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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.32 by root, Thu Apr 19 19:36:24 2012 UTC vs.
Revision 1.53 by root, Wed Apr 25 00:02:25 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h"
7#include "schmorp.h"
8
5#ifdef I_DLFCN 9#ifdef I_DLFCN
6 #include <dlfcn.h> 10 #include <dlfcn.h>
11#endif
12
13// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
14// but then fail to define the api functions unless you ALSO define
15// this. This breaks 100% of the opencl 1.1 apps, for what reason?
16// after all, the functions are deprecated, not removed.
17// in addition, you cannot test for this in any future-proof way.
18// each time a new opencl version comes out, you need to make a new
19// release.
20#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
21
22#ifndef PREFER_1_1
23 #define PREFER_1_1 1
24#endif
25
26#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
7#endif 28#endif
8 29
9#ifdef __APPLE__ 30#ifdef __APPLE__
10 #include <OpenCL/opencl.h> 31 #include <OpenCL/opencl.h>
11#else 32#else
12 #include <CL/opencl.h> 33 #include <CL/opencl.h>
34#endif
35
36#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1
38 #define PREFER_1_1 1
13#endif 39#endif
14 40
15typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
16typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
17typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
89 enum { buffers = 3 }; 115 enum { buffers = 3 };
90 static int idx; 116 static int idx;
91 static void *buf [buffers]; 117 static void *buf [buffers];
92 static size_t len [buffers]; 118 static size_t len [buffers];
93 119
94 idx = ++idx % buffers; 120 idx = (idx + 1) % buffers;
95 121
96 if (len [idx] < size) 122 if (len [idx] < size)
97 { 123 {
98 free (buf [idx]); 124 free (buf [idx]);
99 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 125 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
164 190
165#define NEED_SUCCESS_ARG(retdecl, name, args) \ 191#define NEED_SUCCESS_ARG(retdecl, name, args) \
166 retdecl = cl ## name args; \ 192 retdecl = cl ## name args; \
167 if (res) \ 193 if (res) \
168 FAIL (name); 194 FAIL (name);
169
170/*****************************************************************************/
171
172#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
173#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
174
175static void *
176SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
177{
178 if (SvROK (sv) && sv_derived_from (sv, pkg))
179 return (void *)SvIV (SvRV (sv));
180
181 croak ("%s: %s is not of type %s", func, svname, pkg);
182}
183 195
184/*****************************************************************************/ 196/*****************************************************************************/
185 197
186static cl_context_properties * 198static cl_context_properties *
187SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
248 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
249} 261}
250 262
251/*****************************************************************************/ 263/*****************************************************************************/
252 264
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr))
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{
274 if (SvROK (sv) && sv_derived_from (sv, pkg))
275 return (void *)SvIV (SvRV (sv));
276
277 croak ("%s: %s is not of type %s", func, svname, pkg);
278}
279
280/*****************************************************************************/
281/* callback stuff */
282
283/* default context callback, log to stderr */
284static void CL_CALLBACK
285context_default_notify (const char *msg, const void *info, size_t cb, void *data)
286{
287 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
288}
289
290typedef struct
291{
292 int free_cb;
293 void (*push)(void *data1, void *data2, void *data3);
294} eq_vtbl;
295
296typedef struct eq_item
297{
298 struct eq_item *next;
299 eq_vtbl *vtbl;
300 SV *cb;
301 void *data1, *data2, *data3;
302} eq_item;
303
304static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail;
308
309static void
310eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
311{
312 eq_item *item = malloc (sizeof (eq_item));
313
314 item->next = 0;
315 item->vtbl = vtbl;
316 item->cb = cb;
317 item->data1 = data1;
318 item->data2 = data2;
319 item->data3 = data3;
320
321 X_LOCK (eq_lock);
322
323 *(eq_head ? &eq_tail->next : &eq_head) = item;
324 eq_tail = item;
325
326 X_UNLOCK (eq_lock);
327
328 eq_signal_func (eq_signal_arg, 0);
329}
330
331static eq_item *
332eq_dec (void)
333{
334 eq_item *res;
335
336 X_LOCK (eq_lock);
337
338 res = eq_head;
339 if (res)
340 eq_head = res->next;
341
342 X_UNLOCK (eq_lock);
343
344 return res;
345}
346
347static void
348eq_poll (void)
349{
350 eq_item *item;
351
352 while ((item = eq_dec ()))
353 {
354 ENTER;
355 SAVETMPS;
356
357 dSP;
358 PUSHMARK (SP);
359 EXTEND (SP, 2);
360
361 if (item->vtbl->free_cb)
362 sv_2mortal (item->cb);
363
364 PUTBACK;
365 item->vtbl->push (item->data1, item->data2, item->data3);
366
367 SV *cb = item->cb;
368 free (item);
369
370 call_sv (cb, G_DISCARD | G_VOID);
371
372 FREETMPS;
373 LEAVE;
374 }
375}
376
377static void
378eq_poll_interrupt (pTHX_ void *c_arg, int value)
379{
380 eq_poll ();
381}
382
383/*****************************************************************************/
384/* context notify */
385
386static void
387eq_context_push (void *data1, void *data2, void *data3)
388{
389 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
392 PUTBACK;
393
394 free (data1);
395 free (data2);
396}
397
398static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
399
400static void CL_CALLBACK
401eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
402{
403 void *pvt_copy = malloc (cb);
404 memcpy (pvt_copy, pvt, cb);
405 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
406}
407
408#define CONTEXT_NOTIFY_CALLBACK \
409 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
410 void *user_data = 0; \
411 \
412 if (SvOK (notify)) \
413 { \
414 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \
416 }
417
418static SV *
419new_clobj_context (cl_context ctx, void *user_data)
420{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx);
422
423 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425
426 return sv;
427}
428
429#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
430
431/*****************************************************************************/
432/* build/compile/link notify */
433
434static void
435eq_program_push (void *data1, void *data2, void *data3)
436{
437 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1);
439 PUTBACK;
440}
441
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443
444static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data)
446{
447 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
448}
449
450struct build_args
451{
452 cl_program program;
453 char *options;
454 void *user_data;
455 cl_uint num_devices;
456};
457
458X_THREAD_PROC (build_program_thread)
459{
460 struct build_args *arg = thr_arg;
461
462 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
463
464 if (arg->user_data)
465 eq_program_notify (arg->program, arg->user_data);
466 else
467 clReleaseProgram (arg->program);
468
469 free (arg->options);
470 free (arg);
471}
472
473static void
474build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
475{
476 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
477
478 arg->program = program;
479 arg->options = strdup (options);
480 arg->user_data = user_data;
481 arg->num_devices = num_devices;
482 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
483
484 xthread_t id;
485 thread_create (&id, build_program_thread, arg);
486}
487
488/*****************************************************************************/
489/* event objects */
490
491static void
492eq_event_push (void *data1, void *data2, void *data3)
493{
494 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK;
498}
499
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
501
502static void CL_CALLBACK
503eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
504{
505 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507}
508
509/*****************************************************************************/
510
253static size_t 511static size_t
254img_row_pitch (cl_mem img) 512img_row_pitch (cl_mem img)
255{ 513{
256 size_t res; 514 size_t res;
257 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
258 return res; 516 return res;
259} 517}
260 518
261static cl_event * 519static cl_event *
262event_list (SV **items, int count) 520event_list (SV **items, cl_uint *rcount)
263{ 521{
522 cl_uint count = *rcount;
523
264 if (!count) 524 if (!count)
265 return 0; 525 return 0;
266 526
267 cl_event *list = tmpbuf (sizeof (cl_event) * count); 527 cl_event *list = tmpbuf (sizeof (cl_event) * count);
528 int i = 0;
268 529
530 do
531 {
532 --count;
533 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
535 }
269 while (count--) 536 while (count);
270 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
271 537
538 *rcount = i;
539
272 return list; 540 return i ? list : 0;
273} 541}
274 542
275#define EVENT_LIST(items,count) \ 543#define EVENT_LIST(items,count) \
276 cl_uint event_list_count = (count); \ 544 cl_uint event_list_count = (count); \
277 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count)
278 546
279#define INFO(class) \ 547#define INFO(class) \
280{ \ 548{ \
281 size_t size; \ 549 size_t size; \
282 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
289} 557}
290 558
291MODULE = OpenCL PACKAGE = OpenCL 559MODULE = OpenCL PACKAGE = OpenCL
292 560
293PROTOTYPES: ENABLE 561PROTOTYPES: ENABLE
562
563void
564poll ()
565 CODE:
566 eq_poll ();
567
568void
569_eq_initialise (IV func, IV arg)
570 CODE:
571 eq_signal_func = (void (*)(void *, int))func;
572 eq_signal_arg = (void*)arg;
294 573
295BOOT: 574BOOT:
296{ 575{
297 HV *stash = gv_stashpv ("OpenCL", 1); 576 HV *stash = gv_stashpv ("OpenCL", 1);
298 static const ivstr *civ, const_iv[] = { 577 static const ivstr *civ, const_iv[] = {
307 { sizeof (cl_half ), "SIZEOF_HALF" }, 586 { sizeof (cl_half ), "SIZEOF_HALF" },
308 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 587 { sizeof (cl_float ), "SIZEOF_FLOAT" },
309 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 588 { sizeof (cl_double), "SIZEOF_DOUBLE" },
310#include "constiv.h" 589#include "constiv.h"
311 }; 590 };
591
312 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
313 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 593 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
594
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
314} 596}
315 597
316cl_int 598cl_int
317errno () 599errno ()
318 CODE: 600 CODE:
319 errno = res; 601 RETVAL = res;
602 OUTPUT:
603 RETVAL
320 604
321const char * 605const char *
322err2str (cl_int err) 606err2str (cl_int err)
323 607
324const char * 608const char *
335 list = tmpbuf (sizeof (*list) * count); 619 list = tmpbuf (sizeof (*list) * count);
336 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
337 621
338 EXTEND (SP, count); 622 EXTEND (SP, count);
339 for (i = 0; i < count; ++i) 623 for (i = 0; i < count; ++i)
340 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 624 PUSH_CLOBJ ("OpenCL::Platform", list [i]);
341 625
342void 626void
343context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
344 PPCODE: 628 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK;
345 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
346 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 631 XPUSH_CLOBJ_CONTEXT;
347 632
348void 633void
349context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 634context (FUTURE properties, FUTURE devices, FUTURE notify)
350 PPCODE: 635 PPCODE:
351 /* der Gipfel der Kunst */ 636 /* der Gipfel der Kunst */
352 637
353void 638void
354wait_for_events (...) 639wait_for_events (...)
362 647
363void 648void
364info (OpenCL::Platform self, cl_platform_info name) 649info (OpenCL::Platform self, cl_platform_info name)
365 PPCODE: 650 PPCODE:
366 INFO (Platform) 651 INFO (Platform)
652
653void
654unload_compiler (OpenCL::Platform self)
655 CODE:
656#if CL_VERSION_1_2
657 clUnloadPlatformCompiler (self);
658#endif
367 659
368#BEGIN:platform 660#BEGIN:platform
369 661
370void 662void
371profile (OpenCL::Platform self) 663profile (OpenCL::Platform self)
400 EXTEND (SP, count); 692 EXTEND (SP, count);
401 for (i = 0; i < count; ++i) 693 for (i = 0; i < count; ++i)
402 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i]));
403 695
404void 696void
405context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
406 PPCODE: 698 PPCODE:
407 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
408 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
409 701
410 AV *av = (AV *)SvRV (devices); 702 AV *av = (AV *)SvRV (devices);
411 cl_uint num_devices = av_len (av) + 1; 703 cl_uint num_devices = av_len (av) + 1;
412 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
705
413 int i; 706 int i;
414
415 for (i = num_devices; i--; ) 707 for (i = num_devices; i--; )
416 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
417 709
710 CONTEXT_NOTIFY_CALLBACK;
418 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
419 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 712 XPUSH_CLOBJ_CONTEXT;
420 713
421void 714void
422context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
423 PPCODE: 716 PPCODE:
424 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
425 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719
720 CONTEXT_NOTIFY_CALLBACK;
426 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
427 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 722 XPUSH_CLOBJ_CONTEXT;
428 723
429MODULE = OpenCL PACKAGE = OpenCL::Device 724MODULE = OpenCL PACKAGE = OpenCL::Device
430 725
431void 726void
432info (OpenCL::Device self, cl_device_info name) 727info (OpenCL::Device self, cl_device_info name)
471 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 766 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
472 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 767 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
473 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 768 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
474 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 769 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
475 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 770 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
476 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 771 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
477 PPCODE: 772 PPCODE:
478 cl_uint value [1]; 773 cl_uint value [1];
479 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 774 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
480 EXTEND (SP, 1); 775 EXTEND (SP, 1);
481 const int i = 0; 776 const int i = 0;
605 cl_platform_id value [1]; 900 cl_platform_id value [1];
606 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
607 EXTEND (SP, 1); 902 EXTEND (SP, 1);
608 const int i = 0; 903 const int i = 0;
609 { 904 {
610 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 905 PUSH_CLOBJ ("OpenCL::Platform", value [i]);
611 } 906 }
612 907
613void 908void
614name (OpenCL::Device self) 909name (OpenCL::Device self)
615 ALIAS: 910 ALIAS:
634 cl_device_id value [1]; 929 cl_device_id value [1];
635 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
636 EXTEND (SP, 1); 931 EXTEND (SP, 1);
637 const int i = 0; 932 const int i = 0;
638 { 933 {
639 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 934 PUSH_CLOBJ ("OpenCL::Device", value [i]);
640 } 935 }
641 936
642void 937void
643partition_types_ext (OpenCL::Device self) 938partition_types_ext (OpenCL::Device self)
644 ALIAS: 939 ALIAS:
671 966
672void 967void
673queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
674 PPCODE: 969 PPCODE:
675 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
676 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 971 XPUSH_CLOBJ ("OpenCL::Queue", queue);
677 972
678void 973void
679user_event (OpenCL::Context self) 974user_event (OpenCL::Context self)
680 PPCODE: 975 PPCODE:
681 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
682 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev);
683 978
684void 979void
685buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
686 PPCODE: 981 PPCODE:
687 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
688 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 983 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
689 984
690 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
691 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
692 987
693void 988void
694buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
695 PPCODE: 990 PPCODE:
696 STRLEN len; 991 STRLEN len;
697 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
698 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
699 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 994 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
700 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 995 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
701 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
702 997
998#if CL_VERSION_1_2
999
703void 1000void
704image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1001image (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)
705 PPCODE: 1002 PPCODE:
706 STRLEN len; 1003 STRLEN len;
707 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
708 const cl_image_format format = { channel_order, channel_type }; 1005 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = {
1007 type,
1008 width, height, depth,
1009 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1012 };
709 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
710 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1014 char *klass = "OpenCL::Image";
1015 switch (type)
1016 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break;
1023 }
1024 XPUSH_CLOBJ (klass, mem);
711 1025
1026#endif
1027
712void 1028void
713image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1029image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
714 PPCODE: 1030 PPCODE:
715 STRLEN len; 1031 STRLEN len;
716 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1032 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
717 const cl_image_format format = { channel_order, channel_type }; 1033 const cl_image_format format = { channel_order, channel_type };
1034#if PREFER_1_1
1035 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1038 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
1041
1042void
1043image3d (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)
1044 PPCODE:
1045 STRLEN len;
1046 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1047 const cl_image_format format = { channel_order, channel_type };
1048#if PREFER_1_1
718 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1052 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif
719 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
720 1055
721#if cl_apple_gl_sharing || cl_khr_gl_sharing 1056#if cl_apple_gl_sharing || cl_khr_gl_sharing
722 1057
723void 1058void
724gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
725 PPCODE: 1060 PPCODE:
726 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
727 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
1063
1064void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
1069
1070#if CL_VERSION_1_2
1071
1072void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS:
1075 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory";
1080 switch (type)
1081 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break;
1088 }
1089 XPUSH_CLOBJ (klass, mem);
1090
1091#endif
728 1092
729void 1093void
730gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
731 PPCODE: 1095 PPCODE:
1096#if PREFER_1_1
732 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif
733 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
734 1102
735void 1103void
736gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
737 PPCODE: 1105 PPCODE:
1106#if PREFER_1_1
738 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif
739 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
740
741void
742gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
743 PPCODE:
744 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
745 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
746 1112
747#endif 1113#endif
748 1114
749void 1115void
750supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
770 1136
771void 1137void
772sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
773 PPCODE: 1139 PPCODE:
774 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
775 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler);
776 1142
777void 1143void
778program_with_source (OpenCL::Context self, SV *program) 1144program_with_source (OpenCL::Context self, SV *program)
779 PPCODE: 1145 PPCODE:
780 STRLEN len; 1146 STRLEN len;
781 size_t len2; 1147 size_t len2;
782 const char *ptr = SvPVbyte (program, len); 1148 const char *ptr = SvPVbyte (program, len);
783 1149
784 len2 = len; 1150 len2 = len;
785 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
786 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1152 XPUSH_CLOBJ ("OpenCL::Program", prog);
787 1153
788#BEGIN:context 1154#BEGIN:context
789 1155
790void 1156void
791reference_count (OpenCL::Context self) 1157reference_count (OpenCL::Context self)
808 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
809 int i, n = size / sizeof (*value); 1175 int i, n = size / sizeof (*value);
810 EXTEND (SP, n); 1176 EXTEND (SP, n);
811 for (i = 0; i < n; ++i) 1177 for (i = 0; i < n; ++i)
812 { 1178 {
813 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1179 PUSH_CLOBJ ("OpenCL::Device", value [i]);
814 } 1180 }
815 1181
816void 1182void
817properties (OpenCL::Context self) 1183properties (OpenCL::Context self)
818 PPCODE: 1184 PPCODE:
845 SvPOK_only (data); 1211 SvPOK_only (data);
846 SvCUR_set (data, len); 1212 SvCUR_set (data, len);
847 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1213 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
848 1214
849 if (ev) 1215 if (ev)
850 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1216 XPUSH_CLOBJ ("OpenCL::Event", ev);
851 1217
852void 1218void
853enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
854 PPCODE: 1220 PPCODE:
855 cl_event ev = 0; 1221 cl_event ev = 0;
856 STRLEN len; 1222 STRLEN len;
857 char *ptr = SvPVbyte (data, len); 1223 char *ptr = SvPVbyte (data, len);
858 EVENT_LIST (5, items - 5); 1224 EVENT_LIST (5, items - 5);
859 1225
860 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1226 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
861 1227
862 if (ev) 1228 if (ev)
863 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1229 XPUSH_CLOBJ ("OpenCL::Event", ev);
1230
1231#if CL_VERSION_1_2
1232
1233void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1235 PPCODE:
1236 cl_event ev = 0;
1237 STRLEN len;
1238 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5);
1240
1241 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242
1243 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev);
1245
1246void
1247enqueue_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, ...)
1248 PPCODE:
1249 cl_event ev = 0;
1250 STRLEN len;
1251 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12);
1254
1255 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1259 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1260 cl_image_format format;
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1263 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1264 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]],
1267 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268
1269 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev);
1271
1272#endif
864 1273
865void 1274void
866enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
867 PPCODE: 1276 PPCODE:
868 cl_event ev = 0; 1277 cl_event ev = 0;
869 EVENT_LIST (6, items - 6); 1278 EVENT_LIST (6, items - 6);
870 1279
871 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1280 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
872 1281
873 if (ev) 1282 if (ev)
874 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1283 XPUSH_CLOBJ ("OpenCL::Event", ev);
875 1284
876void 1285void
877enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1286enqueue_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, ...)
878 PPCODE: 1287 PPCODE:
879 cl_event ev = 0; 1288 cl_event ev = 0;
901 SvPOK_only (data); 1310 SvPOK_only (data);
902 SvCUR_set (data, len); 1311 SvCUR_set (data, len);
903 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1312 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
904 1313
905 if (ev) 1314 if (ev)
906 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1315 XPUSH_CLOBJ ("OpenCL::Event", ev);
907 1316
908void 1317void
909enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1318enqueue_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, ...)
910 PPCODE: 1319 PPCODE:
911 cl_event ev = 0; 1320 cl_event ev = 0;
931 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1340 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
932 1341
933 if (len < min_len) 1342 if (len < min_len)
934 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
935 1344
936 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1345 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
937 1346
938 if (ev) 1347 if (ev)
939 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1348 XPUSH_CLOBJ ("OpenCL::Event", ev);
940 1349
941void 1350void
942enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1351enqueue_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, ...)
943 PPCODE: 1352 PPCODE:
944 cl_event ev = 0; 1353 cl_event ev = 0;
948 EVENT_LIST (16, items - 16); 1357 EVENT_LIST (16, items - 16);
949 1358
950 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1359 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
951 1360
952 if (ev) 1361 if (ev)
953 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1362 XPUSH_CLOBJ ("OpenCL::Event", ev);
954 1363
955void 1364void
956enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1365enqueue_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, ...)
957 PPCODE: 1366 PPCODE:
958 cl_event ev = 0; 1367 cl_event ev = 0;
973 SvPOK_only (data); 1382 SvPOK_only (data);
974 SvCUR_set (data, len); 1383 SvCUR_set (data, len);
975 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1384 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
976 1385
977 if (ev) 1386 if (ev)
978 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1387 XPUSH_CLOBJ ("OpenCL::Event", ev);
979 1388
980void 1389void
981enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1390enqueue_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, ...)
982 PPCODE: 1391 PPCODE:
983 cl_event ev = 0; 1392 cl_event ev = 0;
996 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1405 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
997 1406
998 if (len < min_len) 1407 if (len < min_len)
999 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1000 1409
1001 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1410 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1002 1411
1003 if (ev) 1412 if (ev)
1004 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1413 XPUSH_CLOBJ ("OpenCL::Event", ev);
1005 1414
1006void 1415void
1007enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1416enqueue_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, ...)
1008 PPCODE: 1417 PPCODE:
1009 cl_event ev = 0; 1418 cl_event ev = 0;
1013 EVENT_LIST (12, items - 12); 1422 EVENT_LIST (12, items - 12);
1014 1423
1015 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1424 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1016 1425
1017 if (ev) 1426 if (ev)
1018 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1427 XPUSH_CLOBJ ("OpenCL::Event", ev);
1019 1428
1020void 1429void
1021enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1430enqueue_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, ...)
1022 PPCODE: 1431 PPCODE:
1023 cl_event ev = 0; 1432 cl_event ev = 0;
1026 EVENT_LIST (10, items - 10); 1435 EVENT_LIST (10, items - 10);
1027 1436
1028 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1437 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1029 1438
1030 if (ev) 1439 if (ev)
1031 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1440 XPUSH_CLOBJ ("OpenCL::Event", ev);
1032 1441
1033void 1442void
1034enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1443enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1035 PPCODE: 1444 PPCODE:
1036 cl_event ev = 0; 1445 cl_event ev = 0;
1039 EVENT_LIST (10, items - 10); 1448 EVENT_LIST (10, items - 10);
1040 1449
1041 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1450 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1042 1451
1043 if (ev) 1452 if (ev)
1044 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1453 XPUSH_CLOBJ ("OpenCL::Event", ev);
1045 1454
1046void 1455void
1047enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1456enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1048 PPCODE: 1457 PPCODE:
1049 cl_event ev = 0; 1458 cl_event ev = 0;
1050 EVENT_LIST (2, items - 2); 1459 EVENT_LIST (2, items - 2);
1051 1460
1052 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1461 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1053 1462
1054 if (ev) 1463 if (ev)
1055 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
1056 1465
1057void 1466void
1058enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1467enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1059 PPCODE: 1468 PPCODE:
1060 cl_event ev = 0; 1469 cl_event ev = 0;
1088 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1497 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1089 } 1498 }
1090 1499
1091 if (SvOK (local_work_size)) 1500 if (SvOK (local_work_size))
1092 { 1501 {
1093 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1094 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1503 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference");
1095 1504
1096 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1097 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1506 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1098 1507
1102 } 1511 }
1103 1512
1104 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1513 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1105 1514
1106 if (ev) 1515 if (ev)
1107 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1516 XPUSH_CLOBJ ("OpenCL::Event", ev);
1108 1517
1109#if cl_apple_gl_sharing || cl_khr_gl_sharing 1518#if cl_apple_gl_sharing || cl_khr_gl_sharing
1110 1519
1111void 1520void
1112enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1113 ALIAS: 1522 ALIAS:
1114 enqueue_release_gl_objects = 1 1523 enqueue_release_gl_objects = 1
1115 CODE: 1524 PPCODE:
1116 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1117 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1118 1527
1119 cl_event ev = 0; 1528 cl_event ev = 0;
1120 EVENT_LIST (2, items - 2); 1529 EVENT_LIST (2, items - 2);
1122 cl_uint num_objects = av_len (av) + 1; 1531 cl_uint num_objects = av_len (av) + 1;
1123 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1532 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1124 int i; 1533 int i;
1125 1534
1126 for (i = num_objects; i--; ) 1535 for (i = num_objects; i--; )
1127 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1536 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1128 1537
1129 if (ix) 1538 if (ix)
1130 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1131 else 1540 else
1132 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1133 1542
1134 if (ev) 1543 if (ev)
1135 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1544 XPUSH_CLOBJ ("OpenCL::Event", ev);
1136 1545
1137#endif 1546#endif
1138 1547
1139void 1548void
1140enqueue_marker (OpenCL::Queue self)
1141 PPCODE:
1142 cl_event ev;
1143 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1144 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1145
1146void
1147enqueue_wait_for_events (OpenCL::Queue self, ...) 1549enqueue_wait_for_events (OpenCL::Queue self, ...)
1148 CODE: 1550 CODE:
1149 EVENT_LIST (1, items - 1); 1551 EVENT_LIST (1, items - 1);
1552#if PREFER_1_1
1150 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif
1151 1557
1152void 1558void
1559enqueue_marker (OpenCL::Queue self, ...)
1560 PPCODE:
1561 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1);
1563#if PREFER_1_1
1564 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else
1567#if CL_VERSION_1_2
1568 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1569#else
1570 {
1571 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1572 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1573 }
1574#endif
1575#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif
1578 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev);
1580
1581void
1153enqueue_barrier (OpenCL::Queue self) 1582enqueue_barrier (OpenCL::Queue self, ...)
1154 CODE: 1583 PPCODE:
1584 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1);
1586#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID)
1155 NEED_SUCCESS (EnqueueBarrier, (self)); 1588 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else
1590#if CL_VERSION_1_2
1591 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1592#else
1593 {
1594 if (event_list_count)
1595 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1596
1597 if (GIMME_V != G_VOID)
1598 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1599 }
1600#endif
1601#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif
1604 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev);
1156 1606
1157void 1607void
1158flush (OpenCL::Queue self) 1608flush (OpenCL::Queue self)
1159 CODE: 1609 CODE:
1160 NEED_SUCCESS (Flush, (self)); 1610 NEED_SUCCESS (Flush, (self));
1178 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1179 EXTEND (SP, 1); 1629 EXTEND (SP, 1);
1180 const int i = 0; 1630 const int i = 0;
1181 { 1631 {
1182 NEED_SUCCESS (RetainContext, (value [i])); 1632 NEED_SUCCESS (RetainContext, (value [i]));
1183 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1633 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1184 } 1634 }
1185 1635
1186void 1636void
1187device (OpenCL::Queue self) 1637device (OpenCL::Queue self)
1188 PPCODE: 1638 PPCODE:
1189 cl_device_id value [1]; 1639 cl_device_id value [1];
1190 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1191 EXTEND (SP, 1); 1641 EXTEND (SP, 1);
1192 const int i = 0; 1642 const int i = 0;
1193 { 1643 {
1194 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1644 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1195 } 1645 }
1196 1646
1197void 1647void
1198reference_count (OpenCL::Queue self) 1648reference_count (OpenCL::Queue self)
1199 PPCODE: 1649 PPCODE:
1286 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1287 EXTEND (SP, 1); 1737 EXTEND (SP, 1);
1288 const int i = 0; 1738 const int i = 0;
1289 { 1739 {
1290 NEED_SUCCESS (RetainContext, (value [i])); 1740 NEED_SUCCESS (RetainContext, (value [i]));
1291 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1741 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1292 } 1742 }
1293 1743
1294void 1744void
1295associated_memobject (OpenCL::Memory self) 1745associated_memobject (OpenCL::Memory self)
1296 PPCODE: 1746 PPCODE:
1298 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1299 EXTEND (SP, 1); 1749 EXTEND (SP, 1);
1300 const int i = 0; 1750 const int i = 0;
1301 { 1751 {
1302 NEED_SUCCESS (RetainMemObject, (value [i])); 1752 NEED_SUCCESS (RetainMemObject, (value [i]));
1303 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]);
1304 } 1754 }
1305 1755
1306#END:mem 1756#END:mem
1307 1757
1308#if cl_apple_gl_sharing || cl_khr_gl_sharing 1758#if cl_apple_gl_sharing || cl_khr_gl_sharing
1328 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 1778 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1329 1779
1330 cl_buffer_region crdata = { origin, size }; 1780 cl_buffer_region crdata = { origin, size };
1331 1781
1332 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1333 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem);
1334 1784
1335MODULE = OpenCL PACKAGE = OpenCL::Image 1785MODULE = OpenCL PACKAGE = OpenCL::Image
1336 1786
1337void 1787void
1338image_info (OpenCL::Image self, cl_image_info name) 1788image_info (OpenCL::Image self, cl_image_info name)
1339 PPCODE: 1789 PPCODE:
1340 INFO (Image) 1790 INFO (Image)
1791
1792void
1793format (OpenCL::Image self)
1794 PPCODE:
1795 cl_image_format format;
1796 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1797 EXTEND (SP, 2);
1798 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
1799 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1341 1800
1342#BEGIN:image 1801#BEGIN:image
1343 1802
1344void 1803void
1345element_size (OpenCL::Image self) 1804element_size (OpenCL::Image self)
1415 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1416 EXTEND (SP, 1); 1875 EXTEND (SP, 1);
1417 const int i = 0; 1876 const int i = 0;
1418 { 1877 {
1419 NEED_SUCCESS (RetainContext, (value [i])); 1878 NEED_SUCCESS (RetainContext, (value [i]));
1420 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1879 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1421 } 1880 }
1422 1881
1423void 1882void
1424normalized_coords (OpenCL::Sampler self) 1883normalized_coords (OpenCL::Sampler self)
1425 PPCODE: 1884 PPCODE:
1455DESTROY (OpenCL::Program self) 1914DESTROY (OpenCL::Program self)
1456 CODE: 1915 CODE:
1457 clReleaseProgram (self); 1916 clReleaseProgram (self);
1458 1917
1459void 1918void
1460build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 1919build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1920 ALIAS:
1921 build_async = 1
1461 CODE: 1922 CODE:
1923 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1924 void *user_data = 0;
1925 cl_uint num_devices = 0;
1926 cl_device_id *device_list = 0;
1927
1928 if (SvOK (devices))
1929 {
1930 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1931 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
1932
1933 AV *av = (AV *)SvRV (devices);
1934 num_devices = av_len (av) + 1;
1935
1936 if (num_devices)
1937 {
1938 device_list = tmpbuf (sizeof (*device_list) * num_devices);
1939 int count;
1940 for (count = 0; count < num_devices; ++count)
1941 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1942 }
1943 }
1944
1945 if (SvOK (notify))
1946 {
1947 NEED_SUCCESS (RetainProgram, (self));
1948 pfn_notify = eq_program_notify;
1949 user_data = SvREFCNT_inc (s_get_cv (notify));
1950 }
1951
1952 if (ix)
1953 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1954 else
1462 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 1955 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1463 1956
1464void 1957void
1465build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 1958build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1466 PPCODE: 1959 PPCODE:
1467 size_t size; 1960 size_t size;
1502 1995
1503void 1996void
1504kernel (OpenCL::Program program, SV *function) 1997kernel (OpenCL::Program program, SV *function)
1505 PPCODE: 1998 PPCODE:
1506 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1507 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel);
2001
2002void
2003kernels_in_program (OpenCL::Program program)
2004 PPCODE:
2005 cl_uint num_kernels;
2006 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2007 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009
2010 int i;
2011 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]);
1508 2014
1509void 2015void
1510info (OpenCL::Program self, cl_program_info name) 2016info (OpenCL::Program self, cl_program_info name)
1511 PPCODE: 2017 PPCODE:
1512 INFO (Program) 2018 INFO (Program)
1530 { 2036 {
1531 SV *sv = sv_2mortal (newSV (sizes [i])); 2037 SV *sv = sv_2mortal (newSV (sizes [i]));
1532 SvUPGRADE (sv, SVt_PV); 2038 SvUPGRADE (sv, SVt_PV);
1533 SvPOK_only (sv); 2039 SvPOK_only (sv);
1534 SvCUR_set (sv, sizes [i]); 2040 SvCUR_set (sv, sizes [i]);
1535 ptrs [i] = SvPVX (sv); 2041 ptrs [i] = (void *)SvPVX (sv);
1536 PUSHs (sv); 2042 PUSHs (sv);
1537 } 2043 }
1538 2044
1539 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2045 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1540 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2046 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1560 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1561 EXTEND (SP, 1); 2067 EXTEND (SP, 1);
1562 const int i = 0; 2068 const int i = 0;
1563 { 2069 {
1564 NEED_SUCCESS (RetainContext, (value [i])); 2070 NEED_SUCCESS (RetainContext, (value [i]));
1565 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2071 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1566 } 2072 }
1567 2073
1568void 2074void
1569devices (OpenCL::Program self) 2075devices (OpenCL::Program self)
1570 PPCODE: 2076 PPCODE:
1574 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1575 int i, n = size / sizeof (*value); 2081 int i, n = size / sizeof (*value);
1576 EXTEND (SP, n); 2082 EXTEND (SP, n);
1577 for (i = 0; i < n; ++i) 2083 for (i = 0; i < n; ++i)
1578 { 2084 {
1579 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2085 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1580 } 2086 }
1581 2087
1582void 2088void
1583source (OpenCL::Program self) 2089source (OpenCL::Program self)
1584 PPCODE: 2090 PPCODE:
1690set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2196set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1691 CODE: 2197 CODE:
1692 clSetKernelArg (self, idx, sizeof (value), &value); 2198 clSetKernelArg (self, idx, sizeof (value), &value);
1693 2199
1694void 2200void
2201set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2202 CODE:
2203 clSetKernelArg (self, idx, size, 0);
2204
2205void
1695set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2206set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1696 CODE: 2207 CODE:
1697 clSetKernelArg (self, idx, sizeof (value), &value); 2208 clSetKernelArg (self, idx, sizeof (value), &value);
1698 2209
1699void 2210void
1733 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2244 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1734 EXTEND (SP, 1); 2245 EXTEND (SP, 1);
1735 const int i = 0; 2246 const int i = 0;
1736 { 2247 {
1737 NEED_SUCCESS (RetainContext, (value [i])); 2248 NEED_SUCCESS (RetainContext, (value [i]));
1738 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2249 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1739 } 2250 }
1740 2251
1741void 2252void
1742program (OpenCL::Kernel self) 2253program (OpenCL::Kernel self)
1743 PPCODE: 2254 PPCODE:
1745 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2256 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1746 EXTEND (SP, 1); 2257 EXTEND (SP, 1);
1747 const int i = 0; 2258 const int i = 0;
1748 { 2259 {
1749 NEED_SUCCESS (RetainProgram, (value [i])); 2260 NEED_SUCCESS (RetainProgram, (value [i]));
1750 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2261 PUSH_CLOBJ ("OpenCL::Program", value [i]);
1751 } 2262 }
1752 2263
1753#END:kernel 2264#END:kernel
1754 2265
1755void 2266void
1815wait (OpenCL::Event self) 2326wait (OpenCL::Event self)
1816 CODE: 2327 CODE:
1817 clWaitForEvents (1, &self); 2328 clWaitForEvents (1, &self);
1818 2329
1819void 2330void
2331cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2332 CODE:
2333 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2334
2335void
1820info (OpenCL::Event self, cl_event_info name) 2336info (OpenCL::Event self, cl_event_info name)
1821 PPCODE: 2337 PPCODE:
1822 INFO (Event) 2338 INFO (Event)
1823 2339
1824#BEGIN:event 2340#BEGIN:event
1830 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2346 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1831 EXTEND (SP, 1); 2347 EXTEND (SP, 1);
1832 const int i = 0; 2348 const int i = 0;
1833 { 2349 {
1834 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2350 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1835 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2351 PUSH_CLOBJ ("OpenCL::Queue", value [i]);
1836 } 2352 }
1837 2353
1838void 2354void
1839command_type (OpenCL::Event self) 2355command_type (OpenCL::Event self)
1840 PPCODE: 2356 PPCODE:
1863 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2379 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1864 EXTEND (SP, 1); 2380 EXTEND (SP, 1);
1865 const int i = 0; 2381 const int i = 0;
1866 { 2382 {
1867 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
1868 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2384 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1869 } 2385 }
1870 2386
1871#END:event 2387#END:event
1872 2388
1873void 2389void

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines