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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.30 by root, Thu Apr 19 14:34:56 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;
36typedef SV *FUTURE; 62typedef SV *FUTURE;
37 63
38/*****************************************************************************/ 64/*****************************************************************************/
39 65
40// name must include a leading underscore 66// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
68// for perl. doh.
41static void * 69static void *
42getsym (const char *name) 70glsym (const char *name)
43{ 71{
72 void *fun = 0;
73
44 #if defined I_DLFCN && defined RTLD_DEFAULT 74 #if defined I_DLFCN && defined RTLD_DEFAULT
45 #if !DLSYM_NEEDS_UNDERSCORE 75 fun = dlsym (RTLD_DEFAULT, name + 1);
46 ++name; // skip _
47 #endif
48 return dlsym (RTLD_DEFAULT, name); 76 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
49 #else 77
50 return 0; 78 if (!fun)
79 {
80 static void *libgl;
81 static const char *glso[] = {
82 "libGL.so.1",
83 "libGL.so.3",
84 "libGL.so.4.0",
85 "libGL.so",
86 "/usr/lib/libGL.so",
87 "/usr/X11R6/lib/libGL.1.dylib"
88 };
89 int i;
90
91 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
92 {
93 libgl = dlopen (glso [i], RTLD_LAZY);
94 if (libgl)
95 break;
96 }
97
98 if (libgl)
99 {
100 fun = dlsym (libgl, name + 1);
101 if (!fun) fun = dlsym (libgl, name);
102 }
103 }
51 #endif 104 #endif
105
106 return fun;
52} 107}
53 108
54/*****************************************************************************/ 109/*****************************************************************************/
55 110
56/* up to two temporary buffers */ 111/* up to two temporary buffers */
60 enum { buffers = 3 }; 115 enum { buffers = 3 };
61 static int idx; 116 static int idx;
62 static void *buf [buffers]; 117 static void *buf [buffers];
63 static size_t len [buffers]; 118 static size_t len [buffers];
64 119
65 idx = ++idx % buffers; 120 idx = (idx + 1) % buffers;
66 121
67 if (len [idx] < size) 122 if (len [idx] < size)
68 { 123 {
69 free (buf [idx]); 124 free (buf [idx]);
70 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 125 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
138 if (res) \ 193 if (res) \
139 FAIL (name); 194 FAIL (name);
140 195
141/*****************************************************************************/ 196/*****************************************************************************/
142 197
143#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
144#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
145
146static void *
147SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
148{
149 if (SvROK (sv) && sv_derived_from (sv, pkg))
150 return (void *)SvIV (SvRV (sv));
151
152 croak ("%s: %s is not of type %s", func, svname, pkg);
153}
154
155/*****************************************************************************/
156
157static cl_context_properties * 198static cl_context_properties *
158SvCONTEXTPROPERTIES (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)
159{ 200{
160 if (!sv || !SvOK (sv)) 201 if (!sv || !SvOK (sv))
161 if (extra) 202 if (extra)
178 219
179 for (i = 0; i < len; i += 2) 220 for (i = 0; i < len; i += 2)
180 { 221 {
181 cl_context_properties t = SvIV (*av_fetch (av, i , 0)); 222 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
182 SV *p_sv = *av_fetch (av, i + 1, 0); 223 SV *p_sv = *av_fetch (av, i + 1, 0);
183 cl_context_properties v; 224 cl_context_properties v = SvIV (p_sv); // code below can override
184 225
185 switch (t) 226 switch (t)
186 { 227 {
228 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv))
230 {
231 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func)
233 v = (cl_context_properties)((void *(*)(void))func)();
234 }
235 break;
236
237 case CL_GL_CONTEXT_KHR:
238 if (!SvOK (p_sv))
239 {
240 void *func = glsym ("_glXGetCurrentContext");
241 if (func)
242 v = (cl_context_properties)((void *(*)(void))func)();
243 }
244 break;
245
187 default: 246 default:
188 /* unknown property, treat as int */ 247 /* unknown property, treat as int */
189 v = SvIV (p_sv);
190 break; 248 break;
191 } 249 }
192 250
193 *l++ = t; 251 *l++ = t;
194 *l++ = v; 252 *l++ = v;
202 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);
203} 261}
204 262
205/*****************************************************************************/ 263/*****************************************************************************/
206 264
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
207static size_t 511static size_t
208img_row_pitch (cl_mem img) 512img_row_pitch (cl_mem img)
209{ 513{
210 size_t res; 514 size_t res;
211 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
212 return res; 516 return res;
213} 517}
214 518
215static cl_event * 519static cl_event *
216event_list (SV **items, int count) 520event_list (SV **items, cl_uint *rcount)
217{ 521{
522 cl_uint count = *rcount;
523
218 if (!count) 524 if (!count)
219 return 0; 525 return 0;
220 526
221 cl_event *list = tmpbuf (sizeof (cl_event) * count); 527 cl_event *list = tmpbuf (sizeof (cl_event) * count);
528 int i = 0;
222 529
530 do
531 {
532 --count;
533 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
535 }
223 while (count--) 536 while (count);
224 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
225 537
538 *rcount = i;
539
226 return list; 540 return i ? list : 0;
227} 541}
228 542
229#define EVENT_LIST(items,count) \ 543#define EVENT_LIST(items,count) \
230 cl_uint event_list_count = (count); \ 544 cl_uint event_list_count = (count); \
231 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)
232 546
233#define INFO(class) \ 547#define INFO(class) \
234{ \ 548{ \
235 size_t size; \ 549 size_t size; \
236 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
243} 557}
244 558
245MODULE = OpenCL PACKAGE = OpenCL 559MODULE = OpenCL PACKAGE = OpenCL
246 560
247PROTOTYPES: 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;
248 573
249BOOT: 574BOOT:
250{ 575{
251 HV *stash = gv_stashpv ("OpenCL", 1); 576 HV *stash = gv_stashpv ("OpenCL", 1);
252 static const ivstr *civ, const_iv[] = { 577 static const ivstr *civ, const_iv[] = {
261 { sizeof (cl_half ), "SIZEOF_HALF" }, 586 { sizeof (cl_half ), "SIZEOF_HALF" },
262 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 587 { sizeof (cl_float ), "SIZEOF_FLOAT" },
263 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 588 { sizeof (cl_double), "SIZEOF_DOUBLE" },
264#include "constiv.h" 589#include "constiv.h"
265 }; 590 };
591
266 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--)
267 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);
268} 596}
269 597
270cl_int 598cl_int
271errno () 599errno ()
272 CODE: 600 CODE:
273 errno = res; 601 RETVAL = res;
602 OUTPUT:
603 RETVAL
274 604
275const char * 605const char *
276err2str (cl_int err) 606err2str (cl_int err)
277 607
278const char * 608const char *
289 list = tmpbuf (sizeof (*list) * count); 619 list = tmpbuf (sizeof (*list) * count);
290 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
291 621
292 EXTEND (SP, count); 622 EXTEND (SP, count);
293 for (i = 0; i < count; ++i) 623 for (i = 0; i < count; ++i)
294 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 624 PUSH_CLOBJ ("OpenCL::Platform", list [i]);
295 625
296void 626void
297context_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)
298 PPCODE: 628 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK;
299 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));
300 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 631 XPUSH_CLOBJ_CONTEXT;
301 632
302void 633void
303context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 634context (FUTURE properties, FUTURE devices, FUTURE notify)
304 PPCODE: 635 PPCODE:
305 /* der Gipfel der Kunst */ 636 /* der Gipfel der Kunst */
306 637
307void 638void
308wait_for_events (...) 639wait_for_events (...)
316 647
317void 648void
318info (OpenCL::Platform self, cl_platform_info name) 649info (OpenCL::Platform self, cl_platform_info name)
319 PPCODE: 650 PPCODE:
320 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
321 659
322#BEGIN:platform 660#BEGIN:platform
323 661
324void 662void
325profile (OpenCL::Platform self) 663profile (OpenCL::Platform self)
354 EXTEND (SP, count); 692 EXTEND (SP, count);
355 for (i = 0; i < count; ++i) 693 for (i = 0; i < count; ++i)
356 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i]));
357 695
358void 696void
359context (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)
360 PPCODE: 698 PPCODE:
361 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
362 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");
363 701
364 AV *av = (AV *)SvRV (devices); 702 AV *av = (AV *)SvRV (devices);
365 cl_uint num_devices = av_len (av) + 1; 703 cl_uint num_devices = av_len (av) + 1;
366 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
367 int i; 706 int i;
368
369 for (i = num_devices; i--; ) 707 for (i = num_devices; i--; )
370 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");
371 709
710 CONTEXT_NOTIFY_CALLBACK;
372 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));
373 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 712 XPUSH_CLOBJ_CONTEXT;
374 713
375void 714void
376context_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)
377 PPCODE: 716 PPCODE:
378 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
379 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;
380 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));
381 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 722 XPUSH_CLOBJ_CONTEXT;
382 723
383MODULE = OpenCL PACKAGE = OpenCL::Device 724MODULE = OpenCL PACKAGE = OpenCL::Device
384 725
385void 726void
386info (OpenCL::Device self, cl_device_info name) 727info (OpenCL::Device self, cl_device_info name)
425 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 766 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
426 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 767 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
427 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 768 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
428 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 769 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
429 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 770 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
430 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 771 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
431 PPCODE: 772 PPCODE:
432 cl_uint value [1]; 773 cl_uint value [1];
433 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 774 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
434 EXTEND (SP, 1); 775 EXTEND (SP, 1);
435 const int i = 0; 776 const int i = 0;
559 cl_platform_id value [1]; 900 cl_platform_id value [1];
560 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
561 EXTEND (SP, 1); 902 EXTEND (SP, 1);
562 const int i = 0; 903 const int i = 0;
563 { 904 {
564 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 905 PUSH_CLOBJ ("OpenCL::Platform", value [i]);
565 } 906 }
566 907
567void 908void
568name (OpenCL::Device self) 909name (OpenCL::Device self)
569 ALIAS: 910 ALIAS:
588 cl_device_id value [1]; 929 cl_device_id value [1];
589 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));
590 EXTEND (SP, 1); 931 EXTEND (SP, 1);
591 const int i = 0; 932 const int i = 0;
592 { 933 {
593 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 934 PUSH_CLOBJ ("OpenCL::Device", value [i]);
594 } 935 }
595 936
596void 937void
597partition_types_ext (OpenCL::Device self) 938partition_types_ext (OpenCL::Device self)
598 ALIAS: 939 ALIAS:
625 966
626void 967void
627queue (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)
628 PPCODE: 969 PPCODE:
629 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));
630 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 971 XPUSH_CLOBJ ("OpenCL::Queue", queue);
631 972
632void 973void
633user_event (OpenCL::Context self) 974user_event (OpenCL::Context self)
634 PPCODE: 975 PPCODE:
635 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
636 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev);
637 978
638void 979void
639buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
640 PPCODE: 981 PPCODE:
641 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))
642 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?");
643 984
644 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));
645 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
646 987
647void 988void
648buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
649 PPCODE: 990 PPCODE:
650 STRLEN len; 991 STRLEN len;
651 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
652 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)))
653 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?");
654 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));
655 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
656 997
998#if CL_VERSION_1_2
999
657void 1000void
658image2d (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)
659 PPCODE: 1002 PPCODE:
660 STRLEN len; 1003 STRLEN len;
661 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
662 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 };
663 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));
664 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);
665 1025
1026#endif
1027
666void 1028void
667image3d (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)
668 PPCODE: 1030 PPCODE:
669 STRLEN len; 1031 STRLEN len;
670 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1032 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
671 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
672 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
673 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
674 1055
675#if cl_apple_gl_sharing || cl_khr_gl_sharing 1056#if cl_apple_gl_sharing || cl_khr_gl_sharing
676 1057
677void 1058void
678gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
679 PPCODE: 1060 PPCODE:
680 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
681 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
682 1092
683void 1093void
684gl_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)
685 PPCODE: 1095 PPCODE:
1096#if PREFER_1_1
686 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
687 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
688 1102
689void 1103void
690gl_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)
691 PPCODE: 1105 PPCODE:
1106#if PREFER_1_1
692 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
693 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
694
695void
696gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
697 PPCODE:
698 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
699 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
700 1112
701#endif 1113#endif
702 1114
703void 1115void
704supported_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)
724 1136
725void 1137void
726sampler (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)
727 PPCODE: 1139 PPCODE:
728 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));
729 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler);
730 1142
731void 1143void
732program_with_source (OpenCL::Context self, SV *program) 1144program_with_source (OpenCL::Context self, SV *program)
733 PPCODE: 1145 PPCODE:
734 STRLEN len; 1146 STRLEN len;
735 size_t len2; 1147 size_t len2;
736 const char *ptr = SvPVbyte (program, len); 1148 const char *ptr = SvPVbyte (program, len);
737 1149
738 len2 = len; 1150 len2 = len;
739 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));
740 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1152 XPUSH_CLOBJ ("OpenCL::Program", prog);
741 1153
742#BEGIN:context 1154#BEGIN:context
743 1155
744void 1156void
745reference_count (OpenCL::Context self) 1157reference_count (OpenCL::Context self)
762 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
763 int i, n = size / sizeof (*value); 1175 int i, n = size / sizeof (*value);
764 EXTEND (SP, n); 1176 EXTEND (SP, n);
765 for (i = 0; i < n; ++i) 1177 for (i = 0; i < n; ++i)
766 { 1178 {
767 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1179 PUSH_CLOBJ ("OpenCL::Device", value [i]);
768 } 1180 }
769 1181
770void 1182void
771properties (OpenCL::Context self) 1183properties (OpenCL::Context self)
772 PPCODE: 1184 PPCODE:
799 SvPOK_only (data); 1211 SvPOK_only (data);
800 SvCUR_set (data, len); 1212 SvCUR_set (data, len);
801 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));
802 1214
803 if (ev) 1215 if (ev)
804 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1216 XPUSH_CLOBJ ("OpenCL::Event", ev);
805 1217
806void 1218void
807enqueue_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, ...)
808 PPCODE: 1220 PPCODE:
809 cl_event ev = 0; 1221 cl_event ev = 0;
810 STRLEN len; 1222 STRLEN len;
811 char *ptr = SvPVbyte (data, len); 1223 char *ptr = SvPVbyte (data, len);
812 EVENT_LIST (5, items - 5); 1224 EVENT_LIST (5, items - 5);
813 1225
814 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));
815 1227
816 if (ev) 1228 if (ev)
817 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
818 1273
819void 1274void
820enqueue_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, ...)
821 PPCODE: 1276 PPCODE:
822 cl_event ev = 0; 1277 cl_event ev = 0;
823 EVENT_LIST (6, items - 6); 1278 EVENT_LIST (6, items - 6);
824 1279
825 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));
826 1281
827 if (ev) 1282 if (ev)
828 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1283 XPUSH_CLOBJ ("OpenCL::Event", ev);
829 1284
830void 1285void
831enqueue_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, ...)
832 PPCODE: 1287 PPCODE:
833 cl_event ev = 0; 1288 cl_event ev = 0;
855 SvPOK_only (data); 1310 SvPOK_only (data);
856 SvCUR_set (data, len); 1311 SvCUR_set (data, len);
857 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));
858 1313
859 if (ev) 1314 if (ev)
860 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1315 XPUSH_CLOBJ ("OpenCL::Event", ev);
861 1316
862void 1317void
863enqueue_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, ...)
864 PPCODE: 1319 PPCODE:
865 cl_event ev = 0; 1320 cl_event ev = 0;
885 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];
886 1341
887 if (len < min_len) 1342 if (len < min_len)
888 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
889 1344
890 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));
891 1346
892 if (ev) 1347 if (ev)
893 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1348 XPUSH_CLOBJ ("OpenCL::Event", ev);
894 1349
895void 1350void
896enqueue_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, ...)
897 PPCODE: 1352 PPCODE:
898 cl_event ev = 0; 1353 cl_event ev = 0;
902 EVENT_LIST (16, items - 16); 1357 EVENT_LIST (16, items - 16);
903 1358
904 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));
905 1360
906 if (ev) 1361 if (ev)
907 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1362 XPUSH_CLOBJ ("OpenCL::Event", ev);
908 1363
909void 1364void
910enqueue_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, ...)
911 PPCODE: 1366 PPCODE:
912 cl_event ev = 0; 1367 cl_event ev = 0;
927 SvPOK_only (data); 1382 SvPOK_only (data);
928 SvCUR_set (data, len); 1383 SvCUR_set (data, len);
929 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));
930 1385
931 if (ev) 1386 if (ev)
932 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1387 XPUSH_CLOBJ ("OpenCL::Event", ev);
933 1388
934void 1389void
935enqueue_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, ...)
936 PPCODE: 1391 PPCODE:
937 cl_event ev = 0; 1392 cl_event ev = 0;
950 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;
951 1406
952 if (len < min_len) 1407 if (len < min_len)
953 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
954 1409
955 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));
956 1411
957 if (ev) 1412 if (ev)
958 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1413 XPUSH_CLOBJ ("OpenCL::Event", ev);
959 1414
960void 1415void
961enqueue_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, ...)
962 PPCODE: 1417 PPCODE:
963 cl_event ev = 0; 1418 cl_event ev = 0;
967 EVENT_LIST (12, items - 12); 1422 EVENT_LIST (12, items - 12);
968 1423
969 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));
970 1425
971 if (ev) 1426 if (ev)
972 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1427 XPUSH_CLOBJ ("OpenCL::Event", ev);
973 1428
974void 1429void
975enqueue_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, ...)
976 PPCODE: 1431 PPCODE:
977 cl_event ev = 0; 1432 cl_event ev = 0;
980 EVENT_LIST (10, items - 10); 1435 EVENT_LIST (10, items - 10);
981 1436
982 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));
983 1438
984 if (ev) 1439 if (ev)
985 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1440 XPUSH_CLOBJ ("OpenCL::Event", ev);
986 1441
987void 1442void
988enqueue_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, ...)
989 PPCODE: 1444 PPCODE:
990 cl_event ev = 0; 1445 cl_event ev = 0;
993 EVENT_LIST (10, items - 10); 1448 EVENT_LIST (10, items - 10);
994 1449
995 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));
996 1451
997 if (ev) 1452 if (ev)
998 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1453 XPUSH_CLOBJ ("OpenCL::Event", ev);
999 1454
1000void 1455void
1001enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1456enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1002 PPCODE: 1457 PPCODE:
1003 cl_event ev = 0; 1458 cl_event ev = 0;
1004 EVENT_LIST (2, items - 2); 1459 EVENT_LIST (2, items - 2);
1005 1460
1006 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));
1007 1462
1008 if (ev) 1463 if (ev)
1009 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
1010 1465
1011void 1466void
1012enqueue_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, ...)
1013 PPCODE: 1468 PPCODE:
1014 cl_event ev = 0; 1469 cl_event ev = 0;
1042 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1497 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1043 } 1498 }
1044 1499
1045 if (SvOK (local_work_size)) 1500 if (SvOK (local_work_size))
1046 { 1501 {
1047 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)
1048 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");
1049 1504
1050 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1051 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");
1052 1507
1056 } 1511 }
1057 1512
1058 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));
1059 1514
1060 if (ev) 1515 if (ev)
1061 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1516 XPUSH_CLOBJ ("OpenCL::Event", ev);
1062 1517
1063#if cl_apple_gl_sharing || cl_khr_gl_sharing 1518#if cl_apple_gl_sharing || cl_khr_gl_sharing
1064 1519
1065void 1520void
1066enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1067 ALIAS: 1522 ALIAS:
1068 enqueue_release_gl_objects = 1 1523 enqueue_release_gl_objects = 1
1069 CODE: 1524 PPCODE:
1070 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1071 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");
1072 1527
1073 cl_event ev = 0; 1528 cl_event ev = 0;
1074 EVENT_LIST (2, items - 2); 1529 EVENT_LIST (2, items - 2);
1076 cl_uint num_objects = av_len (av) + 1; 1531 cl_uint num_objects = av_len (av) + 1;
1077 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1532 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1078 int i; 1533 int i;
1079 1534
1080 for (i = num_objects; i--; ) 1535 for (i = num_objects; i--; )
1081 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");
1082 1537
1083 if (ix) 1538 if (ix)
1084 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));
1085 else 1540 else
1086 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));
1087 1542
1088 if (ev) 1543 if (ev)
1089 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1544 XPUSH_CLOBJ ("OpenCL::Event", ev);
1090 1545
1091#endif 1546#endif
1092 1547
1093void 1548void
1094enqueue_marker (OpenCL::Queue self)
1095 PPCODE:
1096 cl_event ev;
1097 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1098 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1099
1100void
1101enqueue_wait_for_events (OpenCL::Queue self, ...) 1549enqueue_wait_for_events (OpenCL::Queue self, ...)
1102 CODE: 1550 CODE:
1103 EVENT_LIST (1, items - 1); 1551 EVENT_LIST (1, items - 1);
1552#if PREFER_1_1
1104 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
1105 1557
1106void 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
1107enqueue_barrier (OpenCL::Queue self) 1582enqueue_barrier (OpenCL::Queue self, ...)
1108 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)
1109 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);
1110 1606
1111void 1607void
1112flush (OpenCL::Queue self) 1608flush (OpenCL::Queue self)
1113 CODE: 1609 CODE:
1114 NEED_SUCCESS (Flush, (self)); 1610 NEED_SUCCESS (Flush, (self));
1132 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1133 EXTEND (SP, 1); 1629 EXTEND (SP, 1);
1134 const int i = 0; 1630 const int i = 0;
1135 { 1631 {
1136 NEED_SUCCESS (RetainContext, (value [i])); 1632 NEED_SUCCESS (RetainContext, (value [i]));
1137 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1633 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1138 } 1634 }
1139 1635
1140void 1636void
1141device (OpenCL::Queue self) 1637device (OpenCL::Queue self)
1142 PPCODE: 1638 PPCODE:
1143 cl_device_id value [1]; 1639 cl_device_id value [1];
1144 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1145 EXTEND (SP, 1); 1641 EXTEND (SP, 1);
1146 const int i = 0; 1642 const int i = 0;
1147 { 1643 {
1148 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1644 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1149 } 1645 }
1150 1646
1151void 1647void
1152reference_count (OpenCL::Queue self) 1648reference_count (OpenCL::Queue self)
1153 PPCODE: 1649 PPCODE:
1240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1241 EXTEND (SP, 1); 1737 EXTEND (SP, 1);
1242 const int i = 0; 1738 const int i = 0;
1243 { 1739 {
1244 NEED_SUCCESS (RetainContext, (value [i])); 1740 NEED_SUCCESS (RetainContext, (value [i]));
1245 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1741 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1246 } 1742 }
1247 1743
1248void 1744void
1249associated_memobject (OpenCL::Memory self) 1745associated_memobject (OpenCL::Memory self)
1250 PPCODE: 1746 PPCODE:
1252 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));
1253 EXTEND (SP, 1); 1749 EXTEND (SP, 1);
1254 const int i = 0; 1750 const int i = 0;
1255 { 1751 {
1256 NEED_SUCCESS (RetainMemObject, (value [i])); 1752 NEED_SUCCESS (RetainMemObject, (value [i]));
1257 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]);
1258 } 1754 }
1259 1755
1260#END:mem 1756#END:mem
1261 1757
1262#if cl_apple_gl_sharing || cl_khr_gl_sharing 1758#if cl_apple_gl_sharing || cl_khr_gl_sharing
1264void 1760void
1265gl_object_info (OpenCL::Memory self) 1761gl_object_info (OpenCL::Memory self)
1266 PPCODE: 1762 PPCODE:
1267 cl_gl_object_type type; 1763 cl_gl_object_type type;
1268 cl_GLuint name; 1764 cl_GLuint name;
1269 NEED_SUCCESS (clGetGLObjectInfo, (self, &type, &name)); 1765 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
1270 EXTEND (SP, 2); 1766 EXTEND (SP, 2);
1271 PUSHs (sv_2mortal (newSVuv (type))); 1767 PUSHs (sv_2mortal (newSVuv (type)));
1272 PUSHs (sv_2mortal (newSVuv (name))); 1768 PUSHs (sv_2mortal (newSVuv (name)));
1273 1769
1274#endif 1770#endif
1282 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!");
1283 1779
1284 cl_buffer_region crdata = { origin, size }; 1780 cl_buffer_region crdata = { origin, size };
1285 1781
1286 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));
1287 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem);
1288 1784
1289MODULE = OpenCL PACKAGE = OpenCL::Image 1785MODULE = OpenCL PACKAGE = OpenCL::Image
1290 1786
1291void 1787void
1292image_info (OpenCL::Image self, cl_image_info name) 1788image_info (OpenCL::Image self, cl_image_info name)
1293 PPCODE: 1789 PPCODE:
1294 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)));
1295 1800
1296#BEGIN:image 1801#BEGIN:image
1297 1802
1298void 1803void
1299element_size (OpenCL::Image self) 1804element_size (OpenCL::Image self)
1319 1824
1320void 1825void
1321target (OpenCL::Image self) 1826target (OpenCL::Image self)
1322 PPCODE: 1827 PPCODE:
1323 cl_GLenum value [1]; 1828 cl_GLenum value [1];
1324 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0)); 1829 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
1325 EXTEND (SP, 1); 1830 EXTEND (SP, 1);
1326 const int i = 0; 1831 const int i = 0;
1327 PUSHs (sv_2mortal (newSVuv (value [i]))); 1832 PUSHs (sv_2mortal (newSVuv (value [i])));
1328 1833
1329void 1834void
1330gl_mipmap_level (OpenCL::Image self) 1835gl_mipmap_level (OpenCL::Image self)
1331 PPCODE: 1836 PPCODE:
1332 cl_GLint value [1]; 1837 cl_GLint value [1];
1333 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0)); 1838 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
1334 EXTEND (SP, 1); 1839 EXTEND (SP, 1);
1335 const int i = 0; 1840 const int i = 0;
1336 PUSHs (sv_2mortal (newSViv (value [i]))); 1841 PUSHs (sv_2mortal (newSViv (value [i])));
1337 1842
1338#END:gl_texture 1843#END:gl_texture
1369 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1370 EXTEND (SP, 1); 1875 EXTEND (SP, 1);
1371 const int i = 0; 1876 const int i = 0;
1372 { 1877 {
1373 NEED_SUCCESS (RetainContext, (value [i])); 1878 NEED_SUCCESS (RetainContext, (value [i]));
1374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1879 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1375 } 1880 }
1376 1881
1377void 1882void
1378normalized_coords (OpenCL::Sampler self) 1883normalized_coords (OpenCL::Sampler self)
1379 PPCODE: 1884 PPCODE:
1409DESTROY (OpenCL::Program self) 1914DESTROY (OpenCL::Program self)
1410 CODE: 1915 CODE:
1411 clReleaseProgram (self); 1916 clReleaseProgram (self);
1412 1917
1413void 1918void
1414build (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
1415 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
1416 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));
1417 1956
1418void 1957void
1419build_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)
1420 PPCODE: 1959 PPCODE:
1421 size_t size; 1960 size_t size;
1456 1995
1457void 1996void
1458kernel (OpenCL::Program program, SV *function) 1997kernel (OpenCL::Program program, SV *function)
1459 PPCODE: 1998 PPCODE:
1460 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));
1461 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]);
1462 2014
1463void 2015void
1464info (OpenCL::Program self, cl_program_info name) 2016info (OpenCL::Program self, cl_program_info name)
1465 PPCODE: 2017 PPCODE:
1466 INFO (Program) 2018 INFO (Program)
1484 { 2036 {
1485 SV *sv = sv_2mortal (newSV (sizes [i])); 2037 SV *sv = sv_2mortal (newSV (sizes [i]));
1486 SvUPGRADE (sv, SVt_PV); 2038 SvUPGRADE (sv, SVt_PV);
1487 SvPOK_only (sv); 2039 SvPOK_only (sv);
1488 SvCUR_set (sv, sizes [i]); 2040 SvCUR_set (sv, sizes [i]);
1489 ptrs [i] = SvPVX (sv); 2041 ptrs [i] = (void *)SvPVX (sv);
1490 PUSHs (sv); 2042 PUSHs (sv);
1491 } 2043 }
1492 2044
1493 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));
1494 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2046 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1514 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1515 EXTEND (SP, 1); 2067 EXTEND (SP, 1);
1516 const int i = 0; 2068 const int i = 0;
1517 { 2069 {
1518 NEED_SUCCESS (RetainContext, (value [i])); 2070 NEED_SUCCESS (RetainContext, (value [i]));
1519 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2071 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1520 } 2072 }
1521 2073
1522void 2074void
1523devices (OpenCL::Program self) 2075devices (OpenCL::Program self)
1524 PPCODE: 2076 PPCODE:
1528 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1529 int i, n = size / sizeof (*value); 2081 int i, n = size / sizeof (*value);
1530 EXTEND (SP, n); 2082 EXTEND (SP, n);
1531 for (i = 0; i < n; ++i) 2083 for (i = 0; i < n; ++i)
1532 { 2084 {
1533 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2085 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1534 } 2086 }
1535 2087
1536void 2088void
1537source (OpenCL::Program self) 2089source (OpenCL::Program self)
1538 PPCODE: 2090 PPCODE:
1644set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2196set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1645 CODE: 2197 CODE:
1646 clSetKernelArg (self, idx, sizeof (value), &value); 2198 clSetKernelArg (self, idx, sizeof (value), &value);
1647 2199
1648void 2200void
2201set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2202 CODE:
2203 clSetKernelArg (self, idx, size, 0);
2204
2205void
1649set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2206set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1650 CODE: 2207 CODE:
1651 clSetKernelArg (self, idx, sizeof (value), &value); 2208 clSetKernelArg (self, idx, sizeof (value), &value);
1652 2209
1653void 2210void
1687 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2244 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1688 EXTEND (SP, 1); 2245 EXTEND (SP, 1);
1689 const int i = 0; 2246 const int i = 0;
1690 { 2247 {
1691 NEED_SUCCESS (RetainContext, (value [i])); 2248 NEED_SUCCESS (RetainContext, (value [i]));
1692 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2249 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1693 } 2250 }
1694 2251
1695void 2252void
1696program (OpenCL::Kernel self) 2253program (OpenCL::Kernel self)
1697 PPCODE: 2254 PPCODE:
1699 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2256 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1700 EXTEND (SP, 1); 2257 EXTEND (SP, 1);
1701 const int i = 0; 2258 const int i = 0;
1702 { 2259 {
1703 NEED_SUCCESS (RetainProgram, (value [i])); 2260 NEED_SUCCESS (RetainProgram, (value [i]));
1704 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2261 PUSH_CLOBJ ("OpenCL::Program", value [i]);
1705 } 2262 }
1706 2263
1707#END:kernel 2264#END:kernel
1708 2265
1709void 2266void
1769wait (OpenCL::Event self) 2326wait (OpenCL::Event self)
1770 CODE: 2327 CODE:
1771 clWaitForEvents (1, &self); 2328 clWaitForEvents (1, &self);
1772 2329
1773void 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
1774info (OpenCL::Event self, cl_event_info name) 2336info (OpenCL::Event self, cl_event_info name)
1775 PPCODE: 2337 PPCODE:
1776 INFO (Event) 2338 INFO (Event)
1777 2339
1778#BEGIN:event 2340#BEGIN:event
1784 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));
1785 EXTEND (SP, 1); 2347 EXTEND (SP, 1);
1786 const int i = 0; 2348 const int i = 0;
1787 { 2349 {
1788 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2350 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1789 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2351 PUSH_CLOBJ ("OpenCL::Queue", value [i]);
1790 } 2352 }
1791 2353
1792void 2354void
1793command_type (OpenCL::Event self) 2355command_type (OpenCL::Event self)
1794 PPCODE: 2356 PPCODE:
1817 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2379 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1818 EXTEND (SP, 1); 2380 EXTEND (SP, 1);
1819 const int i = 0; 2381 const int i = 0;
1820 { 2382 {
1821 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
1822 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2384 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1823 } 2385 }
1824 2386
1825#END:event 2387#END:event
1826 2388
1827void 2389void

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines