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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.32 by root, Thu Apr 19 19:36:24 2012 UTC vs.
Revision 1.72 by root, Sat May 5 12:54:04 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 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"
10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
7#endif 30#endif
8 31
9#ifdef __APPLE__ 32#ifdef __APPLE__
10 #include <OpenCL/opencl.h> 33 #include <OpenCL/opencl.h>
11#else 34#else
12 #include <CL/opencl.h> 35 #include <CL/opencl.h>
13#endif 36#endif
14 37
38#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1
40 #define PREFER_1_1 1
41#endif
42
15typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
16typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
17typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
18typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
19typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
20typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
21typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
22typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
23typedef cl_mem OpenCL__Image2D;
24typedef cl_mem OpenCL__Image3D;
25typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
26typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
27typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
28typedef cl_mem OpenCL__Image2D_ornull;
29typedef cl_mem OpenCL__Image3D_ornull;
30typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
31typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
32typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
33typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
34typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
35 60
36typedef SV *FUTURE; 61typedef struct mapped * OpenCL__Mapped;
62
63static HV
64 *stash_platform,
65 *stash_device,
66 *stash_subdevice,
67 *stash_context,
68 *stash_queue,
69 *stash_program,
70 *stash_kernel,
71 *stash_sampler,
72 *stash_event,
73 *stash_userevent,
74 *stash_memory,
75 *stash_buffer,
76 *stash_bufferobj,
77 *stash_image,
78 *stash_image1d,
79 *stash_image1darray,
80 *stash_image1dbuffer,
81 *stash_image2d,
82 *stash_image2darray,
83 *stash_image3d,
84 *stash_mapped,
85 *stash_mappedbuffer,
86 *stash_mappedimage;
37 87
38/*****************************************************************************/ 88/*****************************************************************************/
39 89
40// name must include a leading underscore 90// name must include a leading underscore
41// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 91// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
81} 131}
82 132
83/*****************************************************************************/ 133/*****************************************************************************/
84 134
85/* up to two temporary buffers */ 135/* up to two temporary buffers */
86static void * 136static void * ecb_noinline
87tmpbuf (size_t size) 137tmpbuf (size_t size)
88{ 138{
89 enum { buffers = 3 }; 139 enum { buffers = 4 };
90 static int idx; 140 static int idx;
91 static void *buf [buffers]; 141 static void *buf [buffers];
92 static size_t len [buffers]; 142 static size_t len [buffers];
93 143
94 idx = ++idx % buffers; 144 idx = (idx + 1) % buffers;
95 145
96 if (len [idx] < size) 146 if (len [idx] < size)
97 { 147 {
98 free (buf [idx]); 148 free (buf [idx]);
99 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 149 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
100 buf [idx] = malloc (len [idx]); 150 buf [idx] = malloc (len [idx]);
101 } 151 }
102 152
103 return buf [idx]; 153 return buf [idx];
154}
155
156static const char * ecb_noinline
157cv_get_name (CV *cv)
158{
159 static char fullname [256];
160
161 GV *gv = CvGV (cv); // gv better be nonzero
162
163 HV *stash = GvSTASH (gv);
164 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
165 const char *gvname = GvNAME (gv);
166
167 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
168 return fullname;
104} 169}
105 170
106/*****************************************************************************/ 171/*****************************************************************************/
107 172
108typedef struct 173typedef struct
167 if (res) \ 232 if (res) \
168 FAIL (name); 233 FAIL (name);
169 234
170/*****************************************************************************/ 235/*****************************************************************************/
171 236
172#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
173#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
174
175static void * 237static SV *
238new_clobj (HV *stash, IV id)
239{
240 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
241}
242
243#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
244#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
245
246/* cl objects are either \$iv, or [$iv, ...] */
247/* they can be upgraded at runtime to the array form */
248static void * ecb_noinline
176SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 249SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
177{ 250{
251 // sv_derived_from is quite slow :(
178 if (SvROK (sv) && sv_derived_from (sv, pkg)) 252 if (SvROK (sv) && sv_derived_from (sv, pkg))
179 return (void *)SvIV (SvRV (sv)); 253 return (void *)SvIV (SvRV (sv));
180 254
181 croak ("%s: %s is not of type %s", func, svname, pkg); 255 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
256}
257
258// the "no-inherit" version of the above
259static void * ecb_noinline
260SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
261{
262 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
263 return (void *)SvIV (SvRV (sv));
264
265 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
182} 266}
183 267
184/*****************************************************************************/ 268/*****************************************************************************/
185 269
186static cl_context_properties * 270static cl_context_properties * ecb_noinline
187SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 271SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
188{ 272{
189 if (!sv || !SvOK (sv)) 273 if (!sv || !SvOK (sv))
190 if (extra) 274 if (extra)
191 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 275 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
192 else 276 else
198 int i, len = av_len (av) + 1; 282 int i, len = av_len (av) + 1;
199 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 283 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
200 cl_context_properties *l = p; 284 cl_context_properties *l = p;
201 285
202 if (len & 1) 286 if (len & 1)
203 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 287 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
204 288
205 while (extracount--) 289 while (extracount--)
206 *l++ = *extra++; 290 *l++ = *extra++;
207 291
208 for (i = 0; i < len; i += 2) 292 for (i = 0; i < len; i += 2)
211 SV *p_sv = *av_fetch (av, i + 1, 0); 295 SV *p_sv = *av_fetch (av, i + 1, 0);
212 cl_context_properties v = SvIV (p_sv); // code below can override 296 cl_context_properties v = SvIV (p_sv); // code below can override
213 297
214 switch (t) 298 switch (t)
215 { 299 {
300 case CL_CONTEXT_PLATFORM:
301 if (SvROK (p_sv))
302 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
303 break;
304
216 case CL_GLX_DISPLAY_KHR: 305 case CL_GLX_DISPLAY_KHR:
217 if (!SvOK (p_sv)) 306 if (!SvOK (p_sv))
218 { 307 {
219 void *func = glsym ("_glXGetCurrentDisplay"); 308 void *func = glsym ("_glXGetCurrentDisplay");
220 if (func) 309 if (func)
243 *l = 0; 332 *l = 0;
244 333
245 return p; 334 return p;
246 } 335 }
247 336
248 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 337 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
338}
339
340// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
341// is a pointer (and only on well-behaved systems).
342static void * ecb_noinline
343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
344{
345 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
346 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
347 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
348
349 AV *av = (AV *)SvRV (arg);
350 void **list = 0;
351 cl_uint count = av_len (av) + 1;
352
353 if (count)
354 {
355 list = tmpbuf (sizeof (*list) * count);
356 int i;
357 for (i = 0; i < count; ++i)
358 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
359 }
360
361 if (!count && !or_undef)
362 croak ("%s: '%s' must contain at least one %s object",
363 cv_get_name (cv), argname, klass);
364
365 *rcount = count;
366 return (void *)list;
249} 367}
250 368
251/*****************************************************************************/ 369/*****************************************************************************/
370/* callback stuff */
371
372/* default context callback, log to stderr */
373static void CL_CALLBACK
374context_default_notify (const char *msg, const void *info, size_t cb, void *data)
375{
376 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
377}
378
379typedef struct
380{
381 int free_cb;
382 void (*push)(void *data1, void *data2, void *data3);
383} eq_vtbl;
384
385typedef struct eq_item
386{
387 struct eq_item *next;
388 eq_vtbl *vtbl;
389 SV *cb;
390 void *data1, *data2, *data3;
391} eq_item;
392
393static void (*eq_signal_func)(void *signal_arg, int value);
394static void *eq_signal_arg;
395static xmutex_t eq_lock = X_MUTEX_INIT;
396static eq_item *eq_head, *eq_tail;
397
398static void ecb_noinline
399eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
400{
401 eq_item *item = malloc (sizeof (eq_item));
402
403 item->next = 0;
404 item->vtbl = vtbl;
405 item->cb = cb;
406 item->data1 = data1;
407 item->data2 = data2;
408 item->data3 = data3;
409
410 X_LOCK (eq_lock);
411
412 *(eq_head ? &eq_tail->next : &eq_head) = item;
413 eq_tail = item;
414
415 X_UNLOCK (eq_lock);
416
417 eq_signal_func (eq_signal_arg, 0);
418}
419
420static eq_item *
421eq_dec (void)
422{
423 eq_item *res;
424
425 X_LOCK (eq_lock);
426
427 res = eq_head;
428 if (res)
429 eq_head = res->next;
430
431 X_UNLOCK (eq_lock);
432
433 return res;
434}
435
436static void
437eq_poll (void)
438{
439 eq_item *item;
440
441 while ((item = eq_dec ()))
442 {
443 ENTER;
444 SAVETMPS;
445
446 dSP;
447 PUSHMARK (SP);
448 EXTEND (SP, 2);
449
450 if (item->vtbl->free_cb)
451 sv_2mortal (item->cb);
452
453 PUTBACK;
454 item->vtbl->push (item->data1, item->data2, item->data3);
455
456 SV *cb = item->cb;
457 free (item);
458
459 call_sv (cb, G_DISCARD | G_VOID);
460
461 FREETMPS;
462 LEAVE;
463 }
464}
465
466static void
467eq_poll_interrupt (pTHX_ void *c_arg, int value)
468{
469 eq_poll ();
470}
471
472/*****************************************************************************/
473/* context notify */
474
475static void ecb_noinline
476eq_context_push (void *data1, void *data2, void *data3)
477{
478 dSP;
479 PUSHs (sv_2mortal (newSVpv (data1, 0)));
480 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
481 PUTBACK;
482
483 free (data1);
484 free (data2);
485}
486
487static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
488
489static void CL_CALLBACK
490eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
491{
492 void *pvt_copy = malloc (cb);
493 memcpy (pvt_copy, pvt, cb);
494 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
495}
496
497#define CONTEXT_NOTIFY_CALLBACK \
498 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
499 void *user_data = 0; \
500 \
501 if (SvOK (notify)) \
502 { \
503 pfn_notify = eq_context_notify; \
504 user_data = s_get_cv (notify); \
505 }
506
507static SV * ecb_noinline
508new_clobj_context (cl_context ctx, void *user_data)
509{
510 SV *sv = new_clobj (stash_context, (IV)ctx);
511
512 if (user_data)
513 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
514
515 return sv;
516}
517
518#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
519
520/*****************************************************************************/
521/* build/compile/link notify */
522
523static void
524eq_program_push (void *data1, void *data2, void *data3)
525{
526 dSP;
527 PUSH_CLOBJ (stash_program, data1);
528 PUTBACK;
529}
530
531static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
532
533static void CL_CALLBACK
534eq_program_notify (cl_program program, void *user_data)
535{
536 clRetainProgram (program);
537
538 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
539}
540
541typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
542
543static program_callback ecb_noinline
544make_program_callback (SV *notify, void **ruser_data)
545{
546 if (SvOK (notify))
547 {
548 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
549 return eq_program_notify;
550 }
551 else
552 {
553 *ruser_data = 0;
554 return 0;
555 }
556}
557
558struct build_args
559{
560 cl_program program;
561 char *options;
562 void *user_data;
563 cl_uint num_devices;
564};
565
566X_THREAD_PROC (build_program_thread)
567{
568 struct build_args *arg = thr_arg;
569
570 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
571
572 if (arg->user_data)
573 eq_program_notify (arg->program, arg->user_data);
574 else
575 clReleaseProgram (arg->program);
576
577 free (arg->options);
578 free (arg);
579
580 return 0;
581}
582
583static void
584build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
585{
586 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
587
588 arg->program = program;
589 arg->options = strdup (options);
590 arg->user_data = user_data;
591 arg->num_devices = num_devices;
592 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
593
594 xthread_t id;
595 thread_create (&id, build_program_thread, arg);
596}
597
598/*****************************************************************************/
599/* event objects */
600
601static void
602eq_event_push (void *data1, void *data2, void *data3)
603{
604 dSP;
605 PUSH_CLOBJ (stash_event, data1);
606 PUSHs (sv_2mortal (newSViv ((IV)data2)));
607 PUTBACK;
608}
609
610static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
611
612static void CL_CALLBACK
613eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
614{
615 clRetainEvent (event);
616 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
617}
618
619/*****************************************************************************/
620/* utilities for XS code */
252 621
253static size_t 622static size_t
254img_row_pitch (cl_mem img) 623img_row_pitch (cl_mem img)
255{ 624{
256 size_t res; 625 size_t res;
257 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 626 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
258 return res; 627 return res;
259} 628}
260 629
261static cl_event * 630static cl_event * ecb_noinline
262event_list (SV **items, int count) 631event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
263{ 632{
264 if (!count) 633 cl_uint count = *rcount;
634
635 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
636 *rcount = count = 0;
637
638 if (!count && !extra)
265 return 0; 639 return 0;
266 640
267 cl_event *list = tmpbuf (sizeof (cl_event) * count); 641 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
642 int i = 0;
268 643
269 while (count--) 644 while (count--)
645 if (SvOK (items [count]))
270 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 646 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
271 647
648 if (extra)
649 list [i++] = extra;
650
651 *rcount = i;
652
272 return list; 653 return i ? list : 0;
273} 654}
274 655
275#define EVENT_LIST(items,count) \ 656#define EVENT_LIST(skip) \
276 cl_uint event_list_count = (count); \ 657 cl_uint event_list_count = items - (skip); \
277 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 658 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
278 659
279#define INFO(class) \ 660#define INFO(class) \
280{ \ 661{ \
281 size_t size; \ 662 size_t size; \
282 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 663 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
286 SvCUR_set (sv, size); \ 667 SvCUR_set (sv, size); \
287 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 668 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
288 XPUSHs (sv); \ 669 XPUSHs (sv); \
289} 670}
290 671
672/*****************************************************************************/
673/* mapped_xxx */
674
675static OpenCL__Mapped
676SvMAPPED (SV *sv)
677{
678 // no typechecking atm., keep your fingers crossed
679 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
680}
681
682struct mapped
683{
684 cl_command_queue queue;
685 cl_mem memobj;
686 void *ptr;
687 size_t cb;
688 cl_event event;
689 size_t row_pitch;
690 size_t slice_pitch;
691};
692
693static SV *
694mapped_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)
695{
696 SV *data = newSV (0);
697 SvUPGRADE (data, SVt_PVMG);
698
699 OpenCL__Mapped mapped;
700 New (0, mapped, 1, struct mapped);
701
702 clRetainCommandQueue (queue);
703
704 mapped->queue = queue;
705 mapped->memobj = memobj;
706 mapped->ptr = ptr;
707 mapped->cb = cb;
708 mapped->event = ev;
709 mapped->row_pitch = row_pitch;
710 mapped->slice_pitch = slice_pitch;
711
712 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
713
714 if (SvLEN (data))
715 Safefree (data);
716
717 SvPVX (data) = (char *)ptr;
718 SvCUR_set (data, cb);
719 SvLEN_set (data, 0);
720 SvPOK_only (data);
721
722 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
723
724 if (!(flags & CL_MAP_WRITE))
725 SvREADONLY_on (data);
726
727 return obj;
728}
729
730static void
731mapped_detach (SV *sv, OpenCL__Mapped mapped)
732{
733 SV *data = SvRV (sv);
734
735 // the next check checks both before AND after detach, where SvPVX should STILL be 0
736 if (SvPVX (data) != (char *)mapped->ptr)
737 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
738 else
739 {
740 SvREADONLY_off (data);
741 SvCUR_set (data, 0);
742 SvPVX (data) = 0;
743 SvOK_off (data);
744 }
745
746 mapped->ptr = 0;
747}
748
749static void
750mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
751{
752 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
753 cl_event ev;
754
755 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
756
757 clReleaseEvent (mapped->event);
758 mapped->event = ev;
759
760 mapped_detach (self, mapped);
761}
762
763/*****************************************************************************/
764
291MODULE = OpenCL PACKAGE = OpenCL 765MODULE = OpenCL PACKAGE = OpenCL
292 766
293PROTOTYPES: ENABLE 767PROTOTYPES: ENABLE
768
769void
770poll ()
771 CODE:
772 eq_poll ();
773
774void
775_eq_initialise (IV func, IV arg)
776 CODE:
777 eq_signal_func = (void (*)(void *, int))func;
778 eq_signal_arg = (void*)arg;
294 779
295BOOT: 780BOOT:
296{ 781{
297 HV *stash = gv_stashpv ("OpenCL", 1); 782 HV *stash = gv_stashpv ("OpenCL", 1);
298 static const ivstr *civ, const_iv[] = { 783 static const ivstr *civ, const_iv[] = {
305 { sizeof (cl_long ), "SIZEOF_LONG" }, 790 { sizeof (cl_long ), "SIZEOF_LONG" },
306 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 791 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
307 { sizeof (cl_half ), "SIZEOF_HALF" }, 792 { sizeof (cl_half ), "SIZEOF_HALF" },
308 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 793 { sizeof (cl_float ), "SIZEOF_FLOAT" },
309 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 794 { sizeof (cl_double), "SIZEOF_DOUBLE" },
795 { PREFER_1_1 , "PREFER_1_1" },
310#include "constiv.h" 796#include "constiv.h"
311 }; 797 };
798
312 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 799 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
313 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 800 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
801
802 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
803 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
804 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
805 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
806 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
807 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
808 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
809 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
810 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
811 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
812 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
813 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
814 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
815 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
816 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
817 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
818 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
819 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
820 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
821 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
822 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
823 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
824 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
825
826 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
314} 827}
315 828
316cl_int 829cl_int
317errno () 830errno ()
318 CODE: 831 CODE:
319 errno = res; 832 RETVAL = res;
833 OUTPUT:
834 RETVAL
320 835
321const char * 836const char *
322err2str (cl_int err) 837err2str (cl_int err = res)
323 838
324const char * 839const char *
325enum2str (cl_uint value) 840enum2str (cl_uint value)
326 841
327void 842void
335 list = tmpbuf (sizeof (*list) * count); 850 list = tmpbuf (sizeof (*list) * count);
336 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 851 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
337 852
338 EXTEND (SP, count); 853 EXTEND (SP, count);
339 for (i = 0; i < count; ++i) 854 for (i = 0; i < count; ++i)
340 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 855 PUSH_CLOBJ (stash_platform, list [i]);
341 856
342void 857void
343context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 858context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
344 PPCODE: 859 PPCODE:
860 CONTEXT_NOTIFY_CALLBACK;
345 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 861 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
346 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 862 XPUSH_CLOBJ_CONTEXT;
347 863
348void 864void
349context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 865context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
350 PPCODE: 866 PPCODE:
351 /* der Gipfel der Kunst */ 867 cl_uint device_count;
868 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
869
870 CONTEXT_NOTIFY_CALLBACK;
871 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
872 XPUSH_CLOBJ_CONTEXT;
352 873
353void 874void
354wait_for_events (...) 875wait_for_events (...)
355 CODE: 876 CODE:
356 EVENT_LIST (0, items); 877 EVENT_LIST (0);
357 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 878 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
358 879
359PROTOTYPES: DISABLE 880PROTOTYPES: DISABLE
360 881
361MODULE = OpenCL PACKAGE = OpenCL::Platform 882MODULE = OpenCL PACKAGE = OpenCL::Platform
362 883
363void 884void
364info (OpenCL::Platform self, cl_platform_info name) 885info (OpenCL::Platform self, cl_platform_info name)
365 PPCODE: 886 PPCODE:
366 INFO (Platform) 887 INFO (Platform)
888
889void
890unload_compiler (OpenCL::Platform self)
891 CODE:
892#if CL_VERSION_1_2
893 clUnloadPlatformCompiler (self);
894#endif
367 895
368#BEGIN:platform 896#BEGIN:platform
369 897
370void 898void
371profile (OpenCL::Platform self) 899profile (OpenCL::Platform self)
397 list = tmpbuf (sizeof (*list) * count); 925 list = tmpbuf (sizeof (*list) * count);
398 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 926 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
399 927
400 EXTEND (SP, count); 928 EXTEND (SP, count);
401 for (i = 0; i < count; ++i) 929 for (i = 0; i < count; ++i)
402 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 930 PUSH_CLOBJ (stash_device, list [i]);
403 931
404void 932void
405context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 933context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
406 PPCODE: 934 PPCODE:
935 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
936 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
937
938 cl_uint device_count;
939 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
940
941 CONTEXT_NOTIFY_CALLBACK;
942 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
943 XPUSH_CLOBJ_CONTEXT;
944
945void
946context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
947 PPCODE:
948 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
949 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
950
951 CONTEXT_NOTIFY_CALLBACK;
952 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
953 XPUSH_CLOBJ_CONTEXT;
954
955MODULE = OpenCL PACKAGE = OpenCL::Device
956
957void
958info (OpenCL::Device self, cl_device_info name)
959 PPCODE:
960 INFO (Device)
961
962#if CL_VERSION_1_2
963
964void
965sub_devices (OpenCL::Device self, SV *properties)
966 PPCODE:
407 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 967 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
408 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 968 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
409 969
410 AV *av = (AV *)SvRV (devices); 970 properties = SvRV (properties);
411 cl_uint num_devices = av_len (av) + 1; 971
412 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 972 cl_uint count = av_len ((AV *)properties) + 1;
973 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
974
413 int i; 975 int i;
976 for (i = 0; i < count; ++i)
977 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
414 978
415 for (i = num_devices; i--; ) 979 props [count] = 0;
416 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
417 980
418 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 981 cl_uint num_devices;
419 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 982 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
983 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
984 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
420 985
421void 986 EXTEND (SP, num_devices);
422context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 987 for (i = 0; i < count; ++i)
423 PPCODE: 988 PUSH_CLOBJ (stash_subdevice, list [i]);
424 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
425 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
426 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
427 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
428 989
429MODULE = OpenCL PACKAGE = OpenCL::Device 990#endif
430
431void
432info (OpenCL::Device self, cl_device_info name)
433 PPCODE:
434 INFO (Device)
435 991
436#BEGIN:device 992#BEGIN:device
437 993
438void 994void
439type (OpenCL::Device self) 995type (OpenCL::Device self)
440 PPCODE: 996 PPCODE:
441 cl_device_type value [1]; 997 cl_device_type value [1];
442 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 998 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
443 EXTEND (SP, 1); 999 EXTEND (SP, 1);
444 const int i = 0; 1000 const int i = 0;
445 PUSHs (sv_2mortal (newSViv (value [i]))); 1001 PUSHs (sv_2mortal (newSVuv (value [i])));
446 1002
447void 1003void
448vendor_id (OpenCL::Device self) 1004vendor_id (OpenCL::Device self)
449 ALIAS: 1005 ALIAS:
450 vendor_id = CL_DEVICE_VENDOR_ID 1006 vendor_id = CL_DEVICE_VENDOR_ID
471 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1027 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
472 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1028 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
473 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1029 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
474 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1030 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
475 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1031 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
476 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1032 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
477 PPCODE: 1033 PPCODE:
478 cl_uint value [1]; 1034 cl_uint value [1];
479 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1035 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
480 EXTEND (SP, 1); 1036 EXTEND (SP, 1);
481 const int i = 0; 1037 const int i = 0;
595 PPCODE: 1151 PPCODE:
596 cl_command_queue_properties value [1]; 1152 cl_command_queue_properties value [1];
597 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1153 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
598 EXTEND (SP, 1); 1154 EXTEND (SP, 1);
599 const int i = 0; 1155 const int i = 0;
600 PUSHs (sv_2mortal (newSViv (value [i]))); 1156 PUSHs (sv_2mortal (newSVuv (value [i])));
601 1157
602void 1158void
603platform (OpenCL::Device self) 1159platform (OpenCL::Device self)
604 PPCODE: 1160 PPCODE:
605 cl_platform_id value [1]; 1161 cl_platform_id value [1];
606 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1162 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
607 EXTEND (SP, 1); 1163 EXTEND (SP, 1);
608 const int i = 0; 1164 const int i = 0;
609 { 1165 {
610 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1166 PUSH_CLOBJ (stash_platform, value [i]);
611 } 1167 }
612 1168
613void 1169void
614name (OpenCL::Device self) 1170name (OpenCL::Device self)
615 ALIAS: 1171 ALIAS:
634 cl_device_id value [1]; 1190 cl_device_id value [1];
635 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1191 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
636 EXTEND (SP, 1); 1192 EXTEND (SP, 1);
637 const int i = 0; 1193 const int i = 0;
638 { 1194 {
639 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1195 PUSH_CLOBJ (stash_device, value [i]);
640 } 1196 }
641 1197
642void 1198void
643partition_types_ext (OpenCL::Device self) 1199partition_types_ext (OpenCL::Device self)
644 ALIAS: 1200 ALIAS:
655 for (i = 0; i < n; ++i) 1211 for (i = 0; i < n; ++i)
656 PUSHs (sv_2mortal (newSVuv (value [i]))); 1212 PUSHs (sv_2mortal (newSVuv (value [i])));
657 1213
658#END:device 1214#END:device
659 1215
1216MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1217
1218#if CL_VERSION_1_2
1219
1220void
1221DESTROY (OpenCL::SubDevice self)
1222 CODE:
1223 clReleaseDevice (self);
1224
1225#endif
1226
660MODULE = OpenCL PACKAGE = OpenCL::Context 1227MODULE = OpenCL PACKAGE = OpenCL::Context
661 1228
662void 1229void
663DESTROY (OpenCL::Context context) 1230DESTROY (OpenCL::Context self)
664 CODE: 1231 CODE:
665 clReleaseContext (context); 1232 clReleaseContext (self);
666 1233
667void 1234void
668info (OpenCL::Context self, cl_context_info name) 1235info (OpenCL::Context self, cl_context_info name)
669 PPCODE: 1236 PPCODE:
670 INFO (Context) 1237 INFO (Context)
671 1238
672void 1239void
673queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1240queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
674 PPCODE: 1241 PPCODE:
675 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1242 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
676 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1243 XPUSH_CLOBJ (stash_queue, queue);
677 1244
678void 1245void
679user_event (OpenCL::Context self) 1246user_event (OpenCL::Context self)
680 PPCODE: 1247 PPCODE:
681 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1248 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
682 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1249 XPUSH_CLOBJ (stash_userevent, ev);
683 1250
684void 1251void
685buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1252buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
686 PPCODE: 1253 PPCODE:
687 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1254 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
688 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1255 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
689 1256
690 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1257 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
691 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1258 XPUSH_CLOBJ (stash_bufferobj, mem);
692 1259
693void 1260void
694buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1261buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
695 PPCODE: 1262 PPCODE:
696 STRLEN len; 1263 STRLEN len;
697 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1264 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
698 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1265 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
699 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1266 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
700 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1267 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
701 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1268 XPUSH_CLOBJ (stash_bufferobj, mem);
702 1269
1270#if CL_VERSION_1_2
1271
703void 1272void
704image2d (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) 1273image (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)
705 PPCODE: 1274 PPCODE:
706 STRLEN len; 1275 STRLEN len;
707 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1276 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
708 const cl_image_format format = { channel_order, channel_type }; 1277 const cl_image_format format = { channel_order, channel_type };
1278 const cl_image_desc desc = {
1279 type,
1280 width, height, depth,
1281 array_size, row_pitch, slice_pitch,
1282 num_mip_level, num_samples,
1283 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1284 };
709 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1285 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
710 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1286 HV *stash = stash_image;
1287 switch (type)
1288 {
1289 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1290 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1291 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1292 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1293 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1294 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1295 }
1296 XPUSH_CLOBJ (stash, mem);
711 1297
1298#endif
1299
712void 1300void
713image3d (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) 1301image2d (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)
714 PPCODE: 1302 PPCODE:
715 STRLEN len; 1303 STRLEN len;
716 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1304 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
717 const cl_image_format format = { channel_order, channel_type }; 1305 const cl_image_format format = { channel_order, channel_type };
1306#if PREFER_1_1
1307 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1308#else
1309 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1310 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1311#endif
1312 XPUSH_CLOBJ (stash_image2d, mem);
1313
1314void
1315image3d (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)
1316 PPCODE:
1317 STRLEN len;
1318 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1319 const cl_image_format format = { channel_order, channel_type };
1320#if PREFER_1_1
718 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
719 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1322#else
1323 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1325#endif
1326 XPUSH_CLOBJ (stash_image3d, mem);
720 1327
721#if cl_apple_gl_sharing || cl_khr_gl_sharing 1328#if cl_apple_gl_sharing || cl_khr_gl_sharing
722 1329
723void 1330void
724gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1331gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
725 PPCODE: 1332 PPCODE:
726 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1333 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
727 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1334 XPUSH_CLOBJ (stash_bufferobj, mem);
1335
1336void
1337gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1338 PPCODE:
1339 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1340 XPUSH_CLOBJ (stash_image2d, mem);
1341
1342#if CL_VERSION_1_2
1343
1344void
1345gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1346 ALIAS:
1347 PPCODE:
1348 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1349 cl_gl_object_type type;
1350 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1351 HV *stash = stash_memory;
1352 switch (type)
1353 {
1354 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1355 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1356 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1357 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1358 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1359 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1360 }
1361 XPUSH_CLOBJ (stash, mem);
1362
1363#endif
728 1364
729void 1365void
730gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1366gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
731 PPCODE: 1367 PPCODE:
1368#if PREFER_1_1
732 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1369 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
733 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1370#else
1371 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1372#endif
1373 XPUSH_CLOBJ (stash_image2d, mem);
734 1374
735void 1375void
736gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1376gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
737 PPCODE: 1377 PPCODE:
1378#if PREFER_1_1
738 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1379 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
739 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1380#else
740 1381 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
741void 1382#endif
742gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1383 XPUSH_CLOBJ (stash_image3d, mem);
743 PPCODE:
744 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
745 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
746 1384
747#endif 1385#endif
748 1386
749void 1387void
750supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1388supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
770 1408
771void 1409void
772sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1410sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
773 PPCODE: 1411 PPCODE:
774 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1412 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
775 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1413 XPUSH_CLOBJ (stash_sampler, sampler);
776 1414
777void 1415void
778program_with_source (OpenCL::Context self, SV *program) 1416program_with_source (OpenCL::Context self, SV *program)
779 PPCODE: 1417 PPCODE:
780 STRLEN len; 1418 STRLEN len;
781 size_t len2; 1419 size_t len2;
782 const char *ptr = SvPVbyte (program, len); 1420 const char *ptr = SvPVbyte (program, len);
783 1421
784 len2 = len; 1422 len2 = len;
785 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1423 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
786 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1424 XPUSH_CLOBJ (stash_program, prog);
1425
1426void
1427program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1428 PPCODE:
1429 cl_uint device_count;
1430 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1431
1432 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1433 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1434
1435 binaries = SvRV (binaries);
1436
1437 if (device_count != av_len ((AV *)binaries) + 1)
1438 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1439
1440 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1441 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1442 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1443
1444 int i;
1445 for (i = 0; i < device_count; ++i)
1446 {
1447 STRLEN len;
1448 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1449 length_list [i] = len;
1450 }
1451
1452 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1453 length_list, binary_list,
1454 GIMME_V == G_ARRAY ? status_list : 0, &res));
1455
1456 EXTEND (SP, 2);
1457 PUSH_CLOBJ (stash_program, prog);
1458
1459 if (GIMME_V == G_ARRAY)
1460 {
1461 AV *av = newAV ();
1462 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1463
1464 for (i = device_count; i--; )
1465 av_store (av, i, newSViv (status_list [i]));
1466 }
1467
1468#if CL_VERSION_1_2
1469
1470void
1471program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1472 PPCODE:
1473 cl_uint device_count;
1474 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1475
1476 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1477
1478 XPUSH_CLOBJ (stash_program, prog);
1479
1480void
1481link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1482 CODE:
1483 cl_uint device_count = 0;
1484 cl_device_id *device_list = 0;
1485
1486 if (SvOK (devices))
1487 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1488
1489 cl_uint program_count;
1490 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1491
1492 void *user_data;
1493 program_callback pfn_notify = make_program_callback (notify, &user_data);
1494
1495 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1496 program_count, program_list, pfn_notify, user_data, &res));
1497
1498 XPUSH_CLOBJ (stash_program, prog);
1499
1500#endif
787 1501
788#BEGIN:context 1502#BEGIN:context
789 1503
790void 1504void
791reference_count (OpenCL::Context self) 1505reference_count (OpenCL::Context self)
808 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1522 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
809 int i, n = size / sizeof (*value); 1523 int i, n = size / sizeof (*value);
810 EXTEND (SP, n); 1524 EXTEND (SP, n);
811 for (i = 0; i < n; ++i) 1525 for (i = 0; i < n; ++i)
812 { 1526 {
813 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1527 PUSH_CLOBJ (stash_device, value [i]);
814 } 1528 }
815 1529
816void 1530void
817properties (OpenCL::Context self) 1531properties (OpenCL::Context self)
818 PPCODE: 1532 PPCODE:
833DESTROY (OpenCL::Queue self) 1547DESTROY (OpenCL::Queue self)
834 CODE: 1548 CODE:
835 clReleaseCommandQueue (self); 1549 clReleaseCommandQueue (self);
836 1550
837void 1551void
838enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1552read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1553 ALIAS:
1554 enqueue_read_buffer = 0
839 PPCODE: 1555 PPCODE:
840 cl_event ev = 0;
841 EVENT_LIST (6, items - 6); 1556 EVENT_LIST (6);
842 1557
843 SvUPGRADE (data, SVt_PV); 1558 SvUPGRADE (data, SVt_PV);
844 SvGROW (data, len); 1559 SvGROW (data, len);
845 SvPOK_only (data); 1560 SvPOK_only (data);
846 SvCUR_set (data, len); 1561 SvCUR_set (data, len);
1562
1563 cl_event ev = 0;
847 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1564 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
848 1565
849 if (ev) 1566 if (ev)
850 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1567 XPUSH_CLOBJ (stash_event, ev);
851 1568
852void 1569void
853enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1570write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1571 ALIAS:
1572 enqueue_write_buffer = 0
854 PPCODE: 1573 PPCODE:
855 cl_event ev = 0; 1574 EVENT_LIST (5);
1575
856 STRLEN len; 1576 STRLEN len;
857 char *ptr = SvPVbyte (data, len); 1577 char *ptr = SvPVbyte (data, len);
858 EVENT_LIST (5, items - 5);
859 1578
1579 cl_event ev = 0;
860 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1580 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
861 1581
862 if (ev) 1582 if (ev)
863 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1583 XPUSH_CLOBJ (stash_event, ev);
864 1584
1585#if CL_VERSION_1_2
1586
865void 1587void
866enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1588fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1589 ALIAS:
1590 enqueue_fill_buffer = 0
867 PPCODE: 1591 PPCODE:
1592 EVENT_LIST (5);
1593
1594 STRLEN len;
1595 char *ptr = SvPVbyte (data, len);
1596
868 cl_event ev = 0; 1597 cl_event ev = 0;
1598 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1599
1600 if (ev)
1601 XPUSH_CLOBJ (stash_event, ev);
1602
1603void
1604fill_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, ...)
1605 ALIAS:
1606 enqueue_fill_image = 0
1607 PPCODE:
1608 EVENT_LIST (12);
1609
1610 const size_t origin [3] = { x, y, z };
1611 const size_t region [3] = { width, height, depth };
1612
1613 const cl_float c_f [4] = { r, g, b, a };
1614 const cl_uint c_u [4] = { r, g, b, a };
1615 const cl_int c_s [4] = { r, g, b, a };
1616 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1617 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1618 cl_image_format format;
1619 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1620 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1621 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1622 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1623
1624 cl_event ev = 0;
1625 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1626 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627
1628 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev);
1630
1631#endif
1632
1633void
1634copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1635 ALIAS:
1636 enqueue_copy_buffer = 0
1637 PPCODE:
869 EVENT_LIST (6, items - 6); 1638 EVENT_LIST (6);
870 1639
1640 cl_event ev = 0;
871 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1641 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
872 1642
873 if (ev) 1643 if (ev)
874 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1644 XPUSH_CLOBJ (stash_event, ev);
875 1645
876void 1646void
877enqueue_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, ...) 1647read_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, ...)
1648 ALIAS:
1649 enqueue_read_buffer_rect = 0
878 PPCODE: 1650 PPCODE:
879 cl_event ev = 0; 1651 EVENT_LIST (17);
1652
880 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1653 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
881 const size_t host_origin[3] = { host_x, host_y, host_z }; 1654 const size_t host_origin[3] = { host_x, host_y, host_z };
882 const size_t region[3] = { width, height, depth }; 1655 const size_t region[3] = { width, height, depth };
883 EVENT_LIST (17, items - 17);
884 1656
885 if (!buf_row_pitch) 1657 if (!buf_row_pitch)
886 buf_row_pitch = region [0]; 1658 buf_row_pitch = region [0];
887 1659
888 if (!buf_slice_pitch) 1660 if (!buf_slice_pitch)
898 1670
899 SvUPGRADE (data, SVt_PV); 1671 SvUPGRADE (data, SVt_PV);
900 SvGROW (data, len); 1672 SvGROW (data, len);
901 SvPOK_only (data); 1673 SvPOK_only (data);
902 SvCUR_set (data, len); 1674 SvCUR_set (data, len);
1675
1676 cl_event ev = 0;
903 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)); 1677 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));
904 1678
905 if (ev) 1679 if (ev)
906 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1680 XPUSH_CLOBJ (stash_event, ev);
907 1681
908void 1682void
909enqueue_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, ...) 1683write_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, ...)
1684 ALIAS:
1685 enqueue_write_buffer_rect = 0
910 PPCODE: 1686 PPCODE:
911 cl_event ev = 0; 1687 EVENT_LIST (17);
1688
912 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1689 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
913 const size_t host_origin[3] = { host_x, host_y, host_z }; 1690 const size_t host_origin[3] = { host_x, host_y, host_z };
914 const size_t region[3] = { width, height, depth }; 1691 const size_t region[3] = { width, height, depth };
915 STRLEN len; 1692 STRLEN len;
916 char *ptr = SvPVbyte (data, len); 1693 char *ptr = SvPVbyte (data, len);
917 EVENT_LIST (17, items - 17);
918 1694
919 if (!buf_row_pitch) 1695 if (!buf_row_pitch)
920 buf_row_pitch = region [0]; 1696 buf_row_pitch = region [0];
921 1697
922 if (!buf_slice_pitch) 1698 if (!buf_slice_pitch)
931 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1707 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
932 1708
933 if (len < min_len) 1709 if (len < min_len)
934 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1710 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
935 1711
1712 cl_event ev = 0;
936 NEED_SUCCESS (EnqueueWriteBufferRect, (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)); 1713 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));
937 1714
938 if (ev) 1715 if (ev)
939 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1716 XPUSH_CLOBJ (stash_event, ev);
940 1717
941void 1718void
942enqueue_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, ...) 1719copy_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, ...)
1720 ALIAS:
1721 enqueue_copy_buffer_rect = 0
943 PPCODE: 1722 PPCODE:
944 cl_event ev = 0; 1723 EVENT_LIST (16);
1724
945 const size_t src_origin[3] = { src_x, src_y, src_z }; 1725 const size_t src_origin[3] = { src_x, src_y, src_z };
946 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1726 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
947 const size_t region[3] = { width, height, depth }; 1727 const size_t region[3] = { width, height, depth };
948 EVENT_LIST (16, items - 16);
949 1728
1729 cl_event ev = 0;
950 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)); 1730 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));
951 1731
952 if (ev) 1732 if (ev)
953 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1733 XPUSH_CLOBJ (stash_event, ev);
954 1734
955void 1735void
956enqueue_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, ...) 1736read_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, ...)
1737 ALIAS:
1738 enqueue_read_image = 0
957 PPCODE: 1739 PPCODE:
958 cl_event ev = 0; 1740 EVENT_LIST (12);
1741
959 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 };
960 const size_t region[3] = { width, height, depth }; 1743 const size_t region[3] = { width, height, depth };
961 EVENT_LIST (12, items - 12);
962 1744
963 if (!row_pitch) 1745 if (!row_pitch)
964 row_pitch = img_row_pitch (src); 1746 row_pitch = img_row_pitch (src);
965 1747
966 if (depth > 1 && !slice_pitch) 1748 if (depth > 1 && !slice_pitch)
970 1752
971 SvUPGRADE (data, SVt_PV); 1753 SvUPGRADE (data, SVt_PV);
972 SvGROW (data, len); 1754 SvGROW (data, len);
973 SvPOK_only (data); 1755 SvPOK_only (data);
974 SvCUR_set (data, len); 1756 SvCUR_set (data, len);
1757
1758 cl_event ev = 0;
975 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)); 1759 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));
976 1760
977 if (ev) 1761 if (ev)
978 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1762 XPUSH_CLOBJ (stash_event, ev);
979 1763
980void 1764void
981enqueue_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, ...) 1765write_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, ...)
1766 ALIAS:
1767 enqueue_write_image = 0
982 PPCODE: 1768 PPCODE:
983 cl_event ev = 0; 1769 EVENT_LIST (12);
1770
984 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1771 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
985 const size_t region[3] = { width, height, depth }; 1772 const size_t region[3] = { width, height, depth };
986 STRLEN len; 1773 STRLEN len;
987 char *ptr = SvPVbyte (data, len); 1774 char *ptr = SvPVbyte (data, len);
988 EVENT_LIST (12, items - 12);
989 1775
990 if (!row_pitch) 1776 if (!row_pitch)
991 row_pitch = img_row_pitch (dst); 1777 row_pitch = img_row_pitch (dst);
992 1778
993 if (depth > 1 && !slice_pitch) 1779 if (depth > 1 && !slice_pitch)
996 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1782 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
997 1783
998 if (len < min_len) 1784 if (len < min_len)
999 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1785 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1000 1786
1787 cl_event ev = 0;
1001 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1788 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));
1002 1789
1003 if (ev) 1790 if (ev)
1004 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1791 XPUSH_CLOBJ (stash_event, ev);
1005 1792
1006void 1793void
1007enqueue_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, ...) 1794copy_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, ...)
1795 ALIAS:
1796 enqueue_copy_image = 0
1008 PPCODE: 1797 PPCODE:
1009 cl_event ev = 0; 1798 EVENT_LIST (12);
1799
1010 const size_t src_origin[3] = { src_x, src_y, src_z }; 1800 const size_t src_origin[3] = { src_x, src_y, src_z };
1011 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1801 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1012 const size_t region[3] = { width, height, depth }; 1802 const size_t region[3] = { width, height, depth };
1013 EVENT_LIST (12, items - 12);
1014 1803
1804 cl_event ev = 0;
1015 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1805 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1016 1806
1017 if (ev) 1807 if (ev)
1018 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1019 1809
1020void 1810void
1021enqueue_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, ...) 1811copy_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, ...)
1812 ALIAS:
1813 enqueue_copy_image_to_buffer = 0
1022 PPCODE: 1814 PPCODE:
1815 EVENT_LIST (10);
1816
1817 const size_t src_origin[3] = { src_x, src_y, src_z };
1818 const size_t region [3] = { width, height, depth };
1819
1023 cl_event ev = 0; 1820 cl_event ev = 0;
1821 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1822
1823 if (ev)
1824 XPUSH_CLOBJ (stash_event, ev);
1825
1826void
1827copy_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, ...)
1828 ALIAS:
1829 enqueue_copy_buffer_to_image = 0
1830 PPCODE:
1831 EVENT_LIST (10);
1832
1024 const size_t src_origin[3] = { src_x, src_y, src_z }; 1833 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1834 const size_t region [3] = { width, height, depth };
1835
1836 cl_event ev = 0;
1837 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1838
1839 if (ev)
1840 XPUSH_CLOBJ (stash_event, ev);
1841
1842void
1843map_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, ...)
1844 ALIAS:
1845 enqueue_map_buffer = 0
1846 PPCODE:
1847 EVENT_LIST (6);
1848
1849 size_t cb = SvIV (cb_);
1850
1851 if (!SvOK (cb_))
1852 {
1853 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1854 cb -= offset;
1855 }
1856
1857 cl_event ev;
1858 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1859 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1860
1861void
1862map_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, ...)
1863 ALIAS:
1864 enqueue_map_image = 0
1865 PPCODE:
1866 size_t width = SvIV (width_);
1867 if (!SvOK (width_))
1868 {
1869 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1870 width -= x;
1871 }
1872
1873 size_t height = SvIV (width_);
1874 if (!SvOK (height_))
1875 {
1876 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1877 height -= y;
1878 }
1879
1880 size_t depth = SvIV (width_);
1881 if (!SvOK (depth_))
1882 {
1883 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1884 depth -= z;
1885
1886 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1887 if (!depth)
1888 depth = 1;
1889 }
1890
1891 const size_t origin[3] = { x, y, z };
1025 const size_t region[3] = { width, height, depth }; 1892 const size_t region[3] = { width, height, depth };
1893 size_t row_pitch, slice_pitch;
1026 EVENT_LIST (10, items - 10); 1894 EVENT_LIST (10);
1027 1895
1896 cl_event ev;
1897 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));
1898
1899 size_t cb = slice_pitch ? slice_pitch * region [2]
1900 : row_pitch ? row_pitch * region [1]
1901 : region [0];
1902
1903 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1904
1905void
1906unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1907 PPCODE:
1908 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1909 if (GIMME_V != G_VOID)
1910 {
1911 clRetainEvent (mapped->event);
1912 XPUSH_CLOBJ (stash_event, mapped->event);
1913 }
1914
1915void
1916task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1917 ALIAS:
1918 enqueue_task = 0
1919 PPCODE:
1920 EVENT_LIST (2);
1921
1922 cl_event ev = 0;
1028 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1923 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1029 1924
1030 if (ev) 1925 if (ev)
1031 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1926 XPUSH_CLOBJ (stash_event, ev);
1032 1927
1033void 1928void
1034enqueue_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, ...)
1035 PPCODE:
1036 cl_event ev = 0;
1037 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1038 const size_t region[3] = { width, height, depth };
1039 EVENT_LIST (10, items - 10);
1040
1041 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1042
1043 if (ev)
1044 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1045
1046void
1047enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1048 PPCODE:
1049 cl_event ev = 0;
1050 EVENT_LIST (2, items - 2);
1051
1052 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1053
1054 if (ev)
1055 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1056
1057void
1058enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1929nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1930 ALIAS:
1931 enqueue_nd_range_kernel = 0
1059 PPCODE: 1932 PPCODE:
1060 cl_event ev = 0; 1933 EVENT_LIST (5);
1934
1061 size_t *gwo = 0, *gws, *lws = 0; 1935 size_t *gwo = 0, *gws, *lws = 0;
1062 int gws_len; 1936 int gws_len;
1063 size_t *lists; 1937 size_t *lists;
1064 int i; 1938 int i;
1065 EVENT_LIST (5, items - 5);
1066 1939
1067 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1940 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1068 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1941 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1069 1942
1070 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1943 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1071 1944
1072 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1945 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1073 1946
1074 gws = lists + gws_len * 0; 1947 gws = lists + gws_len * 0;
1075 for (i = 0; i < gws_len; ++i) 1948 for (i = 0; i < gws_len; ++i)
1949 {
1076 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1950 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1951 // at least nvidia crashes for 0-sized work group sizes, work around
1952 if (!gws [i])
1953 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1954 }
1077 1955
1078 if (SvOK (global_work_offset)) 1956 if (SvOK (global_work_offset))
1079 { 1957 {
1080 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1958 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1081 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1959 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1088 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1966 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1089 } 1967 }
1090 1968
1091 if (SvOK (local_work_size)) 1969 if (SvOK (local_work_size))
1092 { 1970 {
1093 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1971 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1094 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1972 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1095 1973
1096 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1974 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1097 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1975 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1098 1976
1099 lws = lists + gws_len * 2; 1977 lws = lists + gws_len * 2;
1100 for (i = 0; i < gws_len; ++i) 1978 for (i = 0; i < gws_len; ++i)
1979 {
1101 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1980 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1981 // at least nvidia crashes for 0-sized work group sizes, work around
1982 if (!lws [i])
1983 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1984 }
1102 } 1985 }
1103 1986
1987 cl_event ev = 0;
1104 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1988 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1105 1989
1106 if (ev) 1990 if (ev)
1107 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1991 XPUSH_CLOBJ (stash_event, ev);
1992
1993#if CL_VERSION_1_2
1994
1995void
1996migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1997 ALIAS:
1998 enqueue_migrate_mem_objects = 0
1999 PPCODE:
2000 EVENT_LIST (3);
2001
2002 cl_uint obj_count;
2003 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2004
2005 cl_event ev = 0;
2006 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2007
2008 if (ev)
2009 XPUSH_CLOBJ (stash_event, ev);
2010
2011#endif
1108 2012
1109#if cl_apple_gl_sharing || cl_khr_gl_sharing 2013#if cl_apple_gl_sharing || cl_khr_gl_sharing
1110 2014
1111void 2015void
1112enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2016acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1113 ALIAS: 2017 ALIAS:
2018 release_gl_objects = 1
2019 enqueue_acquire_gl_objects = 0
1114 enqueue_release_gl_objects = 1 2020 enqueue_release_gl_objects = 1
1115 CODE: 2021 PPCODE:
1116 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2022 EVENT_LIST (2);
1117 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2023
2024 cl_uint obj_count;
2025 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1118 2026
1119 cl_event ev = 0; 2027 cl_event ev = 0;
1120 EVENT_LIST (2, items - 2);
1121 AV *av = (AV *)SvRV (objects);
1122 cl_uint num_objects = av_len (av) + 1;
1123 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1124 int i;
1125
1126 for (i = num_objects; i--; )
1127 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1128 2028
1129 if (ix) 2029 if (ix)
1130 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2030 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1131 else 2031 else
1132 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2032 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1133 2033
1134 if (ev) 2034 if (ev)
1135 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2035 XPUSH_CLOBJ (stash_event, ev);
1136 2036
1137#endif 2037#endif
1138 2038
1139void 2039void
1140enqueue_marker (OpenCL::Queue self)
1141 PPCODE:
1142 cl_event ev;
1143 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1144 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1145
1146void
1147enqueue_wait_for_events (OpenCL::Queue self, ...) 2040wait_for_events (OpenCL::Queue self, ...)
2041 ALIAS:
2042 enqueue_wait_for_events = 0
1148 CODE: 2043 CODE:
1149 EVENT_LIST (1, items - 1); 2044 EVENT_LIST (1);
2045#if PREFER_1_1
1150 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2046 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2047#else
2048 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2049#endif
1151 2050
1152void 2051void
1153enqueue_barrier (OpenCL::Queue self) 2052marker (OpenCL::Queue self, ...)
2053 ALIAS:
2054 enqueue_marker = 0
1154 CODE: 2055 PPCODE:
2056 EVENT_LIST (1);
2057 cl_event ev = 0;
2058#if PREFER_1_1
2059 if (!event_list_count)
2060 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2061 else
2062#if CL_VERSION_1_2
2063 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2064#else
2065 {
2066 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2067 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2068 }
2069#endif
2070#else
2071 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2072#endif
2073 if (ev)
2074 XPUSH_CLOBJ (stash_event, ev);
2075
2076void
2077barrier (OpenCL::Queue self, ...)
2078 ALIAS:
2079 enqueue_barrier = 0
2080 PPCODE:
2081 EVENT_LIST (1);
2082 cl_event ev = 0;
2083#if PREFER_1_1
2084 if (!event_list_count && GIMME_V == G_VOID)
1155 NEED_SUCCESS (EnqueueBarrier, (self)); 2085 NEED_SUCCESS (EnqueueBarrier, (self));
2086 else
2087#if CL_VERSION_1_2
2088 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2089#else
2090 {
2091 if (event_list_count)
2092 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2093
2094 if (GIMME_V != G_VOID)
2095 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2096 }
2097#endif
2098#else
2099 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2100#endif
2101 if (ev)
2102 XPUSH_CLOBJ (stash_event, ev);
1156 2103
1157void 2104void
1158flush (OpenCL::Queue self) 2105flush (OpenCL::Queue self)
1159 CODE: 2106 CODE:
1160 NEED_SUCCESS (Flush, (self)); 2107 NEED_SUCCESS (Flush, (self));
1178 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2125 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1179 EXTEND (SP, 1); 2126 EXTEND (SP, 1);
1180 const int i = 0; 2127 const int i = 0;
1181 { 2128 {
1182 NEED_SUCCESS (RetainContext, (value [i])); 2129 NEED_SUCCESS (RetainContext, (value [i]));
1183 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2130 PUSH_CLOBJ (stash_context, value [i]);
1184 } 2131 }
1185 2132
1186void 2133void
1187device (OpenCL::Queue self) 2134device (OpenCL::Queue self)
1188 PPCODE: 2135 PPCODE:
1189 cl_device_id value [1]; 2136 cl_device_id value [1];
1190 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2137 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1191 EXTEND (SP, 1); 2138 EXTEND (SP, 1);
1192 const int i = 0; 2139 const int i = 0;
1193 { 2140 {
1194 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2141 PUSH_CLOBJ (stash_device, value [i]);
1195 } 2142 }
1196 2143
1197void 2144void
1198reference_count (OpenCL::Queue self) 2145reference_count (OpenCL::Queue self)
1199 PPCODE: 2146 PPCODE:
1208 PPCODE: 2155 PPCODE:
1209 cl_command_queue_properties value [1]; 2156 cl_command_queue_properties value [1];
1210 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2157 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1211 EXTEND (SP, 1); 2158 EXTEND (SP, 1);
1212 const int i = 0; 2159 const int i = 0;
1213 PUSHs (sv_2mortal (newSViv (value [i]))); 2160 PUSHs (sv_2mortal (newSVuv (value [i])));
1214 2161
1215#END:command_queue 2162#END:command_queue
1216 2163
1217MODULE = OpenCL PACKAGE = OpenCL::Memory 2164MODULE = OpenCL PACKAGE = OpenCL::Memory
1218 2165
1233 PPCODE: 2180 PPCODE:
1234 cl_mem_object_type value [1]; 2181 cl_mem_object_type value [1];
1235 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2182 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1236 EXTEND (SP, 1); 2183 EXTEND (SP, 1);
1237 const int i = 0; 2184 const int i = 0;
1238 PUSHs (sv_2mortal (newSViv (value [i]))); 2185 PUSHs (sv_2mortal (newSVuv (value [i])));
1239 2186
1240void 2187void
1241flags (OpenCL::Memory self) 2188flags (OpenCL::Memory self)
1242 PPCODE: 2189 PPCODE:
1243 cl_mem_flags value [1]; 2190 cl_mem_flags value [1];
1244 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2191 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1245 EXTEND (SP, 1); 2192 EXTEND (SP, 1);
1246 const int i = 0; 2193 const int i = 0;
1247 PUSHs (sv_2mortal (newSViv (value [i]))); 2194 PUSHs (sv_2mortal (newSVuv (value [i])));
1248 2195
1249void 2196void
1250size (OpenCL::Memory self) 2197size (OpenCL::Memory self)
1251 ALIAS: 2198 ALIAS:
1252 size = CL_MEM_SIZE 2199 size = CL_MEM_SIZE
1286 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2233 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1287 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1288 const int i = 0; 2235 const int i = 0;
1289 { 2236 {
1290 NEED_SUCCESS (RetainContext, (value [i])); 2237 NEED_SUCCESS (RetainContext, (value [i]));
1291 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2238 PUSH_CLOBJ (stash_context, value [i]);
1292 } 2239 }
1293 2240
1294void 2241void
1295associated_memobject (OpenCL::Memory self) 2242associated_memobject (OpenCL::Memory self)
1296 PPCODE: 2243 PPCODE:
1298 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2245 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1299 EXTEND (SP, 1); 2246 EXTEND (SP, 1);
1300 const int i = 0; 2247 const int i = 0;
1301 { 2248 {
1302 NEED_SUCCESS (RetainMemObject, (value [i])); 2249 NEED_SUCCESS (RetainMemObject, (value [i]));
1303 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2250 PUSH_CLOBJ (stash_memory, value [i]);
1304 } 2251 }
1305 2252
1306#END:mem 2253#END:mem
1307 2254
1308#if cl_apple_gl_sharing || cl_khr_gl_sharing 2255#if cl_apple_gl_sharing || cl_khr_gl_sharing
1328 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2275 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1329 2276
1330 cl_buffer_region crdata = { origin, size }; 2277 cl_buffer_region crdata = { origin, size };
1331 2278
1332 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2279 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1333 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2280 XPUSH_CLOBJ (stash_buffer, mem);
1334 2281
1335MODULE = OpenCL PACKAGE = OpenCL::Image 2282MODULE = OpenCL PACKAGE = OpenCL::Image
1336 2283
1337void 2284void
1338image_info (OpenCL::Image self, cl_image_info name) 2285image_info (OpenCL::Image self, cl_image_info name)
1339 PPCODE: 2286 PPCODE:
1340 INFO (Image) 2287 INFO (Image)
2288
2289void
2290format (OpenCL::Image self)
2291 PPCODE:
2292 cl_image_format format;
2293 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2294 EXTEND (SP, 2);
2295 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2296 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1341 2297
1342#BEGIN:image 2298#BEGIN:image
1343 2299
1344void 2300void
1345element_size (OpenCL::Image self) 2301element_size (OpenCL::Image self)
1415 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1416 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
1417 const int i = 0; 2373 const int i = 0;
1418 { 2374 {
1419 NEED_SUCCESS (RetainContext, (value [i])); 2375 NEED_SUCCESS (RetainContext, (value [i]));
1420 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2376 PUSH_CLOBJ (stash_context, value [i]);
1421 } 2377 }
1422 2378
1423void 2379void
1424normalized_coords (OpenCL::Sampler self) 2380normalized_coords (OpenCL::Sampler self)
1425 PPCODE: 2381 PPCODE:
1426 cl_addressing_mode value [1]; 2382 cl_addressing_mode value [1];
1427 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2383 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1428 EXTEND (SP, 1); 2384 EXTEND (SP, 1);
1429 const int i = 0; 2385 const int i = 0;
1430 PUSHs (sv_2mortal (newSViv (value [i]))); 2386 PUSHs (sv_2mortal (newSVuv (value [i])));
1431 2387
1432void 2388void
1433addressing_mode (OpenCL::Sampler self) 2389addressing_mode (OpenCL::Sampler self)
1434 PPCODE: 2390 PPCODE:
1435 cl_filter_mode value [1]; 2391 cl_filter_mode value [1];
1436 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2392 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1437 EXTEND (SP, 1); 2393 EXTEND (SP, 1);
1438 const int i = 0; 2394 const int i = 0;
1439 PUSHs (sv_2mortal (newSViv (value [i]))); 2395 PUSHs (sv_2mortal (newSVuv (value [i])));
1440 2396
1441void 2397void
1442filter_mode (OpenCL::Sampler self) 2398filter_mode (OpenCL::Sampler self)
1443 PPCODE: 2399 PPCODE:
1444 cl_bool value [1]; 2400 cl_bool value [1];
1455DESTROY (OpenCL::Program self) 2411DESTROY (OpenCL::Program self)
1456 CODE: 2412 CODE:
1457 clReleaseProgram (self); 2413 clReleaseProgram (self);
1458 2414
1459void 2415void
1460build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2416build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2417 ALIAS:
2418 build_async = 1
1461 CODE: 2419 CODE:
2420 cl_uint device_count = 0;
2421 cl_device_id *device_list = 0;
2422
2423 if (SvOK (devices))
2424 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2425
2426 void *user_data;
2427 program_callback pfn_notify = make_program_callback (notify, &user_data);
2428
2429 if (ix)
2430 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2431 else
2432 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2433
2434#if CL_VERSION_1_2
2435
2436void
2437compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2438 CODE:
2439 cl_uint device_count = 0;
2440 cl_device_id *device_list = 0;
2441
2442 if (SvOK (devices))
2443 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2444
2445 cl_uint header_count = 0;
2446 cl_program *header_list = 0;
2447 const char **header_name = 0;
2448
2449 if (SvOK (headers))
2450 {
2451 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2452 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2453
2454 HV *hv = (HV *)SvRV (devices);
2455
2456 header_count = hv_iterinit (hv);
2457 header_list = tmpbuf (sizeof (*header_list) * header_count);
2458 header_name = tmpbuf (sizeof (*header_name) * header_count);
2459
2460 HE *he;
2461 int i = 0;
2462 while (he = hv_iternext (hv))
2463 {
2464 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2465 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2466 ++i;
2467 }
2468 }
2469
2470 void *user_data;
2471 program_callback pfn_notify = make_program_callback (notify, &user_data);
2472
1462 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2473 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2474 header_count, header_list, header_name, pfn_notify, user_data));
2475
2476#endif
1463 2477
1464void 2478void
1465build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2479build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1466 PPCODE: 2480 PPCODE:
1467 size_t size; 2481 size_t size;
1502 2516
1503void 2517void
1504kernel (OpenCL::Program program, SV *function) 2518kernel (OpenCL::Program program, SV *function)
1505 PPCODE: 2519 PPCODE:
1506 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2520 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1507 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2521 XPUSH_CLOBJ (stash_kernel, kernel);
2522
2523void
2524kernels_in_program (OpenCL::Program program)
2525 PPCODE:
2526 cl_uint num_kernels;
2527 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2528 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2529 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2530
2531 int i;
2532 EXTEND (SP, num_kernels);
2533 for (i = 0; i < num_kernels; ++i)
2534 PUSH_CLOBJ (stash_kernel, kernels [i]);
1508 2535
1509void 2536void
1510info (OpenCL::Program self, cl_program_info name) 2537info (OpenCL::Program self, cl_program_info name)
1511 PPCODE: 2538 PPCODE:
1512 INFO (Program) 2539 INFO (Program)
1530 { 2557 {
1531 SV *sv = sv_2mortal (newSV (sizes [i])); 2558 SV *sv = sv_2mortal (newSV (sizes [i]));
1532 SvUPGRADE (sv, SVt_PV); 2559 SvUPGRADE (sv, SVt_PV);
1533 SvPOK_only (sv); 2560 SvPOK_only (sv);
1534 SvCUR_set (sv, sizes [i]); 2561 SvCUR_set (sv, sizes [i]);
1535 ptrs [i] = SvPVX (sv); 2562 ptrs [i] = (void *)SvPVX (sv);
1536 PUSHs (sv); 2563 PUSHs (sv);
1537 } 2564 }
1538 2565
1539 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2566 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1540 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2567 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1560 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2587 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1561 EXTEND (SP, 1); 2588 EXTEND (SP, 1);
1562 const int i = 0; 2589 const int i = 0;
1563 { 2590 {
1564 NEED_SUCCESS (RetainContext, (value [i])); 2591 NEED_SUCCESS (RetainContext, (value [i]));
1565 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2592 PUSH_CLOBJ (stash_context, value [i]);
1566 } 2593 }
1567 2594
1568void 2595void
1569devices (OpenCL::Program self) 2596devices (OpenCL::Program self)
1570 PPCODE: 2597 PPCODE:
1574 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2601 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1575 int i, n = size / sizeof (*value); 2602 int i, n = size / sizeof (*value);
1576 EXTEND (SP, n); 2603 EXTEND (SP, n);
1577 for (i = 0; i < n; ++i) 2604 for (i = 0; i < n; ++i)
1578 { 2605 {
1579 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2606 PUSH_CLOBJ (stash_device, value [i]);
1580 } 2607 }
1581 2608
1582void 2609void
1583source (OpenCL::Program self) 2610source (OpenCL::Program self)
1584 PPCODE: 2611 PPCODE:
1610DESTROY (OpenCL::Kernel self) 2637DESTROY (OpenCL::Kernel self)
1611 CODE: 2638 CODE:
1612 clReleaseKernel (self); 2639 clReleaseKernel (self);
1613 2640
1614void 2641void
2642setf (OpenCL::Kernel self, const char *format, ...)
2643 CODE:
2644 int i;
2645 for (i = 2; ; ++i)
2646 {
2647 while (*format == ' ')
2648 ++format;
2649
2650 char type = *format++;
2651
2652 if (!type)
2653 break;
2654
2655 if (i >= items)
2656 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2657
2658 SV *sv = ST (i);
2659
2660 union
2661 {
2662 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2663 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2664 cl_half ch; cl_float cf; cl_double cd;
2665 cl_mem cm;
2666 cl_sampler ca;
2667 size_t cz;
2668 cl_event ce;
2669 } arg;
2670 size_t size;
2671 int nullarg = 0;
2672
2673 switch (type)
2674 {
2675 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2676 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2677 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2678 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2679 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2680 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2681 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2682 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2683
2684 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2685 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2686 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2687
2688 case 'z': nullarg = 1; size = SvIV (sv); break;
2689
2690 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2691 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2692 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2693
2694 default:
2695 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2696 }
2697
2698 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2699 if (res)
2700 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2701 }
2702
2703 if (i != items)
2704 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2705
2706void
1615set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2707set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1616 CODE: 2708 CODE:
1617 clSetKernelArg (self, idx, sizeof (value), &value); 2709 clSetKernelArg (self, idx, sizeof (value), &value);
1618 2710
1619void 2711void
1667 clSetKernelArg (self, idx, sizeof (value), &value); 2759 clSetKernelArg (self, idx, sizeof (value), &value);
1668 2760
1669void 2761void
1670set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2762set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1671 CODE: 2763 CODE:
2764 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2765
2766void
2767set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2768 CODE:
2769 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2770
2771void
2772set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2773 ALIAS:
2774 set_image2d = 0
2775 set_image3d = 0
2776 CODE:
2777 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2778
2779void
2780set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2781 CODE:
1672 clSetKernelArg (self, idx, sizeof (value), &value); 2782 clSetKernelArg (self, idx, sizeof (value), &value);
1673 2783
1674void 2784void
1675set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2785set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1676 CODE: 2786 CODE:
1677 clSetKernelArg (self, idx, sizeof (value), &value); 2787 clSetKernelArg (self, idx, size, 0);
1678
1679void
1680set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1681 CODE:
1682 clSetKernelArg (self, idx, sizeof (value), &value);
1683
1684void
1685set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1686 CODE:
1687 clSetKernelArg (self, idx, sizeof (value), &value);
1688
1689void
1690set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1691 CODE:
1692 clSetKernelArg (self, idx, sizeof (value), &value);
1693 2788
1694void 2789void
1695set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2790set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1696 CODE: 2791 CODE:
1697 clSetKernelArg (self, idx, sizeof (value), &value); 2792 clSetKernelArg (self, idx, sizeof (value), &value);
1733 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2828 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1734 EXTEND (SP, 1); 2829 EXTEND (SP, 1);
1735 const int i = 0; 2830 const int i = 0;
1736 { 2831 {
1737 NEED_SUCCESS (RetainContext, (value [i])); 2832 NEED_SUCCESS (RetainContext, (value [i]));
1738 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2833 PUSH_CLOBJ (stash_context, value [i]);
1739 } 2834 }
1740 2835
1741void 2836void
1742program (OpenCL::Kernel self) 2837program (OpenCL::Kernel self)
1743 PPCODE: 2838 PPCODE:
1745 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2840 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1746 EXTEND (SP, 1); 2841 EXTEND (SP, 1);
1747 const int i = 0; 2842 const int i = 0;
1748 { 2843 {
1749 NEED_SUCCESS (RetainProgram, (value [i])); 2844 NEED_SUCCESS (RetainProgram, (value [i]));
1750 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2845 PUSH_CLOBJ (stash_program, value [i]);
1751 } 2846 }
1752 2847
1753#END:kernel 2848#END:kernel
1754 2849
1755void 2850void
1802 const int i = 0; 2897 const int i = 0;
1803 PUSHs (sv_2mortal (newSVuv (value [i]))); 2898 PUSHs (sv_2mortal (newSVuv (value [i])));
1804 2899
1805#END:kernel_work_group 2900#END:kernel_work_group
1806 2901
2902#if CL_VERSION_1_2
2903
2904void
2905arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2906 PPCODE:
2907 size_t size;
2908 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2909 SV *sv = sv_2mortal (newSV (size));
2910 SvUPGRADE (sv, SVt_PV);
2911 SvPOK_only (sv);
2912 SvCUR_set (sv, size);
2913 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2914 XPUSHs (sv);
2915
2916#BEGIN:kernel_arg
2917
2918void
2919arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2920 PPCODE:
2921 cl_kernel_arg_address_qualifier value [1];
2922 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2923 EXTEND (SP, 1);
2924 const int i = 0;
2925 PUSHs (sv_2mortal (newSVuv (value [i])));
2926
2927void
2928arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2929 PPCODE:
2930 cl_kernel_arg_access_qualifier value [1];
2931 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2932 EXTEND (SP, 1);
2933 const int i = 0;
2934 PUSHs (sv_2mortal (newSVuv (value [i])));
2935
2936void
2937arg_type_name (OpenCL::Kernel self, cl_uint idx)
2938 ALIAS:
2939 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2940 arg_name = CL_KERNEL_ARG_NAME
2941 PPCODE:
2942 size_t size;
2943 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2944 char *value = tmpbuf (size);
2945 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2946 EXTEND (SP, 1);
2947 const int i = 0;
2948 PUSHs (sv_2mortal (newSVpv (value, 0)));
2949
2950void
2951arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2952 PPCODE:
2953 cl_kernel_arg_type_qualifier value [1];
2954 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2955 EXTEND (SP, 1);
2956 const int i = 0;
2957 PUSHs (sv_2mortal (newSVuv (value [i])));
2958
2959#END:kernel_arg
2960
2961#endif
2962
1807MODULE = OpenCL PACKAGE = OpenCL::Event 2963MODULE = OpenCL PACKAGE = OpenCL::Event
1808 2964
1809void 2965void
1810DESTROY (OpenCL::Event self) 2966DESTROY (OpenCL::Event self)
1811 CODE: 2967 CODE:
1813 2969
1814void 2970void
1815wait (OpenCL::Event self) 2971wait (OpenCL::Event self)
1816 CODE: 2972 CODE:
1817 clWaitForEvents (1, &self); 2973 clWaitForEvents (1, &self);
2974
2975void
2976cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2977 CODE:
2978 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1818 2979
1819void 2980void
1820info (OpenCL::Event self, cl_event_info name) 2981info (OpenCL::Event self, cl_event_info name)
1821 PPCODE: 2982 PPCODE:
1822 INFO (Event) 2983 INFO (Event)
1830 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2991 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1831 EXTEND (SP, 1); 2992 EXTEND (SP, 1);
1832 const int i = 0; 2993 const int i = 0;
1833 { 2994 {
1834 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2995 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1835 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2996 PUSH_CLOBJ (stash_queue, value [i]);
1836 } 2997 }
1837 2998
1838void 2999void
1839command_type (OpenCL::Event self) 3000command_type (OpenCL::Event self)
1840 PPCODE: 3001 PPCODE:
1863 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3024 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1864 EXTEND (SP, 1); 3025 EXTEND (SP, 1);
1865 const int i = 0; 3026 const int i = 0;
1866 { 3027 {
1867 NEED_SUCCESS (RetainContext, (value [i])); 3028 NEED_SUCCESS (RetainContext, (value [i]));
1868 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 3029 PUSH_CLOBJ (stash_context, value [i]);
1869 } 3030 }
1870 3031
1871#END:event 3032#END:event
1872 3033
1873void 3034void
1898void 3059void
1899set_status (OpenCL::UserEvent self, cl_int execution_status) 3060set_status (OpenCL::UserEvent self, cl_int execution_status)
1900 CODE: 3061 CODE:
1901 clSetUserEventStatus (self, execution_status); 3062 clSetUserEventStatus (self, execution_status);
1902 3063
3064MODULE = OpenCL PACKAGE = OpenCL::Mapped
3065
3066void
3067DESTROY (SV *self)
3068 CODE:
3069 OpenCL__Mapped mapped = SvMAPPED (self);
3070
3071 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3072 mapped_detach (self, mapped);
3073
3074 clReleaseCommandQueue (mapped->queue);
3075 clReleaseEvent (mapped->event);
3076 Safefree (mapped);
3077
3078void
3079unmap (OpenCL::Mapped self, ...)
3080 CODE:
3081 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3082
3083bool
3084mapped (OpenCL::Mapped self)
3085 CODE:
3086 RETVAL = !!self->ptr;
3087 OUTPUT:
3088 RETVAL
3089
3090void
3091wait (OpenCL::Mapped self)
3092 PPCODE:
3093 if (self->event)
3094 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3095
3096void
3097event (OpenCL::Mapped self)
3098 PPCODE:
3099 if (!self->event)
3100 XSRETURN_UNDEF;
3101
3102 clRetainEvent (self->event);
3103 XPUSH_CLOBJ (stash_event, self->event);
3104
3105size_t
3106size (OpenCL::Mapped self)
3107 CODE:
3108 RETVAL = self->cb;
3109 OUTPUT:
3110 RETVAL
3111
3112IV
3113ptr (OpenCL::Mapped self)
3114 CODE:
3115 RETVAL = PTR2IV (self->ptr);
3116 OUTPUT:
3117 RETVAL
3118
3119void
3120set (OpenCL::Mapped self, size_t offset, SV *data)
3121 CODE:
3122 STRLEN len;
3123 const char *ptr = SvPVbyte (data, len);
3124
3125 if (offset + len > self->cb)
3126 croak ("OpenCL::Mapped::set out of bound condition detected");
3127
3128 memcpy (offset + (char *)self->ptr, ptr, len);
3129
3130MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3131
3132MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3133
3134IV
3135row_pitch (OpenCL::Mapped self)
3136 ALIAS:
3137 slice_pitch = 1
3138 CODE:
3139 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3140 OUTPUT:
3141 RETVAL
3142
3143
3144

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines