_derived_from (sv, pkg)) 269 if (SvROK (sv) && sv_derived_from (sv, pkg))
201 return (void *)SvIV (SvRV (sv)); 270 return (void *)SvIV (SvRV (sv));
202 271
203 croak ("%s: %s is not of type %s", func, svname, pkg); 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));
204} 283}
205 284
206/*****************************************************************************/ 285/*****************************************************************************/
207 286
208static cl_context_properties * 287static cl_context_properties * ecb_noinline
209SvCONTEXTPROPERTIES (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)
210{ 289{
211 if (!sv || !SvOK (sv)) 290 if (!sv || !SvOK (sv))
212 if (extra) 291 if (extra)
213 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
214 else 293 else
220 int i, len = av_len (av) + 1; 299 int i, len = av_len (av) + 1;
221 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));
222 cl_context_properties *l = p; 301 cl_context_properties *l = p;
223 302
224 if (len & 1) 303 if (len & 1)
225 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);
226 305
227 while (extracount--) 306 while (extracount--)
228 *l++ = *extra++; 307 *l++ = *extra++;
229 308
230 for (i = 0; i < len; i += 2) 309 for (i = 0; i < len; i += 2)
233 SV *p_sv = *av_fetch (av, i + 1, 0); 312 SV *p_sv = *av_fetch (av, i + 1, 0);
234 cl_context_properties v = SvIV (p_sv); // code below can override 313 cl_context_properties v = SvIV (p_sv); // code below can override
235 314
236 switch (t) 315 switch (t)
237 { 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
238 case CL_GLX_DISPLAY_KHR: 322 case CL_GLX_DISPLAY_KHR:
239 if (!SvOK (p_sv)) 323 if (!SvOK (p_sv))
240 { 324 {
241 void *func = glsym ("_glXGetCurrentDisplay"); 325 void *func = glsym ("_glXGetCurrentDisplay");
242 if (func) 326 if (func)
265 *l = 0; 349 *l = 0;
266 350
267 return p; 351 return p;
268 } 352 }
269 353
270 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);
355}
356
357// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
358// is a pointer (and only on well-behaved systems).
359static void * ecb_noinline
360object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
361{
362 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
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);
365
366 AV *av = (AV *)SvRV (arg);
367 void **list = 0;
368 cl_uint count = av_len (av) + 1;
369
370 if (count)
371 {
372 list = tmpbuf (sizeof (*list) * count);
373 int i;
374 for (i = 0; i < count; ++i)
375 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
376 }
377
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;
271} 384}
272 385
273/*****************************************************************************/ 386/*****************************************************************************/
387/* callback stuff */
388
389/* default context callback, log to stderr */
390static void CL_CALLBACK
391context_default_notify (const char *msg, const void *info, size_t cb, void *data)
392{
393 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
394}
395
396typedef struct
397{
398 int free_cb;
399 void (*push)(void *data1, void *data2, void *data3);
400} eq_vtbl;
401
402typedef struct eq_item
403{
404 struct eq_item *next;
405 eq_vtbl *vtbl;
406 SV *cb;
407 void *data1, *data2, *data3;
408} eq_item;
409
410static void (*eq_signal_func)(void *signal_arg, int value);
411static void *eq_signal_arg;
412static xmutex_t eq_lock = X_MUTEX_INIT;
413static eq_item *eq_head, *eq_tail;
414
415static void ecb_noinline
416eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
417{
418 eq_item *item = malloc (sizeof (eq_item));
419
420 item->next = 0;
421 item->vtbl = vtbl;
422 item->cb = cb;
423 item->data1 = data1;
424 item->data2 = data2;
425 item->data3 = data3;
426
427 X_LOCK (eq_lock);
428
429 *(eq_head ? &eq_tail->next : &eq_head) = item;
430 eq_tail = item;
431
432 X_UNLOCK (eq_lock);
433
434 eq_signal_func (eq_signal_arg, 0);
435}
436
437static eq_item *
438eq_dec (void)
439{
440 eq_item *res;
441
442 X_LOCK (eq_lock);
443
444 res = eq_head;
445 if (res)
446 eq_head = res->next;
447
448 X_UNLOCK (eq_lock);
449
450 return res;
451}
452
453static void
454eq_poll (void)
455{
456 eq_item *item;
457
458 while ((item = eq_dec ()))
459 {
460 ENTER;
461 SAVETMPS;
462
463 dSP;
464 PUSHMARK (SP);
465 EXTEND (SP, 2);
466
467 if (item->vtbl->free_cb)
468 sv_2mortal (item->cb);
469
470 PUTBACK;
471 item->vtbl->push (item->data1, item->data2, item->data3);
472
473 SV *cb = item->cb;
474 free (item);
475
476 call_sv (cb, G_DISCARD | G_VOID);
477
478 FREETMPS;
479 LEAVE;
480 }
481}
482
483static void
484eq_poll_interrupt (pTHX_ void *c_arg, int value)
485{
486 eq_poll ();
487}
488
489/*****************************************************************************/
490/* context notify */
491
492static void ecb_noinline
493eq_context_push (void *data1, void *data2, void *data3)
494{
495 dSP;
496 PUSHs (sv_2mortal (newSVpv (data1, 0)));
497 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
498 PUTBACK;
499
500 free (data1);
501 free (data2);
502}
503
504static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
505
506static void CL_CALLBACK
507eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
508{
509 void *pvt_copy = malloc (cb);
510 memcpy (pvt_copy, pvt, cb);
511 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
512}
513
514#define CONTEXT_NOTIFY_CALLBACK \
515 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
516 void *user_data = 0; \
517 \
518 if (SvOK (notify)) \
519 { \
520 pfn_notify = eq_context_notify; \
521 user_data = s_get_cv (notify); \
522 }
523
524static SV * ecb_noinline
525new_clobj_context (cl_context ctx, void *user_data)
526{
527 SV *sv = new_clobj (stash_context, (IV)ctx);
528
529 if (user_data)
530 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
531
532 return sv;
533}
534
535#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
536
537/*****************************************************************************/
538/* build/compile/link notify */
539
540static void
541eq_program_push (void *data1, void *data2, void *data3)
542{
543 dSP;
544 PUSH_CLOBJ (stash_program, data1);
545 PUTBACK;
546}
547
548static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
549
550static void CL_CALLBACK
551eq_program_notify (cl_program program, void *user_data)
552{
553 clRetainProgram (program);
554
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 }
573}
574
575struct build_args
576{
577 cl_program program;
578 char *options;
579 void *user_data;
580 cl_uint num_devices;
581};
582
583X_THREAD_PROC (build_program_thread)
584{
585 struct build_args *arg = thr_arg;
586
587 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
588
589 if (arg->user_data)
590 eq_program_notify (arg->program, arg->user_data);
591 else
592 clReleaseProgram (arg->program);
593
594 free (arg->options);
595 free (arg);
596
597 return 0;
598}
599
600static void
601build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
602{
603 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
604
605 arg->program = program;
606 arg->options = strdup (options);
607 arg->user_data = user_data;
608 arg->num_devices = num_devices;
609 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
610
611 xthread_t id;
612 thread_create (&id, build_program_thread, arg);
613}
614
615/*****************************************************************************/
616/* event objects */
617
618static void
619eq_event_push (void *data1, void *data2, void *data3)
620{
621 dSP;
622 PUSH_CLOBJ (stash_event, data1);
623 PUSHs (sv_2mortal (newSViv ((IV)data2)));
624 PUTBACK;
625}
626
627static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
628
629static void CL_CALLBACK
630eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
631{
632 clRetainEvent (event);
633 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
634}
635
636/*****************************************************************************/
637/* utilities for XS code */
274 638
275static size_t 639static size_t
276img_row_pitch (cl_mem img) 640img_row_pitch (cl_mem img)
277{ 641{
278 size_t res; 642 size_t res;
279 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 643 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
280 return res; 644 return res;
281} 645}
282 646
283static cl_event * 647static cl_event * ecb_noinline
284event_list (SV **items, cl_uint *rcount) 648event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
285{ 649{
286 cl_uint count = *rcount; 650 cl_uint count = *rcount;
287 651
288 if (!count) 652 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
653 *rcount = count = 0;
654
655 if (!count && !extra)
289 return 0; 656 return 0;
290 657
291 cl_event *list = tmpbuf (sizeof (cl_event) * count); 658 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
292 int i = 0; 659 int i = 0;
293 660
294 do 661 while (count--)
295 {
296 --count;
297 if (SvOK (items [count])) 662 if (SvOK (items [count]))
298 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 663 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
299 } 664
300 while (count); 665 if (extra)
666 list [i++] = extra;
301 667
302 *rcount = i; 668 *rcount = i;
303 669
304 return i ? list : 0; 670 return i ? list : 0;
305} 671}
306 672
307#define EVENT_LIST(items,count) \ 673#define EVENT_LIST(skip) \
308 cl_uint event_list_count = (count); \ 674 cl_uint event_list_count = items - (skip); \
309 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 675 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
310 676
311#define INFO(class) \ 677#define INFO(class) \
312{ \ 678{ \
313 size_t size; \ 679 size_t size; \
314 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 680 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
318 SvCUR_set (sv, size); \ 684 SvCUR_set (sv, size); \
319 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 685 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
320 XPUSHs (sv); \ 686 XPUSHs (sv); \
321} 687}
322 688
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/*****************************************************************************/
781
323MODULE = OpenCL PACKAGE = OpenCL 782MODULE = OpenCL PACKAGE = OpenCL
324 783
325PROTOTYPES: ENABLE 784PROTOTYPES: ENABLE
785
786void
787poll ()
788 CODE:
789 eq_poll ();
790
791void
792_eq_initialise (IV func, IV arg)
793 CODE:
794 eq_signal_func = (void (*)(void *, int))func;
795 eq_signal_arg = (void*)arg;
326 796
327BOOT: 797BOOT:
328{ 798{
329 HV *stash = gv_stashpv ("OpenCL", 1); 799 HV *stash = gv_stashpv ("OpenCL", 1);
330 static const ivstr *civ, const_iv[] = { 800 static const ivstr *civ, const_iv[] = {
337 { sizeof (cl_long ), "SIZEOF_LONG" }, 807 { sizeof (cl_long ), "SIZEOF_LONG" },
338 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 808 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
339 { sizeof (cl_half ), "SIZEOF_HALF" }, 809 { sizeof (cl_half ), "SIZEOF_HALF" },
340 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 810 { sizeof (cl_float ), "SIZEOF_FLOAT" },
341 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 811 { sizeof (cl_double), "SIZEOF_DOUBLE" },
812 { PREFER_1_1 , "PREFER_1_1" },
342#include "constiv.h" 813#include "constiv.h"
343 }; 814 };
815
344 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--)
345 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 817 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
818
819 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
820 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
821 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
822 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
823 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
824 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
825 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
826 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
827 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
828 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
829 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
830 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
831 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
832 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
833 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
834 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
835 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
836 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
837 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
838 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
839 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
840 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
841 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
842
843 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
346} 844}
347 845
348cl_int 846cl_int
349errno () 847errno ()
350 CODE: 848 CODE:
351 RETVAL = res; 849 RETVAL = res;
352 OUTPUT: 850 OUTPUT:
353 RETVAL 851 RETVAL
354 852
355const char * 853const char *
356err2str (cl_int err) 854err2str (cl_int err = res)
357 855
358const char * 856const char *
359enum2str (cl_uint value) 857enum2str (cl_uint value)
360 858
361void 859void
369 list = tmpbuf (sizeof (*list) * count); 867 list = tmpbuf (sizeof (*list) * count);
370 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 868 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
371 869
372 EXTEND (SP, count); 870 EXTEND (SP, count);
373 for (i = 0; i < count; ++i) 871 for (i = 0; i < count; ++i)
374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 872 PUSH_CLOBJ (stash_platform, list [i]);
375 873
376void 874void
377context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 875context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
378 PPCODE: 876 PPCODE:
877 CONTEXT_NOTIFY_CALLBACK;
379 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));
380 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 879 XPUSH_CLOBJ_CONTEXT;
381 880
382#if 0
383
384void 881void
385context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 882context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
386 PPCODE: 883 PPCODE:
387 /* 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);
388 886
389#endif 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;
390 890
391void 891void
392wait_for_events (...) 892wait_for_events (...)
393 CODE: 893 CODE:
394 EVENT_LIST (0, items); 894 EVENT_LIST (0);
395 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 895 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
396 896
397PROTOTYPES: DISABLE 897PROTOTYPES: DISABLE
398 898
399MODULE = OpenCL PACKAGE = OpenCL::Platform 899MODULE = OpenCL PACKAGE = OpenCL::Platform
442 list = tmpbuf (sizeof (*list) * count); 942 list = tmpbuf (sizeof (*list) * count);
443 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 943 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
444 944
445 EXTEND (SP, count); 945 EXTEND (SP, count);
446 for (i = 0; i < count; ++i) 946 for (i = 0; i < count; ++i)
447 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 947 PUSH_CLOBJ (stash_device, list [i]);
448 948
449void 949void
450context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 950context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
451 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:
452 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 984 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
453 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");
454 986
455 AV *av = (AV *)SvRV (devices); 987 properties = SvRV (properties);
456 cl_uint num_devices = av_len (av) + 1; 988
457 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);
991
458 int i; 992 int i;
993 for (i = 0; i < count; ++i)
994 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
459 995
460 for (i = num_devices; i--; ) 996 props [count] = 0;
461 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
462 997
463 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 998 cl_uint num_devices;
464 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 999 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1000 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1001 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
465 1002
466void 1003 EXTEND (SP, num_devices);
467context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1004 for (i = 0; i < count; ++i)
468 PPCODE: 1005 PUSH_CLOBJ (stash_subdevice, list [i]);
469 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
470 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
471 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
472 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
473 1006
474MODULE = OpenCL PACKAGE = OpenCL::Device 1007#endif
475
476void
477info (OpenCL::Device self, cl_device_info name)
478 PPCODE:
479 INFO (Device)
480 1008
481#BEGIN:device 1009#BEGIN:device
482 1010
483void 1011void
484type (OpenCL::Device self) 1012type (OpenCL::Device self)
485 PPCODE: 1013 PPCODE:
486 cl_device_type value [1]; 1014 cl_device_type value [1];
487 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1015 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
488 EXTEND (SP, 1); 1016 EXTEND (SP, 1);
489 const int i = 0; 1017 const int i = 0;
490 PUSHs (sv_2mortal (newSViv (value [i]))); 1018 PUSHs (sv_2mortal (newSVuv (value [i])));
491 1019
492void 1020void
493vendor_id (OpenCL::Device self) 1021vendor_id (OpenCL::Device self)
494 ALIAS: 1022 ALIAS:
495 vendor_id = CL_DEVICE_VENDOR_ID 1023 vendor_id = CL_DEVICE_VENDOR_ID
640 PPCODE: 1168 PPCODE:
641 cl_command_queue_properties value [1]; 1169 cl_command_queue_properties value [1];
642 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1170 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
643 EXTEND (SP, 1); 1171 EXTEND (SP, 1);
644 const int i = 0; 1172 const int i = 0;
645 PUSHs (sv_2mortal (newSViv (value [i]))); 1173 PUSHs (sv_2mortal (newSVuv (value [i])));
646 1174
647void 1175void
648platform (OpenCL::Device self) 1176platform (OpenCL::Device self)
649 PPCODE: 1177 PPCODE:
650 cl_platform_id value [1]; 1178 cl_platform_id value [1];
651 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1179 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
652 EXTEND (SP, 1); 1180 EXTEND (SP, 1);
653 const int i = 0; 1181 const int i = 0;
654 { 1182 {
655 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1183 PUSH_CLOBJ (stash_platform, value [i]);
656 } 1184 }
657 1185
658void 1186void
659name (OpenCL::Device self) 1187name (OpenCL::Device self)
660 ALIAS: 1188 ALIAS:
679 cl_device_id value [1]; 1207 cl_device_id value [1];
680 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1208 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
681 EXTEND (SP, 1); 1209 EXTEND (SP, 1);
682 const int i = 0; 1210 const int i = 0;
683 { 1211 {
684 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1212 PUSH_CLOBJ (stash_device, value [i]);
685 } 1213 }
686 1214
687void 1215void
688partition_types_ext (OpenCL::Device self) 1216partition_types_ext (OpenCL::Device self)
689 ALIAS: 1217 ALIAS:
700 for (i = 0; i < n; ++i) 1228 for (i = 0; i < n; ++i)
701 PUSHs (sv_2mortal (newSVuv (value [i]))); 1229 PUSHs (sv_2mortal (newSVuv (value [i])));
702 1230
703#END:device 1231#END:device
704 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
705MODULE = OpenCL PACKAGE = OpenCL::Context 1244MODULE = OpenCL PACKAGE = OpenCL::Context
706 1245
707void 1246void
708DESTROY (OpenCL::Context context) 1247DESTROY (OpenCL::Context self)
709 CODE: 1248 CODE:
710 clReleaseContext (context); 1249 clReleaseContext (self);
711 1250
712void 1251void
713info (OpenCL::Context self, cl_context_info name) 1252info (OpenCL::Context self, cl_context_info name)
714 PPCODE: 1253 PPCODE:
715 INFO (Context) 1254 INFO (Context)
716 1255
717void 1256void
718queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1257queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
719 PPCODE: 1258 PPCODE:
720 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1259 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
721 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1260 XPUSH_CLOBJ (stash_queue, queue);
722 1261
723void 1262void
724user_event (OpenCL::Context self) 1263user_event (OpenCL::Context self)
725 PPCODE: 1264 PPCODE:
726 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1265 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
727 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1266 XPUSH_CLOBJ (stash_userevent, ev);
728 1267
729void 1268void
730buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1269buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
731 PPCODE: 1270 PPCODE:
732 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1271 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
733 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1272 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
734 1273
735 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1274 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
736 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1275 XPUSH_CLOBJ (stash_bufferobj, mem);
737 1276
738void 1277void
739buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1278buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
740 PPCODE: 1279 PPCODE:
741 STRLEN len; 1280 STRLEN len;
742 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1281 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
743 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1282 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
744 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1283 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
745 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1284 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
746 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1285 XPUSH_CLOBJ (stash_bufferobj, mem);
747 1286
748#if CL_VERSION_1_2 1287#if CL_VERSION_1_2
749 1288
750void 1289void
751image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1290image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
752 PPCODE: 1291 PPCODE:
753 STRLEN len; 1292 STRLEN len;
754 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1293 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
755 const cl_image_format format = { channel_order, channel_type }; 1294 const cl_image_format format = { channel_order, channel_type };
756 const cl_image_desc desc = { 1295 const cl_image_desc desc = {
757 type, 1296 type,
758 width, height, depth, 1297 width, height, depth,
759 array_size, row_pitch, slice_pitch, 1298 array_size, row_pitch, slice_pitch,
760 num_mip_level, num_samples, 1299 num_mip_level, num_samples,
761 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvPTROBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1300 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
762 }; 1301 };
763 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));
764 char *klass = "OpenCL::Image"; 1303 HV *stash = stash_image;
765 switch (type) 1304 switch (type)
766 { 1305 {
767 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1306 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
768 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1307 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
769 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1308 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
770 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1309 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
771 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1310 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
772 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1311 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
773 } 1312 }
774 XPUSH_NEW_OBJ (klass, mem); 1313 XPUSH_CLOBJ (stash, mem);
775 1314
776#endif 1315#endif
777 1316
778void 1317void
779image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1318image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
785 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
786#else 1325#else
787 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1326 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
788 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1327 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
789#endif 1328#endif
790 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1329 XPUSH_CLOBJ (stash_image2d, mem);
791 1330
792void 1331void
793image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1332image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
794 PPCODE: 1333 PPCODE:
795 STRLEN len; 1334 STRLEN len;
799 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
800#else 1339#else
801 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1340 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
802 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
803#endif 1342#endif
804 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1343 XPUSH_CLOBJ (stash_image3d, mem);
805 1344
806#if cl_apple_gl_sharing || cl_khr_gl_sharing 1345#if cl_apple_gl_sharing || cl_khr_gl_sharing
807 1346
808void 1347void
809gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1348gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
810 PPCODE: 1349 PPCODE:
811 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1350 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
812 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1351 XPUSH_CLOBJ (stash_bufferobj, mem);
813 1352
814void 1353void
815gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1354gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
816 PPCODE: 1355 PPCODE:
817 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1356 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
818 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1357 XPUSH_CLOBJ (stash_image2d, mem);
819 1358
820#if CL_VERSION_1_2 1359#if CL_VERSION_1_2
821 1360
822void 1361void
823gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1362gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
824 ALIAS: 1363 ALIAS:
825 PPCODE: 1364 PPCODE:
826 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1365 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
827 cl_gl_object_type type; 1366 cl_gl_object_type type;
828 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1367 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
829 char *klass = "OpenCL::Memory"; 1368 HV *stash = stash_memory;
830 switch (type) 1369 switch (type)
831 { 1370 {
832 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1371 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
833 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1372 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
834 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1373 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
835 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1374 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
836 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1375 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
837 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1376 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
838 } 1377 }
839 XPUSH_NEW_OBJ (klass, mem); 1378 XPUSH_CLOBJ (stash, mem);
840 1379
841#endif 1380#endif
842 1381
843void 1382void
844gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1383gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
846#if PREFER_1_1 1385#if PREFER_1_1
847 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1386 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
848#else 1387#else
849 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1388 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
850#endif 1389#endif
851 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1390 XPUSH_CLOBJ (stash_image2d, mem);
852 1391
853void 1392void
854gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1393gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
855 PPCODE: 1394 PPCODE:
856#if PREFER_1_1 1395#if PREFER_1_1
857 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1396 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
858#else 1397#else
859 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1398 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
860#endif 1399#endif
861 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1400 XPUSH_CLOBJ (stash_image3d, mem);
862 1401
863#endif 1402#endif
864 1403
865void 1404void
866supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1405supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
886 1425
887void 1426void
888sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1427sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
889 PPCODE: 1428 PPCODE:
890 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1429 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
891 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1430 XPUSH_CLOBJ (stash_sampler, sampler);
892 1431
893void 1432void
894program_with_source (OpenCL::Context self, SV *program) 1433program_with_source (OpenCL::Context self, SV *program)
895 PPCODE: 1434 PPCODE:
896 STRLEN len; 1435 STRLEN len;
897 size_t len2; 1436 size_t len2;
898 const char *ptr = SvPVbyte (program, len); 1437 const char *ptr = SvPVbyte (program, len);
899 1438
900 len2 = len; 1439 len2 = len;
901 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));
902 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1441 XPUSH_CLOBJ (stash_program, prog);
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
903 1518
904#BEGIN:context 1519#BEGIN:context
905 1520
906void 1521void
907reference_count (OpenCL::Context self) 1522reference_count (OpenCL::Context self)
924 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1539 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
925 int i, n = size / sizeof (*value); 1540 int i, n = size / sizeof (*value);
926 EXTEND (SP, n); 1541 EXTEND (SP, n);
927 for (i = 0; i < n; ++i) 1542 for (i = 0; i < n; ++i)
928 { 1543 {
929 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1544 PUSH_CLOBJ (stash_device, value [i]);
930 } 1545 }
931 1546
932void 1547void
933properties (OpenCL::Context self) 1548properties (OpenCL::Context self)
934 PPCODE: 1549 PPCODE:
949DESTROY (OpenCL::Queue self) 1564DESTROY (OpenCL::Queue self)
950 CODE: 1565 CODE:
951 clReleaseCommandQueue (self); 1566 clReleaseCommandQueue (self);
952 1567
953void 1568void
954enqueue_read_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, ...)
1570 ALIAS:
1571 enqueue_read_buffer = 0
955 PPCODE: 1572 PPCODE:
956 cl_event ev = 0;
957 EVENT_LIST (6, items - 6); 1573 EVENT_LIST (6);
958 1574
959 SvUPGRADE (data, SVt_PV); 1575 SvUPGRADE (data, SVt_PV);
960 SvGROW (data, len); 1576 SvGROW (data, len);
961 SvPOK_only (data); 1577 SvPOK_only (data);
962 SvCUR_set (data, len); 1578 SvCUR_set (data, len);
1579
1580 cl_event ev = 0;
963 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));
964 1582
965 if (ev) 1583 if (ev)
966 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1584 XPUSH_CLOBJ (stash_event, ev);
967 1585
968void 1586void
969enqueue_write_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, ...)
1588 ALIAS:
1589 enqueue_write_buffer = 0
970 PPCODE: 1590 PPCODE:
971 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
972 STRLEN len; 1593 STRLEN len;
973 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
974 EVENT_LIST (5, items - 5);
975 1595
1596 cl_event ev = 0;
976 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));
977 1598
978 if (ev) 1599 if (ev)
979 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1600 XPUSH_CLOBJ (stash_event, ev);
980 1601
1602#if CL_VERSION_1_2
1603
981void 1604void
982enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1605fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1606 ALIAS:
1607 enqueue_fill_buffer = 0
983 PPCODE: 1608 PPCODE:
1609 EVENT_LIST (5);
1610
1611 STRLEN len;
1612 char *ptr = SvPVbyte (data, len);
1613
984 cl_event ev = 0; 1614 cl_event ev = 0;
1615 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1616
1617 if (ev)
1618 XPUSH_CLOBJ (stash_event, ev);
1619
1620void
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, ...)
1622 ALIAS:
1623 enqueue_fill_image = 0
1624 PPCODE:
1625 EVENT_LIST (12);
1626
1627 const size_t origin [3] = { x, y, z };
1628 const size_t region [3] = { width, height, depth };
1629
1630 const cl_float c_f [4] = { r, g, b, a };
1631 const cl_uint c_u [4] = { r, g, b, a };
1632 const cl_int c_s [4] = { r, g, b, a };
1633 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1634 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1635 cl_image_format format;
1636 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1637 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1638 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1639 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1640
1641 cl_event ev = 0;
1642 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1643 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1644
1645 if (ev)
1646 XPUSH_CLOBJ (stash_event, ev);
1647
1648#endif
1649
1650void
1651copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1652 ALIAS:
1653 enqueue_copy_buffer = 0
1654 PPCODE:
985 EVENT_LIST (6, items - 6); 1655 EVENT_LIST (6);
986 1656
1657 cl_event ev = 0;
987 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));
988 1659
989 if (ev) 1660 if (ev)
990 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1661 XPUSH_CLOBJ (stash_event, ev);
991 1662
992void 1663void
993enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 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, ...)
1665 ALIAS:
1666 enqueue_read_buffer_rect = 0
994 PPCODE: 1667 PPCODE:
995 cl_event ev = 0; 1668 EVENT_LIST (17);
1669
996 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 };
997 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 };
998 const size_t region[3] = { width, height, depth }; 1672 const size_t region[3] = { width, height, depth };
999 EVENT_LIST (17, items - 17);
1000 1673
1001 if (!buf_row_pitch) 1674 if (!buf_row_pitch)
1002 buf_row_pitch = region [0]; 1675 buf_row_pitch = region [0];
1003 1676
1004 if (!buf_slice_pitch) 1677 if (!buf_slice_pitch)
1014 1687
1015 SvUPGRADE (data, SVt_PV); 1688 SvUPGRADE (data, SVt_PV);
1016 SvGROW (data, len); 1689 SvGROW (data, len);
1017 SvPOK_only (data); 1690 SvPOK_only (data);
1018 SvCUR_set (data, len); 1691 SvCUR_set (data, len);
1692
1693 cl_event ev = 0;
1019 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));
1020 1695
1021 if (ev) 1696 if (ev)
1022 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1697 XPUSH_CLOBJ (stash_event, ev);
1023 1698
1024void 1699void
1025enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 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, ...)
1701 ALIAS:
1702 enqueue_write_buffer_rect = 0
1026 PPCODE: 1703 PPCODE:
1027 cl_event ev = 0; 1704 EVENT_LIST (17);
1705
1028 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 };
1029 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 };
1030 const size_t region[3] = { width, height, depth }; 1708 const size_t region[3] = { width, height, depth };
1031 STRLEN len; 1709 STRLEN len;
1032 char *ptr = SvPVbyte (data, len); 1710 char *ptr = SvPVbyte (data, len);
1033 EVENT_LIST (17, items - 17);
1034 1711
1035 if (!buf_row_pitch) 1712 if (!buf_row_pitch)
1036 buf_row_pitch = region [0]; 1713 buf_row_pitch = region [0];
1037 1714
1038 if (!buf_slice_pitch) 1715 if (!buf_slice_pitch)
1047 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];
1048 1725
1049 if (len < min_len) 1726 if (len < min_len)
1050 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1727 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1051 1728
1729 cl_event ev = 0;
1052 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));
1053 1731
1054 if (ev) 1732 if (ev)
1055 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1733 XPUSH_CLOBJ (stash_event, ev);
1056 1734
1057void 1735void
1058enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 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, ...)
1737 ALIAS:
1738 enqueue_copy_buffer_rect = 0
1059 PPCODE: 1739 PPCODE:
1060 cl_event ev = 0; 1740 EVENT_LIST (16);
1741
1061 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 };
1062 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 };
1063 const size_t region[3] = { width, height, depth }; 1744 const size_t region[3] = { width, height, depth };
1064 EVENT_LIST (16, items - 16);
1065 1745
1746 cl_event ev = 0;
1066 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));
1067 1748
1068 if (ev) 1749 if (ev)
1069 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1070 1751
1071void 1752void
1072enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 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, ...)
1754 ALIAS:
1755 enqueue_read_image = 0
1073 PPCODE: 1756 PPCODE:
1074 cl_event ev = 0; 1757 EVENT_LIST (12);
1758
1075 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 };
1076 const size_t region[3] = { width, height, depth }; 1760 const size_t region[3] = { width, height, depth };
1077 EVENT_LIST (12, items - 12);
1078 1761
1079 if (!row_pitch) 1762 if (!row_pitch)
1080 row_pitch = img_row_pitch (src); 1763 row_pitch = img_row_pitch (src);
1081 1764
1082 if (depth > 1 && !slice_pitch) 1765 if (depth > 1 && !slice_pitch)
1086 1769
1087 SvUPGRADE (data, SVt_PV); 1770 SvUPGRADE (data, SVt_PV);
1088 SvGROW (data, len); 1771 SvGROW (data, len);
1089 SvPOK_only (data); 1772 SvPOK_only (data);
1090 SvCUR_set (data, len); 1773 SvCUR_set (data, len);
1774
1775 cl_event ev = 0;
1091 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));
1092 1777
1093 if (ev) 1778 if (ev)
1094 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1779 XPUSH_CLOBJ (stash_event, ev);
1095 1780
1096void 1781void
1097enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 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, ...)
1783 ALIAS:
1784 enqueue_write_image = 0
1098 PPCODE: 1785 PPCODE:
1099 cl_event ev = 0; 1786 EVENT_LIST (12);
1787
1100 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 };
1101 const size_t region[3] = { width, height, depth }; 1789 const size_t region[3] = { width, height, depth };
1102 STRLEN len; 1790 STRLEN len;
1103 char *ptr = SvPVbyte (data, len); 1791 char *ptr = SvPVbyte (data, len);
1104 EVENT_LIST (12, items - 12);
1105 1792
1106 if (!row_pitch) 1793 if (!row_pitch)
1107 row_pitch = img_row_pitch (dst); 1794 row_pitch = img_row_pitch (dst);
1108 1795
1109 if (depth > 1 && !slice_pitch) 1796 if (depth > 1 && !slice_pitch)
1112 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;
1113 1800
1114 if (len < min_len) 1801 if (len < min_len)
1115 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1802 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1116 1803
1804 cl_event ev = 0;
1117 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));
1118 1806
1119 if (ev) 1807 if (ev)
1120 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1121 1809
1122void 1810void
1123enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 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, ...)
1812 ALIAS:
1813 enqueue_copy_image = 0
1124 PPCODE: 1814 PPCODE:
1125 cl_event ev = 0; 1815 EVENT_LIST (12);
1816
1126 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 };
1127 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 };
1128 const size_t region[3] = { width, height, depth }; 1819 const size_t region[3] = { width, height, depth };
1129 EVENT_LIST (12, items - 12);
1130 1820
1821 cl_event ev = 0;
1131 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));
1132 1823
1133 if (ev) 1824 if (ev)
1134 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1135 1826
1136void 1827void
1137enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 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, ...)
1829 ALIAS:
1830 enqueue_copy_image_to_buffer = 0
1138 PPCODE: 1831 PPCODE:
1832 EVENT_LIST (10);
1833
1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1835 const size_t region [3] = { width, height, depth };
1836
1139 cl_event ev = 0; 1837 cl_event 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));
1839
1840 if (ev)
1841 XPUSH_CLOBJ (stash_event, ev);
1842
1843void
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, ...)
1845 ALIAS:
1846 enqueue_copy_buffer_to_image = 0
1847 PPCODE:
1848 EVENT_LIST (10);
1849
1140 const size_t src_origin[3] = { src_x, src_y, src_z }; 1850 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1851 const size_t region [3] = { width, height, depth };
1852
1853 cl_event 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));
1855
1856 if (ev)
1857 XPUSH_CLOBJ (stash_event, ev);
1858
1859void
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, ...)
1861 ALIAS:
1862 enqueue_map_buffer = 0
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
1874 cl_event ev;
1875 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1876 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1877
1878void
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, ...)
1880 ALIAS:
1881 enqueue_map_image = 0
1882 PPCODE:
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
1908 const size_t origin[3] = { x, y, z };
1141 const size_t region[3] = { width, height, depth }; 1909 const size_t region[3] = { width, height, depth };
1910 size_t row_pitch, slice_pitch;
1142 EVENT_LIST (10, items - 10); 1911 EVENT_LIST (10);
1143 1912
1913 cl_event ev;
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));
1915
1916 size_t cb = slice_pitch ? slice_pitch * region [2]
1917 : row_pitch ? row_pitch * region [1]
1918 : region [0];
1919
1920 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1921
1922void
1923unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1924 PPCODE:
1925 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1926 if (GIMME_V != G_VOID)
1927 {
1928 clRetainEvent (mapped->event);
1929 XPUSH_CLOBJ (stash_event, mapped->event);
1930 }
1931
1932void
1933task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1934 ALIAS:
1935 enqueue_task = 0
1936 PPCODE:
1937 EVENT_LIST (2);
1938
1939 cl_event ev = 0;
1144 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, 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));
1145 1941
1146 if (ev) 1942 if (ev)
1147 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1943 XPUSH_CLOBJ (stash_event, ev);
1148 1944
1149void 1945void
1150enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1151 PPCODE:
1152 cl_event ev = 0;
1153 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1154 const size_t region[3] = { width, height, depth };
1155 EVENT_LIST (10, items - 10);
1156
1157 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1158
1159 if (ev)
1160 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1161
1162void
1163enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1164 PPCODE:
1165 cl_event ev = 0;
1166 EVENT_LIST (2, items - 2);
1167
1168 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1169
1170 if (ev)
1171 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1172
1173void
1174enqueue_nd_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, ...)
1947 ALIAS:
1948 enqueue_nd_range_kernel = 0
1175 PPCODE: 1949 PPCODE:
1176 cl_event ev = 0; 1950 EVENT_LIST (5);
1951
1177 size_t *gwo = 0, *gws, *lws = 0; 1952 size_t *gwo = 0, *gws, *lws = 0;
1178 int gws_len; 1953 int gws_len;
1179 size_t *lists; 1954 size_t *lists;
1180 int i; 1955 int i;
1181 EVENT_LIST (5, items - 5);
1182 1956
1183 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)
1184 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1958 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1185 1959
1186 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1960 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1187 1961
1188 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1962 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1189 1963
1190 gws = lists + gws_len * 0; 1964 gws = lists + gws_len * 0;
1191 for (i = 0; i < gws_len; ++i) 1965 for (i = 0; i < gws_len; ++i)
1966 {
1192 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1967 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1968 // at least nvidia crashes for 0-sized work group sizes, work around
1969 if (!gws [i])
1970 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1971 }
1193 1972
1194 if (SvOK (global_work_offset)) 1973 if (SvOK (global_work_offset))
1195 { 1974 {
1196 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1975 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1197 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1976 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1205 } 1984 }
1206 1985
1207 if (SvOK (local_work_size)) 1986 if (SvOK (local_work_size))
1208 { 1987 {
1209 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1988 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1210 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1989 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1211 1990
1212 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1991 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1213 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1992 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1214 1993
1215 lws = lists + gws_len * 2; 1994 lws = lists + gws_len * 2;
1216 for (i = 0; i < gws_len; ++i) 1995 for (i = 0; i < gws_len; ++i)
1996 {
1217 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1997 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1998 // at least nvidia crashes for 0-sized work group sizes, work around
1999 if (!lws [i])
2000 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2001 }
1218 } 2002 }
1219 2003
2004 cl_event ev = 0;
1220 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));
1221 2006
1222 if (ev) 2007 if (ev)
1223 XPUSH_NEW_OBJ ("OpenCL::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
1224 2029
1225#if cl_apple_gl_sharing || cl_khr_gl_sharing 2030#if cl_apple_gl_sharing || cl_khr_gl_sharing
1226 2031
1227void 2032void
1228enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2033acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1229 ALIAS: 2034 ALIAS:
2035 release_gl_objects = 1
2036 enqueue_acquire_gl_objects = 0
1230 enqueue_release_gl_objects = 1 2037 enqueue_release_gl_objects = 1
1231 PPCODE: 2038 PPCODE:
1232 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2039 EVENT_LIST (2);
1233 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);
1234 2043
1235 cl_event ev = 0; 2044 cl_event ev = 0;
1236 EVENT_LIST (2, items - 2);
1237 AV *av = (AV *)SvRV (objects);
1238 cl_uint num_objects = av_len (av) + 1;
1239 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1240 int i;
1241
1242 for (i = num_objects; i--; )
1243 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1244 2045
1245 if (ix) 2046 if (ix)
1246 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));
1247 else 2048 else
1248 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));
1249 2050
1250 if (ev) 2051 if (ev)
1251 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2052 XPUSH_CLOBJ (stash_event, ev);
1252 2053
1253#endif 2054#endif
1254 2055
1255void 2056void
1256enqueue_wait_for_events (OpenCL::Queue self, ...) 2057wait_for_events (OpenCL::Queue self, ...)
2058 ALIAS:
2059 enqueue_wait_for_events = 0
1257 CODE: 2060 CODE:
1258 EVENT_LIST (1, items - 1); 2061 EVENT_LIST (1);
1259#if PREFER_1_1 2062#if PREFER_1_1
1260 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2063 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1261#else 2064#else
1262 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2065 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1263#endif 2066#endif
1264 2067
1265void 2068void
1266enqueue_marker (OpenCL::Queue self, ...) 2069marker (OpenCL::Queue self, ...)
2070 ALIAS:
2071 enqueue_marker = 0
1267 PPCODE: 2072 PPCODE:
2073 EVENT_LIST (1);
1268 cl_event ev = 0; 2074 cl_event ev = 0;
1269 EVENT_LIST (1, items - 1);
1270#if PREFER_1_1 2075#if PREFER_1_1
1271 if (!event_list_count) 2076 if (!event_list_count)
1272 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2077 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1273 else 2078 else
1274#if CL_VERSION_1_2 2079#if CL_VERSION_1_2
1275 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2080 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1276#else 2081#else
2082 {
1277 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); // also a barrier 2083 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2084 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2085 }
1278#endif 2086#endif
1279#else 2087#else
1280 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2088 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281#endif 2089#endif
1282 if (ev) 2090 if (ev)
1283 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2091 XPUSH_CLOBJ (stash_event, ev);
1284 2092
1285void 2093void
1286enqueue_barrier (OpenCL::Queue self, ...) 2094barrier (OpenCL::Queue self, ...)
2095 ALIAS:
2096 enqueue_barrier = 0
1287 PPCODE: 2097 PPCODE:
2098 EVENT_LIST (1);
1288 cl_event ev = 0; 2099 cl_event ev = 0;
1289 EVENT_LIST (1, items - 1);
1290#if PREFER_1_1 2100#if PREFER_1_1
1291 if (!event_list_count && GIMME_V == G_VOID) 2101 if (!event_list_count && GIMME_V == G_VOID)
1292 NEED_SUCCESS (EnqueueBarrier, (self)); 2102 NEED_SUCCESS (EnqueueBarrier, (self));
1293 else 2103 else
1294#if CL_VERSION_1_2 2104#if CL_VERSION_1_2
1295 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2105 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1296#else 2106#else
1297 { 2107 {
1298 if (event_list_count) 2108 if (event_list_count)
1299 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2109 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1300 2110
1301 if (GIMME_V != G_VOID) 2111 if (GIMME_V != G_VOID)
1302 NEED_SUCCESS (EnqueueMarker, (self, &ev)); 2112 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1303 } 2113 }
1304#endif 2114#endif
1305#else 2115#else
1306 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2116 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1307#endif 2117#endif
1308 if (ev) 2118 if (ev)
1309 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2119 XPUSH_CLOBJ (stash_event, ev);
1310 2120
1311void 2121void
1312flush (OpenCL::Queue self) 2122flush (OpenCL::Queue self)
1313 CODE: 2123 CODE:
1314 NEED_SUCCESS (Flush, (self)); 2124 NEED_SUCCESS (Flush, (self));
1332 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2142 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1333 EXTEND (SP, 1); 2143 EXTEND (SP, 1);
1334 const int i = 0; 2144 const int i = 0;
1335 { 2145 {
1336 NEED_SUCCESS (RetainContext, (value [i])); 2146 NEED_SUCCESS (RetainContext, (value [i]));
1337 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2147 PUSH_CLOBJ (stash_context, value [i]);
1338 } 2148 }
1339 2149
1340void 2150void
1341device (OpenCL::Queue self) 2151device (OpenCL::Queue self)
1342 PPCODE: 2152 PPCODE:
1343 cl_device_id value [1]; 2153 cl_device_id value [1];
1344 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2154 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1345 EXTEND (SP, 1); 2155 EXTEND (SP, 1);
1346 const int i = 0; 2156 const int i = 0;
1347 { 2157 {
1348 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2158 PUSH_CLOBJ (stash_device, value [i]);
1349 } 2159 }
1350 2160
1351void 2161void
1352reference_count (OpenCL::Queue self) 2162reference_count (OpenCL::Queue self)
1353 PPCODE: 2163 PPCODE:
1362 PPCODE: 2172 PPCODE:
1363 cl_command_queue_properties value [1]; 2173 cl_command_queue_properties value [1];
1364 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2174 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1365 EXTEND (SP, 1); 2175 EXTEND (SP, 1);
1366 const int i = 0; 2176 const int i = 0;
1367 PUSHs (sv_2mortal (newSViv (value [i]))); 2177 PUSHs (sv_2mortal (newSVuv (value [i])));
1368 2178
1369#END:command_queue 2179#END:command_queue
1370 2180
1371MODULE = OpenCL PACKAGE = OpenCL::Memory 2181MODULE = OpenCL PACKAGE = OpenCL::Memory
1372 2182
1387 PPCODE: 2197 PPCODE:
1388 cl_mem_object_type value [1]; 2198 cl_mem_object_type value [1];
1389 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2199 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1390 EXTEND (SP, 1); 2200 EXTEND (SP, 1);
1391 const int i = 0; 2201 const int i = 0;
1392 PUSHs (sv_2mortal (newSViv (value [i]))); 2202 PUSHs (sv_2mortal (newSVuv (value [i])));
1393 2203
1394void 2204void
1395flags (OpenCL::Memory self) 2205flags (OpenCL::Memory self)
1396 PPCODE: 2206 PPCODE:
1397 cl_mem_flags value [1]; 2207 cl_mem_flags value [1];
1398 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2208 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1399 EXTEND (SP, 1); 2209 EXTEND (SP, 1);
1400 const int i = 0; 2210 const int i = 0;
1401 PUSHs (sv_2mortal (newSViv (value [i]))); 2211 PUSHs (sv_2mortal (newSVuv (value [i])));
1402 2212
1403void 2213void
1404size (OpenCL::Memory self) 2214size (OpenCL::Memory self)
1405 ALIAS: 2215 ALIAS:
1406 size = CL_MEM_SIZE 2216 size = CL_MEM_SIZE
1440 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1441 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1442 const int i = 0; 2252 const int i = 0;
1443 { 2253 {
1444 NEED_SUCCESS (RetainContext, (value [i])); 2254 NEED_SUCCESS (RetainContext, (value [i]));
1445 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2255 PUSH_CLOBJ (stash_context, value [i]);
1446 } 2256 }
1447 2257
1448void 2258void
1449associated_memobject (OpenCL::Memory self) 2259associated_memobject (OpenCL::Memory self)
1450 PPCODE: 2260 PPCODE:
1452 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2262 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1453 EXTEND (SP, 1); 2263 EXTEND (SP, 1);
1454 const int i = 0; 2264 const int i = 0;
1455 { 2265 {
1456 NEED_SUCCESS (RetainMemObject, (value [i])); 2266 NEED_SUCCESS (RetainMemObject, (value [i]));
1457 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2267 PUSH_CLOBJ (stash_memory, value [i]);
1458 } 2268 }
1459 2269
1460#END:mem 2270#END:mem
1461 2271
1462#if cl_apple_gl_sharing || cl_khr_gl_sharing 2272#if cl_apple_gl_sharing || cl_khr_gl_sharing
1482 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2292 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1483 2293
1484 cl_buffer_region crdata = { origin, size }; 2294 cl_buffer_region crdata = { origin, size };
1485 2295
1486 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2296 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1487 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2297 XPUSH_CLOBJ (stash_buffer, mem);
1488 2298
1489MODULE = OpenCL PACKAGE = OpenCL::Image 2299MODULE = OpenCL PACKAGE = OpenCL::Image
1490 2300
1491void 2301void
1492image_info (OpenCL::Image self, cl_image_info name) 2302image_info (OpenCL::Image self, cl_image_info name)
1493 PPCODE: 2303 PPCODE:
1494 INFO (Image) 2304 INFO (Image)
2305
2306void
2307format (OpenCL::Image self)
2308 PPCODE:
2309 cl_image_format format;
2310 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2311 EXTEND (SP, 2);
2312 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2313 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1495 2314
1496#BEGIN:image 2315#BEGIN:image
1497 2316
1498void 2317void
1499element_size (OpenCL::Image self) 2318element_size (OpenCL::Image self)
1569 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1570 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1571 const int i = 0; 2390 const int i = 0;
1572 { 2391 {
1573 NEED_SUCCESS (RetainContext, (value [i])); 2392 NEED_SUCCESS (RetainContext, (value [i]));
1574 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2393 PUSH_CLOBJ (stash_context, value [i]);
1575 } 2394 }
1576 2395
1577void 2396void
1578normalized_coords (OpenCL::Sampler self) 2397normalized_coords (OpenCL::Sampler self)
1579 PPCODE: 2398 PPCODE:
1580 cl_addressing_mode value [1]; 2399 cl_addressing_mode value [1];
1581 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2400 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1582 EXTEND (SP, 1); 2401 EXTEND (SP, 1);
1583 const int i = 0; 2402 const int i = 0;
1584 PUSHs (sv_2mortal (newSViv (value [i]))); 2403 PUSHs (sv_2mortal (newSVuv (value [i])));
1585 2404
1586void 2405void
1587addressing_mode (OpenCL::Sampler self) 2406addressing_mode (OpenCL::Sampler self)
1588 PPCODE: 2407 PPCODE:
1589 cl_filter_mode value [1]; 2408 cl_filter_mode value [1];
1590 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2409 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1591 EXTEND (SP, 1); 2410 EXTEND (SP, 1);
1592 const int i = 0; 2411 const int i = 0;
1593 PUSHs (sv_2mortal (newSViv (value [i]))); 2412 PUSHs (sv_2mortal (newSVuv (value [i])));
1594 2413
1595void 2414void
1596filter_mode (OpenCL::Sampler self) 2415filter_mode (OpenCL::Sampler self)
1597 PPCODE: 2416 PPCODE:
1598 cl_bool value [1]; 2417 cl_bool value [1];
1609DESTROY (OpenCL::Program self) 2428DESTROY (OpenCL::Program self)
1610 CODE: 2429 CODE:
1611 clReleaseProgram (self); 2430 clReleaseProgram (self);
1612 2431
1613void 2432void
1614build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2433build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2434 ALIAS:
2435 build_async = 1
1615 CODE: 2436 CODE:
2437 cl_uint device_count = 0;
2438 cl_device_id *device_list = 0;
2439
2440 if (SvOK (devices))
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 {
2468 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2469 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2470
2471 HV *hv = (HV *)SvRV (devices);
2472
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))
2480 {
2481 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2482 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2483 ++i;
2484 }
2485 }
2486
2487 void *user_data;
2488 program_callback pfn_notify = make_program_callback (notify, &user_data);
2489
1616 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 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
1617 2494
1618void 2495void
1619build_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)
1620 PPCODE: 2497 PPCODE:
1621 size_t size; 2498 size_t size;
1650 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2527 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1651 EXTEND (SP, 1); 2528 EXTEND (SP, 1);
1652 const int i = 0; 2529 const int i = 0;
1653 PUSHs (sv_2mortal (newSVpv (value, 0))); 2530 PUSHs (sv_2mortal (newSVpv (value, 0)));
1654 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])));
2540
1655#END:program_build 2541#END:program_build
1656 2542
1657void 2543void
1658kernel (OpenCL::Program program, SV *function) 2544kernel (OpenCL::Program program, SV *function)
1659 PPCODE: 2545 PPCODE:
1660 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2546 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1661 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2547 XPUSH_CLOBJ (stash_kernel, kernel);
1662 2548
1663void 2549void
1664kernels_in_program (OpenCL::Program program) 2550kernels_in_program (OpenCL::Program program)
1665 PPCODE: 2551 PPCODE:
1666 cl_uint num_kernels; 2552 cl_uint num_kernels;
1669 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2555 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
1670 2556
1671 int i; 2557 int i;
1672 EXTEND (SP, num_kernels); 2558 EXTEND (SP, num_kernels);
1673 for (i = 0; i < num_kernels; ++i) 2559 for (i = 0; i < num_kernels; ++i)
1674 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Kernel", kernels [i])); 2560 PUSH_CLOBJ (stash_kernel, kernels [i]);
1675 2561
1676void 2562void
1677info (OpenCL::Program self, cl_program_info name) 2563info (OpenCL::Program self, cl_program_info name)
1678 PPCODE: 2564 PPCODE:
1679 INFO (Program) 2565 INFO (Program)
1727 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2613 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1728 EXTEND (SP, 1); 2614 EXTEND (SP, 1);
1729 const int i = 0; 2615 const int i = 0;
1730 { 2616 {
1731 NEED_SUCCESS (RetainContext, (value [i])); 2617 NEED_SUCCESS (RetainContext, (value [i]));
1732 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2618 PUSH_CLOBJ (stash_context, value [i]);
1733 } 2619 }
1734 2620
1735void 2621void
1736devices (OpenCL::Program self) 2622devices (OpenCL::Program self)
1737 PPCODE: 2623 PPCODE:
1741 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2627 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1742 int i, n = size / sizeof (*value); 2628 int i, n = size / sizeof (*value);
1743 EXTEND (SP, n); 2629 EXTEND (SP, n);
1744 for (i = 0; i < n; ++i) 2630 for (i = 0; i < n; ++i)
1745 { 2631 {
1746 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2632 PUSH_CLOBJ (stash_device, value [i]);
1747 } 2633 }
1748 2634
1749void 2635void
1750source (OpenCL::Program self) 2636source (OpenCL::Program self)
1751 PPCODE: 2637 PPCODE:
1777DESTROY (OpenCL::Kernel self) 2663DESTROY (OpenCL::Kernel self)
1778 CODE: 2664 CODE:
1779 clReleaseKernel (self); 2665 clReleaseKernel (self);
1780 2666
1781void 2667void
2668setf (OpenCL::Kernel self, const char *format, ...)
2669 CODE:
2670 int i;
2671 for (i = 2; ; ++i)
2672 {
2673 while (*format == ' ')
2674 ++format;
2675
2676 char type = *format++;
2677
2678 if (!type)
2679 break;
2680
2681 if (i >= items)
2682 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2683
2684 SV *sv = ST (i);
2685
2686 union
2687 {
2688 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2689 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2690 cl_half ch; cl_float cf; cl_double cd;
2691 cl_mem cm;
2692 cl_sampler ca;
2693 size_t cz;
2694 cl_event ce;
2695 } arg;
2696 size_t size;
2697 int nullarg = 0;
2698
2699 switch (type)
2700 {
2701 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2702 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2703 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2704 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2705 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2706 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2707 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2708 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2709
2710 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2711 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2712 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2713
2714 case 'z': nullarg = 1; size = SvIV (sv); break;
2715
2716 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2717 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2718 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2719
2720 default:
2721 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2722 }
2723
2724 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2725 if (res)
2726 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2727 }
2728
2729 if (i != items)
2730 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2731
2732void
1782set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2733set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1783 CODE: 2734 CODE:
1784 clSetKernelArg (self, idx, sizeof (value), &value); 2735 clSetKernelArg (self, idx, sizeof (value), &value);
1785 2736
1786void 2737void
1834 clSetKernelArg (self, idx, sizeof (value), &value); 2785 clSetKernelArg (self, idx, sizeof (value), &value);
1835 2786
1836void 2787void
1837set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2788set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1838 CODE: 2789 CODE:
1839 clSetKernelArg (self, idx, sizeof (value), &value); 2790 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1840 2791
1841void 2792void
1842set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2793set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1843 CODE: 2794 CODE:
1844 clSetKernelArg (self, idx, sizeof (value), &value); 2795 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1845 2796
1846void 2797void
1847set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2798set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
1848 CODE: 2799 CODE:
1849 clSetKernelArg (self, idx, sizeof (value), &value); 2800 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1850
1851void
1852set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1853 CODE:
1854 clSetKernelArg (self, idx, sizeof (value), &value);
1855 2801
1856void 2802void
1857set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2803set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1858 CODE: 2804 CODE:
1859 clSetKernelArg (self, idx, sizeof (value), &value); 2805 clSetKernelArg (self, idx, sizeof (value), &value);
1905 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2851 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1906 EXTEND (SP, 1); 2852 EXTEND (SP, 1);
1907 const int i = 0; 2853 const int i = 0;
1908 { 2854 {
1909 NEED_SUCCESS (RetainContext, (value [i])); 2855 NEED_SUCCESS (RetainContext, (value [i]));
1910 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2856 PUSH_CLOBJ (stash_context, value [i]);
1911 } 2857 }
1912 2858
1913void 2859void
1914program (OpenCL::Kernel self) 2860program (OpenCL::Kernel self)
1915 PPCODE: 2861 PPCODE:
1917 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2863 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1918 EXTEND (SP, 1); 2864 EXTEND (SP, 1);
1919 const int i = 0; 2865 const int i = 0;
1920 { 2866 {
1921 NEED_SUCCESS (RetainProgram, (value [i])); 2867 NEED_SUCCESS (RetainProgram, (value [i]));
1922 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2868 PUSH_CLOBJ (stash_program, value [i]);
1923 } 2869 }
1924 2870
1925#END:kernel 2871#END:kernel
1926 2872
1927void 2873void
1974 const int i = 0; 2920 const int i = 0;
1975 PUSHs (sv_2mortal (newSVuv (value [i]))); 2921 PUSHs (sv_2mortal (newSVuv (value [i])));
1976 2922
1977#END:kernel_work_group 2923#END:kernel_work_group
1978 2924
2925#if CL_VERSION_1_2
2926
2927void
2928arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2929 PPCODE:
2930 size_t size;
2931 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2932 SV *sv = sv_2mortal (newSV (size));
2933 SvUPGRADE (sv, SVt_PV);
2934 SvPOK_only (sv);
2935 SvCUR_set (sv, size);
2936 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2937 XPUSHs (sv);
2938
2939#BEGIN:kernel_arg
2940
2941void
2942arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2943 PPCODE:
2944 cl_kernel_arg_address_qualifier value [1];
2945 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2946 EXTEND (SP, 1);
2947 const int i = 0;
2948 PUSHs (sv_2mortal (newSVuv (value [i])));
2949
2950void
2951arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2952 PPCODE:
2953 cl_kernel_arg_access_qualifier value [1];
2954 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2955 EXTEND (SP, 1);
2956 const int i = 0;
2957 PUSHs (sv_2mortal (newSVuv (value [i])));
2958
2959void
2960arg_type_name (OpenCL::Kernel self, cl_uint idx)
2961 ALIAS:
2962 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2963 arg_name = CL_KERNEL_ARG_NAME
2964 PPCODE:
2965 size_t size;
2966 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2967 char *value = tmpbuf (size);
2968 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2969 EXTEND (SP, 1);
2970 const int i = 0;
2971 PUSHs (sv_2mortal (newSVpv (value, 0)));
2972
2973void
2974arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2975 PPCODE:
2976 cl_kernel_arg_type_qualifier value [1];
2977 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2978 EXTEND (SP, 1);
2979 const int i = 0;
2980 PUSHs (sv_2mortal (newSVuv (value [i])));
2981
2982#END:kernel_arg
2983
2984#endif
2985
1979MODULE = OpenCL PACKAGE = OpenCL::Event 2986MODULE = OpenCL PACKAGE = OpenCL::Event
1980 2987
1981void 2988void
1982DESTROY (OpenCL::Event self) 2989DESTROY (OpenCL::Event self)
1983 CODE: 2990 CODE:
1985 2992
1986void 2993void
1987wait (OpenCL::Event self) 2994wait (OpenCL::Event self)
1988 CODE: 2995 CODE:
1989 clWaitForEvents (1, &self); 2996 clWaitForEvents (1, &self);
2997
2998void
2999cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
3000 CODE:
3001 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1990 3002
1991void 3003void
1992info (OpenCL::Event self, cl_event_info name) 3004info (OpenCL::Event self, cl_event_info name)
1993 PPCODE: 3005 PPCODE:
1994 INFO (Event) 3006 INFO (Event)
2002 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3014 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2003 EXTEND (SP, 1); 3015 EXTEND (SP, 1);
2004 const int i = 0; 3016 const int i = 0;
2005 { 3017 {
2006 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3018 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2007 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 3019 PUSH_CLOBJ (stash_queue, value [i]);
2008 } 3020 }
2009 3021
2010void 3022void
2011command_type (OpenCL::Event self) 3023command_type (OpenCL::Event self)
2012 PPCODE: 3024 PPCODE:
2035 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3047 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2036 EXTEND (SP, 1); 3048 EXTEND (SP, 1);
2037 const int i = 0; 3049 const int i = 0;
2038 { 3050 {
2039 NEED_SUCCESS (RetainContext, (value [i])); 3051 NEED_SUCCESS (RetainContext, (value [i]));
2040 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 3052 PUSH_CLOBJ (stash_context, value [i]);
2041 } 3053 }
2042 3054
2043#END:event 3055#END:event
2044 3056
2045void 3057void
2070void 3082void
2071set_status (OpenCL::UserEvent self, cl_int execution_status) 3083set_status (OpenCL::UserEvent self, cl_int execution_status)
2072 CODE: 3084 CODE:
2073 clSetUserEventStatus (self, execution_status); 3085 clSetUserEventStatus (self, execution_status);
2074 3086
3087MODULE = OpenCL PACKAGE = OpenCL::Mapped
3088
3089void
3090DESTROY (SV *self)
3091 CODE:
3092 OpenCL__Mapped mapped = SvMAPPED (self);
3093
3094 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3095 mapped_detach (self, mapped);
3096
3097 clReleaseCommandQueue (mapped->queue);
3098 clReleaseEvent (mapped->event);
3099 Safefree (mapped);
3100
3101void
3102unmap (OpenCL::Mapped self, ...)
3103 CODE:
3104 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3105
3106bool
3107mapped (OpenCL::Mapped self)
3108 CODE:
3109 RETVAL = !!self->ptr;
3110 OUTPUT:
3111 RETVAL
3112
3113void
3114wait (OpenCL::Mapped self)
3115 PPCODE:
3116 if (self->event)
3117 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3118
3119void
3120event (OpenCL::Mapped self)
3121 PPCODE:
3122 if (!self->event)
3123 XSRETURN_UNDEF;
3124
3125 clRetainEvent (self->event);
3126 XPUSH_CLOBJ (stash_event, self->event);
3127
3128size_t
3129size (OpenCL::Mapped self)
3130 CODE:
3131 RETVAL = self->cb;
3132 OUTPUT:
3133 RETVAL
3134
3135IV
3136ptr (OpenCL::Mapped self)
3137 CODE:
3138 RETVAL = PTR2IV (self->ptr);
3139 OUTPUT:
3140 RETVAL
3141
3142void
3143set (OpenCL::Mapped self, size_t offset, SV *data)
3144 CODE:
3145 STRLEN len;
3146 const char *ptr = SvPVbyte (data, len);
3147
3148 if (offset + len > self->cb)
3149 croak ("OpenCL::Mapped::set out of bound condition detected");
3150
3151 memcpy (offset + (char *)self->ptr, ptr, len);
3152
3153MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3154
3155MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3156
3157IV
3158row_pitch (OpenCL::Mapped self)
3159 ALIAS:
3160 slice_pitch = 1
3161 CODE:
3162 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3163 OUTPUT:
3164 RETVAL
3165
3166
3167

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines