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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.6 by root, Thu Nov 17 01:36:52 2011 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 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
29
30#ifdef __APPLE__
31 #include <OpenCL/opencl.h>
32#else
5#include <CL/opencl.h> 33 #include <CL/opencl.h>
34#endif
6 35
36#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1
38 #define PREFER_1_1 1
39#endif
40
7typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
8typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
9typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
10typedef cl_command_queue OpenCL__Queue; 44typedef cl_command_queue OpenCL__Queue;
11typedef cl_mem OpenCL__Memory; 45typedef cl_mem OpenCL__Memory;
12typedef cl_mem OpenCL__Buffer; 46typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj;
13typedef cl_mem OpenCL__Image; 48typedef cl_mem OpenCL__Image;
14typedef cl_mem OpenCL__Image2D; 49typedef cl_mem OpenCL__Image2D;
15typedef cl_mem OpenCL__Image3D; 50typedef cl_mem OpenCL__Image3D;
16typedef cl_mem OpenCL__Memory_ornull; 51typedef cl_mem OpenCL__Memory_ornull;
17typedef cl_mem OpenCL__Buffer_ornull; 52typedef cl_mem OpenCL__Buffer_ornull;
18typedef cl_mem OpenCL__Image_ornull; 53typedef cl_mem OpenCL__Image_ornull;
19typedef cl_mem OpenCL__Image2D_ornull; 54typedef cl_mem OpenCL__Image2D_ornull;
20typedef cl_mem OpenCL__Image3D_ornull; 55typedef cl_mem OpenCL__Image3D_ornull;
21typedef cl_sampler OpenCL__Sampler; 56typedef cl_sampler OpenCL__Sampler;
22typedef cl_program OpenCL__Program; 57typedef cl_program OpenCL__Program;
23typedef cl_kernel OpenCL__Kernel; 58typedef cl_kernel OpenCL__Kernel;
24typedef cl_event OpenCL__Event; 59typedef cl_event OpenCL__Event;
25typedef cl_event OpenCL__UserEvent; 60typedef cl_event OpenCL__UserEvent;
61
62typedef SV *FUTURE;
63
64/*****************************************************************************/
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}
26 108
27/*****************************************************************************/ 109/*****************************************************************************/
28 110
29/* up to two temporary buffers */ 111/* up to two temporary buffers */
30static void * 112static void *
31tmpbuf (size_t size) 113tmpbuf (size_t size)
32{ 114{
115 enum { buffers = 3 };
33 static int idx; 116 static int idx;
34 static void *buf [2]; 117 static void *buf [buffers];
35 static size_t len [2]; 118 static size_t len [buffers];
36 119
37 idx ^= 1; 120 idx = (idx + 1) % buffers;
38 121
39 if (len [idx] < size) 122 if (len [idx] < size)
40 { 123 {
41 free (buf [idx]); 124 free (buf [idx]);
42 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 125 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
90 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)"); 173 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)");
91} 174}
92 175
93/*****************************************************************************/ 176/*****************************************************************************/
94 177
95static cl_int last_error; 178static cl_int res;
96 179
97#define FAIL(name,err) \ 180#define FAIL(name) \
98 croak ("cl" # name ": %s", err2str (last_error = err)); 181 croak ("cl" # name ": %s", err2str (res));
99 182
100#define NEED_SUCCESS(name,args) \ 183#define NEED_SUCCESS(name,args) \
101 do { \ 184 do { \
102 cl_int res = cl ## name args; \ 185 res = cl ## name args; \
103 \ 186 \
104 if (res) \ 187 if (res) \
105 FAIL (name, res); \ 188 FAIL (name); \
106 } while (0) 189 } while (0)
107 190
191#define NEED_SUCCESS_ARG(retdecl, name, args) \
192 retdecl = cl ## name args; \
193 if (res) \
194 FAIL (name);
195
108/*****************************************************************************/ 196/*****************************************************************************/
109 197
198static cl_context_properties *
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{
201 if (!sv || !SvOK (sv))
202 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else
205 return 0;
206
207 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
208 {
209 AV *av = (AV *)SvRV (sv);
210 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p;
213
214 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname);
216
217 while (extracount--)
218 *l++ = *extra++;
219
220 for (i = 0; i < len; i += 2)
221 {
222 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
223 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override
225
226 switch (t)
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
246 default:
247 /* unknown property, treat as int */
248 break;
249 }
250
251 *l++ = t;
252 *l++ = v;
253 }
254
255 *l = 0;
256
257 return p;
258 }
259
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
261}
262
263/*****************************************************************************/
264
110#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr))
111#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr)) 267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
112 268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
113static void * 271static void *
114SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
115{ 273{
116 if (SvROK (sv) && sv_derived_from (sv, pkg)) 274 if (SvROK (sv) && sv_derived_from (sv, pkg))
117 return (void *)SvIV (SvRV (sv)); 275 return (void *)SvIV (SvRV (sv));
118 276
119 croak ("%s: %s is not of type %s", func, svname, pkg); 277 croak ("%s: %s is not of type %s", func, svname, pkg);
120} 278}
121 279
122/*****************************************************************************/ 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
511static size_t
512img_row_pitch (cl_mem img)
513{
514 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res;
517}
123 518
124static cl_event * 519static cl_event *
125event_list (SV **items, int count) 520event_list (SV **items, cl_uint *rcount)
126{ 521{
522 cl_uint count = *rcount;
523
524 if (!count)
525 return 0;
526
127 cl_event *list = tmpbuf (sizeof (cl_event) * count); 527 cl_event *list = tmpbuf (sizeof (cl_event) * count);
528 int i = 0;
128 529
530 do
531 {
532 --count;
533 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
535 }
129 while (count--) 536 while (count);
130 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
131 537
538 *rcount = i;
539
132 return list; 540 return i ? list : 0;
133} 541}
134 542
135#define EVENT_LIST(items,count) \ 543#define EVENT_LIST(items,count) \
136 cl_uint event_list_count = (count); \ 544 cl_uint event_list_count = (count); \
137 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)
138 546
139#define INFO(class) \ 547#define INFO(class) \
140{ \ 548{ \
141 size_t size; \ 549 size_t size; \
142 SV *sv; \
143 \
144 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
145 sv = sv_2mortal (newSV (size)); \ 551 SV *sv = sv_2mortal (newSV (size)); \
146 SvUPGRADE (sv, SVt_PV); \ 552 SvUPGRADE (sv, SVt_PV); \
147 SvPOK_only (sv); \ 553 SvPOK_only (sv); \
148 SvCUR_set (sv, size); \ 554 SvCUR_set (sv, size); \
149 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
150 XPUSHs (sv); \ 556 XPUSHs (sv); \
151} 557}
152 558
153MODULE = OpenCL PACKAGE = OpenCL 559MODULE = OpenCL PACKAGE = OpenCL
154 560
155PROTOTYPES: ENABLE 561PROTOTYPES: ENABLE
156 562
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;
573
157BOOT: 574BOOT:
158{ 575{
159 HV *stash = gv_stashpv ("OpenCL", 1); 576 HV *stash = gv_stashpv ("OpenCL", 1);
160 static const ivstr *civ, const_iv[] = { 577 static const ivstr *civ, const_iv[] = {
161 { sizeof (cl_char ), "SIZEOF_CHAR" }, 578 { sizeof (cl_char ), "SIZEOF_CHAR" },
162 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 579 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
163 { sizeof (cl_short ), "SIZEOF_SHORT" }, 580 { sizeof (cl_short ), "SIZEOF_SHORT" },
164 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 581 { sizeof (cl_ushort), "SIZEOF_USHORT" },
165 { sizeof (cl_int ), "SIZEOF_INT" }, 582 { sizeof (cl_int ), "SIZEOF_INT" },
166 { sizeof (cl_uint ), "SIZEOF_UINT" }, 583 { sizeof (cl_uint ), "SIZEOF_UINT" },
167 { sizeof (cl_long ), "SIZEOF_LONG" }, 584 { sizeof (cl_long ), "SIZEOF_LONG" },
168 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 585 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
169 { sizeof (cl_half ), "SIZEOF_HALF" }, 586 { sizeof (cl_half ), "SIZEOF_HALF" },
170 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 587 { sizeof (cl_float ), "SIZEOF_FLOAT" },
171 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 588 { sizeof (cl_double), "SIZEOF_DOUBLE" },
172#include "constiv.h" 589#include "constiv.h"
173 }; 590 };
591
174 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--)
175 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);
176} 596}
177 597
178cl_int 598cl_int
179errno () 599errno ()
180 CODE: 600 CODE:
181 errno = last_error; 601 RETVAL = res;
602 OUTPUT:
603 RETVAL
182 604
183const char * 605const char *
184err2str (cl_int err) 606err2str (cl_int err = res)
185 607
186const char * 608const char *
187enum2str (cl_uint value) 609enum2str (cl_uint value)
188 610
189void 611void
190platforms () 612platforms ()
191 PPCODE: 613 PPCODE:
192{
193 cl_platform_id *list; 614 cl_platform_id *list;
194 cl_uint count; 615 cl_uint count;
195 int i; 616 int i;
196 617
197 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count)); 618 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count));
198 list = tmpbuf (sizeof (*list) * count); 619 list = tmpbuf (sizeof (*list) * count);
199 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
200 621
201 EXTEND (SP, count); 622 EXTEND (SP, count);
202 for (i = 0; i < count; ++i) 623 for (i = 0; i < count; ++i)
203 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 624 PUSH_CLOBJ ("OpenCL::Platform", list [i]);
204}
205 625
206void 626void
207context_from_type_simple (cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
208 PPCODE: 628 PPCODE:
209{ 629 CONTEXT_NOTIFY_CALLBACK;
210 cl_int res; 630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
211 cl_context ctx = clCreateContextFromType (0, type, 0, 0, &res); 631 XPUSH_CLOBJ_CONTEXT;
212 632
213 if (res) 633void
214 FAIL (CreateContextFromType, res); 634context (FUTURE properties, FUTURE devices, FUTURE notify)
215 635 PPCODE:
216 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 636 /* der Gipfel der Kunst */
217}
218 637
219void 638void
220wait_for_events (...) 639wait_for_events (...)
221 CODE: 640 CODE:
222{
223 EVENT_LIST (0, items); 641 EVENT_LIST (0, items);
224 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
225}
226 643
227PROTOTYPES: DISABLE 644PROTOTYPES: DISABLE
228 645
229MODULE = OpenCL PACKAGE = OpenCL::Platform 646MODULE = OpenCL PACKAGE = OpenCL::Platform
230 647
231void 648void
232info (OpenCL::Platform this, cl_platform_info name) 649info (OpenCL::Platform self, cl_platform_info name)
233 PPCODE: 650 PPCODE:
234 INFO (Platform) 651 INFO (Platform)
235 652
236void 653void
654unload_compiler (OpenCL::Platform self)
655 CODE:
656#if CL_VERSION_1_2
657 clUnloadPlatformCompiler (self);
658#endif
659
660#BEGIN:platform
661
662void
663profile (OpenCL::Platform self)
664 ALIAS:
665 profile = CL_PLATFORM_PROFILE
666 version = CL_PLATFORM_VERSION
667 name = CL_PLATFORM_NAME
668 vendor = CL_PLATFORM_VENDOR
669 extensions = CL_PLATFORM_EXTENSIONS
670 PPCODE:
671 size_t size;
672 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
673 char *value = tmpbuf (size);
674 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
675 EXTEND (SP, 1);
676 const int i = 0;
677 PUSHs (sv_2mortal (newSVpv (value, 0)));
678
679#END:platform
680
681void
237devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 682devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
238 PPCODE: 683 PPCODE:
239{
240 cl_device_id *list; 684 cl_device_id *list;
241 cl_uint count; 685 cl_uint count;
242 int i; 686 int i;
243 687
244 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 688 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
245 list = tmpbuf (sizeof (*list) * count); 689 list = tmpbuf (sizeof (*list) * count);
246 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
247 691
248 EXTEND (SP, count); 692 EXTEND (SP, count);
249 for (i = 0; i < count; ++i) 693 for (i = 0; i < count; ++i)
250 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i]));
251}
252 695
253void 696void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
701
702 AV *av = (AV *)SvRV (devices);
703 cl_uint num_devices = av_len (av) + 1;
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
705
706 int i;
707 for (i = num_devices; i--; )
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
709
710 CONTEXT_NOTIFY_CALLBACK;
711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
712 XPUSH_CLOBJ_CONTEXT;
713
714void
254context_from_type_simple (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
255 PPCODE: 716 PPCODE:
256{
257 cl_int res;
258 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719
720 CONTEXT_NOTIFY_CALLBACK;
259 cl_context ctx = clCreateContextFromType (props, type, 0, 0, &res); 721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
260 722 XPUSH_CLOBJ_CONTEXT;
261 if (res)
262 FAIL (CreateContextFromType, res);
263
264 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
265}
266 723
267MODULE = OpenCL PACKAGE = OpenCL::Device 724MODULE = OpenCL PACKAGE = OpenCL::Device
268 725
269void 726void
270info (OpenCL::Device this, cl_device_info name) 727info (OpenCL::Device self, cl_device_info name)
271 PPCODE: 728 PPCODE:
272 INFO (Device) 729 INFO (Device)
273 730
731#BEGIN:device
732
274void 733void
275context_simple (OpenCL::Device this) 734type (OpenCL::Device self)
276 PPCODE: 735 PPCODE:
277{ 736 cl_device_type value [1];
278 cl_int res; 737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
279 cl_context ctx = clCreateContext (0, 1, &this, 0, 0, &res); 738 EXTEND (SP, 1);
739 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i])));
280 741
281 if (res) 742void
282 FAIL (CreateContext, res); 743vendor_id (OpenCL::Device self)
744 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID
746 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
747 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
748 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
749 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
750 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
751 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
752 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
753 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
754 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
755 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
756 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
757 image_support = CL_DEVICE_IMAGE_SUPPORT
758 max_samplers = CL_DEVICE_MAX_SAMPLERS
759 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
760 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
761 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
762 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
763 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
764 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
765 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
766 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
767 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
768 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
769 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
770 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
771 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
772 PPCODE:
773 cl_uint value [1];
774 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
775 EXTEND (SP, 1);
776 const int i = 0;
777 PUSHs (sv_2mortal (newSVuv (value [i])));
283 778
284 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 779void
285} 780max_work_group_size (OpenCL::Device self)
781 ALIAS:
782 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
783 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
784 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
785 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
786 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
787 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
788 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
789 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
790 PPCODE:
791 size_t value [1];
792 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
793 EXTEND (SP, 1);
794 const int i = 0;
795 PUSHs (sv_2mortal (newSVuv (value [i])));
796
797void
798max_work_item_sizes (OpenCL::Device self)
799 PPCODE:
800 size_t size;
801 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
802 size_t *value = tmpbuf (size);
803 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
804 int i, n = size / sizeof (*value);
805 EXTEND (SP, n);
806 for (i = 0; i < n; ++i)
807 PUSHs (sv_2mortal (newSVuv (value [i])));
808
809void
810address_bits (OpenCL::Device self)
811 PPCODE:
812 cl_bitfield value [1];
813 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
814 EXTEND (SP, 1);
815 const int i = 0;
816 PUSHs (sv_2mortal (newSVuv (value [i])));
817
818void
819max_mem_alloc_size (OpenCL::Device self)
820 ALIAS:
821 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
822 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
823 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
824 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
825 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
826 PPCODE:
827 cl_ulong value [1];
828 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
829 EXTEND (SP, 1);
830 const int i = 0;
831 PUSHs (sv_2mortal (newSVuv (value [i])));
832
833void
834single_fp_config (OpenCL::Device self)
835 ALIAS:
836 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
837 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
838 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
839 PPCODE:
840 cl_device_fp_config value [1];
841 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
842 EXTEND (SP, 1);
843 const int i = 0;
844 PUSHs (sv_2mortal (newSVuv (value [i])));
845
846void
847global_mem_cache_type (OpenCL::Device self)
848 PPCODE:
849 cl_device_mem_cache_type value [1];
850 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
851 EXTEND (SP, 1);
852 const int i = 0;
853 PUSHs (sv_2mortal (newSVuv (value [i])));
854
855void
856local_mem_type (OpenCL::Device self)
857 PPCODE:
858 cl_device_local_mem_type value [1];
859 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
860 EXTEND (SP, 1);
861 const int i = 0;
862 PUSHs (sv_2mortal (newSVuv (value [i])));
863
864void
865error_correction_support (OpenCL::Device self)
866 ALIAS:
867 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
868 endian_little = CL_DEVICE_ENDIAN_LITTLE
869 available = CL_DEVICE_AVAILABLE
870 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
871 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
872 PPCODE:
873 cl_bool value [1];
874 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
875 EXTEND (SP, 1);
876 const int i = 0;
877 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
878
879void
880execution_capabilities (OpenCL::Device self)
881 PPCODE:
882 cl_device_exec_capabilities value [1];
883 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
884 EXTEND (SP, 1);
885 const int i = 0;
886 PUSHs (sv_2mortal (newSVuv (value [i])));
887
888void
889properties (OpenCL::Device self)
890 PPCODE:
891 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1);
894 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i])));
896
897void
898platform (OpenCL::Device self)
899 PPCODE:
900 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1);
903 const int i = 0;
904 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]);
906 }
907
908void
909name (OpenCL::Device self)
910 ALIAS:
911 name = CL_DEVICE_NAME
912 vendor = CL_DEVICE_VENDOR
913 driver_version = CL_DRIVER_VERSION
914 profile = CL_DEVICE_PROFILE
915 version = CL_DEVICE_VERSION
916 extensions = CL_DEVICE_EXTENSIONS
917 PPCODE:
918 size_t size;
919 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
920 char *value = tmpbuf (size);
921 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
922 EXTEND (SP, 1);
923 const int i = 0;
924 PUSHs (sv_2mortal (newSVpv (value, 0)));
925
926void
927parent_device_ext (OpenCL::Device self)
928 PPCODE:
929 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1);
932 const int i = 0;
933 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]);
935 }
936
937void
938partition_types_ext (OpenCL::Device self)
939 ALIAS:
940 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
941 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
942 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
943 PPCODE:
944 size_t size;
945 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
946 cl_device_partition_property_ext *value = tmpbuf (size);
947 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
948 int i, n = size / sizeof (*value);
949 EXTEND (SP, n);
950 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i])));
952
953#END:device
286 954
287MODULE = OpenCL PACKAGE = OpenCL::Context 955MODULE = OpenCL PACKAGE = OpenCL::Context
288 956
289void 957void
290DESTROY (OpenCL::Context context) 958DESTROY (OpenCL::Context context)
291 CODE: 959 CODE:
292 clReleaseContext (context); 960 clReleaseContext (context);
293 961
294void 962void
295info (OpenCL::Context this, cl_context_info name) 963info (OpenCL::Context self, cl_context_info name)
296 PPCODE: 964 PPCODE:
297 INFO (Context) 965 INFO (Context)
298 966
299void 967void
300queue (OpenCL::Context this, OpenCL::Device device, cl_command_queue_properties properties) 968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
301 PPCODE: 969 PPCODE:
302{
303 cl_int res;
304 cl_command_queue queue = clCreateCommandQueue (this, device, properties, &res); 970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
305
306 if (res)
307 FAIL (CreateCommandQueue, res);
308
309 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 971 XPUSH_CLOBJ ("OpenCL::Queue", queue);
310}
311 972
312void 973void
313user_event (OpenCL::Context this) 974user_event (OpenCL::Context self)
314 PPCODE: 975 PPCODE:
315{ 976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
316 cl_int res;
317 cl_event ev = clCreateUserEvent (this, &res);
318
319 if (res)
320 FAIL (CreateUserevent, res);
321
322 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev);
323}
324 978
325void 979void
326buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
327 PPCODE: 981 PPCODE:
328{
329 cl_int res;
330 cl_mem mem;
331
332 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))
333 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?");
334 984
335 mem = clCreateBuffer (this, flags, len, 0, &res); 985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
336
337 if (res)
338 FAIL (CreateBuffer, res);
339
340 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
341}
342 987
343void 988void
344buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
345 PPCODE: 990 PPCODE:
346{
347 STRLEN len; 991 STRLEN len;
348 char *ptr = SvPVbyte (data, len); 992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
349 cl_int res;
350 cl_mem mem;
351
352 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)))
353 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?");
354 995 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
355 mem = clCreateBuffer (this, flags, len, ptr, &res);
356
357 if (res)
358 FAIL (CreateBuffer, res);
359
360 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem);
361}
362 997
998#if CL_VERSION_1_2
999
363void 1000void
364image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, SV *data) 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)
365 PPCODE: 1002 PPCODE:
366{
367 STRLEN len; 1003 STRLEN len;
368 char *ptr = SvPVbyte (data, len); 1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
369 const cl_image_format format = { channel_order, channel_type }; 1005 const cl_image_format format = { channel_order, channel_type };
370 cl_int res; 1006 const cl_image_desc desc = {
371 cl_mem mem = clCreateImage2D (this, flags, &format, width, height, len / height, ptr, &res); 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 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
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);
372 1025
373 if (res) 1026#endif
374 FAIL (CreateImage2D, res);
375 1027
376 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
377}
378
379void 1028void
380image3d (OpenCL::Context this, 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 slice_pitch, SV *data) 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)
381 PPCODE: 1030 PPCODE:
382{
383 STRLEN len; 1031 STRLEN len;
384 char *ptr = SvPVbyte (data, len); 1032 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
385 const cl_image_format format = { channel_order, channel_type }; 1033 const cl_image_format format = { channel_order, channel_type };
386 cl_int res; 1034#if PREFER_1_1
387 cl_mem mem = clCreateImage3D (this, flags, &format, width, height, 1035 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
388 depth, len / (height * slice_pitch), slice_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);
389 1041
390 if (res) 1042void
391 FAIL (CreateImage3D, res); 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)
392 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
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
393 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
394}
395 1055
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057
396void 1058void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
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
1092
1093void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1095 PPCODE:
1096#if PREFER_1_1
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
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem);
1102
1103void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE:
1106#if PREFER_1_1
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
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem);
1112
1113#endif
1114
1115void
397supported_image_formats (OpenCL::Context this, 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)
398 PPCODE: 1117 PPCODE:
399{ 1118{
400 cl_uint count; 1119 cl_uint count;
401 cl_image_format *list; 1120 cl_image_format *list;
402 int i; 1121 int i;
403 1122
404 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1123 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
405 Newx (list, count, cl_image_format); 1124 Newx (list, count, cl_image_format);
406 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1125 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
407 1126
408 EXTEND (SP, count); 1127 EXTEND (SP, count);
409 for (i = 0; i < count; ++i) 1128 for (i = 0; i < count; ++i)
410 { 1129 {
411 AV *av = newAV (); 1130 AV *av = newAV ();
414 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1133 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
415 } 1134 }
416} 1135}
417 1136
418void 1137void
419sampler (OpenCL::Context this, 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)
420 PPCODE: 1139 PPCODE:
421{
422 cl_int res;
423 cl_sampler sampler = clCreateSampler (this, normalized_coords, addressing_mode, filter_mode, &res); 1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
424
425 if (res)
426 FAIL (CreateSampler, res);
427
428 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler);
429}
430 1142
431void 1143void
432program_with_source (OpenCL::Context this, SV *program) 1144program_with_source (OpenCL::Context self, SV *program)
433 PPCODE: 1145 PPCODE:
434{
435 STRLEN len; 1146 STRLEN len;
436 size_t len2; 1147 size_t len2;
437 const char *ptr = SvPVbyte (program, len); 1148 const char *ptr = SvPVbyte (program, len);
438 cl_int res;
439 cl_program prog;
440 1149
441 len2 = len; 1150 len2 = len;
442 prog = clCreateProgramWithSource (this, 1, &ptr, &len2, &res); 1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
443
444 if (res)
445 FAIL (CreateProgramWithSource, res);
446
447 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1152 XPUSH_CLOBJ ("OpenCL::Program", prog);
448} 1153
1154#BEGIN:context
1155
1156void
1157reference_count (OpenCL::Context self)
1158 ALIAS:
1159 reference_count = CL_CONTEXT_REFERENCE_COUNT
1160 num_devices = CL_CONTEXT_NUM_DEVICES
1161 PPCODE:
1162 cl_uint value [1];
1163 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1164 EXTEND (SP, 1);
1165 const int i = 0;
1166 PUSHs (sv_2mortal (newSVuv (value [i])));
1167
1168void
1169devices (OpenCL::Context self)
1170 PPCODE:
1171 size_t size;
1172 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1173 cl_device_id *value = tmpbuf (size);
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i)
1178 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1180 }
1181
1182void
1183properties (OpenCL::Context self)
1184 PPCODE:
1185 size_t size;
1186 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1187 cl_context_properties *value = tmpbuf (size);
1188 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1189 int i, n = size / sizeof (*value);
1190 EXTEND (SP, n);
1191 for (i = 0; i < n; ++i)
1192 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1193
1194#END:context
449 1195
450MODULE = OpenCL PACKAGE = OpenCL::Queue 1196MODULE = OpenCL PACKAGE = OpenCL::Queue
451 1197
452void 1198void
453DESTROY (OpenCL::Queue this) 1199DESTROY (OpenCL::Queue self)
454 CODE: 1200 CODE:
455 clReleaseCommandQueue (this); 1201 clReleaseCommandQueue (self);
456 1202
457void 1203void
458info (OpenCL::Queue this, cl_command_queue_info name)
459 PPCODE:
460 INFO (CommandQueue)
461
462void
463enqueue_read_buffer (OpenCL::Queue this, 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
464 PPCODE: 1207 PPCODE:
465{
466 cl_event ev = 0; 1208 cl_event ev = 0;
467 EVENT_LIST (6, items - 6); 1209 EVENT_LIST (6, items - 6);
468 1210
469 SvUPGRADE (data, SVt_PV); 1211 SvUPGRADE (data, SVt_PV);
470 SvGROW (data, len); 1212 SvGROW (data, len);
471 SvPOK_only (data); 1213 SvPOK_only (data);
472 SvCUR_set (data, len); 1214 SvCUR_set (data, len);
473 NEED_SUCCESS (EnqueueReadBuffer, (this, 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));
474 1216
475 if (ev) 1217 if (ev)
476 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1218 XPUSH_CLOBJ ("OpenCL::Event", ev);
477}
478 1219
479void 1220void
480enqueue_write_buffer (OpenCL::Queue this, 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
481 PPCODE: 1224 PPCODE:
482{
483 cl_event ev = 0; 1225 cl_event ev = 0;
484 STRLEN len; 1226 STRLEN len;
485 char *ptr = SvPVbyte (data, len); 1227 char *ptr = SvPVbyte (data, len);
486 EVENT_LIST (5, items - 5); 1228 EVENT_LIST (5, items - 5);
487 1229
488 NEED_SUCCESS (EnqueueReadBuffer, (this, 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));
489 1231
490 if (ev) 1232 if (ev)
491 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1233 XPUSH_CLOBJ ("OpenCL::Event", ev);
492}
493 1234
1235#if CL_VERSION_1_2
1236
494void 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
495enqueue_copy_buffer (OpenCL::Queue this, 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
496 PPCODE: 1286 PPCODE:
497{
498 cl_event ev = 0; 1287 cl_event ev = 0;
499 EVENT_LIST (6, items - 6); 1288 EVENT_LIST (6, items - 6);
500 1289
501 NEED_SUCCESS (EnqueueCopyBuffer, (this, 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));
502 1291
503 if (ev) 1292 if (ev)
504 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1293 XPUSH_CLOBJ ("OpenCL::Event", ev);
505}
506 1294
507 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html */
508 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html */
509
510void 1295void
511enqueue_read_image (OpenCL::Queue this, 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, ...) 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
512 PPCODE: 1299 PPCODE:
513{
514 cl_event ev = 0; 1300 cl_event ev = 0;
1301 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
515 const size_t src_origin[3] = { src_x, src_y, src_z }; 1302 const size_t host_origin[3] = { host_x, host_y, host_z };
516 const size_t region[3] = { width, height, depth }; 1303 const size_t region[3] = { width, height, depth };
517 size_t len = row_pitch * slice_pitch * depth;
518 EVENT_LIST (11, items - 11); 1304 EVENT_LIST (17, items - 17);
1305
1306 if (!buf_row_pitch)
1307 buf_row_pitch = region [0];
1308
1309 if (!buf_slice_pitch)
1310 buf_slice_pitch = region [1] * buf_row_pitch;
1311
1312 if (!host_row_pitch)
1313 host_row_pitch = region [0];
1314
1315 if (!host_slice_pitch)
1316 host_slice_pitch = region [1] * host_row_pitch;
1317
1318 size_t len = host_row_pitch * host_slice_pitch * region [2];
519 1319
520 SvUPGRADE (data, SVt_PV); 1320 SvUPGRADE (data, SVt_PV);
521 SvGROW (data, len); 1321 SvGROW (data, len);
522 SvPOK_only (data); 1322 SvPOK_only (data);
523 SvCUR_set (data, len); 1323 SvCUR_set (data, len);
524 NEED_SUCCESS (EnqueueReadImage, (this, src, blocking, src_origin, region, row_pitch, 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));
525 1325
526 if (ev) 1326 if (ev)
527 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1327 XPUSH_CLOBJ ("OpenCL::Event", ev);
528}
529 1328
530void 1329void
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
1333 PPCODE:
1334 cl_event ev = 0;
1335 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1336 const size_t host_origin[3] = { host_x, host_y, host_z };
1337 const size_t region[3] = { width, height, depth };
1338 STRLEN len;
1339 char *ptr = SvPVbyte (data, len);
1340 EVENT_LIST (17, items - 17);
1341
1342 if (!buf_row_pitch)
1343 buf_row_pitch = region [0];
1344
1345 if (!buf_slice_pitch)
1346 buf_slice_pitch = region [1] * buf_row_pitch;
1347
1348 if (!host_row_pitch)
1349 host_row_pitch = region [0];
1350
1351 if (!host_slice_pitch)
1352 host_slice_pitch = region [1] * host_row_pitch;
1353
1354 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1355
1356 if (len < min_len)
1357 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1358
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));
1360
1361 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev);
1363
1364void
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
1368 PPCODE:
1369 cl_event ev = 0;
1370 const size_t src_origin[3] = { src_x, src_y, src_z };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16);
1374
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));
1376
1377 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev);
1379
1380void
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
1384 PPCODE:
1385 cl_event ev = 0;
1386 const size_t src_origin[3] = { src_x, src_y, src_z };
1387 const size_t region[3] = { width, height, depth };
1388 EVENT_LIST (12, items - 12);
1389
1390 if (!row_pitch)
1391 row_pitch = img_row_pitch (src);
1392
1393 if (depth > 1 && !slice_pitch)
1394 slice_pitch = row_pitch * height;
1395
1396 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1397
1398 SvUPGRADE (data, SVt_PV);
1399 SvGROW (data, len);
1400 SvPOK_only (data);
1401 SvCUR_set (data, len);
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));
1403
1404 if (ev)
1405 XPUSH_CLOBJ ("OpenCL::Event", ev);
1406
1407void
531enqueue_write_image (OpenCL::Queue this, 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, 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
532 PPCODE: 1411 PPCODE:
533{
534 cl_event ev = 0; 1412 cl_event ev = 0;
535 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 };
536 const size_t region[3] = { width, height, depth }; 1414 const size_t region[3] = { width, height, depth };
537 STRLEN len; 1415 STRLEN len;
538 char *ptr = SvPVbyte (data, len); 1416 char *ptr = SvPVbyte (data, len);
539 size_t slice_pitch = len / (row_pitch * height);
540 EVENT_LIST (11, items - 11); 1417 EVENT_LIST (12, items - 12);
541 1418
1419 if (!row_pitch)
1420 row_pitch = img_row_pitch (dst);
1421
1422 if (depth > 1 && !slice_pitch)
1423 slice_pitch = row_pitch * height;
1424
1425 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1426
1427 if (len < min_len)
1428 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1429
542 NEED_SUCCESS (EnqueueWriteImage, (this, 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));
543 1431
544 if (ev) 1432 if (ev)
545 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1433 XPUSH_CLOBJ ("OpenCL::Event", ev);
546}
547 1434
548void 1435void
549enqueue_copy_buffer_rect (OpenCL::Queue this, 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, ...)
550 PPCODE:
551{
552 cl_event ev = 0;
553 const size_t src_origin[3] = { src_x, src_y, src_z };
554 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
555 const size_t region[3] = { width, height, depth };
556 EVENT_LIST (16, items - 16);
557
558 NEED_SUCCESS (EnqueueCopyBufferRect, (this, 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));
559
560 if (ev)
561 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
562}
563
564void
565enqueue_copy_buffer_to_image (OpenCL::Queue this, 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, ...)
566 PPCODE:
567{
568 cl_event ev = 0;
569 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
570 const size_t region[3] = { width, height, depth };
571 EVENT_LIST (10, items - 10);
572
573 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
574
575 if (ev)
576 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
577}
578
579void
580enqueue_copy_image (OpenCL::Queue this, OpenCL::Image 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, ...) 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
581 PPCODE: 1439 PPCODE:
582{
583 cl_event ev = 0; 1440 cl_event ev = 0;
584 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 };
585 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 };
586 const size_t region[3] = { width, height, depth }; 1443 const size_t region[3] = { width, height, depth };
587 EVENT_LIST (12, items - 12); 1444 EVENT_LIST (12, items - 12);
588 1445
589 NEED_SUCCESS (EnqueueCopyImage, (this, 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));
590 1447
591 if (ev) 1448 if (ev)
592 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1449 XPUSH_CLOBJ ("OpenCL::Event", ev);
593}
594 1450
595void 1451void
596enqueue_copy_image_to_buffer (OpenCL::Queue this, 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
597 PPCODE: 1455 PPCODE:
598{
599 cl_event ev = 0; 1456 cl_event ev = 0;
600 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 };
601 const size_t region[3] = { width, height, depth }; 1458 const size_t region[3] = { width, height, depth };
602 EVENT_LIST (10, items - 10); 1459 EVENT_LIST (10, items - 10);
603 1460
604 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, 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));
605 1462
606 if (ev) 1463 if (ev)
607 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
608}
609 1465
610void 1466void
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
1470 PPCODE:
1471 cl_event ev = 0;
1472 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1473 const size_t region[3] = { width, height, depth };
1474 EVENT_LIST (10, items - 10);
1475
1476 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1477
1478 if (ev)
1479 XPUSH_CLOBJ ("OpenCL::Event", ev);
1480
1481void
611enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...) 1482task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1483 ALIAS:
1484 enqueue_task = 0
612 PPCODE: 1485 PPCODE:
613{
614 cl_event ev = 0; 1486 cl_event ev = 0;
615 EVENT_LIST (2, items - 2); 1487 EVENT_LIST (2, items - 2);
616 1488
617 NEED_SUCCESS (EnqueueTask, (this, 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));
618 1490
619 if (ev) 1491 if (ev)
620 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1492 XPUSH_CLOBJ ("OpenCL::Event", ev);
621}
622 1493
623void 1494void
624enqueue_nd_range_kernel (OpenCL::Queue this, 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
625 PPCODE: 1498 PPCODE:
626{
627 cl_event ev = 0; 1499 cl_event ev = 0;
628 size_t *gwo = 0, *gws, *lws = 0; 1500 size_t *gwo = 0, *gws, *lws = 0;
629 int gws_len; 1501 int gws_len;
630 size_t *lists; 1502 size_t *lists;
631 int i; 1503 int i;
638 1510
639 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1511 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
640 1512
641 gws = lists + gws_len * 0; 1513 gws = lists + gws_len * 0;
642 for (i = 0; i < gws_len; ++i) 1514 for (i = 0; i < gws_len; ++i)
1515 {
643 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 }
644 1521
645 if (SvOK (global_work_offset)) 1522 if (SvOK (global_work_offset))
646 { 1523 {
647 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)
648 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");
655 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1532 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
656 } 1533 }
657 1534
658 if (SvOK (local_work_size)) 1535 if (SvOK (local_work_size))
659 { 1536 {
660 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)
661 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");
662 1539
663 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1540 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
664 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");
665 1542
666 lws = lists + gws_len * 2; 1543 lws = lists + gws_len * 2;
667 for (i = 0; i < gws_len; ++i) 1544 for (i = 0; i < gws_len; ++i)
1545 {
668 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 }
669 } 1551 }
670 1552
671 NEED_SUCCESS (EnqueueNDRangeKernel, (this, 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));
672 1554
673 if (ev) 1555 if (ev)
674 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1556 XPUSH_CLOBJ ("OpenCL::Event", ev);
675}
676 1557
1558#if cl_apple_gl_sharing || cl_khr_gl_sharing
1559
677void 1560void
678enqueue_marker (OpenCL::Queue this) 1561acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1562 ALIAS:
1563 enqueue_acquire_gl_objects = 0
1564 ALIAS:
1565 enqueue_release_gl_objects = 1
679 PPCODE: 1566 PPCODE:
680{ 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
681 cl_event ev; 1570 cl_event ev = 0;
682 NEED_SUCCESS (EnqueueMarker, (this, &ev)); 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;
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
1583 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1584
1585 if (ev)
683 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1586 XPUSH_CLOBJ ("OpenCL::Event", ev);
684}
685 1587
1588#endif
1589
686void 1590void
687enqueue_wait_for_events (OpenCL::Queue this, ...) 1591wait_for_events (OpenCL::Queue self, ...)
1592 ALIAS:
1593 enqueue_wait_for_events = 0
688 CODE: 1594 CODE:
689{
690 EVENT_LIST (1, items - 1); 1595 EVENT_LIST (1, items - 1);
1596#if PREFER_1_1
691 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 1597 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
692} 1598#else
1599 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1600#endif
693 1601
694void 1602void
695enqueue_barrier (OpenCL::Queue this) 1603marker (OpenCL::Queue self, ...)
1604 ALIAS:
1605 enqueue_marker = 0
696 CODE: 1606 PPCODE:
1607 cl_event ev = 0;
1608 EVENT_LIST (1, items - 1);
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
1622 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1623#endif
1624 if (ev)
1625 XPUSH_CLOBJ ("OpenCL::Event", ev);
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)
697 NEED_SUCCESS (EnqueueBarrier, (this)); 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));
698 1644
1645 if (GIMME_V != G_VOID)
1646 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1647 }
1648#endif
1649#else
1650 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1651#endif
1652 if (ev)
1653 XPUSH_CLOBJ ("OpenCL::Event", ev);
1654
699void 1655void
700flush (OpenCL::Queue this) 1656flush (OpenCL::Queue self)
701 CODE: 1657 CODE:
702 NEED_SUCCESS (Flush, (this)); 1658 NEED_SUCCESS (Flush, (self));
703 1659
704void 1660void
705finish (OpenCL::Queue this) 1661finish (OpenCL::Queue self)
706 CODE: 1662 CODE:
707 NEED_SUCCESS (Finish, (this)); 1663 NEED_SUCCESS (Finish, (self));
1664
1665void
1666info (OpenCL::Queue self, cl_command_queue_info name)
1667 PPCODE:
1668 INFO (CommandQueue)
1669
1670#BEGIN:command_queue
1671
1672void
1673context (OpenCL::Queue self)
1674 PPCODE:
1675 cl_context value [1];
1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1677 EXTEND (SP, 1);
1678 const int i = 0;
1679 {
1680 NEED_SUCCESS (RetainContext, (value [i]));
1681 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1682 }
1683
1684void
1685device (OpenCL::Queue self)
1686 PPCODE:
1687 cl_device_id value [1];
1688 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1689 EXTEND (SP, 1);
1690 const int i = 0;
1691 {
1692 PUSH_CLOBJ ("OpenCL::Device", value [i]);
1693 }
1694
1695void
1696reference_count (OpenCL::Queue self)
1697 PPCODE:
1698 cl_uint value [1];
1699 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
1700 EXTEND (SP, 1);
1701 const int i = 0;
1702 PUSHs (sv_2mortal (newSVuv (value [i])));
1703
1704void
1705properties (OpenCL::Queue self)
1706 PPCODE:
1707 cl_command_queue_properties value [1];
1708 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1709 EXTEND (SP, 1);
1710 const int i = 0;
1711 PUSHs (sv_2mortal (newSViv (value [i])));
1712
1713#END:command_queue
708 1714
709MODULE = OpenCL PACKAGE = OpenCL::Memory 1715MODULE = OpenCL PACKAGE = OpenCL::Memory
710 1716
711void 1717void
712DESTROY (OpenCL::Memory this) 1718DESTROY (OpenCL::Memory self)
713 CODE: 1719 CODE:
714 clReleaseMemObject (this); 1720 clReleaseMemObject (self);
715 1721
716void 1722void
717info (OpenCL::Memory this, cl_mem_info name) 1723info (OpenCL::Memory self, cl_mem_info name)
718 PPCODE: 1724 PPCODE:
719 INFO (MemObject) 1725 INFO (MemObject)
720 1726
1727#BEGIN:mem
1728
1729void
1730type (OpenCL::Memory self)
1731 PPCODE:
1732 cl_mem_object_type value [1];
1733 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1734 EXTEND (SP, 1);
1735 const int i = 0;
1736 PUSHs (sv_2mortal (newSViv (value [i])));
1737
1738void
1739flags (OpenCL::Memory self)
1740 PPCODE:
1741 cl_mem_flags value [1];
1742 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1743 EXTEND (SP, 1);
1744 const int i = 0;
1745 PUSHs (sv_2mortal (newSViv (value [i])));
1746
1747void
1748size (OpenCL::Memory self)
1749 ALIAS:
1750 size = CL_MEM_SIZE
1751 offset = CL_MEM_OFFSET
1752 PPCODE:
1753 size_t value [1];
1754 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1755 EXTEND (SP, 1);
1756 const int i = 0;
1757 PUSHs (sv_2mortal (newSVuv (value [i])));
1758
1759void
1760host_ptr (OpenCL::Memory self)
1761 PPCODE:
1762 void * value [1];
1763 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
1764 EXTEND (SP, 1);
1765 const int i = 0;
1766 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1767
1768void
1769map_count (OpenCL::Memory self)
1770 ALIAS:
1771 map_count = CL_MEM_MAP_COUNT
1772 reference_count = CL_MEM_REFERENCE_COUNT
1773 PPCODE:
1774 cl_uint value [1];
1775 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1776 EXTEND (SP, 1);
1777 const int i = 0;
1778 PUSHs (sv_2mortal (newSVuv (value [i])));
1779
1780void
1781context (OpenCL::Memory self)
1782 PPCODE:
1783 cl_context value [1];
1784 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1785 EXTEND (SP, 1);
1786 const int i = 0;
1787 {
1788 NEED_SUCCESS (RetainContext, (value [i]));
1789 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1790 }
1791
1792void
1793associated_memobject (OpenCL::Memory self)
1794 PPCODE:
1795 cl_mem value [1];
1796 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1797 EXTEND (SP, 1);
1798 const int i = 0;
1799 {
1800 NEED_SUCCESS (RetainMemObject, (value [i]));
1801 PUSH_CLOBJ ("OpenCL::Memory", value [i]);
1802 }
1803
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
1819
1820MODULE = OpenCL PACKAGE = OpenCL::BufferObj
1821
1822void
1823sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
1824 PPCODE:
1825 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
1826 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1827
1828 cl_buffer_region crdata = { origin, size };
1829
1830 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1831 XPUSH_CLOBJ ("OpenCL::Buffer", mem);
1832
1833MODULE = OpenCL PACKAGE = OpenCL::Image
1834
1835void
1836image_info (OpenCL::Image self, cl_image_info name)
1837 PPCODE:
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)));
1848
1849#BEGIN:image
1850
1851void
1852element_size (OpenCL::Image self)
1853 ALIAS:
1854 element_size = CL_IMAGE_ELEMENT_SIZE
1855 row_pitch = CL_IMAGE_ROW_PITCH
1856 slice_pitch = CL_IMAGE_SLICE_PITCH
1857 width = CL_IMAGE_WIDTH
1858 height = CL_IMAGE_HEIGHT
1859 depth = CL_IMAGE_DEPTH
1860 PPCODE:
1861 size_t value [1];
1862 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
1863 EXTEND (SP, 1);
1864 const int i = 0;
1865 PUSHs (sv_2mortal (newSVuv (value [i])));
1866
1867#END:image
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
721MODULE = OpenCL PACKAGE = OpenCL::Sampler 1895MODULE = OpenCL PACKAGE = OpenCL::Sampler
722 1896
723void 1897void
724DESTROY (OpenCL::Sampler this) 1898DESTROY (OpenCL::Sampler self)
725 CODE: 1899 CODE:
726 clReleaseSampler (this); 1900 clReleaseSampler (self);
727 1901
728void 1902void
729info (OpenCL::Sampler this, cl_sampler_info name) 1903info (OpenCL::Sampler self, cl_sampler_info name)
730 PPCODE: 1904 PPCODE:
731 INFO (Sampler) 1905 INFO (Sampler)
732 1906
1907#BEGIN:sampler
1908
1909void
1910reference_count (OpenCL::Sampler self)
1911 PPCODE:
1912 cl_uint value [1];
1913 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0));
1914 EXTEND (SP, 1);
1915 const int i = 0;
1916 PUSHs (sv_2mortal (newSVuv (value [i])));
1917
1918void
1919context (OpenCL::Sampler self)
1920 PPCODE:
1921 cl_context value [1];
1922 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1923 EXTEND (SP, 1);
1924 const int i = 0;
1925 {
1926 NEED_SUCCESS (RetainContext, (value [i]));
1927 PUSH_CLOBJ ("OpenCL::Context", value [i]);
1928 }
1929
1930void
1931normalized_coords (OpenCL::Sampler self)
1932 PPCODE:
1933 cl_addressing_mode value [1];
1934 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1935 EXTEND (SP, 1);
1936 const int i = 0;
1937 PUSHs (sv_2mortal (newSViv (value [i])));
1938
1939void
1940addressing_mode (OpenCL::Sampler self)
1941 PPCODE:
1942 cl_filter_mode value [1];
1943 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1944 EXTEND (SP, 1);
1945 const int i = 0;
1946 PUSHs (sv_2mortal (newSViv (value [i])));
1947
1948void
1949filter_mode (OpenCL::Sampler self)
1950 PPCODE:
1951 cl_bool value [1];
1952 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1953 EXTEND (SP, 1);
1954 const int i = 0;
1955 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1956
1957#END:sampler
1958
733MODULE = OpenCL PACKAGE = OpenCL::Program 1959MODULE = OpenCL PACKAGE = OpenCL::Program
734 1960
735void 1961void
736DESTROY (OpenCL::Program this) 1962DESTROY (OpenCL::Program self)
737 CODE: 1963 CODE:
738 clReleaseProgram (this); 1964 clReleaseProgram (self);
739 1965
740void 1966void
741info (OpenCL::Program this, cl_program_info name) 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
742 PPCODE: 1970 CODE:
743 INFO (Program) 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;
744 1975
745void 1976 if (SvOK (devices))
746build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 1977 {
747 CODE: 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
748 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2003 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
749 2004
750void 2005void
751build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2006build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
752 PPCODE: 2007 PPCODE:
753{
754 size_t size; 2008 size_t size;
755 SV *sv;
756
757 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2009 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
758 sv = sv_2mortal (newSV (size)); 2010 SV *sv = sv_2mortal (newSV (size));
759 SvUPGRADE (sv, SVt_PV); 2011 SvUPGRADE (sv, SVt_PV);
760 SvPOK_only (sv); 2012 SvPOK_only (sv);
761 SvCUR_set (sv, size); 2013 SvCUR_set (sv, size);
762 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2014 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
763 XPUSHs (sv); 2015 XPUSHs (sv);
764} 2016
2017#BEGIN:program_build
2018
2019void
2020build_status (OpenCL::Program self, OpenCL::Device device)
2021 PPCODE:
2022 cl_build_status value [1];
2023 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2024 EXTEND (SP, 1);
2025 const int i = 0;
2026 PUSHs (sv_2mortal (newSViv (value [i])));
2027
2028void
2029build_options (OpenCL::Program self, OpenCL::Device device)
2030 ALIAS:
2031 build_options = CL_PROGRAM_BUILD_OPTIONS
2032 build_log = CL_PROGRAM_BUILD_LOG
2033 PPCODE:
2034 size_t size;
2035 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2036 char *value = tmpbuf (size);
2037 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2038 EXTEND (SP, 1);
2039 const int i = 0;
2040 PUSHs (sv_2mortal (newSVpv (value, 0)));
2041
2042#END:program_build
765 2043
766void 2044void
767kernel (OpenCL::Program program, SV *function) 2045kernel (OpenCL::Program program, SV *function)
768 PPCODE: 2046 PPCODE:
769{
770 cl_int res;
771 cl_kernel kernel = clCreateKernel (program, SvPVbyte_nolen (function), &res); 2047 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
772
773 if (res)
774 FAIL (CreateKernel, res);
775
776 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2048 XPUSH_CLOBJ ("OpenCL::Kernel", kernel);
777} 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]);
2062
2063void
2064info (OpenCL::Program self, cl_program_info name)
2065 PPCODE:
2066 INFO (Program)
2067
2068void
2069binaries (OpenCL::Program self)
2070 PPCODE:
2071 cl_uint n, i;
2072 size_t size;
2073
2074 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2075 if (!n) XSRETURN_EMPTY;
2076
2077 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2078 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2079 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2080 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2081
2082 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i)
2084 {
2085 SV *sv = sv_2mortal (newSV (sizes [i]));
2086 SvUPGRADE (sv, SVt_PV);
2087 SvPOK_only (sv);
2088 SvCUR_set (sv, sizes [i]);
2089 ptrs [i] = (void *)SvPVX (sv);
2090 PUSHs (sv);
2091 }
2092
2093 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2094 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2095
2096#BEGIN:program
2097
2098void
2099reference_count (OpenCL::Program self)
2100 ALIAS:
2101 reference_count = CL_PROGRAM_REFERENCE_COUNT
2102 num_devices = CL_PROGRAM_NUM_DEVICES
2103 PPCODE:
2104 cl_uint value [1];
2105 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2106 EXTEND (SP, 1);
2107 const int i = 0;
2108 PUSHs (sv_2mortal (newSVuv (value [i])));
2109
2110void
2111context (OpenCL::Program self)
2112 PPCODE:
2113 cl_context value [1];
2114 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2115 EXTEND (SP, 1);
2116 const int i = 0;
2117 {
2118 NEED_SUCCESS (RetainContext, (value [i]));
2119 PUSH_CLOBJ ("OpenCL::Context", value [i]);
2120 }
2121
2122void
2123devices (OpenCL::Program self)
2124 PPCODE:
2125 size_t size;
2126 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2127 cl_device_id *value = tmpbuf (size);
2128 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2129 int i, n = size / sizeof (*value);
2130 EXTEND (SP, n);
2131 for (i = 0; i < n; ++i)
2132 {
2133 PUSH_CLOBJ ("OpenCL::Device", value [i]);
2134 }
2135
2136void
2137source (OpenCL::Program self)
2138 PPCODE:
2139 size_t size;
2140 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2141 char *value = tmpbuf (size);
2142 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2143 EXTEND (SP, 1);
2144 const int i = 0;
2145 PUSHs (sv_2mortal (newSVpv (value, 0)));
2146
2147void
2148binary_sizes (OpenCL::Program self)
2149 PPCODE:
2150 size_t size;
2151 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2152 size_t *value = tmpbuf (size);
2153 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2154 int i, n = size / sizeof (*value);
2155 EXTEND (SP, n);
2156 for (i = 0; i < n; ++i)
2157 PUSHs (sv_2mortal (newSVuv (value [i])));
2158
2159#END:program
778 2160
779MODULE = OpenCL PACKAGE = OpenCL::Kernel 2161MODULE = OpenCL PACKAGE = OpenCL::Kernel
780 2162
781void 2163void
782DESTROY (OpenCL::Kernel this) 2164DESTROY (OpenCL::Kernel self)
783 CODE: 2165 CODE:
784 clReleaseKernel (this); 2166 clReleaseKernel (self);
785 2167
786void 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
2234set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2235 CODE:
2236 clSetKernelArg (self, idx, sizeof (value), &value);
2237
2238void
2239set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2240 CODE:
2241 clSetKernelArg (self, idx, sizeof (value), &value);
2242
2243void
2244set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2245 CODE:
2246 clSetKernelArg (self, idx, sizeof (value), &value);
2247
2248void
2249set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2250 CODE:
2251 clSetKernelArg (self, idx, sizeof (value), &value);
2252
2253void
2254set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2255 CODE:
2256 clSetKernelArg (self, idx, sizeof (value), &value);
2257
2258void
2259set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2260 CODE:
2261 clSetKernelArg (self, idx, sizeof (value), &value);
2262
2263void
2264set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2265 CODE:
2266 clSetKernelArg (self, idx, sizeof (value), &value);
2267
2268void
2269set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2270 CODE:
2271 clSetKernelArg (self, idx, sizeof (value), &value);
2272
2273void
2274set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2275 CODE:
2276 clSetKernelArg (self, idx, sizeof (value), &value);
2277
2278void
2279set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2280 CODE:
2281 clSetKernelArg (self, idx, sizeof (value), &value);
2282
2283void
2284set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2285 CODE:
2286 clSetKernelArg (self, idx, sizeof (value), &value);
2287
2288void
2289set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
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:
2309 clSetKernelArg (self, idx, sizeof (value), &value);
2310
2311void
2312set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2313 CODE:
2314 clSetKernelArg (self, idx, size, 0);
2315
2316void
2317set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2318 CODE:
2319 clSetKernelArg (self, idx, sizeof (value), &value);
2320
2321void
787info (OpenCL::Kernel this, cl_kernel_info name) 2322info (OpenCL::Kernel self, cl_kernel_info name)
788 PPCODE: 2323 PPCODE:
789 INFO (Kernel) 2324 INFO (Kernel)
790 2325
2326#BEGIN:kernel
2327
791void 2328void
792set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2329function_name (OpenCL::Kernel self)
2330 PPCODE:
2331 size_t size;
2332 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2333 char *value = tmpbuf (size);
2334 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2335 EXTEND (SP, 1);
2336 const int i = 0;
2337 PUSHs (sv_2mortal (newSVpv (value, 0)));
2338
2339void
2340num_args (OpenCL::Kernel self)
2341 ALIAS:
2342 num_args = CL_KERNEL_NUM_ARGS
2343 reference_count = CL_KERNEL_REFERENCE_COUNT
2344 PPCODE:
2345 cl_uint value [1];
2346 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2347 EXTEND (SP, 1);
2348 const int i = 0;
2349 PUSHs (sv_2mortal (newSVuv (value [i])));
2350
2351void
2352context (OpenCL::Kernel self)
2353 PPCODE:
2354 cl_context value [1];
2355 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2356 EXTEND (SP, 1);
2357 const int i = 0;
2358 {
2359 NEED_SUCCESS (RetainContext, (value [i]));
2360 PUSH_CLOBJ ("OpenCL::Context", value [i]);
2361 }
2362
2363void
2364program (OpenCL::Kernel self)
2365 PPCODE:
2366 cl_program value [1];
2367 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2368 EXTEND (SP, 1);
2369 const int i = 0;
2370 {
2371 NEED_SUCCESS (RetainProgram, (value [i]));
2372 PUSH_CLOBJ ("OpenCL::Program", value [i]);
2373 }
2374
2375#END:kernel
2376
2377void
2378work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
793 CODE: 2379 PPCODE:
794 clSetKernelArg (this, idx, sizeof (value), &value); 2380 size_t size;
2381 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2382 SV *sv = sv_2mortal (newSV (size));
2383 SvUPGRADE (sv, SVt_PV);
2384 SvPOK_only (sv);
2385 SvCUR_set (sv, size);
2386 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2387 XPUSHs (sv);
795 2388
796void 2389#BEGIN:kernel_work_group
797set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value)
798 CODE:
799 clSetKernelArg (this, idx, sizeof (value), &value);
800 2390
801void 2391void
802set_short (OpenCL::Kernel this, cl_uint idx, cl_short value) 2392work_group_size (OpenCL::Kernel self, OpenCL::Device device)
803 CODE: 2393 ALIAS:
804 clSetKernelArg (this, idx, sizeof (value), &value); 2394 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2395 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2396 PPCODE:
2397 size_t value [1];
2398 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2399 EXTEND (SP, 1);
2400 const int i = 0;
2401 PUSHs (sv_2mortal (newSVuv (value [i])));
805 2402
806void 2403void
807set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2404compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
808 CODE: 2405 PPCODE:
809 clSetKernelArg (this, idx, sizeof (value), &value); 2406 size_t size;
2407 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2408 size_t *value = tmpbuf (size);
2409 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2410 int i, n = size / sizeof (*value);
2411 EXTEND (SP, n);
2412 for (i = 0; i < n; ++i)
2413 PUSHs (sv_2mortal (newSVuv (value [i])));
810 2414
811void 2415void
812set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2416local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
813 CODE: 2417 ALIAS:
814 clSetKernelArg (this, idx, sizeof (value), &value); 2418 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2419 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2420 PPCODE:
2421 cl_ulong value [1];
2422 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2423 EXTEND (SP, 1);
2424 const int i = 0;
2425 PUSHs (sv_2mortal (newSVuv (value [i])));
815 2426
816void 2427#END:kernel_work_group
817set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value)
818 CODE:
819 clSetKernelArg (this, idx, sizeof (value), &value);
820
821void
822set_long (OpenCL::Kernel this, cl_uint idx, cl_long value)
823 CODE:
824 clSetKernelArg (this, idx, sizeof (value), &value);
825
826void
827set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
828 CODE:
829 clSetKernelArg (this, idx, sizeof (value), &value);
830
831void
832set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
833 CODE:
834 clSetKernelArg (this, idx, sizeof (value), &value);
835
836void
837set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
838 CODE:
839 clSetKernelArg (this, idx, sizeof (value), &value);
840
841void
842set_double (OpenCL::Kernel this, cl_uint idx, cl_double value)
843 CODE:
844 clSetKernelArg (this, idx, sizeof (value), &value);
845
846void
847set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
848 CODE:
849 clSetKernelArg (this, idx, sizeof (value), &value);
850
851void
852set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
853 CODE:
854 clSetKernelArg (this, idx, sizeof (value), &value);
855
856void
857set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
858 CODE:
859 clSetKernelArg (this, idx, sizeof (value), &value);
860
861void
862set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
863 CODE:
864 clSetKernelArg (this, idx, sizeof (value), &value);
865
866void
867set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
868 CODE:
869 clSetKernelArg (this, idx, sizeof (value), &value);
870
871void
872set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
873 CODE:
874 clSetKernelArg (this, idx, sizeof (value), &value);
875 2428
876MODULE = OpenCL PACKAGE = OpenCL::Event 2429MODULE = OpenCL PACKAGE = OpenCL::Event
877 2430
878void 2431void
879DESTROY (OpenCL::Event this) 2432DESTROY (OpenCL::Event self)
880 CODE: 2433 CODE:
881 clReleaseEvent (this); 2434 clReleaseEvent (self);
882 2435
883void 2436void
2437wait (OpenCL::Event self)
2438 CODE:
2439 clWaitForEvents (1, &self);
2440
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
884info (OpenCL::Event this, cl_event_info name) 2447info (OpenCL::Event self, cl_event_info name)
885 PPCODE: 2448 PPCODE:
886 INFO (Event) 2449 INFO (Event)
887 2450
2451#BEGIN:event
2452
888void 2453void
889wait (OpenCL::Event this) 2454command_queue (OpenCL::Event self)
2455 PPCODE:
2456 cl_command_queue value [1];
2457 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2458 EXTEND (SP, 1);
2459 const int i = 0;
2460 {
2461 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2462 PUSH_CLOBJ ("OpenCL::Queue", value [i]);
2463 }
2464
2465void
2466command_type (OpenCL::Event self)
2467 PPCODE:
2468 cl_command_type value [1];
2469 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2470 EXTEND (SP, 1);
2471 const int i = 0;
2472 PUSHs (sv_2mortal (newSVuv (value [i])));
2473
2474void
2475reference_count (OpenCL::Event self)
2476 ALIAS:
2477 reference_count = CL_EVENT_REFERENCE_COUNT
2478 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2479 PPCODE:
2480 cl_uint value [1];
2481 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2482 EXTEND (SP, 1);
2483 const int i = 0;
2484 PUSHs (sv_2mortal (newSVuv (value [i])));
2485
2486void
2487context (OpenCL::Event self)
2488 PPCODE:
2489 cl_context value [1];
2490 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2491 EXTEND (SP, 1);
2492 const int i = 0;
2493 {
2494 NEED_SUCCESS (RetainContext, (value [i]));
2495 PUSH_CLOBJ ("OpenCL::Context", value [i]);
2496 }
2497
2498#END:event
2499
2500void
2501profiling_info (OpenCL::Event self, cl_profiling_info name)
890 CODE: 2502 PPCODE:
891 clWaitForEvents (1, &this); 2503 INFO (EventProfiling)
2504
2505#BEGIN:profiling
2506
2507void
2508profiling_command_queued (OpenCL::Event self)
2509 ALIAS:
2510 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
2511 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
2512 profiling_command_start = CL_PROFILING_COMMAND_START
2513 profiling_command_end = CL_PROFILING_COMMAND_END
2514 PPCODE:
2515 cl_ulong value [1];
2516 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
2517 EXTEND (SP, 1);
2518 const int i = 0;
2519 PUSHs (sv_2mortal (newSVuv (value [i])));
2520
2521#END:profiling
892 2522
893MODULE = OpenCL PACKAGE = OpenCL::UserEvent 2523MODULE = OpenCL PACKAGE = OpenCL::UserEvent
894 2524
895void 2525void
896set_status (OpenCL::UserEvent this, cl_int execution_status) 2526set_status (OpenCL::UserEvent self, cl_int execution_status)
897 CODE: 2527 CODE:
898 clSetUserEventStatus (this, execution_status); 2528 clSetUserEventStatus (self, execution_status);
899 2529

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines