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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.25 by root, Thu Apr 19 12:55:30 2012 UTC vs.
Revision 1.59 by root, Sun Apr 29 19:38:05 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h"
7#include "schmorp.h"
8
9#ifdef I_DLFCN
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
28#endif
4 29
5#ifdef __APPLE__ 30#ifdef __APPLE__
6 #include <OpenCL/opencl.h> 31 #include <OpenCL/opencl.h>
7#else 32#else
8 #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
9#endif 39#endif
10 40
11typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
12typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
13typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
31 61
32typedef SV *FUTURE; 62typedef SV *FUTURE;
33 63
34/*****************************************************************************/ 64/*****************************************************************************/
35 65
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.
69static void *
70glsym (const char *name)
71{
72 void *fun = 0;
73
74 #if defined I_DLFCN && defined RTLD_DEFAULT
75 fun = dlsym (RTLD_DEFAULT, name + 1);
76 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
77
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 }
104 #endif
105
106 return fun;
107}
108
109/*****************************************************************************/
110
36/* up to two temporary buffers */ 111/* up to two temporary buffers */
37static void * 112static void *
38tmpbuf (size_t size) 113tmpbuf (size_t size)
39{ 114{
40 enum { buffers = 3 }; 115 enum { buffers = 3 };
41 static int idx; 116 static int idx;
42 static void *buf [buffers]; 117 static void *buf [buffers];
43 static size_t len [buffers]; 118 static size_t len [buffers];
44 119
45 idx = ++idx % buffers; 120 idx = (idx + 1) % buffers;
46 121
47 if (len [idx] < size) 122 if (len [idx] < size)
48 { 123 {
49 free (buf [idx]); 124 free (buf [idx]);
50 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 125 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
118 if (res) \ 193 if (res) \
119 FAIL (name); 194 FAIL (name);
120 195
121/*****************************************************************************/ 196/*****************************************************************************/
122 197
123#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
124#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
125
126static void *
127SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
128{
129 if (SvROK (sv) && sv_derived_from (sv, pkg))
130 return (void *)SvIV (SvRV (sv));
131
132 croak ("%s: %s is not of type %s", func, svname, pkg);
133}
134
135/*****************************************************************************/
136
137static cl_context_properties * 198static cl_context_properties *
138SvCONTEXTPROPERTIES (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)
139{ 200{
140 if (!sv || !SvOK (sv)) 201 if (!sv || !SvOK (sv))
141 if (extra) 202 if (extra)
144 return 0; 205 return 0;
145 206
146 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV) 207 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
147 { 208 {
148 AV *av = (AV *)SvRV (sv); 209 AV *av = (AV *)SvRV (sv);
149 int i, len = av_len (av); 210 int i, len = av_len (av) + 1;
150 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
151 cl_context_properties *l = p; 212 cl_context_properties *l = p;
152 213
153 if (len & 1) 214 if (len & 1)
154 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname);
156 while (extracount--) 217 while (extracount--)
157 *l++ = *extra++; 218 *l++ = *extra++;
158 219
159 for (i = 0; i < len; i += 2) 220 for (i = 0; i < len; i += 2)
160 { 221 {
161 cl_context_properties t = SvIV (*av_fetch (av, i, 0)); 222 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
162 cl_context_properties v; 223 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override
163 225
164 switch (t) 226 switch (t)
165 { 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
166 default: 246 default:
167 /* unknown property, treat as int */ 247 /* unknown property, treat as int */
168 v = SvIV (*av_fetch (av, i, 0));
169 break; 248 break;
170 } 249 }
171 250
172 *l++ = t; 251 *l++ = t;
173 *l++ = v; 252 *l++ = v;
177 256
178 return p; 257 return p;
179 } 258 }
180 259
181 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);
261}
262
263/*****************************************************************************/
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);
182} 507}
183 508
184/*****************************************************************************/ 509/*****************************************************************************/
185 510
186static size_t 511static size_t
190 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
191 return res; 516 return res;
192} 517}
193 518
194static cl_event * 519static cl_event *
195event_list (SV **items, int count) 520event_list (SV **items, cl_uint *rcount)
196{ 521{
522 cl_uint count = *rcount;
523
197 if (!count) 524 if (!count)
198 return 0; 525 return 0;
199 526
200 cl_event *list = tmpbuf (sizeof (cl_event) * count); 527 cl_event *list = tmpbuf (sizeof (cl_event) * count);
528 int i = 0;
201 529
530 do
531 {
532 --count;
533 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
535 }
202 while (count--) 536 while (count);
203 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
204 537
538 *rcount = i;
539
205 return list; 540 return i ? list : 0;
206} 541}
207 542
208#define EVENT_LIST(items,count) \ 543#define EVENT_LIST(items,count) \
209 cl_uint event_list_count = (count); \ 544 cl_uint event_list_count = (count); \
210 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)
211 546
212#define INFO(class) \ 547#define INFO(class) \
213{ \ 548{ \
214 size_t size; \ 549 size_t size; \
215 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
222} 557}
223 558
224MODULE = OpenCL PACKAGE = OpenCL 559MODULE = OpenCL PACKAGE = OpenCL
225 560
226PROTOTYPES: 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;
227 573
228BOOT: 574BOOT:
229{ 575{
230 HV *stash = gv_stashpv ("OpenCL", 1); 576 HV *stash = gv_stashpv ("OpenCL", 1);
231 static const ivstr *civ, const_iv[] = { 577 static const ivstr *civ, const_iv[] = {
240 { sizeof (cl_half ), "SIZEOF_HALF" }, 586 { sizeof (cl_half ), "SIZEOF_HALF" },
241 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 587 { sizeof (cl_float ), "SIZEOF_FLOAT" },
242 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 588 { sizeof (cl_double), "SIZEOF_DOUBLE" },
243#include "constiv.h" 589#include "constiv.h"
244 }; 590 };
591
245 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--)
246 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);
247} 596}
248 597
249cl_int 598cl_int
250errno () 599errno ()
251 CODE: 600 CODE:
252 errno = res; 601 RETVAL = res;
602 OUTPUT:
603 RETVAL
253 604
254const char * 605const char *
255err2str (cl_int err) 606err2str (cl_int err = res)
256 607
257const char * 608const char *
258enum2str (cl_uint value) 609enum2str (cl_uint value)
259 610
260void 611void
268 list = tmpbuf (sizeof (*list) * count); 619 list = tmpbuf (sizeof (*list) * count);
269 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
270 621
271 EXTEND (SP, count); 622 EXTEND (SP, count);
272 for (i = 0; i < count; ++i) 623 for (i = 0; i < count; ++i)
273 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 624 PUSH_CLOBJ ("OpenCL::Platform", list [i]);
274 625
275void 626void
276context_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)
277 PPCODE: 628 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK;
278 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));
279 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 631 XPUSH_CLOBJ_CONTEXT;
280 632
281void 633void
282context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 634context (FUTURE properties, FUTURE devices, FUTURE notify)
283 PPCODE: 635 PPCODE:
284 /* der Gipfel der Kunst */ 636 /* der Gipfel der Kunst */
285 637
286void 638void
287wait_for_events (...) 639wait_for_events (...)
295 647
296void 648void
297info (OpenCL::Platform self, cl_platform_info name) 649info (OpenCL::Platform self, cl_platform_info name)
298 PPCODE: 650 PPCODE:
299 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
300 659
301#BEGIN:platform 660#BEGIN:platform
302 661
303void 662void
304profile (OpenCL::Platform self) 663profile (OpenCL::Platform self)
333 EXTEND (SP, count); 692 EXTEND (SP, count);
334 for (i = 0; i < count; ++i) 693 for (i = 0; i < count; ++i)
335 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i]));
336 695
337void 696void
338context (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)
339 PPCODE: 698 PPCODE:
340 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
341 croak ("OpenCL::Platform 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");
342 701
343 AV *av = (AV *)SvRV (devices); 702 AV *av = (AV *)SvRV (devices);
344 cl_uint num_devices = av_len (av) + 1; 703 cl_uint num_devices = av_len (av) + 1;
345 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
346 int i; 706 int i;
347
348 for (i = num_devices; i--; ) 707 for (i = num_devices; i--; )
349 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");
350 709
710 CONTEXT_NOTIFY_CALLBACK;
351 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));
352 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 712 XPUSH_CLOBJ_CONTEXT;
353 713
354void 714void
355context_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)
356 PPCODE: 716 PPCODE:
357 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
358 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;
359 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));
360 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 722 XPUSH_CLOBJ_CONTEXT;
361 723
362MODULE = OpenCL PACKAGE = OpenCL::Device 724MODULE = OpenCL PACKAGE = OpenCL::Device
363 725
364void 726void
365info (OpenCL::Device self, cl_device_info name) 727info (OpenCL::Device self, cl_device_info name)
404 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 766 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
405 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 767 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
406 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 768 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
407 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 769 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
408 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 770 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
409 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 771 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
410 PPCODE: 772 PPCODE:
411 cl_uint value [1]; 773 cl_uint value [1];
412 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 774 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
413 EXTEND (SP, 1); 775 EXTEND (SP, 1);
414 const int i = 0; 776 const int i = 0;
538 cl_platform_id value [1]; 900 cl_platform_id value [1];
539 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
540 EXTEND (SP, 1); 902 EXTEND (SP, 1);
541 const int i = 0; 903 const int i = 0;
542 { 904 {
543 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 905 PUSH_CLOBJ ("OpenCL::Platform", value [i]);
544 } 906 }
545 907
546void 908void
547name (OpenCL::Device self) 909name (OpenCL::Device self)
548 ALIAS: 910 ALIAS:
567 cl_device_id value [1]; 929 cl_device_id value [1];
568 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));
569 EXTEND (SP, 1); 931 EXTEND (SP, 1);
570 const int i = 0; 932 const int i = 0;
571 { 933 {
572 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 934 PUSH_CLOBJ ("OpenCL::Device", value [i]);
573 } 935 }
574 936
575void 937void
576partition_types_ext (OpenCL::Device self) 938partition_types_ext (OpenCL::Device self)
577 ALIAS: 939 ALIAS:
604 966
605void 967void
606queue (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)
607 PPCODE: 969 PPCODE:
608 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));
609 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 971 XPUSH_CLOBJ ("OpenCL::Queue", queue);
610 972
611void 973void
612user_event (OpenCL::Context self) 974user_event (OpenCL::Context self)
613 PPCODE: 975 PPCODE:
614 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
615 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev);
616 978
617void 979void
618buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
619 PPCODE: 981 PPCODE:
620 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))
621 croak ("clCreateBuffer: 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?");
622 984
623 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));
624 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
625 987
626void 988void
627buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
628 PPCODE: 990 PPCODE:
629 STRLEN len; 991 STRLEN len;
630 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
631 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)))
632 croak ("clCreateBuffer: 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?");
633 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));
634 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
635 997
998#if CL_VERSION_1_2
999
636void 1000void
637image2d (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 = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
638 PPCODE: 1002 PPCODE:
639 STRLEN len; 1003 STRLEN len;
640 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
641 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 };
642 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));
643 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);
644 1025
1026#endif
1027
645void 1028void
646image3d (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)
647 PPCODE: 1030 PPCODE:
648 STRLEN len; 1031 STRLEN len;
649 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1032 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
650 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
651 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
652 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
653 1055
654#if cl_apple_gl_sharing || cl_khr_gl_sharing 1056#if cl_apple_gl_sharing || cl_khr_gl_sharing
655 1057
656void 1058void
657gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
658 PPCODE: 1060 PPCODE:
659 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
660 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
661 1092
662void 1093void
663gl_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)
664 PPCODE: 1095 PPCODE:
1096#if PREFER_1_1
665 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
666 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
667 1102
668void 1103void
669gl_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)
670 PPCODE: 1105 PPCODE:
1106#if PREFER_1_1
671 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
672 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
673
674void
675gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
676 PPCODE:
677 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderBuffer, (self, flags, renderbuffer, &res));
678 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
679 1112
680#endif 1113#endif
681 1114
682void 1115void
683supported_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)
703 1136
704void 1137void
705sampler (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)
706 PPCODE: 1139 PPCODE:
707 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));
708 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler);
709 1142
710void 1143void
711program_with_source (OpenCL::Context self, SV *program) 1144program_with_source (OpenCL::Context self, SV *program)
712 PPCODE: 1145 PPCODE:
713 STRLEN len; 1146 STRLEN len;
714 size_t len2; 1147 size_t len2;
715 const char *ptr = SvPVbyte (program, len); 1148 const char *ptr = SvPVbyte (program, len);
716 1149
717 len2 = len; 1150 len2 = len;
718 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));
719 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1152 XPUSH_CLOBJ ("OpenCL::Program", prog);
720 1153
721#BEGIN:context 1154#BEGIN:context
722 1155
723void 1156void
724reference_count (OpenCL::Context self) 1157reference_count (OpenCL::Context self)
741 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
742 int i, n = size / sizeof (*value); 1175 int i, n = size / sizeof (*value);
743 EXTEND (SP, n); 1176 EXTEND (SP, n);
744 for (i = 0; i < n; ++i) 1177 for (i = 0; i < n; ++i)
745 { 1178 {
746 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1179 PUSH_CLOBJ ("OpenCL::Device", value [i]);
747 } 1180 }
748 1181
749void 1182void
750properties (OpenCL::Context self) 1183properties (OpenCL::Context self)
751 PPCODE: 1184 PPCODE:
766DESTROY (OpenCL::Queue self) 1199DESTROY (OpenCL::Queue self)
767 CODE: 1200 CODE:
768 clReleaseCommandQueue (self); 1201 clReleaseCommandQueue (self);
769 1202
770void 1203void
771enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1204read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1205 ALIAS:
1206 enqueue_read_buffer = 0
772 PPCODE: 1207 PPCODE:
773 cl_event ev = 0; 1208 cl_event ev = 0;
774 EVENT_LIST (6, items - 6); 1209 EVENT_LIST (6, items - 6);
775 1210
776 SvUPGRADE (data, SVt_PV); 1211 SvUPGRADE (data, SVt_PV);
778 SvPOK_only (data); 1213 SvPOK_only (data);
779 SvCUR_set (data, len); 1214 SvCUR_set (data, len);
780 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1215 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
781 1216
782 if (ev) 1217 if (ev)
783 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1218 XPUSH_CLOBJ ("OpenCL::Event", ev);
784 1219
785void 1220void
786enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1221write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1222 ALIAS:
1223 enqueue_write_buffer = 0
787 PPCODE: 1224 PPCODE:
788 cl_event ev = 0; 1225 cl_event ev = 0;
789 STRLEN len; 1226 STRLEN len;
790 char *ptr = SvPVbyte (data, len); 1227 char *ptr = SvPVbyte (data, len);
791 EVENT_LIST (5, items - 5); 1228 EVENT_LIST (5, items - 5);
792 1229
793 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1230 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
794 1231
795 if (ev) 1232 if (ev)
796 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1233 XPUSH_CLOBJ ("OpenCL::Event", ev);
797 1234
1235#if CL_VERSION_1_2
1236
798void 1237void
1238fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1239 ALIAS:
1240 enqueue_fill_buffer = 0
1241 PPCODE:
1242 cl_event ev = 0;
1243 STRLEN len;
1244 char *ptr = SvPVbyte (data, len);
1245 EVENT_LIST (5, items - 5);
1246
1247 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1248
1249 if (ev)
1250 XPUSH_CLOBJ ("OpenCL::Event", ev);
1251
1252void
1253fill_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, ...)
1254 ALIAS:
1255 enqueue_fill_image = 0
1256 PPCODE:
1257 cl_event ev = 0;
1258 STRLEN len;
1259 const size_t origin [3] = { x, y, z };
1260 const size_t region [3] = { width, height, depth };
1261 EVENT_LIST (12, items - 12);
1262
1263 const cl_float c_f [4] = { r, g, b, a };
1264 const cl_uint c_u [4] = { r, g, b, a };
1265 const cl_int c_s [4] = { r, g, b, a };
1266 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1267 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1268 cl_image_format format;
1269 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1270 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1271 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1272 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1273
1274 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1275 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1276
1277 if (ev)
1278 XPUSH_CLOBJ ("OpenCL::Event", ev);
1279
1280#endif
1281
1282void
799enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1283copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1284 ALIAS:
1285 enqueue_copy_buffer = 0
800 PPCODE: 1286 PPCODE:
801 cl_event ev = 0; 1287 cl_event ev = 0;
802 EVENT_LIST (6, items - 6); 1288 EVENT_LIST (6, items - 6);
803 1289
804 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1290 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
805 1291
806 if (ev) 1292 if (ev)
807 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1293 XPUSH_CLOBJ ("OpenCL::Event", ev);
808 1294
809void 1295void
810enqueue_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, ...) 1296read_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, ...)
1297 ALIAS:
1298 enqueue_read_buffer_rect = 0
811 PPCODE: 1299 PPCODE:
812 cl_event ev = 0; 1300 cl_event ev = 0;
813 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1301 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
814 const size_t host_origin[3] = { host_x, host_y, host_z }; 1302 const size_t host_origin[3] = { host_x, host_y, host_z };
815 const size_t region[3] = { width, height, depth }; 1303 const size_t region[3] = { width, height, depth };
834 SvPOK_only (data); 1322 SvPOK_only (data);
835 SvCUR_set (data, len); 1323 SvCUR_set (data, len);
836 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)); 1324 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));
837 1325
838 if (ev) 1326 if (ev)
839 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1327 XPUSH_CLOBJ ("OpenCL::Event", ev);
840 1328
841void 1329void
842enqueue_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, ...) 1330write_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, ...)
1331 ALIAS:
1332 enqueue_write_buffer_rect = 0
843 PPCODE: 1333 PPCODE:
844 cl_event ev = 0; 1334 cl_event ev = 0;
845 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1335 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
846 const size_t host_origin[3] = { host_x, host_y, host_z }; 1336 const size_t host_origin[3] = { host_x, host_y, host_z };
847 const size_t region[3] = { width, height, depth }; 1337 const size_t region[3] = { width, height, depth };
864 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1354 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
865 1355
866 if (len < min_len) 1356 if (len < min_len)
867 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1357 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
868 1358
869 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)); 1359 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));
870 1360
871 if (ev) 1361 if (ev)
872 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1362 XPUSH_CLOBJ ("OpenCL::Event", ev);
873 1363
874void 1364void
875enqueue_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, ...) 1365copy_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, ...)
1366 ALIAS:
1367 enqueue_copy_buffer_rect = 0
876 PPCODE: 1368 PPCODE:
877 cl_event ev = 0; 1369 cl_event ev = 0;
878 const size_t src_origin[3] = { src_x, src_y, src_z }; 1370 const size_t src_origin[3] = { src_x, src_y, src_z };
879 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
880 const size_t region[3] = { width, height, depth }; 1372 const size_t region[3] = { width, height, depth };
881 EVENT_LIST (16, items - 16); 1373 EVENT_LIST (16, items - 16);
882 1374
883 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)); 1375 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));
884 1376
885 if (ev) 1377 if (ev)
886 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1378 XPUSH_CLOBJ ("OpenCL::Event", ev);
887 1379
888void 1380void
889enqueue_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, ...) 1381read_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, ...)
1382 ALIAS:
1383 enqueue_read_image = 0
890 PPCODE: 1384 PPCODE:
891 cl_event ev = 0; 1385 cl_event ev = 0;
892 const size_t src_origin[3] = { src_x, src_y, src_z }; 1386 const size_t src_origin[3] = { src_x, src_y, src_z };
893 const size_t region[3] = { width, height, depth }; 1387 const size_t region[3] = { width, height, depth };
894 EVENT_LIST (12, items - 12); 1388 EVENT_LIST (12, items - 12);
906 SvPOK_only (data); 1400 SvPOK_only (data);
907 SvCUR_set (data, len); 1401 SvCUR_set (data, len);
908 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)); 1402 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));
909 1403
910 if (ev) 1404 if (ev)
911 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1405 XPUSH_CLOBJ ("OpenCL::Event", ev);
912 1406
913void 1407void
914enqueue_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, ...) 1408write_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, ...)
1409 ALIAS:
1410 enqueue_write_image = 0
915 PPCODE: 1411 PPCODE:
916 cl_event ev = 0; 1412 cl_event ev = 0;
917 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1413 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
918 const size_t region[3] = { width, height, depth }; 1414 const size_t region[3] = { width, height, depth };
919 STRLEN len; 1415 STRLEN len;
929 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1425 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
930 1426
931 if (len < min_len) 1427 if (len < min_len)
932 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1428 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
933 1429
934 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)); 1430 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));
935 1431
936 if (ev) 1432 if (ev)
937 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1433 XPUSH_CLOBJ ("OpenCL::Event", ev);
938 1434
939void 1435void
940enqueue_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, ...) 1436copy_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, ...)
1437 ALIAS:
1438 enqueue_copy_image = 0
941 PPCODE: 1439 PPCODE:
942 cl_event ev = 0; 1440 cl_event ev = 0;
943 const size_t src_origin[3] = { src_x, src_y, src_z }; 1441 const size_t src_origin[3] = { src_x, src_y, src_z };
944 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1442 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
945 const size_t region[3] = { width, height, depth }; 1443 const size_t region[3] = { width, height, depth };
946 EVENT_LIST (12, items - 12); 1444 EVENT_LIST (12, items - 12);
947 1445
948 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1446 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
949 1447
950 if (ev) 1448 if (ev)
951 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1449 XPUSH_CLOBJ ("OpenCL::Event", ev);
952 1450
953void 1451void
954enqueue_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, ...) 1452copy_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, ...)
1453 ALIAS:
1454 enqueue_copy_image_to_buffer = 0
955 PPCODE: 1455 PPCODE:
956 cl_event ev = 0; 1456 cl_event ev = 0;
957 const size_t src_origin[3] = { src_x, src_y, src_z }; 1457 const size_t src_origin[3] = { src_x, src_y, src_z };
958 const size_t region[3] = { width, height, depth }; 1458 const size_t region[3] = { width, height, depth };
959 EVENT_LIST (10, items - 10); 1459 EVENT_LIST (10, items - 10);
960 1460
961 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1461 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
962 1462
963 if (ev) 1463 if (ev)
964 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
965 1465
966void 1466void
967enqueue_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, ...) 1467copy_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, ...)
1468 ALIAS:
1469 enqueue_copy_buffer_to_image = 0
968 PPCODE: 1470 PPCODE:
969 cl_event ev = 0; 1471 cl_event ev = 0;
970 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1472 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
971 const size_t region[3] = { width, height, depth }; 1473 const size_t region[3] = { width, height, depth };
972 EVENT_LIST (10, items - 10); 1474 EVENT_LIST (10, items - 10);
973 1475
974 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1476 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
975 1477
976 if (ev) 1478 if (ev)
977 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1479 XPUSH_CLOBJ ("OpenCL::Event", ev);
978 1480
979void 1481void
980enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1482task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1483 ALIAS:
1484 enqueue_task = 0
981 PPCODE: 1485 PPCODE:
982 cl_event ev = 0; 1486 cl_event ev = 0;
983 EVENT_LIST (2, items - 2); 1487 EVENT_LIST (2, items - 2);
984 1488
985 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1489 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
986 1490
987 if (ev) 1491 if (ev)
988 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1492 XPUSH_CLOBJ ("OpenCL::Event", ev);
989 1493
990void 1494void
991enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1495nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1496 ALIAS:
1497 enqueue_nd_range_kernel = 0
992 PPCODE: 1498 PPCODE:
993 cl_event ev = 0; 1499 cl_event ev = 0;
994 size_t *gwo = 0, *gws, *lws = 0; 1500 size_t *gwo = 0, *gws, *lws = 0;
995 int gws_len; 1501 int gws_len;
996 size_t *lists; 1502 size_t *lists;
1004 1510
1005 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1511 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1006 1512
1007 gws = lists + gws_len * 0; 1513 gws = lists + gws_len * 0;
1008 for (i = 0; i < gws_len; ++i) 1514 for (i = 0; i < gws_len; ++i)
1515 {
1009 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1516 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1517 // at least nvidia crashes for 0-sized work group sizes, work around
1518 if (!gws [i])
1519 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1520 }
1010 1521
1011 if (SvOK (global_work_offset)) 1522 if (SvOK (global_work_offset))
1012 { 1523 {
1013 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1524 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1014 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1525 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1021 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1532 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1022 } 1533 }
1023 1534
1024 if (SvOK (local_work_size)) 1535 if (SvOK (local_work_size))
1025 { 1536 {
1026 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1537 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1027 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1538 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1028 1539
1029 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1540 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1030 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1541 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1031 1542
1032 lws = lists + gws_len * 2; 1543 lws = lists + gws_len * 2;
1033 for (i = 0; i < gws_len; ++i) 1544 for (i = 0; i < gws_len; ++i)
1545 {
1034 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1546 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1547 // at least nvidia crashes for 0-sized work group sizes, work around
1548 if (!lws [i])
1549 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1550 }
1035 } 1551 }
1036 1552
1037 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1553 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1038 1554
1039 if (ev) 1555 if (ev)
1040 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1556 XPUSH_CLOBJ ("OpenCL::Event", ev);
1041 1557
1042#if 0
1043#if cl_apple_gl_sharing || cl_khr_gl_sharing 1558#if cl_apple_gl_sharing || cl_khr_gl_sharing
1044 1559
1045void 1560void
1046enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1561acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1562 ALIAS:
1563 enqueue_acquire_gl_objects = 0
1564 ALIAS:
1565 enqueue_release_gl_objects = 1
1047 CODE: 1566 PPCODE:
1567 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1568 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1569
1048 cl_event ev = 0; 1570 cl_event ev = 0;
1049 EVENT_LIST (2, items - 2); 1571 EVENT_LIST (2, items - 2);
1572 AV *av = (AV *)SvRV (objects);
1573 cl_uint num_objects = av_len (av) + 1;
1574 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1575 int i;
1050 1576
1577 for (i = num_objects; i--; )
1578 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1579
1580 if (ix)
1581 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1582 else
1051 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, ..., event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1583 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1052 1584
1053 if (ev) 1585 if (ev)
1054 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1586 XPUSH_CLOBJ ("OpenCL::Event", ev);
1055 1587
1588#endif
1589
1056void 1590void
1057enqueue_release_gl_objects (OpenCL::Queue self, SV *objects, ...) 1591wait_for_events (OpenCL::Queue self, ...)
1592 ALIAS:
1593 enqueue_wait_for_events = 0
1594 CODE:
1595 EVENT_LIST (1, items - 1);
1596#if PREFER_1_1
1597 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1598#else
1599 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1600#endif
1601
1602void
1603marker (OpenCL::Queue self, ...)
1604 ALIAS:
1605 enqueue_marker = 0
1058 CODE: 1606 PPCODE:
1059 cl_event ev = 0; 1607 cl_event ev = 0;
1060 EVENT_LIST (2, items - 2); 1608 EVENT_LIST (1, items - 1);
1061 1609#if PREFER_1_1
1610 if (!event_list_count)
1611 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1612 else
1613#if CL_VERSION_1_2
1614 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1615#else
1616 {
1617 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1618 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1619 }
1620#endif
1621#else
1062 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, ..., event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1622 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1063 1623#endif
1064 if (ev) 1624 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1625 XPUSH_CLOBJ ("OpenCL::Event", ev);
1066 1626
1627void
1628barrier (OpenCL::Queue self, ...)
1629 ALIAS:
1630 enqueue_barrier = 0
1631 PPCODE:
1632 cl_event ev = 0;
1633 EVENT_LIST (1, items - 1);
1634#if PREFER_1_1
1635 if (!event_list_count && GIMME_V == G_VOID)
1636 NEED_SUCCESS (EnqueueBarrier, (self));
1637 else
1638#if CL_VERSION_1_2
1639 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1640#else
1641 {
1642 if (event_list_count)
1643 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1644
1645 if (GIMME_V != G_VOID)
1646 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1647 }
1067#endif 1648#endif
1649#else
1650 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1068#endif 1651#endif
1069 1652 if (ev)
1070void
1071enqueue_marker (OpenCL::Queue self)
1072 PPCODE:
1073 cl_event ev;
1074 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1075 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1653 XPUSH_CLOBJ ("OpenCL::Event", ev);
1076
1077void
1078enqueue_wait_for_events (OpenCL::Queue self, ...)
1079 CODE:
1080 EVENT_LIST (1, items - 1);
1081 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1082
1083void
1084enqueue_barrier (OpenCL::Queue self)
1085 CODE:
1086 NEED_SUCCESS (EnqueueBarrier, (self));
1087 1654
1088void 1655void
1089flush (OpenCL::Queue self) 1656flush (OpenCL::Queue self)
1090 CODE: 1657 CODE:
1091 NEED_SUCCESS (Flush, (self)); 1658 NEED_SUCCESS (Flush, (self));
1109 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1110 EXTEND (SP, 1); 1677 EXTEND (SP, 1);
1111 const int i = 0; 1678 const int i = 0;
1112 { 1679 {
1113 NEED_SUCCESS (RetainContext, (value [i])); 1680 NEED_SUCCESS (RetainContext, (value [i]));
1114 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1681 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1115 } 1682 }
1116 1683
1117void 1684void
1118device (OpenCL::Queue self) 1685device (OpenCL::Queue self)
1119 PPCODE: 1686 PPCODE:
1120 cl_device_id value [1]; 1687 cl_device_id value [1];
1121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1688 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1122 EXTEND (SP, 1); 1689 EXTEND (SP, 1);
1123 const int i = 0; 1690 const int i = 0;
1124 { 1691 {
1125 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1692 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1126 } 1693 }
1127 1694
1128void 1695void
1129reference_count (OpenCL::Queue self) 1696reference_count (OpenCL::Queue self)
1130 PPCODE: 1697 PPCODE:
1217 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1784 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1218 EXTEND (SP, 1); 1785 EXTEND (SP, 1);
1219 const int i = 0; 1786 const int i = 0;
1220 { 1787 {
1221 NEED_SUCCESS (RetainContext, (value [i])); 1788 NEED_SUCCESS (RetainContext, (value [i]));
1222 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1789 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1223 } 1790 }
1224 1791
1225void 1792void
1226associated_memobject (OpenCL::Memory self) 1793associated_memobject (OpenCL::Memory self)
1227 PPCODE: 1794 PPCODE:
1229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1796 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1230 EXTEND (SP, 1); 1797 EXTEND (SP, 1);
1231 const int i = 0; 1798 const int i = 0;
1232 { 1799 {
1233 NEED_SUCCESS (RetainMemObject, (value [i])); 1800 NEED_SUCCESS (RetainMemObject, (value [i]));
1234 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 1801 PUSH_CLOBJ ("OpenCL::Memory", value [i]);
1235 } 1802 }
1236 1803
1237#END:mem 1804#END:mem
1805
1806#if cl_apple_gl_sharing || cl_khr_gl_sharing
1807
1808void
1809gl_object_info (OpenCL::Memory self)
1810 PPCODE:
1811 cl_gl_object_type type;
1812 cl_GLuint name;
1813 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
1814 EXTEND (SP, 2);
1815 PUSHs (sv_2mortal (newSVuv (type)));
1816 PUSHs (sv_2mortal (newSVuv (name)));
1817
1818#endif
1238 1819
1239MODULE = OpenCL PACKAGE = OpenCL::BufferObj 1820MODULE = OpenCL PACKAGE = OpenCL::BufferObj
1240 1821
1241void 1822void
1242sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size) 1823sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
1245 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 1826 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1246 1827
1247 cl_buffer_region crdata = { origin, size }; 1828 cl_buffer_region crdata = { origin, size };
1248 1829
1249 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 1830 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1250 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1831 XPUSH_CLOBJ ("OpenCL::Buffer", mem);
1251 1832
1252MODULE = OpenCL PACKAGE = OpenCL::Image 1833MODULE = OpenCL PACKAGE = OpenCL::Image
1253 1834
1254void 1835void
1255image_info (OpenCL::Image self, cl_image_info name) 1836image_info (OpenCL::Image self, cl_image_info name)
1256 PPCODE: 1837 PPCODE:
1257 INFO (Image) 1838 INFO (Image)
1839
1840void
1841format (OpenCL::Image self)
1842 PPCODE:
1843 cl_image_format format;
1844 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1845 EXTEND (SP, 2);
1846 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
1847 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1258 1848
1259#BEGIN:image 1849#BEGIN:image
1260 1850
1261void 1851void
1262element_size (OpenCL::Image self) 1852element_size (OpenCL::Image self)
1274 const int i = 0; 1864 const int i = 0;
1275 PUSHs (sv_2mortal (newSVuv (value [i]))); 1865 PUSHs (sv_2mortal (newSVuv (value [i])));
1276 1866
1277#END:image 1867#END:image
1278 1868
1869#if cl_apple_gl_sharing || cl_khr_gl_sharing
1870
1871#BEGIN:gl_texture
1872
1873void
1874target (OpenCL::Image self)
1875 PPCODE:
1876 cl_GLenum value [1];
1877 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
1878 EXTEND (SP, 1);
1879 const int i = 0;
1880 PUSHs (sv_2mortal (newSVuv (value [i])));
1881
1882void
1883gl_mipmap_level (OpenCL::Image self)
1884 PPCODE:
1885 cl_GLint value [1];
1886 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
1887 EXTEND (SP, 1);
1888 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i])));
1890
1891#END:gl_texture
1892
1893#endif
1894
1279MODULE = OpenCL PACKAGE = OpenCL::Sampler 1895MODULE = OpenCL PACKAGE = OpenCL::Sampler
1280 1896
1281void 1897void
1282DESTROY (OpenCL::Sampler self) 1898DESTROY (OpenCL::Sampler self)
1283 CODE: 1899 CODE:
1306 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 1922 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1307 EXTEND (SP, 1); 1923 EXTEND (SP, 1);
1308 const int i = 0; 1924 const int i = 0;
1309 { 1925 {
1310 NEED_SUCCESS (RetainContext, (value [i])); 1926 NEED_SUCCESS (RetainContext, (value [i]));
1311 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1927 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1312 } 1928 }
1313 1929
1314void 1930void
1315normalized_coords (OpenCL::Sampler self) 1931normalized_coords (OpenCL::Sampler self)
1316 PPCODE: 1932 PPCODE:
1346DESTROY (OpenCL::Program self) 1962DESTROY (OpenCL::Program self)
1347 CODE: 1963 CODE:
1348 clReleaseProgram (self); 1964 clReleaseProgram (self);
1349 1965
1350void 1966void
1351build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 1967build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1968 ALIAS:
1969 build_async = 1
1352 CODE: 1970 CODE:
1971 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1972 void *user_data = 0;
1973 cl_uint num_devices = 0;
1974 cl_device_id *device_list = 0;
1975
1976 if (SvOK (devices))
1977 {
1978 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1979 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
1980
1981 AV *av = (AV *)SvRV (devices);
1982 num_devices = av_len (av) + 1;
1983
1984 if (num_devices)
1985 {
1986 device_list = tmpbuf (sizeof (*device_list) * num_devices);
1987 int count;
1988 for (count = 0; count < num_devices; ++count)
1989 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1990 }
1991 }
1992
1993 if (SvOK (notify))
1994 {
1995 NEED_SUCCESS (RetainProgram, (self));
1996 pfn_notify = eq_program_notify;
1997 user_data = SvREFCNT_inc (s_get_cv (notify));
1998 }
1999
2000 if (ix)
2001 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2002 else
1353 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2003 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1354 2004
1355void 2005void
1356build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2006build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1357 PPCODE: 2007 PPCODE:
1358 size_t size; 2008 size_t size;
1393 2043
1394void 2044void
1395kernel (OpenCL::Program program, SV *function) 2045kernel (OpenCL::Program program, SV *function)
1396 PPCODE: 2046 PPCODE:
1397 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2047 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1398 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2048 XPUSH_CLOBJ ("OpenCL::Kernel", kernel);
2049
2050void
2051kernels_in_program (OpenCL::Program program)
2052 PPCODE:
2053 cl_uint num_kernels;
2054 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2055 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2056 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2057
2058 int i;
2059 EXTEND (SP, num_kernels);
2060 for (i = 0; i < num_kernels; ++i)
2061 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]);
1399 2062
1400void 2063void
1401info (OpenCL::Program self, cl_program_info name) 2064info (OpenCL::Program self, cl_program_info name)
1402 PPCODE: 2065 PPCODE:
1403 INFO (Program) 2066 INFO (Program)
1421 { 2084 {
1422 SV *sv = sv_2mortal (newSV (sizes [i])); 2085 SV *sv = sv_2mortal (newSV (sizes [i]));
1423 SvUPGRADE (sv, SVt_PV); 2086 SvUPGRADE (sv, SVt_PV);
1424 SvPOK_only (sv); 2087 SvPOK_only (sv);
1425 SvCUR_set (sv, sizes [i]); 2088 SvCUR_set (sv, sizes [i]);
1426 ptrs [i] = SvPVX (sv); 2089 ptrs [i] = (void *)SvPVX (sv);
1427 PUSHs (sv); 2090 PUSHs (sv);
1428 } 2091 }
1429 2092
1430 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2093 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1431 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2094 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1451 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2114 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1452 EXTEND (SP, 1); 2115 EXTEND (SP, 1);
1453 const int i = 0; 2116 const int i = 0;
1454 { 2117 {
1455 NEED_SUCCESS (RetainContext, (value [i])); 2118 NEED_SUCCESS (RetainContext, (value [i]));
1456 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2119 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1457 } 2120 }
1458 2121
1459void 2122void
1460devices (OpenCL::Program self) 2123devices (OpenCL::Program self)
1461 PPCODE: 2124 PPCODE:
1465 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2128 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1466 int i, n = size / sizeof (*value); 2129 int i, n = size / sizeof (*value);
1467 EXTEND (SP, n); 2130 EXTEND (SP, n);
1468 for (i = 0; i < n; ++i) 2131 for (i = 0; i < n; ++i)
1469 { 2132 {
1470 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2133 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1471 } 2134 }
1472 2135
1473void 2136void
1474source (OpenCL::Program self) 2137source (OpenCL::Program self)
1475 PPCODE: 2138 PPCODE:
1501DESTROY (OpenCL::Kernel self) 2164DESTROY (OpenCL::Kernel self)
1502 CODE: 2165 CODE:
1503 clReleaseKernel (self); 2166 clReleaseKernel (self);
1504 2167
1505void 2168void
2169setf (OpenCL::Kernel self, const char *format, ...)
2170 CODE:
2171 int i;
2172 for (i = 2; ; ++i)
2173 {
2174 while (*format == ' ')
2175 ++format;
2176
2177 char type = *format++;
2178
2179 if (!type)
2180 break;
2181
2182 if (i >= items)
2183 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2184
2185 SV *sv = ST (i);
2186
2187 union
2188 {
2189 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2190 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2191 cl_half ch; cl_float cf; cl_double cd;
2192 cl_mem cm;
2193 cl_sampler ca;
2194 size_t cz;
2195 cl_event ce;
2196 } arg;
2197 size_t size;
2198 int nullarg = 0;
2199
2200 switch (type)
2201 {
2202 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2203 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2204 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2205 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2206 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2207 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2208 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2209 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2210
2211 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2212 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2213 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2214
2215 case 'z': nullarg = 1; size = SvIV (sv); break;
2216
2217 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2218 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2219 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2220
2221 default:
2222 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2223 }
2224
2225 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2226 if (res)
2227 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2228 }
2229
2230 if (i != items)
2231 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2232
2233void
1506set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2234set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1507 CODE: 2235 CODE:
1508 clSetKernelArg (self, idx, sizeof (value), &value); 2236 clSetKernelArg (self, idx, sizeof (value), &value);
1509 2237
1510void 2238void
1558 clSetKernelArg (self, idx, sizeof (value), &value); 2286 clSetKernelArg (self, idx, sizeof (value), &value);
1559 2287
1560void 2288void
1561set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2289set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1562 CODE: 2290 CODE:
2291 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2292
2293void
2294set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2295 CODE:
2296 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2297
2298void
2299set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2300 ALIAS:
2301 set_image2d = 0
2302 set_image3d = 0
2303 CODE:
2304 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2305
2306void
2307set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2308 CODE:
1563 clSetKernelArg (self, idx, sizeof (value), &value); 2309 clSetKernelArg (self, idx, sizeof (value), &value);
1564 2310
1565void 2311void
1566set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2312set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1567 CODE: 2313 CODE:
1568 clSetKernelArg (self, idx, sizeof (value), &value); 2314 clSetKernelArg (self, idx, size, 0);
1569
1570void
1571set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1572 CODE:
1573 clSetKernelArg (self, idx, sizeof (value), &value);
1574
1575void
1576set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1577 CODE:
1578 clSetKernelArg (self, idx, sizeof (value), &value);
1579
1580void
1581set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1582 CODE:
1583 clSetKernelArg (self, idx, sizeof (value), &value);
1584 2315
1585void 2316void
1586set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2317set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1587 CODE: 2318 CODE:
1588 clSetKernelArg (self, idx, sizeof (value), &value); 2319 clSetKernelArg (self, idx, sizeof (value), &value);
1624 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2355 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1625 EXTEND (SP, 1); 2356 EXTEND (SP, 1);
1626 const int i = 0; 2357 const int i = 0;
1627 { 2358 {
1628 NEED_SUCCESS (RetainContext, (value [i])); 2359 NEED_SUCCESS (RetainContext, (value [i]));
1629 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2360 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1630 } 2361 }
1631 2362
1632void 2363void
1633program (OpenCL::Kernel self) 2364program (OpenCL::Kernel self)
1634 PPCODE: 2365 PPCODE:
1636 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2367 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1637 EXTEND (SP, 1); 2368 EXTEND (SP, 1);
1638 const int i = 0; 2369 const int i = 0;
1639 { 2370 {
1640 NEED_SUCCESS (RetainProgram, (value [i])); 2371 NEED_SUCCESS (RetainProgram, (value [i]));
1641 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2372 PUSH_CLOBJ ("OpenCL::Program", value [i]);
1642 } 2373 }
1643 2374
1644#END:kernel 2375#END:kernel
1645 2376
1646void 2377void
1706wait (OpenCL::Event self) 2437wait (OpenCL::Event self)
1707 CODE: 2438 CODE:
1708 clWaitForEvents (1, &self); 2439 clWaitForEvents (1, &self);
1709 2440
1710void 2441void
2442cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2443 CODE:
2444 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2445
2446void
1711info (OpenCL::Event self, cl_event_info name) 2447info (OpenCL::Event self, cl_event_info name)
1712 PPCODE: 2448 PPCODE:
1713 INFO (Event) 2449 INFO (Event)
1714 2450
1715#BEGIN:event 2451#BEGIN:event
1721 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2457 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1722 EXTEND (SP, 1); 2458 EXTEND (SP, 1);
1723 const int i = 0; 2459 const int i = 0;
1724 { 2460 {
1725 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2461 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1726 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2462 PUSH_CLOBJ ("OpenCL::Queue", value [i]);
1727 } 2463 }
1728 2464
1729void 2465void
1730command_type (OpenCL::Event self) 2466command_type (OpenCL::Event self)
1731 PPCODE: 2467 PPCODE:
1754 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2490 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1755 EXTEND (SP, 1); 2491 EXTEND (SP, 1);
1756 const int i = 0; 2492 const int i = 0;
1757 { 2493 {
1758 NEED_SUCCESS (RetainContext, (value [i])); 2494 NEED_SUCCESS (RetainContext, (value [i]));
1759 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2495 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1760 } 2496 }
1761 2497
1762#END:event 2498#END:event
1763 2499
1764void 2500void

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines