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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.50 by root, Tue Apr 24 14:57:06 2012 UTC vs.
Revision 1.82 by root, Thu Oct 11 05:02:54 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
6
7#include "ecb.h"//D
8
9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
10#include "xthread.h"
11#include "schmorp.h"
4 12
5#ifdef I_DLFCN 13#ifdef I_DLFCN
6 #include <dlfcn.h> 14 #include <dlfcn.h>
7#endif 15#endif
8 16
22#if PREFER_1_1 30#if PREFER_1_1
23 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
24#endif 32#endif
25 33
26#ifdef __APPLE__ 34#ifdef __APPLE__
27 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
28#else 36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
29 #include <CL/opencl.h> 40#include CLHDR(opencl.h)
41
42#if 0
43#ifndef CL_VERSION_1_2
44 #include CLHDR(cl_d3d9.h)
45#endif
46#endif
47
48#if _WIN32
49 #include CLHDR(cl_d3d10.h)
50 #if CL_VERSION_1_2
51 #include CLHDR<cl_d3d11.h>
52 #endif
53 #include CLHDR<cl_dx9_media_sharing.h.h>
30#endif 54#endif
31 55
32#ifndef CL_VERSION_1_2 56#ifndef CL_VERSION_1_2
33 #undef PREFER_1_1 57 #undef PREFER_1_1
34 #define PREFER_1_1 1 58 #define PREFER_1_1 1
35#endif 59#endif
36 60
61// make sure all constants we might use are actually defined
62#include "default.h"
63
37typedef cl_platform_id OpenCL__Platform; 64typedef cl_platform_id OpenCL__Platform;
38typedef cl_device_id OpenCL__Device; 65typedef cl_device_id OpenCL__Device;
66typedef cl_device_id OpenCL__SubDevice;
39typedef cl_context OpenCL__Context; 67typedef cl_context OpenCL__Context;
40typedef cl_command_queue OpenCL__Queue; 68typedef cl_command_queue OpenCL__Queue;
41typedef cl_mem OpenCL__Memory; 69typedef cl_mem OpenCL__Memory;
42typedef cl_mem OpenCL__Buffer; 70typedef cl_mem OpenCL__Buffer;
43typedef cl_mem OpenCL__BufferObj; 71typedef cl_mem OpenCL__BufferObj;
44typedef cl_mem OpenCL__Image; 72typedef cl_mem OpenCL__Image;
45typedef cl_mem OpenCL__Image2D;
46typedef cl_mem OpenCL__Image3D;
47typedef cl_mem OpenCL__Memory_ornull; 73typedef cl_mem OpenCL__Memory_ornull;
48typedef cl_mem OpenCL__Buffer_ornull; 74typedef cl_mem OpenCL__Buffer_ornull;
49typedef cl_mem OpenCL__Image_ornull; 75typedef cl_mem OpenCL__Image_ornull;
50typedef cl_mem OpenCL__Image2D_ornull;
51typedef cl_mem OpenCL__Image3D_ornull;
52typedef cl_sampler OpenCL__Sampler; 76typedef cl_sampler OpenCL__Sampler;
53typedef cl_program OpenCL__Program; 77typedef cl_program OpenCL__Program;
54typedef cl_kernel OpenCL__Kernel; 78typedef cl_kernel OpenCL__Kernel;
55typedef cl_event OpenCL__Event; 79typedef cl_event OpenCL__Event;
56typedef cl_event OpenCL__UserEvent; 80typedef cl_event OpenCL__UserEvent;
57 81
58typedef SV *FUTURE; 82typedef struct mapped * OpenCL__Mapped;
83
84static HV
85 *stash_platform,
86 *stash_device,
87 *stash_subdevice,
88 *stash_context,
89 *stash_queue,
90 *stash_program,
91 *stash_kernel,
92 *stash_sampler,
93 *stash_event,
94 *stash_userevent,
95 *stash_memory,
96 *stash_buffer,
97 *stash_bufferobj,
98 *stash_image,
99 *stash_image1d,
100 *stash_image1darray,
101 *stash_image1dbuffer,
102 *stash_image2d,
103 *stash_image2darray,
104 *stash_image3d,
105 *stash_mapped,
106 *stash_mappedbuffer,
107 *stash_mappedimage;
59 108
60/*****************************************************************************/ 109/*****************************************************************************/
61 110
62// name must include a leading underscore 111// name must include a leading underscore
63// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 112// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
103} 152}
104 153
105/*****************************************************************************/ 154/*****************************************************************************/
106 155
107/* up to two temporary buffers */ 156/* up to two temporary buffers */
108static void * 157static void * ecb_noinline
109tmpbuf (size_t size) 158tmpbuf (size_t size)
110{ 159{
111 enum { buffers = 3 }; 160 enum { buffers = 4 };
112 static int idx; 161 static int idx;
113 static void *buf [buffers]; 162 static void *buf [buffers];
114 static size_t len [buffers]; 163 static size_t len [buffers];
115 164
116 idx = (idx + 1) % buffers; 165 idx = (idx + 1) % buffers;
123 } 172 }
124 173
125 return buf [idx]; 174 return buf [idx];
126} 175}
127 176
177static const char * ecb_noinline
178cv_get_name (CV *cv)
179{
180 static char fullname [256];
181
182 GV *gv = CvGV (cv); // gv better be nonzero
183
184 HV *stash = GvSTASH (gv);
185 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
186 const char *gvname = GvNAME (gv);
187
188 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
189 return fullname;
190}
191
128/*****************************************************************************/ 192/*****************************************************************************/
129 193
130typedef struct 194typedef struct
131{ 195{
132 IV iv; 196 IV iv;
133 const char *name; 197 const char *name;
134 #define const_iv(name) { (IV)CL_ ## name, # name }, 198 #define const_iv(name) { (IV)CL_ ## name, # name },
135} ivstr; 199} ivstr;
200
201typedef struct
202{
203 NV nv;
204 const char *name;
205 #define const_nv(name) { (NV)CL_ ## name, # name },
206} nvstr;
136 207
137static const char * 208static const char *
138iv2str (IV value, const ivstr *base, int count, const char *fallback) 209iv2str (IV value, const ivstr *base, int count, const char *fallback)
139{ 210{
140 int i; 211 int i;
189 if (res) \ 260 if (res) \
190 FAIL (name); 261 FAIL (name);
191 262
192/*****************************************************************************/ 263/*****************************************************************************/
193 264
194#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
195#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
196
197static void * 265static SV *
266new_clobj (HV *stash, IV id)
267{
268 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
269}
270
271#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
272#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
273
274/* cl objects are either \$iv, or [$iv, ...] */
275/* they can be upgraded at runtime to the array form */
276static void * ecb_noinline
198SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 277SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
199{ 278{
279 // sv_derived_from is quite slow :(
200 if (SvROK (sv) && sv_derived_from (sv, pkg)) 280 if (SvROK (sv) && sv_derived_from (sv, pkg))
201 return (void *)SvIV (SvRV (sv)); 281 return (void *)SvIV (SvRV (sv));
202 282
203 croak ("%s: %s is not of type %s", func, svname, pkg); 283 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
284}
285
286// the "no-inherit" version of the above
287static void * ecb_noinline
288SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
289{
290 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
291 return (void *)SvIV (SvRV (sv));
292
293 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
204} 294}
205 295
206/*****************************************************************************/ 296/*****************************************************************************/
207 297
208static cl_context_properties * 298static cl_context_properties * ecb_noinline
209SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 299SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
210{ 300{
211 if (!sv || !SvOK (sv)) 301 if (!sv || !SvOK (sv))
212 if (extra) 302 if (extra)
213 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 303 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
214 else 304 else
220 int i, len = av_len (av) + 1; 310 int i, len = av_len (av) + 1;
221 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 311 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
222 cl_context_properties *l = p; 312 cl_context_properties *l = p;
223 313
224 if (len & 1) 314 if (len & 1)
225 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 315 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
226 316
227 while (extracount--) 317 while (extracount--)
228 *l++ = *extra++; 318 *l++ = *extra++;
229 319
230 for (i = 0; i < len; i += 2) 320 for (i = 0; i < len; i += 2)
233 SV *p_sv = *av_fetch (av, i + 1, 0); 323 SV *p_sv = *av_fetch (av, i + 1, 0);
234 cl_context_properties v = SvIV (p_sv); // code below can override 324 cl_context_properties v = SvIV (p_sv); // code below can override
235 325
236 switch (t) 326 switch (t)
237 { 327 {
328 case CL_CONTEXT_PLATFORM:
329 if (SvROK (p_sv))
330 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
331 break;
332
238 case CL_GLX_DISPLAY_KHR: 333 case CL_GLX_DISPLAY_KHR:
239 if (!SvOK (p_sv)) 334 if (!SvOK (p_sv))
240 { 335 {
241 void *func = glsym ("_glXGetCurrentDisplay"); 336 void *func = glsym ("_glXGetCurrentDisplay");
242 if (func) 337 if (func)
265 *l = 0; 360 *l = 0;
266 361
267 return p; 362 return p;
268 } 363 }
269 364
270 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 365 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
366}
367
368// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
369// is a pointer (and only on well-behaved systems).
370static void * ecb_noinline
371object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
372{
373 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
374 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
375 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
376
377 AV *av = (AV *)SvRV (arg);
378 void **list = 0;
379 cl_uint count = av_len (av) + 1;
380
381 if (count)
382 {
383 list = tmpbuf (sizeof (*list) * count);
384 int i;
385 for (i = 0; i < count; ++i)
386 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
387 }
388
389 if (!count && !or_undef)
390 croak ("%s: '%s' must contain at least one %s object",
391 cv_get_name (cv), argname, klass);
392
393 *rcount = count;
394 return (void *)list;
271} 395}
272 396
273/*****************************************************************************/ 397/*****************************************************************************/
398/* callback stuff */
399
400/* default context callback, log to stderr */
401static void CL_CALLBACK
402context_default_notify (const char *msg, const void *info, size_t cb, void *data)
403{
404 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
405}
406
407typedef struct
408{
409 int free_cb;
410 void (*push)(void *data1, void *data2, void *data3);
411} eq_vtbl;
412
413typedef struct eq_item
414{
415 struct eq_item *next;
416 eq_vtbl *vtbl;
417 SV *cb;
418 void *data1, *data2, *data3;
419} eq_item;
420
421static void (*eq_signal_func)(void *signal_arg, int value);
422static void *eq_signal_arg;
423static xmutex_t eq_lock = X_MUTEX_INIT;
424static eq_item *eq_head, *eq_tail;
425
426static void ecb_noinline
427eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
428{
429 eq_item *item = malloc (sizeof (eq_item));
430
431 item->next = 0;
432 item->vtbl = vtbl;
433 item->cb = cb;
434 item->data1 = data1;
435 item->data2 = data2;
436 item->data3 = data3;
437
438 X_LOCK (eq_lock);
439
440 *(eq_head ? &eq_tail->next : &eq_head) = item;
441 eq_tail = item;
442
443 X_UNLOCK (eq_lock);
444
445 eq_signal_func (eq_signal_arg, 0);
446}
447
448static eq_item *
449eq_dec (void)
450{
451 eq_item *res;
452
453 X_LOCK (eq_lock);
454
455 res = eq_head;
456 if (res)
457 eq_head = res->next;
458
459 X_UNLOCK (eq_lock);
460
461 return res;
462}
463
464static void
465eq_poll (void)
466{
467 eq_item *item;
468
469 while ((item = eq_dec ()))
470 {
471 ENTER;
472 SAVETMPS;
473
474 dSP;
475 PUSHMARK (SP);
476 EXTEND (SP, 2);
477
478 if (item->vtbl->free_cb)
479 sv_2mortal (item->cb);
480
481 PUTBACK;
482 item->vtbl->push (item->data1, item->data2, item->data3);
483
484 SV *cb = item->cb;
485 free (item);
486
487 call_sv (cb, G_DISCARD | G_VOID);
488
489 FREETMPS;
490 LEAVE;
491 }
492}
493
494static void
495eq_poll_interrupt (pTHX_ void *c_arg, int value)
496{
497 eq_poll ();
498}
499
500/*****************************************************************************/
501/* context notify */
502
503static void ecb_noinline
504eq_context_push (void *data1, void *data2, void *data3)
505{
506 dSP;
507 PUSHs (sv_2mortal (newSVpv (data1, 0)));
508 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
509 PUTBACK;
510
511 free (data1);
512 free (data2);
513}
514
515static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
516
517static void CL_CALLBACK
518eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
519{
520 void *pvt_copy = malloc (cb);
521 memcpy (pvt_copy, pvt, cb);
522 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
523}
524
525#define CONTEXT_NOTIFY_CALLBACK \
526 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
527 void *user_data = 0; \
528 \
529 if (SvOK (notify)) \
530 { \
531 pfn_notify = eq_context_notify; \
532 user_data = s_get_cv (notify); \
533 }
534
535static SV * ecb_noinline
536new_clobj_context (cl_context ctx, void *user_data)
537{
538 SV *sv = new_clobj (stash_context, (IV)ctx);
539
540 if (user_data)
541 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
542
543 return sv;
544}
545
546#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
547
548/*****************************************************************************/
549/* build/compile/link notify */
550
551static void
552eq_program_push (void *data1, void *data2, void *data3)
553{
554 dSP;
555 PUSH_CLOBJ (stash_program, data1);
556 PUTBACK;
557}
558
559static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
560
561static void CL_CALLBACK
562eq_program_notify (cl_program program, void *user_data)
563{
564 clRetainProgram (program);
565
566 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
567}
568
569typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
570
571static program_callback ecb_noinline
572make_program_callback (SV *notify, void **ruser_data)
573{
574 if (SvOK (notify))
575 {
576 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
577 return eq_program_notify;
578 }
579 else
580 {
581 *ruser_data = 0;
582 return 0;
583 }
584}
585
586struct build_args
587{
588 cl_program program;
589 char *options;
590 void *user_data;
591 cl_uint num_devices;
592};
593
594X_THREAD_PROC (build_program_thread)
595{
596 struct build_args *arg = thr_arg;
597
598 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
599
600 if (arg->user_data)
601 eq_program_notify (arg->program, arg->user_data);
602 else
603 clReleaseProgram (arg->program);
604
605 free (arg->options);
606 free (arg);
607
608 return 0;
609}
610
611static void
612build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
613{
614 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
615
616 arg->program = program;
617 arg->options = strdup (options);
618 arg->user_data = user_data;
619 arg->num_devices = num_devices;
620 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
621
622 xthread_t id;
623 xthread_create (&id, build_program_thread, arg);
624}
625
626/*****************************************************************************/
627/* mem object destructor notify */
628
629static void ecb_noinline
630eq_destructor_push (void *data1, void *data2, void *data3)
631{
632}
633
634static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
635
636static void CL_CALLBACK
637eq_destructor_notify (cl_mem memobj, void *user_data)
638{
639 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
640}
641
642/*****************************************************************************/
643/* event objects */
644
645static void
646eq_event_push (void *data1, void *data2, void *data3)
647{
648 dSP;
649 PUSH_CLOBJ (stash_event, data1);
650 PUSHs (sv_2mortal (newSViv ((IV)data2)));
651 PUTBACK;
652}
653
654static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
655
656static void CL_CALLBACK
657eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
658{
659 clRetainEvent (event);
660 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
661}
662
663/*****************************************************************************/
664/* utilities for XS code */
274 665
275static size_t 666static size_t
276img_row_pitch (cl_mem img) 667img_row_pitch (cl_mem img)
277{ 668{
278 size_t res; 669 size_t res;
279 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 670 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
280 return res; 671 return res;
281} 672}
282 673
283static cl_event * 674static cl_event * ecb_noinline
284event_list (SV **items, cl_uint *rcount) 675event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
285{ 676{
286 cl_uint count = *rcount; 677 cl_uint count = *rcount;
287 678
288 if (!count) 679 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
680 *rcount = count = 0;
681
682 if (!count && !extra)
289 return 0; 683 return 0;
290 684
291 cl_event *list = tmpbuf (sizeof (cl_event) * count); 685 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
292 int i = 0; 686 int i = 0;
293 687
294 do 688 while (count--)
295 {
296 --count;
297 if (SvOK (items [count])) 689 if (SvOK (items [count]))
298 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 690 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
299 } 691
300 while (count); 692 if (extra)
693 list [i++] = extra;
301 694
302 *rcount = i; 695 *rcount = i;
303 696
304 return i ? list : 0; 697 return i ? list : 0;
305} 698}
306 699
307#define EVENT_LIST(items,count) \ 700#define EVENT_LIST(skip) \
308 cl_uint event_list_count = (count); \ 701 cl_uint event_list_count = items - (skip); \
309 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 702 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
310 703
311#define INFO(class) \ 704#define INFO(class) \
312{ \ 705{ \
313 size_t size; \ 706 size_t size; \
314 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 707 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
318 SvCUR_set (sv, size); \ 711 SvCUR_set (sv, size); \
319 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 712 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
320 XPUSHs (sv); \ 713 XPUSHs (sv); \
321} 714}
322 715
716/*****************************************************************************/
717/* mapped_xxx */
718
719static OpenCL__Mapped
720SvMAPPED (SV *sv)
721{
722 // no typechecking atm., keep your fingers crossed
723 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
724}
725
726struct mapped
727{
728 cl_command_queue queue;
729 cl_mem memobj;
730 void *ptr;
731 size_t cb;
732 cl_event event;
733 size_t row_pitch;
734 size_t slice_pitch;
735
736 size_t element_size;
737 size_t width, height, depth;
738};
739
740static SV *
741mapped_new (
742 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
743 void *ptr, size_t cb, cl_event ev,
744 size_t row_pitch, size_t slice_pitch, size_t element_size,
745 size_t width, size_t height, size_t depth
746)
747{
748 SV *data = newSV (0);
749 SvUPGRADE (data, SVt_PVMG);
750
751 OpenCL__Mapped mapped;
752 New (0, mapped, 1, struct mapped);
753
754 clRetainCommandQueue (queue);
755
756 mapped->queue = queue;
757 mapped->memobj = memobj;
758 mapped->ptr = ptr;
759 mapped->cb = cb;
760 mapped->event = ev;
761 mapped->row_pitch = row_pitch;
762 mapped->slice_pitch = slice_pitch;
763
764 mapped->element_size = element_size;
765 mapped->width = width;
766 mapped->height = height;
767 mapped->depth = depth;
768
769 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
770
771 if (SvLEN (data))
772 Safefree (data);
773
774 SvPVX (data) = (char *)ptr;
775 SvCUR_set (data, cb);
776 SvLEN_set (data, 0);
777 SvPOK_only (data);
778
779 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
780
781 if (!(flags & CL_MAP_WRITE))
782 SvREADONLY_on (data);
783
784 return obj;
785}
786
787static void
788mapped_detach (SV *sv, OpenCL__Mapped mapped)
789{
790 SV *data = SvRV (sv);
791
792 // the next check checks both before AND after detach, where SvPVX should STILL be 0
793 if (SvPVX (data) != (char *)mapped->ptr)
794 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
795 else
796 {
797 SvREADONLY_off (data);
798 SvCUR_set (data, 0);
799 SvPVX (data) = 0;
800 SvOK_off (data);
801 }
802
803 mapped->ptr = 0;
804}
805
806static void
807mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
808{
809 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
810 cl_event ev;
811
812 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
813
814 clReleaseEvent (mapped->event);
815 mapped->event = ev;
816
817 mapped_detach (self, mapped);
818}
819
820static size_t
821mapped_element_size (OpenCL__Mapped self)
822{
823 if (!self->element_size)
824 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
825
826 return self->element_size;
827}
828
829/*****************************************************************************/
830
323MODULE = OpenCL PACKAGE = OpenCL 831MODULE = OpenCL PACKAGE = OpenCL
324 832
325PROTOTYPES: ENABLE 833PROTOTYPES: ENABLE
326 834
835void
836poll ()
837 CODE:
838 eq_poll ();
839
840void
841_eq_initialise (IV func, IV arg)
842 CODE:
843 eq_signal_func = (void (*)(void *, int))func;
844 eq_signal_arg = (void*)arg;
845
327BOOT: 846BOOT:
328{ 847{
329 HV *stash = gv_stashpv ("OpenCL", 1); 848 HV *stash = gv_stashpv ("OpenCL", 1);
849
330 static const ivstr *civ, const_iv[] = { 850 static const ivstr *civ, const_iv[] = {
331 { sizeof (cl_char ), "SIZEOF_CHAR" }, 851 { sizeof (cl_char ), "SIZEOF_CHAR" },
332 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 852 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
333 { sizeof (cl_short ), "SIZEOF_SHORT" }, 853 { sizeof (cl_short ), "SIZEOF_SHORT" },
334 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 854 { sizeof (cl_ushort), "SIZEOF_USHORT" },
337 { sizeof (cl_long ), "SIZEOF_LONG" }, 857 { sizeof (cl_long ), "SIZEOF_LONG" },
338 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 858 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
339 { sizeof (cl_half ), "SIZEOF_HALF" }, 859 { sizeof (cl_half ), "SIZEOF_HALF" },
340 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 860 { sizeof (cl_float ), "SIZEOF_FLOAT" },
341 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 861 { sizeof (cl_double), "SIZEOF_DOUBLE" },
862 { PREFER_1_1 , "PREFER_1_1" },
342#include "constiv.h" 863#include "constiv.h"
343 }; 864 };
865
344 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 866 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
345 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 867 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
868
869 static const nvstr *cnv, const_nv[] = {
870#include "constnv.h"
871 };
872
873 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
874 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
875
876 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
877
878 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
879 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
880 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
881 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
882 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
883 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
884 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
885 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
886 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
887 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
888 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
889 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
890 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
891 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
892 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
893 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
894 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
895 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
896 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
897 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
898 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
899 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
900 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
901
902 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
346} 903}
347 904
348cl_int 905cl_int
349errno () 906errno ()
350 CODE: 907 CODE:
351 RETVAL = res; 908 RETVAL = res;
352 OUTPUT: 909 OUTPUT:
353 RETVAL 910 RETVAL
354 911
355const char * 912const char *
356err2str (cl_int err) 913err2str (cl_int err = res)
357 914
358const char * 915const char *
359enum2str (cl_uint value) 916enum2str (cl_uint value)
360 917
361void 918void
369 list = tmpbuf (sizeof (*list) * count); 926 list = tmpbuf (sizeof (*list) * count);
370 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 927 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
371 928
372 EXTEND (SP, count); 929 EXTEND (SP, count);
373 for (i = 0; i < count; ++i) 930 for (i = 0; i < count; ++i)
374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 931 PUSH_CLOBJ (stash_platform, list [i]);
375 932
376void 933void
377context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 934context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
378 PPCODE: 935 PPCODE:
936 CONTEXT_NOTIFY_CALLBACK;
379 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 937 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
380 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 938 XPUSH_CLOBJ_CONTEXT;
381 939
382#if 0
383
384void 940void
385context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 941context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
386 PPCODE: 942 PPCODE:
387 /* der Gipfel der Kunst */ 943 cl_uint device_count;
944 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
388 945
389#endif 946 CONTEXT_NOTIFY_CALLBACK;
947 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
948 XPUSH_CLOBJ_CONTEXT;
390 949
391void 950void
392wait_for_events (...) 951wait_for_events (...)
393 CODE: 952 CODE:
394 EVENT_LIST (0, items); 953 EVENT_LIST (0);
395 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 954 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
396 955
397PROTOTYPES: DISABLE 956PROTOTYPES: DISABLE
398 957
399MODULE = OpenCL PACKAGE = OpenCL::Platform 958MODULE = OpenCL PACKAGE = OpenCL::Platform
442 list = tmpbuf (sizeof (*list) * count); 1001 list = tmpbuf (sizeof (*list) * count);
443 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 1002 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
444 1003
445 EXTEND (SP, count); 1004 EXTEND (SP, count);
446 for (i = 0; i < count; ++i) 1005 for (i = 0; i < count; ++i)
447 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1006 PUSH_CLOBJ (stash_device, list [i]);
448 1007
449void 1008void
450context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 1009context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
451 PPCODE: 1010 PPCODE:
1011 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1012 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1013
1014 cl_uint device_count;
1015 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1016
1017 CONTEXT_NOTIFY_CALLBACK;
1018 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1019 XPUSH_CLOBJ_CONTEXT;
1020
1021void
1022context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1023 PPCODE:
1024 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1025 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1026
1027 CONTEXT_NOTIFY_CALLBACK;
1028 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1029 XPUSH_CLOBJ_CONTEXT;
1030
1031MODULE = OpenCL PACKAGE = OpenCL::Device
1032
1033void
1034info (OpenCL::Device self, cl_device_info name)
1035 PPCODE:
1036 INFO (Device)
1037
1038#if CL_VERSION_1_2
1039
1040void
1041sub_devices (OpenCL::Device self, SV *properties)
1042 PPCODE:
452 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1043 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
453 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1044 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
454 1045
455 AV *av = (AV *)SvRV (devices); 1046 properties = SvRV (properties);
456 cl_uint num_devices = av_len (av) + 1; 1047
457 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1048 cl_uint count = av_len ((AV *)properties) + 1;
1049 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1050
458 int i; 1051 int i;
1052 for (i = 0; i < count; ++i)
1053 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
459 1054
460 for (i = num_devices; i--; ) 1055 props [count] = 0;
461 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
462 1056
463 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 1057 cl_uint num_devices;
464 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1058 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1059 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1060 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
465 1061
466void 1062 EXTEND (SP, num_devices);
467context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1063 for (i = 0; i < count; ++i)
468 PPCODE: 1064 PUSH_CLOBJ (stash_subdevice, list [i]);
469 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
470 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
471 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
472 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
473 1065
474MODULE = OpenCL PACKAGE = OpenCL::Device 1066#endif
475
476void
477info (OpenCL::Device self, cl_device_info name)
478 PPCODE:
479 INFO (Device)
480 1067
481#BEGIN:device 1068#BEGIN:device
482 1069
483void 1070void
484type (OpenCL::Device self) 1071type (OpenCL::Device self)
1072 ALIAS:
1073 type = CL_DEVICE_TYPE
1074 address_bits = CL_DEVICE_ADDRESS_BITS
1075 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1076 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1077 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1078 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1079 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1080 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1081 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1082 properties = CL_DEVICE_QUEUE_PROPERTIES
1083 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1084 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
485 PPCODE: 1085 PPCODE:
486 cl_device_type value [1]; 1086 cl_ulong value [1];
487 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
488 EXTEND (SP, 1); 1088 EXTEND (SP, 1);
489 const int i = 0; 1089 const int i = 0;
490 PUSHs (sv_2mortal (newSViv (value [i]))); 1090 PUSHs (sv_2mortal (newSVuv (value [i])));
491 1091
492void 1092void
493vendor_id (OpenCL::Device self) 1093vendor_id (OpenCL::Device self)
494 ALIAS: 1094 ALIAS:
495 vendor_id = CL_DEVICE_VENDOR_ID 1095 vendor_id = CL_DEVICE_VENDOR_ID
506 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1106 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
507 image_support = CL_DEVICE_IMAGE_SUPPORT 1107 image_support = CL_DEVICE_IMAGE_SUPPORT
508 max_samplers = CL_DEVICE_MAX_SAMPLERS 1108 max_samplers = CL_DEVICE_MAX_SAMPLERS
509 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1109 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
510 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1110 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1111 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
511 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1112 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
512 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1113 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1114 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
513 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1115 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
514 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1116 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
515 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1117 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
516 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1118 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
517 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1119 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
555 EXTEND (SP, n); 1157 EXTEND (SP, n);
556 for (i = 0; i < n; ++i) 1158 for (i = 0; i < n; ++i)
557 PUSHs (sv_2mortal (newSVuv (value [i]))); 1159 PUSHs (sv_2mortal (newSVuv (value [i])));
558 1160
559void 1161void
560address_bits (OpenCL::Device self)
561 PPCODE:
562 cl_bitfield value [1];
563 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
564 EXTEND (SP, 1);
565 const int i = 0;
566 PUSHs (sv_2mortal (newSVuv (value [i])));
567
568void
569max_mem_alloc_size (OpenCL::Device self)
570 ALIAS:
571 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
572 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
573 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
574 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
575 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
576 PPCODE:
577 cl_ulong value [1];
578 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
579 EXTEND (SP, 1);
580 const int i = 0;
581 PUSHs (sv_2mortal (newSVuv (value [i])));
582
583void
584single_fp_config (OpenCL::Device self)
585 ALIAS:
586 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
587 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
588 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
589 PPCODE:
590 cl_device_fp_config value [1];
591 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
592 EXTEND (SP, 1);
593 const int i = 0;
594 PUSHs (sv_2mortal (newSVuv (value [i])));
595
596void
597global_mem_cache_type (OpenCL::Device self)
598 PPCODE:
599 cl_device_mem_cache_type value [1];
600 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
601 EXTEND (SP, 1);
602 const int i = 0;
603 PUSHs (sv_2mortal (newSVuv (value [i])));
604
605void
606local_mem_type (OpenCL::Device self)
607 PPCODE:
608 cl_device_local_mem_type value [1];
609 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
610 EXTEND (SP, 1);
611 const int i = 0;
612 PUSHs (sv_2mortal (newSVuv (value [i])));
613
614void
615error_correction_support (OpenCL::Device self) 1162error_correction_support (OpenCL::Device self)
616 ALIAS: 1163 ALIAS:
617 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1164 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
618 endian_little = CL_DEVICE_ENDIAN_LITTLE 1165 endian_little = CL_DEVICE_ENDIAN_LITTLE
619 available = CL_DEVICE_AVAILABLE 1166 available = CL_DEVICE_AVAILABLE
620 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1167 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
621 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1168 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
622 PPCODE: 1169 PPCODE:
623 cl_bool value [1]; 1170 cl_uint value [1];
624 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1171 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
625 EXTEND (SP, 1); 1172 EXTEND (SP, 1);
626 const int i = 0; 1173 const int i = 0;
627 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1174 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
628
629void
630execution_capabilities (OpenCL::Device self)
631 PPCODE:
632 cl_device_exec_capabilities value [1];
633 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
634 EXTEND (SP, 1);
635 const int i = 0;
636 PUSHs (sv_2mortal (newSVuv (value [i])));
637
638void
639properties (OpenCL::Device self)
640 PPCODE:
641 cl_command_queue_properties value [1];
642 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
643 EXTEND (SP, 1);
644 const int i = 0;
645 PUSHs (sv_2mortal (newSViv (value [i])));
646 1175
647void 1176void
648platform (OpenCL::Device self) 1177platform (OpenCL::Device self)
649 PPCODE: 1178 PPCODE:
650 cl_platform_id value [1]; 1179 cl_platform_id value [1];
651 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1180 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
652 EXTEND (SP, 1); 1181 EXTEND (SP, 1);
653 const int i = 0; 1182 const int i = 0;
654 { 1183 PUSH_CLOBJ (stash_platform, value [i]);
655 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i]));
656 }
657 1184
658void 1185void
659name (OpenCL::Device self) 1186name (OpenCL::Device self)
660 ALIAS: 1187 ALIAS:
661 name = CL_DEVICE_NAME 1188 name = CL_DEVICE_NAME
678 PPCODE: 1205 PPCODE:
679 cl_device_id value [1]; 1206 cl_device_id value [1];
680 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1207 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
681 EXTEND (SP, 1); 1208 EXTEND (SP, 1);
682 const int i = 0; 1209 const int i = 0;
683 { 1210 PUSH_CLOBJ (stash_device, value [i]);
684 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
685 }
686 1211
687void 1212void
688partition_types_ext (OpenCL::Device self) 1213partition_types_ext (OpenCL::Device self)
689 ALIAS: 1214 ALIAS:
690 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1215 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
700 for (i = 0; i < n; ++i) 1225 for (i = 0; i < n; ++i)
701 PUSHs (sv_2mortal (newSVuv (value [i]))); 1226 PUSHs (sv_2mortal (newSVuv (value [i])));
702 1227
703#END:device 1228#END:device
704 1229
1230MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1231
1232#if CL_VERSION_1_2
1233
1234void
1235DESTROY (OpenCL::SubDevice self)
1236 CODE:
1237 clReleaseDevice (self);
1238
1239#endif
1240
705MODULE = OpenCL PACKAGE = OpenCL::Context 1241MODULE = OpenCL PACKAGE = OpenCL::Context
706 1242
707void 1243void
708DESTROY (OpenCL::Context context) 1244DESTROY (OpenCL::Context self)
709 CODE: 1245 CODE:
710 clReleaseContext (context); 1246 clReleaseContext (self);
711 1247
712void 1248void
713info (OpenCL::Context self, cl_context_info name) 1249info (OpenCL::Context self, cl_context_info name)
714 PPCODE: 1250 PPCODE:
715 INFO (Context) 1251 INFO (Context)
716 1252
717void 1253void
718queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1254queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
719 PPCODE: 1255 PPCODE:
720 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1256 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
721 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1257 XPUSH_CLOBJ (stash_queue, queue);
722 1258
723void 1259void
724user_event (OpenCL::Context self) 1260user_event (OpenCL::Context self)
725 PPCODE: 1261 PPCODE:
726 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1262 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
727 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1263 XPUSH_CLOBJ (stash_userevent, ev);
728 1264
729void 1265void
730buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1266buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
731 PPCODE: 1267 PPCODE:
732 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1268 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
733 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1269 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
734 1270
735 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
736 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1272 XPUSH_CLOBJ (stash_bufferobj, mem);
737 1273
738void 1274void
739buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1275buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
740 PPCODE: 1276 PPCODE:
741 STRLEN len; 1277 STRLEN len;
742 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1278 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
743 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1279 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
744 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1280 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
745 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
746 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1282 XPUSH_CLOBJ (stash_bufferobj, mem);
747 1283
748#if CL_VERSION_1_2 1284#if CL_VERSION_1_2
749 1285
750void 1286void
751image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1287image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
752 PPCODE: 1288 PPCODE:
753 STRLEN len; 1289 STRLEN len;
754 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1290 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
755 const cl_image_format format = { channel_order, channel_type }; 1291 const cl_image_format format = { channel_order, channel_type };
756 const cl_image_desc desc = { 1292 const cl_image_desc desc = {
757 type, 1293 type,
758 width, height, depth, 1294 width, height, depth,
759 array_size, row_pitch, slice_pitch, 1295 array_size, row_pitch, slice_pitch,
760 num_mip_level, num_samples, 1296 num_mip_level, num_samples,
761 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvPTROBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1297 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
762 }; 1298 };
763 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1299 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
764 char *klass = "OpenCL::Image"; 1300 HV *stash = stash_image;
765 switch (type) 1301 switch (type)
766 { 1302 {
767 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1303 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
768 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1304 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
769 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1305 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
770 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1306 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
771 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1307 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
772 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1308 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
773 } 1309 }
774 XPUSH_NEW_OBJ (klass, mem); 1310 XPUSH_CLOBJ (stash, mem);
775 1311
776#endif 1312#endif
777 1313
778void 1314void
779image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1315image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
785 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
786#else 1322#else
787 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1323 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
788 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
789#endif 1325#endif
790 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1326 XPUSH_CLOBJ (stash_image2d, mem);
791 1327
792void 1328void
793image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1329image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
794 PPCODE: 1330 PPCODE:
795 STRLEN len; 1331 STRLEN len;
799 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
800#else 1336#else
801 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1337 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
802 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
803#endif 1339#endif
804 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1340 XPUSH_CLOBJ (stash_image3d, mem);
805 1341
806#if cl_apple_gl_sharing || cl_khr_gl_sharing 1342#if cl_apple_gl_sharing || cl_khr_gl_sharing
807 1343
808void 1344void
809gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1345gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
810 PPCODE: 1346 PPCODE:
811 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
812 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1348 XPUSH_CLOBJ (stash_bufferobj, mem);
813 1349
814void 1350void
815gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1351gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
816 PPCODE: 1352 PPCODE:
817 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1353 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
818 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1354 XPUSH_CLOBJ (stash_image2d, mem);
819 1355
820#if CL_VERSION_1_2 1356#if CL_VERSION_1_2
821 1357
822void 1358void
823gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1359gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
824 ALIAS: 1360 ALIAS:
825 PPCODE: 1361 PPCODE:
826 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1362 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
827 cl_gl_object_type type; 1363 cl_gl_object_type type;
828 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1364 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
829 char *klass = "OpenCL::Memory"; 1365 HV *stash = stash_memory;
830 switch (type) 1366 switch (type)
831 { 1367 {
832 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1368 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
833 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1369 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
834 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1370 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
835 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1371 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
836 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1372 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
837 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1373 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
838 } 1374 }
839 XPUSH_NEW_OBJ (klass, mem); 1375 XPUSH_CLOBJ (stash, mem);
840 1376
841#endif 1377#endif
842 1378
843void 1379void
844gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1380gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
846#if PREFER_1_1 1382#if PREFER_1_1
847 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
848#else 1384#else
849 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1385 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
850#endif 1386#endif
851 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1387 XPUSH_CLOBJ (stash_image2d, mem);
852 1388
853void 1389void
854gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1390gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
855 PPCODE: 1391 PPCODE:
856#if PREFER_1_1 1392#if PREFER_1_1
857 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
858#else 1394#else
859 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1395 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
860#endif 1396#endif
861 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1397 XPUSH_CLOBJ (stash_image3d, mem);
862 1398
863#endif 1399#endif
864 1400
865void 1401void
866supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1402supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
886 1422
887void 1423void
888sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1424sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
889 PPCODE: 1425 PPCODE:
890 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1426 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
891 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1427 XPUSH_CLOBJ (stash_sampler, sampler);
892 1428
893void 1429void
894program_with_source (OpenCL::Context self, SV *program) 1430program_with_source (OpenCL::Context self, SV *program)
895 PPCODE: 1431 PPCODE:
896 STRLEN len; 1432 STRLEN len;
897 size_t len2; 1433 size_t len2;
898 const char *ptr = SvPVbyte (program, len); 1434 const char *ptr = SvPVbyte (program, len);
899 1435
900 len2 = len; 1436 len2 = len;
901 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1437 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
902 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1438 XPUSH_CLOBJ (stash_program, prog);
1439
1440void
1441program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1442 PPCODE:
1443 cl_uint device_count;
1444 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1445
1446 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1447 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1448
1449 binaries = SvRV (binaries);
1450
1451 if (device_count != av_len ((AV *)binaries) + 1)
1452 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1453
1454 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1455 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1456 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1457
1458 int i;
1459 for (i = 0; i < device_count; ++i)
1460 {
1461 STRLEN len;
1462 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1463 length_list [i] = len;
1464 }
1465
1466 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1467 length_list, binary_list,
1468 GIMME_V == G_ARRAY ? status_list : 0, &res));
1469
1470 EXTEND (SP, 2);
1471 PUSH_CLOBJ (stash_program, prog);
1472
1473 if (GIMME_V == G_ARRAY)
1474 {
1475 AV *av = newAV ();
1476 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1477
1478 for (i = device_count; i--; )
1479 av_store (av, i, newSViv (status_list [i]));
1480 }
1481
1482#if CL_VERSION_1_2
1483
1484void
1485program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1486 PPCODE:
1487 cl_uint device_count;
1488 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1489
1490 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1491
1492 XPUSH_CLOBJ (stash_program, prog);
1493
1494void
1495link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1496 CODE:
1497 cl_uint device_count = 0;
1498 cl_device_id *device_list = 0;
1499
1500 if (SvOK (devices))
1501 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1502
1503 cl_uint program_count;
1504 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1505
1506 void *user_data;
1507 program_callback pfn_notify = make_program_callback (notify, &user_data);
1508
1509 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1510 program_count, program_list, pfn_notify, user_data, &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514#endif
903 1515
904#BEGIN:context 1516#BEGIN:context
905 1517
906void 1518void
907reference_count (OpenCL::Context self) 1519reference_count (OpenCL::Context self)
923 cl_device_id *value = tmpbuf (size); 1535 cl_device_id *value = tmpbuf (size);
924 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1536 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
925 int i, n = size / sizeof (*value); 1537 int i, n = size / sizeof (*value);
926 EXTEND (SP, n); 1538 EXTEND (SP, n);
927 for (i = 0; i < n; ++i) 1539 for (i = 0; i < n; ++i)
928 { 1540 PUSH_CLOBJ (stash_device, value [i]);
929 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
930 }
931 1541
932void 1542void
933properties (OpenCL::Context self) 1543properties (OpenCL::Context self)
934 PPCODE: 1544 PPCODE:
935 size_t size; 1545 size_t size;
949DESTROY (OpenCL::Queue self) 1559DESTROY (OpenCL::Queue self)
950 CODE: 1560 CODE:
951 clReleaseCommandQueue (self); 1561 clReleaseCommandQueue (self);
952 1562
953void 1563void
954enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1564read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1565 ALIAS:
1566 enqueue_read_buffer = 0
955 PPCODE: 1567 PPCODE:
956 cl_event ev = 0;
957 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
958 1569
959 SvUPGRADE (data, SVt_PV); 1570 SvUPGRADE (data, SVt_PV);
960 SvGROW (data, len); 1571 SvGROW (data, len);
961 SvPOK_only (data); 1572 SvPOK_only (data);
962 SvCUR_set (data, len); 1573 SvCUR_set (data, len);
1574
1575 cl_event ev = 0;
963 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
964 1577
965 if (ev) 1578 if (ev)
966 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
967 1580
968void 1581void
969enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1582write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1583 ALIAS:
1584 enqueue_write_buffer = 0
970 PPCODE: 1585 PPCODE:
971 cl_event ev = 0; 1586 EVENT_LIST (5);
1587
972 STRLEN len; 1588 STRLEN len;
973 char *ptr = SvPVbyte (data, len); 1589 char *ptr = SvPVbyte (data, len);
974 EVENT_LIST (5, items - 5);
975 1590
1591 cl_event ev = 0;
976 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1592 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
977 1593
978 if (ev) 1594 if (ev)
979 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1595 XPUSH_CLOBJ (stash_event, ev);
980 1596
981#if CL_VERSION_1_2 1597#if CL_VERSION_1_2
982 1598
983void 1599void
984enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1600fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1601 ALIAS:
1602 enqueue_fill_buffer = 0
985 PPCODE: 1603 PPCODE:
986 cl_event ev = 0; 1604 EVENT_LIST (5);
1605
987 STRLEN len; 1606 STRLEN len;
988 char *ptr = SvPVbyte (data, len); 1607 char *ptr = SvPVbyte (data, len);
989 EVENT_LIST (5, items - 5);
990 1608
1609 cl_event ev = 0;
991 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1610 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
992 1611
993 if (ev) 1612 if (ev)
994 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1613 XPUSH_CLOBJ (stash_event, ev);
995 1614
996void 1615void
997enqueue_fill_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, ...) 1616fill_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, ...)
1617 ALIAS:
1618 enqueue_fill_image = 0
998 PPCODE: 1619 PPCODE:
999 cl_event ev = 0; 1620 EVENT_LIST (12);
1000 STRLEN len; 1621
1001 const size_t origin [3] = { x, y, z }; 1622 const size_t origin [3] = { x, y, z };
1002 const size_t region [3] = { width, height, depth }; 1623 const size_t region [3] = { width, height, depth };
1003 EVENT_LIST (12, items - 12);
1004 1624
1005 const cl_float c_f [4] = { r, g, b, a }; 1625 const cl_float c_f [4] = { r, g, b, a };
1006 const cl_uint c_u [4] = { r, g, b, a }; 1626 const cl_uint c_u [4] = { r, g, b, a };
1007 const cl_int c_s [4] = { r, g, b, a }; 1627 const cl_int c_s [4] = { r, g, b, a };
1008 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1628 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1011 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1631 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1012 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1632 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1013 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1633 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1014 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1634 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1015 1635
1636 cl_event ev = 0;
1016 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1637 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1017 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1638 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1018 1639
1019 if (ev) 1640 if (ev)
1020 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1641 XPUSH_CLOBJ (stash_event, ev);
1021 1642
1022#endif 1643#endif
1023 1644
1024void 1645void
1025enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1646copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1647 ALIAS:
1648 enqueue_copy_buffer = 0
1026 PPCODE: 1649 PPCODE:
1650 EVENT_LIST (6);
1651
1027 cl_event ev = 0; 1652 cl_event ev = 0;
1028 EVENT_LIST (6, items - 6);
1029
1030 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1653 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1031 1654
1032 if (ev) 1655 if (ev)
1033 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1656 XPUSH_CLOBJ (stash_event, ev);
1034 1657
1035void 1658void
1036enqueue_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, ...) 1659read_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, ...)
1660 ALIAS:
1661 enqueue_read_buffer_rect = 0
1037 PPCODE: 1662 PPCODE:
1038 cl_event ev = 0; 1663 EVENT_LIST (17);
1664
1039 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1665 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1040 const size_t host_origin[3] = { host_x, host_y, host_z }; 1666 const size_t host_origin[3] = { host_x, host_y, host_z };
1041 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1042 EVENT_LIST (17, items - 17);
1043 1668
1044 if (!buf_row_pitch) 1669 if (!buf_row_pitch)
1045 buf_row_pitch = region [0]; 1670 buf_row_pitch = region [0];
1046 1671
1047 if (!buf_slice_pitch) 1672 if (!buf_slice_pitch)
1057 1682
1058 SvUPGRADE (data, SVt_PV); 1683 SvUPGRADE (data, SVt_PV);
1059 SvGROW (data, len); 1684 SvGROW (data, len);
1060 SvPOK_only (data); 1685 SvPOK_only (data);
1061 SvCUR_set (data, len); 1686 SvCUR_set (data, len);
1687
1688 cl_event ev = 0;
1062 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)); 1689 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));
1063 1690
1064 if (ev) 1691 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1692 XPUSH_CLOBJ (stash_event, ev);
1066 1693
1067void 1694void
1068enqueue_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, ...) 1695write_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, ...)
1696 ALIAS:
1697 enqueue_write_buffer_rect = 0
1069 PPCODE: 1698 PPCODE:
1070 cl_event ev = 0; 1699 EVENT_LIST (17);
1700
1071 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1701 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1072 const size_t host_origin[3] = { host_x, host_y, host_z }; 1702 const size_t host_origin[3] = { host_x, host_y, host_z };
1073 const size_t region[3] = { width, height, depth }; 1703 const size_t region[3] = { width, height, depth };
1074 STRLEN len; 1704 STRLEN len;
1075 char *ptr = SvPVbyte (data, len); 1705 char *ptr = SvPVbyte (data, len);
1076 EVENT_LIST (17, items - 17);
1077 1706
1078 if (!buf_row_pitch) 1707 if (!buf_row_pitch)
1079 buf_row_pitch = region [0]; 1708 buf_row_pitch = region [0];
1080 1709
1081 if (!buf_slice_pitch) 1710 if (!buf_slice_pitch)
1090 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1719 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1091 1720
1092 if (len < min_len) 1721 if (len < min_len)
1093 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1722 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1094 1723
1724 cl_event ev = 0;
1095 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)); 1725 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));
1096 1726
1097 if (ev) 1727 if (ev)
1098 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1728 XPUSH_CLOBJ (stash_event, ev);
1099 1729
1100void 1730void
1101enqueue_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, ...) 1731copy_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, ...)
1732 ALIAS:
1733 enqueue_copy_buffer_rect = 0
1102 PPCODE: 1734 PPCODE:
1103 cl_event ev = 0; 1735 EVENT_LIST (16);
1736
1104 const size_t src_origin[3] = { src_x, src_y, src_z }; 1737 const size_t src_origin[3] = { src_x, src_y, src_z };
1105 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1738 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1106 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
1107 EVENT_LIST (16, items - 16);
1108 1740
1741 cl_event ev = 0;
1109 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)); 1742 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));
1110 1743
1111 if (ev) 1744 if (ev)
1112 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1745 XPUSH_CLOBJ (stash_event, ev);
1113 1746
1114void 1747void
1115enqueue_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, ...) 1748read_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, ...)
1749 ALIAS:
1750 enqueue_read_image = 0
1116 PPCODE: 1751 PPCODE:
1117 cl_event ev = 0; 1752 EVENT_LIST (12);
1753
1118 const size_t src_origin[3] = { src_x, src_y, src_z }; 1754 const size_t src_origin[3] = { src_x, src_y, src_z };
1119 const size_t region[3] = { width, height, depth }; 1755 const size_t region[3] = { width, height, depth };
1120 EVENT_LIST (12, items - 12);
1121 1756
1122 if (!row_pitch) 1757 if (!row_pitch)
1123 row_pitch = img_row_pitch (src); 1758 row_pitch = img_row_pitch (src);
1124 1759
1125 if (depth > 1 && !slice_pitch) 1760 if (depth > 1 && !slice_pitch)
1129 1764
1130 SvUPGRADE (data, SVt_PV); 1765 SvUPGRADE (data, SVt_PV);
1131 SvGROW (data, len); 1766 SvGROW (data, len);
1132 SvPOK_only (data); 1767 SvPOK_only (data);
1133 SvCUR_set (data, len); 1768 SvCUR_set (data, len);
1769
1770 cl_event ev = 0;
1134 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)); 1771 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));
1135 1772
1136 if (ev) 1773 if (ev)
1137 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1774 XPUSH_CLOBJ (stash_event, ev);
1138 1775
1139void 1776void
1140enqueue_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, ...) 1777write_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, ...)
1778 ALIAS:
1779 enqueue_write_image = 0
1141 PPCODE: 1780 PPCODE:
1142 cl_event ev = 0; 1781 EVENT_LIST (12);
1782
1143 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1783 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1144 const size_t region[3] = { width, height, depth }; 1784 const size_t region[3] = { width, height, depth };
1145 STRLEN len; 1785 STRLEN len;
1146 char *ptr = SvPVbyte (data, len); 1786 char *ptr = SvPVbyte (data, len);
1147 EVENT_LIST (12, items - 12);
1148 1787
1149 if (!row_pitch) 1788 if (!row_pitch)
1150 row_pitch = img_row_pitch (dst); 1789 row_pitch = img_row_pitch (dst);
1151 1790
1152 if (depth > 1 && !slice_pitch) 1791 if (depth > 1 && !slice_pitch)
1155 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1794 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1156 1795
1157 if (len < min_len) 1796 if (len < min_len)
1158 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1797 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1159 1798
1799 cl_event ev = 0;
1160 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)); 1800 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));
1161 1801
1162 if (ev) 1802 if (ev)
1163 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1803 XPUSH_CLOBJ (stash_event, ev);
1164 1804
1165void 1805void
1166enqueue_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, ...) 1806copy_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, ...)
1807 ALIAS:
1808 enqueue_copy_image = 0
1167 PPCODE: 1809 PPCODE:
1168 cl_event ev = 0; 1810 EVENT_LIST (12);
1811
1169 const size_t src_origin[3] = { src_x, src_y, src_z }; 1812 const size_t src_origin[3] = { src_x, src_y, src_z };
1170 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1813 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1171 const size_t region[3] = { width, height, depth }; 1814 const size_t region[3] = { width, height, depth };
1172 EVENT_LIST (12, items - 12);
1173 1815
1816 cl_event ev = 0;
1174 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1817 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1175 1818
1176 if (ev) 1819 if (ev)
1177 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1820 XPUSH_CLOBJ (stash_event, ev);
1178 1821
1179void 1822void
1180enqueue_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, ...) 1823copy_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, ...)
1824 ALIAS:
1825 enqueue_copy_image_to_buffer = 0
1181 PPCODE: 1826 PPCODE:
1827 EVENT_LIST (10);
1828
1829 const size_t src_origin[3] = { src_x, src_y, src_z };
1830 const size_t region [3] = { width, height, depth };
1831
1182 cl_event ev = 0; 1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839copy_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, ...)
1840 ALIAS:
1841 enqueue_copy_buffer_to_image = 0
1842 PPCODE:
1843 EVENT_LIST (10);
1844
1183 const size_t src_origin[3] = { src_x, src_y, src_z }; 1845 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1846 const size_t region [3] = { width, height, depth };
1847
1848 cl_event ev = 0;
1849 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1850
1851 if (ev)
1852 XPUSH_CLOBJ (stash_event, ev);
1853
1854void
1855map_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, ...)
1856 ALIAS:
1857 enqueue_map_buffer = 0
1858 PPCODE:
1859 EVENT_LIST (6);
1860
1861 size_t cb = SvIV (cb_);
1862
1863 if (!SvOK (cb_))
1864 {
1865 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1866 cb -= offset;
1867 }
1868
1869 cl_event ev;
1870 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1871 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1872
1873void
1874map_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, ...)
1875 ALIAS:
1876 enqueue_map_image = 0
1877 PPCODE:
1878 size_t width = SvIV (width_);
1879 if (!SvOK (width_))
1880 {
1881 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1882 width -= x;
1883 }
1884
1885 size_t height = SvIV (width_);
1886 if (!SvOK (height_))
1887 {
1888 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1889 height -= y;
1890
1891 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1892 if (!height)
1893 height = 1;
1894 }
1895
1896 size_t depth = SvIV (width_);
1897 if (!SvOK (depth_))
1898 {
1899 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1900 depth -= z;
1901
1902 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1903 if (!depth)
1904 depth = 1;
1905 }
1906
1907 const size_t origin[3] = { x, y, z };
1184 const size_t region[3] = { width, height, depth }; 1908 const size_t region[3] = { width, height, depth };
1909 size_t row_pitch, slice_pitch;
1185 EVENT_LIST (10, items - 10); 1910 EVENT_LIST (10);
1186 1911
1912 cl_event ev;
1913 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1914
1915 size_t cb = slice_pitch ? slice_pitch * region [2]
1916 : row_pitch ? row_pitch * region [1]
1917 : region [0];
1918
1919 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1920
1921void
1922unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1923 PPCODE:
1924 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1925 if (GIMME_V != G_VOID)
1926 {
1927 clRetainEvent (mapped->event);
1928 XPUSH_CLOBJ (stash_event, mapped->event);
1929 }
1930
1931void
1932task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1933 ALIAS:
1934 enqueue_task = 0
1935 PPCODE:
1936 EVENT_LIST (2);
1937
1938 cl_event ev = 0;
1187 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1939 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1188 1940
1189 if (ev) 1941 if (ev)
1190 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1942 XPUSH_CLOBJ (stash_event, ev);
1191 1943
1192void 1944void
1193enqueue_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, ...)
1194 PPCODE:
1195 cl_event ev = 0;
1196 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1197 const size_t region[3] = { width, height, depth };
1198 EVENT_LIST (10, items - 10);
1199
1200 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1201
1202 if (ev)
1203 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1204
1205void
1206enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1207 PPCODE:
1208 cl_event ev = 0;
1209 EVENT_LIST (2, items - 2);
1210
1211 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1212
1213 if (ev)
1214 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1215
1216void
1217enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1945nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1946 ALIAS:
1947 enqueue_nd_range_kernel = 0
1218 PPCODE: 1948 PPCODE:
1219 cl_event ev = 0; 1949 EVENT_LIST (5);
1950
1220 size_t *gwo = 0, *gws, *lws = 0; 1951 size_t *gwo = 0, *gws, *lws = 0;
1221 int gws_len; 1952 int gws_len;
1222 size_t *lists; 1953 size_t *lists;
1223 int i; 1954 int i;
1224 EVENT_LIST (5, items - 5);
1225 1955
1226 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1956 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1227 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1957 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1228 1958
1229 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1959 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1230 1960
1231 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1961 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1232 1962
1233 gws = lists + gws_len * 0; 1963 gws = lists + gws_len * 0;
1234 for (i = 0; i < gws_len; ++i) 1964 for (i = 0; i < gws_len; ++i)
1965 {
1235 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1966 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1967 // at least nvidia crashes for 0-sized work group sizes, work around
1968 if (!gws [i])
1969 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1970 }
1236 1971
1237 if (SvOK (global_work_offset)) 1972 if (SvOK (global_work_offset))
1238 { 1973 {
1239 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1974 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1240 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1248 } 1983 }
1249 1984
1250 if (SvOK (local_work_size)) 1985 if (SvOK (local_work_size))
1251 { 1986 {
1252 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1987 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1253 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1988 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1254 1989
1255 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1990 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1256 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1991 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1257 1992
1258 lws = lists + gws_len * 2; 1993 lws = lists + gws_len * 2;
1259 for (i = 0; i < gws_len; ++i) 1994 for (i = 0; i < gws_len; ++i)
1995 {
1260 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1996 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1997 // at least nvidia crashes for 0-sized work group sizes, work around
1998 if (!lws [i])
1999 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2000 }
1261 } 2001 }
1262 2002
2003 cl_event ev = 0;
1263 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2004 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1264 2005
1265 if (ev) 2006 if (ev)
1266 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2007 XPUSH_CLOBJ (stash_event, ev);
2008
2009#if CL_VERSION_1_2
2010
2011void
2012migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2013 ALIAS:
2014 enqueue_migrate_mem_objects = 0
2015 PPCODE:
2016 EVENT_LIST (3);
2017
2018 cl_uint obj_count;
2019 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2020
2021 cl_event ev = 0;
2022 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2023
2024 if (ev)
2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#endif
1267 2028
1268#if cl_apple_gl_sharing || cl_khr_gl_sharing 2029#if cl_apple_gl_sharing || cl_khr_gl_sharing
1269 2030
1270void 2031void
1271enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2032acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1272 ALIAS: 2033 ALIAS:
2034 release_gl_objects = 1
2035 enqueue_acquire_gl_objects = 0
1273 enqueue_release_gl_objects = 1 2036 enqueue_release_gl_objects = 1
1274 PPCODE: 2037 PPCODE:
1275 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2038 EVENT_LIST (2);
1276 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2039
2040 cl_uint obj_count;
2041 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1277 2042
1278 cl_event ev = 0; 2043 cl_event ev = 0;
1279 EVENT_LIST (2, items - 2);
1280 AV *av = (AV *)SvRV (objects);
1281 cl_uint num_objects = av_len (av) + 1;
1282 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1283 int i;
1284
1285 for (i = num_objects; i--; )
1286 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1287 2044
1288 if (ix) 2045 if (ix)
1289 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1290 else 2047 else
1291 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2048 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1292 2049
1293 if (ev) 2050 if (ev)
1294 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2051 XPUSH_CLOBJ (stash_event, ev);
1295 2052
1296#endif 2053#endif
1297 2054
1298void 2055void
1299enqueue_wait_for_events (OpenCL::Queue self, ...) 2056wait_for_events (OpenCL::Queue self, ...)
2057 ALIAS:
2058 enqueue_wait_for_events = 0
1300 CODE: 2059 CODE:
1301 EVENT_LIST (1, items - 1); 2060 EVENT_LIST (1);
1302#if PREFER_1_1 2061#if PREFER_1_1
1303 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1304#else 2063#else
1305 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2064 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1306#endif 2065#endif
1307 2066
1308void 2067void
1309enqueue_marker (OpenCL::Queue self, ...) 2068marker (OpenCL::Queue self, ...)
2069 ALIAS:
2070 enqueue_marker = 0
1310 PPCODE: 2071 PPCODE:
2072 EVENT_LIST (1);
1311 cl_event ev = 0; 2073 cl_event ev = 0;
1312 EVENT_LIST (1, items - 1);
1313#if PREFER_1_1 2074#if PREFER_1_1
1314 if (!event_list_count) 2075 if (!event_list_count)
1315 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2076 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1316 else 2077 else
1317#if CL_VERSION_1_2 2078#if CL_VERSION_1_2
1324#endif 2085#endif
1325#else 2086#else
1326 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2087 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1327#endif 2088#endif
1328 if (ev) 2089 if (ev)
1329 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2090 XPUSH_CLOBJ (stash_event, ev);
1330 2091
1331void 2092void
1332enqueue_barrier (OpenCL::Queue self, ...) 2093barrier (OpenCL::Queue self, ...)
2094 ALIAS:
2095 enqueue_barrier = 0
1333 PPCODE: 2096 PPCODE:
2097 EVENT_LIST (1);
1334 cl_event ev = 0; 2098 cl_event ev = 0;
1335 EVENT_LIST (1, items - 1);
1336#if PREFER_1_1 2099#if PREFER_1_1
1337 if (!event_list_count && GIMME_V == G_VOID) 2100 if (!event_list_count && GIMME_V == G_VOID)
1338 NEED_SUCCESS (EnqueueBarrier, (self)); 2101 NEED_SUCCESS (EnqueueBarrier, (self));
1339 else 2102 else
1340#if CL_VERSION_1_2 2103#if CL_VERSION_1_2
1350#endif 2113#endif
1351#else 2114#else
1352 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2115 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1353#endif 2116#endif
1354 if (ev) 2117 if (ev)
1355 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2118 XPUSH_CLOBJ (stash_event, ev);
1356 2119
1357void 2120void
1358flush (OpenCL::Queue self) 2121flush (OpenCL::Queue self)
1359 CODE: 2122 CODE:
1360 NEED_SUCCESS (Flush, (self)); 2123 NEED_SUCCESS (Flush, (self));
1376 PPCODE: 2139 PPCODE:
1377 cl_context value [1]; 2140 cl_context value [1];
1378 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1379 EXTEND (SP, 1); 2142 EXTEND (SP, 1);
1380 const int i = 0; 2143 const int i = 0;
1381 {
1382 NEED_SUCCESS (RetainContext, (value [i])); 2144 NEED_SUCCESS (RetainContext, (value [i]));
1383 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2145 PUSH_CLOBJ (stash_context, value [i]);
1384 }
1385 2146
1386void 2147void
1387device (OpenCL::Queue self) 2148device (OpenCL::Queue self)
1388 PPCODE: 2149 PPCODE:
1389 cl_device_id value [1]; 2150 cl_device_id value [1];
1390 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2151 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1391 EXTEND (SP, 1); 2152 EXTEND (SP, 1);
1392 const int i = 0; 2153 const int i = 0;
1393 { 2154 PUSH_CLOBJ (stash_device, value [i]);
1394 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1395 }
1396 2155
1397void 2156void
1398reference_count (OpenCL::Queue self) 2157reference_count (OpenCL::Queue self)
1399 PPCODE: 2158 PPCODE:
1400 cl_uint value [1]; 2159 cl_uint value [1];
1404 PUSHs (sv_2mortal (newSVuv (value [i]))); 2163 PUSHs (sv_2mortal (newSVuv (value [i])));
1405 2164
1406void 2165void
1407properties (OpenCL::Queue self) 2166properties (OpenCL::Queue self)
1408 PPCODE: 2167 PPCODE:
1409 cl_command_queue_properties value [1]; 2168 cl_ulong value [1];
1410 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2169 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1411 EXTEND (SP, 1); 2170 EXTEND (SP, 1);
1412 const int i = 0; 2171 const int i = 0;
1413 PUSHs (sv_2mortal (newSViv (value [i]))); 2172 PUSHs (sv_2mortal (newSVuv (value [i])));
1414 2173
1415#END:command_queue 2174#END:command_queue
1416 2175
1417MODULE = OpenCL PACKAGE = OpenCL::Memory 2176MODULE = OpenCL PACKAGE = OpenCL::Memory
1418 2177
1424void 2183void
1425info (OpenCL::Memory self, cl_mem_info name) 2184info (OpenCL::Memory self, cl_mem_info name)
1426 PPCODE: 2185 PPCODE:
1427 INFO (MemObject) 2186 INFO (MemObject)
1428 2187
2188void
2189destructor_callback (OpenCL::Memory self, SV *notify)
2190 PPCODE:
2191 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2192
1429#BEGIN:mem 2193#BEGIN:mem
1430 2194
1431void 2195void
1432type (OpenCL::Memory self) 2196type (OpenCL::Memory self)
2197 ALIAS:
2198 type = CL_MEM_TYPE
2199 map_count = CL_MEM_MAP_COUNT
2200 reference_count = CL_MEM_REFERENCE_COUNT
1433 PPCODE: 2201 PPCODE:
1434 cl_mem_object_type value [1]; 2202 cl_uint value [1];
1435 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2203 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1436 EXTEND (SP, 1); 2204 EXTEND (SP, 1);
1437 const int i = 0; 2205 const int i = 0;
1438 PUSHs (sv_2mortal (newSViv (value [i]))); 2206 PUSHs (sv_2mortal (newSVuv (value [i])));
1439 2207
1440void 2208void
1441flags (OpenCL::Memory self) 2209flags (OpenCL::Memory self)
1442 PPCODE: 2210 PPCODE:
1443 cl_mem_flags value [1]; 2211 cl_ulong value [1];
1444 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1445 EXTEND (SP, 1); 2213 EXTEND (SP, 1);
1446 const int i = 0; 2214 const int i = 0;
1447 PUSHs (sv_2mortal (newSViv (value [i]))); 2215 PUSHs (sv_2mortal (newSVuv (value [i])));
1448 2216
1449void 2217void
1450size (OpenCL::Memory self) 2218size (OpenCL::Memory self)
1451 ALIAS: 2219 ALIAS:
1452 size = CL_MEM_SIZE 2220 size = CL_MEM_SIZE
1466 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1467 const int i = 0; 2235 const int i = 0;
1468 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2236 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1469 2237
1470void 2238void
1471map_count (OpenCL::Memory self)
1472 ALIAS:
1473 map_count = CL_MEM_MAP_COUNT
1474 reference_count = CL_MEM_REFERENCE_COUNT
1475 PPCODE:
1476 cl_uint value [1];
1477 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1478 EXTEND (SP, 1);
1479 const int i = 0;
1480 PUSHs (sv_2mortal (newSVuv (value [i])));
1481
1482void
1483context (OpenCL::Memory self) 2239context (OpenCL::Memory self)
1484 PPCODE: 2240 PPCODE:
1485 cl_context value [1]; 2241 cl_context value [1];
1486 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2242 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1487 EXTEND (SP, 1); 2243 EXTEND (SP, 1);
1488 const int i = 0; 2244 const int i = 0;
1489 {
1490 NEED_SUCCESS (RetainContext, (value [i])); 2245 NEED_SUCCESS (RetainContext, (value [i]));
1491 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2246 PUSH_CLOBJ (stash_context, value [i]);
1492 }
1493 2247
1494void 2248void
1495associated_memobject (OpenCL::Memory self) 2249associated_memobject (OpenCL::Memory self)
1496 PPCODE: 2250 PPCODE:
1497 cl_mem value [1]; 2251 cl_mem value [1];
1498 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1499 EXTEND (SP, 1); 2253 EXTEND (SP, 1);
1500 const int i = 0; 2254 const int i = 0;
1501 {
1502 NEED_SUCCESS (RetainMemObject, (value [i])); 2255 NEED_SUCCESS (RetainMemObject, (value [i]));
1503 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2256 PUSH_CLOBJ (stash_memory, value [i]);
1504 }
1505 2257
1506#END:mem 2258#END:mem
1507 2259
1508#if cl_apple_gl_sharing || cl_khr_gl_sharing 2260#if cl_apple_gl_sharing || cl_khr_gl_sharing
1509 2261
1528 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2280 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1529 2281
1530 cl_buffer_region crdata = { origin, size }; 2282 cl_buffer_region crdata = { origin, size };
1531 2283
1532 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2284 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1533 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2285 XPUSH_CLOBJ (stash_buffer, mem);
1534 2286
1535MODULE = OpenCL PACKAGE = OpenCL::Image 2287MODULE = OpenCL PACKAGE = OpenCL::Image
1536 2288
1537void 2289void
1538image_info (OpenCL::Image self, cl_image_info name) 2290image_info (OpenCL::Image self, cl_image_info name)
1608 2360
1609#BEGIN:sampler 2361#BEGIN:sampler
1610 2362
1611void 2363void
1612reference_count (OpenCL::Sampler self) 2364reference_count (OpenCL::Sampler self)
2365 ALIAS:
2366 reference_count = CL_SAMPLER_REFERENCE_COUNT
2367 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2368 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1613 PPCODE: 2369 PPCODE:
1614 cl_uint value [1]; 2370 cl_uint value [1];
1615 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1616 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
1617 const int i = 0; 2373 const int i = 0;
1618 PUSHs (sv_2mortal (newSVuv (value [i]))); 2374 PUSHs (sv_2mortal (newSVuv (value [i])));
1619 2375
1620void 2376void
1622 PPCODE: 2378 PPCODE:
1623 cl_context value [1]; 2379 cl_context value [1];
1624 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2380 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1625 EXTEND (SP, 1); 2381 EXTEND (SP, 1);
1626 const int i = 0; 2382 const int i = 0;
1627 {
1628 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
1629 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2384 PUSH_CLOBJ (stash_context, value [i]);
1630 }
1631
1632void
1633normalized_coords (OpenCL::Sampler self)
1634 PPCODE:
1635 cl_addressing_mode value [1];
1636 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1637 EXTEND (SP, 1);
1638 const int i = 0;
1639 PUSHs (sv_2mortal (newSViv (value [i])));
1640
1641void
1642addressing_mode (OpenCL::Sampler self)
1643 PPCODE:
1644 cl_filter_mode value [1];
1645 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1646 EXTEND (SP, 1);
1647 const int i = 0;
1648 PUSHs (sv_2mortal (newSViv (value [i])));
1649 2385
1650void 2386void
1651filter_mode (OpenCL::Sampler self) 2387filter_mode (OpenCL::Sampler self)
1652 PPCODE: 2388 PPCODE:
1653 cl_bool value [1]; 2389 cl_uint value [1];
1654 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1655 EXTEND (SP, 1); 2391 EXTEND (SP, 1);
1656 const int i = 0; 2392 const int i = 0;
1657 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2393 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1658 2394
1664DESTROY (OpenCL::Program self) 2400DESTROY (OpenCL::Program self)
1665 CODE: 2401 CODE:
1666 clReleaseProgram (self); 2402 clReleaseProgram (self);
1667 2403
1668void 2404void
1669build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2405build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2406 ALIAS:
2407 build_async = 1
1670 CODE: 2408 CODE:
2409 cl_uint device_count = 0;
2410 cl_device_id *device_list = 0;
2411
2412 if (SvOK (devices))
2413 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2414
2415 void *user_data;
2416 program_callback pfn_notify = make_program_callback (notify, &user_data);
2417
2418 if (ix)
2419 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2420 else
2421 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2422
2423#if CL_VERSION_1_2
2424
2425void
2426compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2427 CODE:
2428 cl_uint device_count = 0;
2429 cl_device_id *device_list = 0;
2430
2431 if (SvOK (devices))
2432 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2433
2434 cl_uint header_count = 0;
2435 cl_program *header_list = 0;
2436 const char **header_name = 0;
2437
2438 if (SvOK (headers))
2439 {
2440 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2441 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2442
2443 HV *hv = (HV *)SvRV (devices);
2444
2445 header_count = hv_iterinit (hv);
2446 header_list = tmpbuf (sizeof (*header_list) * header_count);
2447 header_name = tmpbuf (sizeof (*header_name) * header_count);
2448
2449 HE *he;
2450 int i = 0;
2451 while (he = hv_iternext (hv))
2452 {
2453 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2454 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2455 ++i;
2456 }
2457 }
2458
2459 void *user_data;
2460 program_callback pfn_notify = make_program_callback (notify, &user_data);
2461
1671 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2462 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2463 header_count, header_list, header_name, pfn_notify, user_data));
2464
2465#endif
1672 2466
1673void 2467void
1674build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2468build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1675 PPCODE: 2469 PPCODE:
1676 size_t size; 2470 size_t size;
1685#BEGIN:program_build 2479#BEGIN:program_build
1686 2480
1687void 2481void
1688build_status (OpenCL::Program self, OpenCL::Device device) 2482build_status (OpenCL::Program self, OpenCL::Device device)
1689 PPCODE: 2483 PPCODE:
1690 cl_build_status value [1]; 2484 cl_int value [1];
1691 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2485 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1692 EXTEND (SP, 1); 2486 EXTEND (SP, 1);
1693 const int i = 0; 2487 const int i = 0;
1694 PUSHs (sv_2mortal (newSViv (value [i]))); 2488 PUSHs (sv_2mortal (newSViv (value [i])));
1695 2489
1705 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2499 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1706 EXTEND (SP, 1); 2500 EXTEND (SP, 1);
1707 const int i = 0; 2501 const int i = 0;
1708 PUSHs (sv_2mortal (newSVpv (value, 0))); 2502 PUSHs (sv_2mortal (newSVpv (value, 0)));
1709 2503
2504void
2505binary_type (OpenCL::Program self, OpenCL::Device device)
2506 PPCODE:
2507 cl_uint value [1];
2508 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2509 EXTEND (SP, 1);
2510 const int i = 0;
2511 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2512
1710#END:program_build 2513#END:program_build
1711 2514
1712void 2515void
1713kernel (OpenCL::Program program, SV *function) 2516kernel (OpenCL::Program program, SV *function)
1714 PPCODE: 2517 PPCODE:
1715 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2518 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1716 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2519 XPUSH_CLOBJ (stash_kernel, kernel);
1717 2520
1718void 2521void
1719kernels_in_program (OpenCL::Program program) 2522kernels_in_program (OpenCL::Program program)
1720 PPCODE: 2523 PPCODE:
1721 cl_uint num_kernels; 2524 cl_uint num_kernels;
1724 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2527 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
1725 2528
1726 int i; 2529 int i;
1727 EXTEND (SP, num_kernels); 2530 EXTEND (SP, num_kernels);
1728 for (i = 0; i < num_kernels; ++i) 2531 for (i = 0; i < num_kernels; ++i)
1729 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Kernel", kernels [i])); 2532 PUSH_CLOBJ (stash_kernel, kernels [i]);
1730 2533
1731void 2534void
1732info (OpenCL::Program self, cl_program_info name) 2535info (OpenCL::Program self, cl_program_info name)
1733 PPCODE: 2536 PPCODE:
1734 INFO (Program) 2537 INFO (Program)
1780 PPCODE: 2583 PPCODE:
1781 cl_context value [1]; 2584 cl_context value [1];
1782 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2585 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1783 EXTEND (SP, 1); 2586 EXTEND (SP, 1);
1784 const int i = 0; 2587 const int i = 0;
1785 {
1786 NEED_SUCCESS (RetainContext, (value [i])); 2588 NEED_SUCCESS (RetainContext, (value [i]));
1787 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2589 PUSH_CLOBJ (stash_context, value [i]);
1788 }
1789 2590
1790void 2591void
1791devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
1792 PPCODE: 2593 PPCODE:
1793 size_t size; 2594 size_t size;
1795 cl_device_id *value = tmpbuf (size); 2596 cl_device_id *value = tmpbuf (size);
1796 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1797 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
1798 EXTEND (SP, n); 2599 EXTEND (SP, n);
1799 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
1800 { 2601 PUSH_CLOBJ (stash_device, value [i]);
1801 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1802 }
1803 2602
1804void 2603void
1805source (OpenCL::Program self) 2604source (OpenCL::Program self)
1806 PPCODE: 2605 PPCODE:
1807 size_t size; 2606 size_t size;
1832DESTROY (OpenCL::Kernel self) 2631DESTROY (OpenCL::Kernel self)
1833 CODE: 2632 CODE:
1834 clReleaseKernel (self); 2633 clReleaseKernel (self);
1835 2634
1836void 2635void
2636setf (OpenCL::Kernel self, const char *format, ...)
2637 CODE:
2638 int i;
2639 for (i = 2; ; ++i)
2640 {
2641 while (*format == ' ')
2642 ++format;
2643
2644 char type = *format++;
2645
2646 if (!type)
2647 break;
2648
2649 if (i >= items)
2650 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2651
2652 SV *sv = ST (i);
2653
2654 union
2655 {
2656 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2657 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2658 cl_half ch; cl_float cf; cl_double cd;
2659 cl_mem cm;
2660 cl_sampler ca;
2661 size_t cz;
2662 cl_event ce;
2663 } arg;
2664 size_t size;
2665 int nullarg = 0;
2666
2667 switch (type)
2668 {
2669 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2670 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2671 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2672 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2673 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2674 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2675 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2676 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2677
2678 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2679 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2680 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2681
2682 case 'z': nullarg = 1; size = SvIV (sv); break;
2683
2684 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2685 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2686 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2687
2688 default:
2689 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2690 }
2691
2692 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2693 if (res)
2694 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2695 }
2696
2697 if (i != items)
2698 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2699
2700void
1837set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2701set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1838 CODE: 2702 CODE:
1839 clSetKernelArg (self, idx, sizeof (value), &value); 2703 clSetKernelArg (self, idx, sizeof (value), &value);
1840 2704
1841void 2705void
1889 clSetKernelArg (self, idx, sizeof (value), &value); 2753 clSetKernelArg (self, idx, sizeof (value), &value);
1890 2754
1891void 2755void
1892set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2756set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1893 CODE: 2757 CODE:
1894 clSetKernelArg (self, idx, sizeof (value), &value); 2758 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1895 2759
1896void 2760void
1897set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2761set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1898 CODE: 2762 CODE:
1899 clSetKernelArg (self, idx, sizeof (value), &value); 2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1900 2764
1901void 2765void
1902set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2766set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
1903 CODE: 2767 CODE:
1904 clSetKernelArg (self, idx, sizeof (value), &value); 2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1905
1906void
1907set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1908 CODE:
1909 clSetKernelArg (self, idx, sizeof (value), &value);
1910 2769
1911void 2770void
1912set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2771set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1913 CODE: 2772 CODE:
1914 clSetKernelArg (self, idx, sizeof (value), &value); 2773 clSetKernelArg (self, idx, sizeof (value), &value);
1958 PPCODE: 2817 PPCODE:
1959 cl_context value [1]; 2818 cl_context value [1];
1960 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2819 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1961 EXTEND (SP, 1); 2820 EXTEND (SP, 1);
1962 const int i = 0; 2821 const int i = 0;
1963 {
1964 NEED_SUCCESS (RetainContext, (value [i])); 2822 NEED_SUCCESS (RetainContext, (value [i]));
1965 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2823 PUSH_CLOBJ (stash_context, value [i]);
1966 }
1967 2824
1968void 2825void
1969program (OpenCL::Kernel self) 2826program (OpenCL::Kernel self)
1970 PPCODE: 2827 PPCODE:
1971 cl_program value [1]; 2828 cl_program value [1];
1972 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2829 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1973 EXTEND (SP, 1); 2830 EXTEND (SP, 1);
1974 const int i = 0; 2831 const int i = 0;
1975 {
1976 NEED_SUCCESS (RetainProgram, (value [i])); 2832 NEED_SUCCESS (RetainProgram, (value [i]));
1977 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2833 PUSH_CLOBJ (stash_program, value [i]);
1978 }
1979 2834
1980#END:kernel 2835#END:kernel
1981 2836
1982void 2837void
1983work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2838work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2029 const int i = 0; 2884 const int i = 0;
2030 PUSHs (sv_2mortal (newSVuv (value [i]))); 2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2031 2886
2032#END:kernel_work_group 2887#END:kernel_work_group
2033 2888
2889#if CL_VERSION_1_2
2890
2891void
2892arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2893 PPCODE:
2894 size_t size;
2895 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2896 SV *sv = sv_2mortal (newSV (size));
2897 SvUPGRADE (sv, SVt_PV);
2898 SvPOK_only (sv);
2899 SvCUR_set (sv, size);
2900 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2901 XPUSHs (sv);
2902
2903#BEGIN:kernel_arg
2904
2905void
2906arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2907 ALIAS:
2908 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2909 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2910 PPCODE:
2911 cl_uint value [1];
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2913 EXTEND (SP, 1);
2914 const int i = 0;
2915 PUSHs (sv_2mortal (newSVuv (value [i])));
2916
2917void
2918arg_type_name (OpenCL::Kernel self, cl_uint idx)
2919 ALIAS:
2920 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2921 arg_name = CL_KERNEL_ARG_NAME
2922 PPCODE:
2923 size_t size;
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2925 char *value = tmpbuf (size);
2926 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2927 EXTEND (SP, 1);
2928 const int i = 0;
2929 PUSHs (sv_2mortal (newSVpv (value, 0)));
2930
2931void
2932arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2933 PPCODE:
2934 cl_ulong value [1];
2935 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2936 EXTEND (SP, 1);
2937 const int i = 0;
2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2939
2940#END:kernel_arg
2941
2942#endif
2943
2034MODULE = OpenCL PACKAGE = OpenCL::Event 2944MODULE = OpenCL PACKAGE = OpenCL::Event
2035 2945
2036void 2946void
2037DESTROY (OpenCL::Event self) 2947DESTROY (OpenCL::Event self)
2038 CODE: 2948 CODE:
2040 2950
2041void 2951void
2042wait (OpenCL::Event self) 2952wait (OpenCL::Event self)
2043 CODE: 2953 CODE:
2044 clWaitForEvents (1, &self); 2954 clWaitForEvents (1, &self);
2955
2956void
2957cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2958 CODE:
2959 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2045 2960
2046void 2961void
2047info (OpenCL::Event self, cl_event_info name) 2962info (OpenCL::Event self, cl_event_info name)
2048 PPCODE: 2963 PPCODE:
2049 INFO (Event) 2964 INFO (Event)
2055 PPCODE: 2970 PPCODE:
2056 cl_command_queue value [1]; 2971 cl_command_queue value [1];
2057 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2972 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2058 EXTEND (SP, 1); 2973 EXTEND (SP, 1);
2059 const int i = 0; 2974 const int i = 0;
2060 {
2061 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2975 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2062 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2976 PUSH_CLOBJ (stash_queue, value [i]);
2063 }
2064 2977
2065void 2978void
2066command_type (OpenCL::Event self) 2979command_type (OpenCL::Event self)
2067 PPCODE:
2068 cl_command_type value [1];
2069 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2070 EXTEND (SP, 1);
2071 const int i = 0;
2072 PUSHs (sv_2mortal (newSVuv (value [i])));
2073
2074void
2075reference_count (OpenCL::Event self)
2076 ALIAS: 2980 ALIAS:
2981 command_type = CL_EVENT_COMMAND_TYPE
2077 reference_count = CL_EVENT_REFERENCE_COUNT 2982 reference_count = CL_EVENT_REFERENCE_COUNT
2078 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2983 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2079 PPCODE: 2984 PPCODE:
2080 cl_uint value [1]; 2985 cl_uint value [1];
2081 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2986 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2088 PPCODE: 2993 PPCODE:
2089 cl_context value [1]; 2994 cl_context value [1];
2090 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2995 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2091 EXTEND (SP, 1); 2996 EXTEND (SP, 1);
2092 const int i = 0; 2997 const int i = 0;
2093 {
2094 NEED_SUCCESS (RetainContext, (value [i])); 2998 NEED_SUCCESS (RetainContext, (value [i]));
2095 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2999 PUSH_CLOBJ (stash_context, value [i]);
2096 }
2097 3000
2098#END:event 3001#END:event
2099 3002
2100void 3003void
2101profiling_info (OpenCL::Event self, cl_profiling_info name) 3004profiling_info (OpenCL::Event self, cl_profiling_info name)
2125void 3028void
2126set_status (OpenCL::UserEvent self, cl_int execution_status) 3029set_status (OpenCL::UserEvent self, cl_int execution_status)
2127 CODE: 3030 CODE:
2128 clSetUserEventStatus (self, execution_status); 3031 clSetUserEventStatus (self, execution_status);
2129 3032
3033MODULE = OpenCL PACKAGE = OpenCL::Mapped
3034
3035void
3036DESTROY (SV *self)
3037 CODE:
3038 OpenCL__Mapped mapped = SvMAPPED (self);
3039
3040 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3041 mapped_detach (self, mapped);
3042
3043 clReleaseCommandQueue (mapped->queue);
3044 clReleaseEvent (mapped->event);
3045 Safefree (mapped);
3046
3047void
3048unmap (OpenCL::Mapped self, ...)
3049 CODE:
3050 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3051
3052bool
3053mapped (OpenCL::Mapped self)
3054 CODE:
3055 RETVAL = !!self->ptr;
3056 OUTPUT:
3057 RETVAL
3058
3059void
3060wait (OpenCL::Mapped self)
3061 PPCODE:
3062 if (self->event)
3063 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3064
3065void
3066event (OpenCL::Mapped self)
3067 PPCODE:
3068 if (!self->event)
3069 XSRETURN_UNDEF;
3070
3071 clRetainEvent (self->event);
3072 XPUSH_CLOBJ (stash_event, self->event);
3073
3074#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3075#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3076#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3077#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3078#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3079#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3080
3081IV
3082size (OpenCL::Mapped self)
3083 ALIAS:
3084 size = MAPPED_OFFSET_CB
3085 row_pitch = MAPPED_OFFSET_ROW_PITCH
3086 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3087 width = MAPPED_OFFSET_WIDTH
3088 height = MAPPED_OFFSET_HEIGHT
3089 depth = MAPPED_OFFSET_DEPTH
3090 CODE:
3091 RETVAL = *(size_t *)((char *)self + ix);
3092 OUTPUT:
3093 RETVAL
3094
3095IV
3096ptr (OpenCL::Mapped self)
3097 CODE:
3098 RETVAL = PTR2IV (self->ptr);
3099 OUTPUT:
3100 RETVAL
3101
3102void
3103set (OpenCL::Mapped self, size_t offset, SV *data)
3104 CODE:
3105 STRLEN len;
3106 const char *ptr = SvPVbyte (data, len);
3107
3108 if (offset + len > self->cb)
3109 croak ("OpenCL::Mapped::set out of bound condition detected");
3110
3111 memcpy (offset + (char *)self->ptr, ptr, len);
3112
3113void
3114get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3115 PPCODE:
3116 if (!SvOK (ST (1)))
3117 count = self->width - x;
3118
3119 if (x + count > self->width)
3120 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3121
3122 if (y >= self->height)
3123 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3124
3125 if (z >= self->depth)
3126 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3127
3128 size_t element = mapped_element_size (self);
3129
3130 count *= element;
3131 x *= element;
3132
3133 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3134 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3135
3136void
3137set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3138 PPCODE:
3139 STRLEN count;
3140 char *dataptr = SvPVbyte (data, count);
3141 size_t element = mapped_element_size (self);
3142
3143 x *= element;
3144
3145 if (x + count > self->width * element)
3146 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3147
3148 if (y >= self->height)
3149 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3150
3151 if (z >= self->depth)
3152 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3153
3154 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3155 memcpy (ptr, dataptr, count);
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3158
3159MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3160
3161IV
3162element_size (OpenCL::Mapped self)
3163 CODE:
3164 RETVAL = mapped_element_size (self);
3165 OUTPUT:
3166 RETVAL
3167

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines