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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.51 by root, Tue Apr 24 22:45:38 2012 UTC vs.
Revision 1.81 by root, Sat May 5 17:09:19 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
6
7#include "ecb.h"//D
4 8
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 10#include "xthread.h"
7#include "schmorp.h" 11#include "schmorp.h"
8 12
26#if PREFER_1_1 30#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 32#endif
29 33
30#ifdef __APPLE__ 34#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
32#else 36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
33 #include <CL/opencl.h> 40#include CLHDR(opencl.h)
41
42#if 0
43#ifndef CL_VERSION_1_2
44 #include CLHDR(cl_d3d9.h)
45#endif
46#endif
47
48#if _WIN32
49 #include CLHDR(cl_d3d10.h)
50 #if CL_VERSION_1_2
51 #include CLHDR<cl_d3d11.h>
52 #endif
53 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 54#endif
35 55
36#ifndef CL_VERSION_1_2 56#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 57 #undef PREFER_1_1
38 #define PREFER_1_1 1 58 #define PREFER_1_1 1
39#endif 59#endif
40 60
61// make sure all constants we might use are actually defined
62#include "default.h"
63
41typedef cl_platform_id OpenCL__Platform; 64typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 65typedef cl_device_id OpenCL__Device;
66typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 67typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 68typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 69typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 70typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 71typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 72typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 73typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 74typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 75typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 76typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 77typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 78typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 79typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 80typedef cl_event OpenCL__UserEvent;
61 81
62typedef SV *FUTURE; 82typedef struct mapped * OpenCL__Mapped;
83
84static HV
85 *stash_platform,
86 *stash_device,
87 *stash_subdevice,
88 *stash_context,
89 *stash_queue,
90 *stash_program,
91 *stash_kernel,
92 *stash_sampler,
93 *stash_event,
94 *stash_userevent,
95 *stash_memory,
96 *stash_buffer,
97 *stash_bufferobj,
98 *stash_image,
99 *stash_image1d,
100 *stash_image1darray,
101 *stash_image1dbuffer,
102 *stash_image2d,
103 *stash_image2darray,
104 *stash_image3d,
105 *stash_mapped,
106 *stash_mappedbuffer,
107 *stash_mappedimage;
63 108
64/*****************************************************************************/ 109/*****************************************************************************/
65 110
66// name must include a leading underscore 111// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 112// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 152}
108 153
109/*****************************************************************************/ 154/*****************************************************************************/
110 155
111/* up to two temporary buffers */ 156/* up to two temporary buffers */
112static void * 157static void * ecb_noinline
113tmpbuf (size_t size) 158tmpbuf (size_t size)
114{ 159{
115 enum { buffers = 3 }; 160 enum { buffers = 4 };
116 static int idx; 161 static int idx;
117 static void *buf [buffers]; 162 static void *buf [buffers];
118 static size_t len [buffers]; 163 static size_t len [buffers];
119 164
120 idx = (idx + 1) % buffers; 165 idx = (idx + 1) % buffers;
127 } 172 }
128 173
129 return buf [idx]; 174 return buf [idx];
130} 175}
131 176
177static const char * ecb_noinline
178cv_get_name (CV *cv)
179{
180 static char fullname [256];
181
182 GV *gv = CvGV (cv); // gv better be nonzero
183
184 HV *stash = GvSTASH (gv);
185 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
186 const char *gvname = GvNAME (gv);
187
188 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
189 return fullname;
190}
191
132/*****************************************************************************/ 192/*****************************************************************************/
133 193
134typedef struct 194typedef struct
135{ 195{
136 IV iv; 196 IV iv;
137 const char *name; 197 const char *name;
138 #define const_iv(name) { (IV)CL_ ## name, # name }, 198 #define const_iv(name) { (IV)CL_ ## name, # name },
139} ivstr; 199} ivstr;
200
201typedef struct
202{
203 NV nv;
204 const char *name;
205 #define const_nv(name) { (NV)CL_ ## name, # name },
206} nvstr;
140 207
141static const char * 208static const char *
142iv2str (IV value, const ivstr *base, int count, const char *fallback) 209iv2str (IV value, const ivstr *base, int count, const char *fallback)
143{ 210{
144 int i; 211 int i;
193 if (res) \ 260 if (res) \
194 FAIL (name); 261 FAIL (name);
195 262
196/*****************************************************************************/ 263/*****************************************************************************/
197 264
265static SV *
266new_clobj (HV *stash, IV id)
267{
268 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
269}
270
271#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
272#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
273
274/* cl objects are either \$iv, or [$iv, ...] */
275/* they can be upgraded at runtime to the array form */
276static void * ecb_noinline
277SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
278{
279 // sv_derived_from is quite slow :(
280 if (SvROK (sv) && sv_derived_from (sv, pkg))
281 return (void *)SvIV (SvRV (sv));
282
283 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
284}
285
286// the "no-inherit" version of the above
287static void * ecb_noinline
288SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
289{
290 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
291 return (void *)SvIV (SvRV (sv));
292
293 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
294}
295
296/*****************************************************************************/
297
198static cl_context_properties * 298static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 299SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 300{
201 if (!sv || !SvOK (sv)) 301 if (!sv || !SvOK (sv))
202 if (extra) 302 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 303 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 304 else
210 int i, len = av_len (av) + 1; 310 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 311 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 312 cl_context_properties *l = p;
213 313
214 if (len & 1) 314 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 315 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 316
217 while (extracount--) 317 while (extracount--)
218 *l++ = *extra++; 318 *l++ = *extra++;
219 319
220 for (i = 0; i < len; i += 2) 320 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 323 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 324 cl_context_properties v = SvIV (p_sv); // code below can override
225 325
226 switch (t) 326 switch (t)
227 { 327 {
328 case CL_CONTEXT_PLATFORM:
329 if (SvROK (p_sv))
330 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
331 break;
332
228 case CL_GLX_DISPLAY_KHR: 333 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 334 if (!SvOK (p_sv))
230 { 335 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 336 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 337 if (func)
255 *l = 0; 360 *l = 0;
256 361
257 return p; 362 return p;
258 } 363 }
259 364
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 365 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 366}
262 367
263/*****************************************************************************/ 368// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 369// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 370static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 371object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{ 372{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 373 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
374 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
375 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
376
377 AV *av = (AV *)SvRV (arg);
378 void **list = 0;
379 cl_uint count = av_len (av) + 1;
380
381 if (count)
275 { 382 {
276 SV *rv = SvRV (sv); 383 list = tmpbuf (sizeof (*list) * count);
277 384 int i;
278 if (SvTYPE (rv) == SVt_PVAV) 385 for (i = 0; i < count; ++i)
279 rv = AvARRAY (rv)[0]; 386 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
280
281 return (void *)SvIV (SvRV (sv));
282 } 387 }
283 388
284 croak ("%s: %s is not of type %s", func, svname, pkg); 389 if (!count && !or_undef)
285} 390 croak ("%s: '%s' must contain at least one %s object",
391 cv_get_name (cv), argname, klass);
286 392
287static void 393 *rcount = count;
288CLOBJ_push (SV *self, SV *data) 394 return (void *)list;
289{
290 SV *rv = SvRV (self);
291
292 if (SvTYPE (rv) != SVt_PVAV)
293 {
294 AV *av = newAV ();
295 av_push (av, rv);
296 rv = (SV *)av;
297 SvRV_set (self, rv);
298 }
299
300 av_push ((AV *)rv, data);
301}
302
303static SV *
304sv_struct (STRLEN size)
305{
306 SV *sv = newSV (size);
307 SvPOK_only (sv);
308 return sv;
309}
310
311static void *
312CLOBJ_push_struct (SV *self, STRLEN size)
313{
314 SV *sv = sv_struct (size);
315 CLOBJ_push (self, sv);
316 return SvPVX (sv);
317} 395}
318 396
319/*****************************************************************************/ 397/*****************************************************************************/
320/* callback stuff */ 398/* callback stuff */
321 399
343static void (*eq_signal_func)(void *signal_arg, int value); 421static void (*eq_signal_func)(void *signal_arg, int value);
344static void *eq_signal_arg; 422static void *eq_signal_arg;
345static xmutex_t eq_lock = X_MUTEX_INIT; 423static xmutex_t eq_lock = X_MUTEX_INIT;
346static eq_item *eq_head, *eq_tail; 424static eq_item *eq_head, *eq_tail;
347 425
348static void 426static void ecb_noinline
349eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 427eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
350{ 428{
351 eq_item *item = malloc (sizeof (eq_item)); 429 eq_item *item = malloc (sizeof (eq_item));
352
353 printf ("enq(%p,%p,%p,%p,%p)\n", vtbl, cb, data1, data2, data3);//D
354 430
355 item->next = 0; 431 item->next = 0;
356 item->vtbl = vtbl; 432 item->vtbl = vtbl;
357 item->cb = cb; 433 item->cb = cb;
358 item->data1 = data1; 434 item->data1 = data1;
383 X_UNLOCK (eq_lock); 459 X_UNLOCK (eq_lock);
384 460
385 return res; 461 return res;
386} 462}
387 463
388#if 0
389static void
390mem_free (pTHX_ void *p)
391{
392 free (p);
393}
394//SAVEDESTRUCTOR_X (mem_free, item);
395#endif
396
397static void 464static void
398eq_poll (void) 465eq_poll (void)
399{ 466{
400 eq_item *item; 467 eq_item *item;
401 468
428eq_poll_interrupt (pTHX_ void *c_arg, int value) 495eq_poll_interrupt (pTHX_ void *c_arg, int value)
429{ 496{
430 eq_poll (); 497 eq_poll ();
431} 498}
432 499
500/*****************************************************************************/
433/* context notify */ 501/* context notify */
434 502
435static void 503static void ecb_noinline
436eq_context_push (void *data1, void *data2, void *data3) 504eq_context_push (void *data1, void *data2, void *data3)
437{ 505{
438 dSP; 506 dSP;
439 PUSHs (sv_2mortal (newSVpv (data1, 0))); 507 PUSHs (sv_2mortal (newSVpv (data1, 0)));
440 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 508 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
441 PUTBACK; 509 PUTBACK;
510
511 free (data1);
512 free (data2);
442} 513}
443 514
444static eq_vtbl eq_context_vtbl = { 0, eq_context_push }; 515static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
445 516
517static void CL_CALLBACK
518eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
519{
520 void *pvt_copy = malloc (cb);
521 memcpy (pvt_copy, pvt, cb);
522 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
523}
524
525#define CONTEXT_NOTIFY_CALLBACK \
526 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
527 void *user_data = 0; \
528 \
529 if (SvOK (notify)) \
530 { \
531 pfn_notify = eq_context_notify; \
532 user_data = s_get_cv (notify); \
533 }
534
535static SV * ecb_noinline
536new_clobj_context (cl_context ctx, void *user_data)
537{
538 SV *sv = new_clobj (stash_context, (IV)ctx);
539
540 if (user_data)
541 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
542
543 return sv;
544}
545
546#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
547
548/*****************************************************************************/
446/* build/compile/link notify */ 549/* build/compile/link notify */
447 550
448static void 551static void
449eq_program_push (void *data1, void *data2, void *data3) 552eq_program_push (void *data1, void *data2, void *data3)
450{ 553{
451 dSP; 554 dSP;
452 PUSH_CLOBJ ("OpenCL::Program", data1); 555 PUSH_CLOBJ (stash_program, data1);
453 PUTBACK; 556 PUTBACK;
454} 557}
455 558
456static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 559static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
457 560
458static void CL_CALLBACK 561static void CL_CALLBACK
459eq_program_notify (cl_program program, void *user_data) 562eq_program_notify (cl_program program, void *user_data)
460{ 563{
564 clRetainProgram (program);
565
461 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 566 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
567}
568
569typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
570
571static program_callback ecb_noinline
572make_program_callback (SV *notify, void **ruser_data)
573{
574 if (SvOK (notify))
575 {
576 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
577 return eq_program_notify;
578 }
579 else
580 {
581 *ruser_data = 0;
582 return 0;
583 }
462} 584}
463 585
464struct build_args 586struct build_args
465{ 587{
466 cl_program program; 588 cl_program program;
480 else 602 else
481 clReleaseProgram (arg->program); 603 clReleaseProgram (arg->program);
482 604
483 free (arg->options); 605 free (arg->options);
484 free (arg); 606 free (arg);
607
608 return 0;
485} 609}
486 610
487static void 611static void
488build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 612build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
489{ 613{
497 621
498 xthread_t id; 622 xthread_t id;
499 thread_create (&id, build_program_thread, arg); 623 thread_create (&id, build_program_thread, arg);
500} 624}
501 625
626/*****************************************************************************/
627/* mem object destructor notify */
628
629static void ecb_noinline
630eq_destructor_push (void *data1, void *data2, void *data3)
631{
632}
633
634static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
635
636static void CL_CALLBACK
637eq_destructor_notify (cl_mem memobj, void *user_data)
638{
639 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
640}
641
642/*****************************************************************************/
502/* event objects */ 643/* event objects */
503 644
504static void 645static void
505eq_event_push (void *data1, void *data2, void *data3) 646eq_event_push (void *data1, void *data2, void *data3)
506{ 647{
507 dSP; 648 dSP;
508 PUSH_CLOBJ ("OpenCL::Event", data1); 649 PUSH_CLOBJ (stash_event, data1);
509 PUSHs (sv_2mortal (newSViv ((IV)data2))); 650 PUSHs (sv_2mortal (newSViv ((IV)data2)));
510 PUTBACK; 651 PUTBACK;
511} 652}
512 653
513static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 654static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
514 655
515static void CL_CALLBACK 656static void CL_CALLBACK
516eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data) 657eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
517{ 658{
518 clRetainEvent (event); 659 clRetainEvent (event);
519 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)event_command_exec_status, 0); 660 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
520} 661}
521 662
522/*****************************************************************************/ 663/*****************************************************************************/
664/* utilities for XS code */
523 665
524static size_t 666static size_t
525img_row_pitch (cl_mem img) 667img_row_pitch (cl_mem img)
526{ 668{
527 size_t res; 669 size_t res;
528 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 670 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
529 return res; 671 return res;
530} 672}
531 673
532static cl_event * 674static cl_event * ecb_noinline
533event_list (SV **items, cl_uint *rcount) 675event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
534{ 676{
535 cl_uint count = *rcount; 677 cl_uint count = *rcount;
536 678
537 if (!count) 679 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
680 *rcount = count = 0;
681
682 if (!count && !extra)
538 return 0; 683 return 0;
539 684
540 cl_event *list = tmpbuf (sizeof (cl_event) * count); 685 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
541 int i = 0; 686 int i = 0;
542 687
543 do 688 while (count--)
544 {
545 --count;
546 if (SvOK (items [count])) 689 if (SvOK (items [count]))
547 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 690 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
548 } 691
549 while (count); 692 if (extra)
693 list [i++] = extra;
550 694
551 *rcount = i; 695 *rcount = i;
552 696
553 return i ? list : 0; 697 return i ? list : 0;
554} 698}
555 699
556#define EVENT_LIST(items,count) \ 700#define EVENT_LIST(skip) \
557 cl_uint event_list_count = (count); \ 701 cl_uint event_list_count = items - (skip); \
558 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 702 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
559 703
560#define INFO(class) \ 704#define INFO(class) \
561{ \ 705{ \
562 size_t size; \ 706 size_t size; \
563 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 707 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
567 SvCUR_set (sv, size); \ 711 SvCUR_set (sv, size); \
568 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 712 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
569 XPUSHs (sv); \ 713 XPUSHs (sv); \
570} 714}
571 715
716/*****************************************************************************/
717/* mapped_xxx */
718
719static OpenCL__Mapped
720SvMAPPED (SV *sv)
721{
722 // no typechecking atm., keep your fingers crossed
723 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
724}
725
726struct mapped
727{
728 cl_command_queue queue;
729 cl_mem memobj;
730 void *ptr;
731 size_t cb;
732 cl_event event;
733 size_t row_pitch;
734 size_t slice_pitch;
735
736 size_t element_size;
737 size_t width, height, depth;
738};
739
740static SV *
741mapped_new (
742 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
743 void *ptr, size_t cb, cl_event ev,
744 size_t row_pitch, size_t slice_pitch, size_t element_size,
745 size_t width, size_t height, size_t depth
746)
747{
748 SV *data = newSV (0);
749 SvUPGRADE (data, SVt_PVMG);
750
751 OpenCL__Mapped mapped;
752 New (0, mapped, 1, struct mapped);
753
754 clRetainCommandQueue (queue);
755
756 mapped->queue = queue;
757 mapped->memobj = memobj;
758 mapped->ptr = ptr;
759 mapped->cb = cb;
760 mapped->event = ev;
761 mapped->row_pitch = row_pitch;
762 mapped->slice_pitch = slice_pitch;
763
764 mapped->element_size = element_size;
765 mapped->width = width;
766 mapped->height = height;
767 mapped->depth = depth;
768
769 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
770
771 if (SvLEN (data))
772 Safefree (data);
773
774 SvPVX (data) = (char *)ptr;
775 SvCUR_set (data, cb);
776 SvLEN_set (data, 0);
777 SvPOK_only (data);
778
779 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
780
781 if (!(flags & CL_MAP_WRITE))
782 SvREADONLY_on (data);
783
784 return obj;
785}
786
787static void
788mapped_detach (SV *sv, OpenCL__Mapped mapped)
789{
790 SV *data = SvRV (sv);
791
792 // the next check checks both before AND after detach, where SvPVX should STILL be 0
793 if (SvPVX (data) != (char *)mapped->ptr)
794 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
795 else
796 {
797 SvREADONLY_off (data);
798 SvCUR_set (data, 0);
799 SvPVX (data) = 0;
800 SvOK_off (data);
801 }
802
803 mapped->ptr = 0;
804}
805
806static void
807mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
808{
809 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
810 cl_event ev;
811
812 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
813
814 clReleaseEvent (mapped->event);
815 mapped->event = ev;
816
817 mapped_detach (self, mapped);
818}
819
820static size_t
821mapped_element_size (OpenCL__Mapped self)
822{
823 if (!self->element_size)
824 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
825
826 return self->element_size;
827}
828
829/*****************************************************************************/
830
572MODULE = OpenCL PACKAGE = OpenCL 831MODULE = OpenCL PACKAGE = OpenCL
573 832
574PROTOTYPES: ENABLE 833PROTOTYPES: ENABLE
575 834
576void 835void
585 eq_signal_arg = (void*)arg; 844 eq_signal_arg = (void*)arg;
586 845
587BOOT: 846BOOT:
588{ 847{
589 HV *stash = gv_stashpv ("OpenCL", 1); 848 HV *stash = gv_stashpv ("OpenCL", 1);
849
590 static const ivstr *civ, const_iv[] = { 850 static const ivstr *civ, const_iv[] = {
591 { sizeof (cl_char ), "SIZEOF_CHAR" }, 851 { sizeof (cl_char ), "SIZEOF_CHAR" },
592 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 852 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
593 { sizeof (cl_short ), "SIZEOF_SHORT" }, 853 { sizeof (cl_short ), "SIZEOF_SHORT" },
594 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 854 { sizeof (cl_ushort), "SIZEOF_USHORT" },
597 { sizeof (cl_long ), "SIZEOF_LONG" }, 857 { sizeof (cl_long ), "SIZEOF_LONG" },
598 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 858 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
599 { sizeof (cl_half ), "SIZEOF_HALF" }, 859 { sizeof (cl_half ), "SIZEOF_HALF" },
600 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 860 { sizeof (cl_float ), "SIZEOF_FLOAT" },
601 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 861 { sizeof (cl_double), "SIZEOF_DOUBLE" },
862 { PREFER_1_1 , "PREFER_1_1" },
602#include "constiv.h" 863#include "constiv.h"
603 }; 864 };
604 865
605 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 866 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
606 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 867 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
868
869 static const nvstr *cnv, const_nv[] = {
870#include "constnv.h"
871 };
872
873 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
874 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
875
876 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
877
878 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
879 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
880 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
881 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
882 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
883 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
884 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
885 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
886 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
887 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
888 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
889 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
890 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
891 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
892 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
893 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
894 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
895 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
896 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
897 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
898 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
899 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
900 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
607 901
608 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 902 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
609} 903}
610 904
611cl_int 905cl_int
614 RETVAL = res; 908 RETVAL = res;
615 OUTPUT: 909 OUTPUT:
616 RETVAL 910 RETVAL
617 911
618const char * 912const char *
619err2str (cl_int err) 913err2str (cl_int err = res)
620 914
621const char * 915const char *
622enum2str (cl_uint value) 916enum2str (cl_uint value)
623 917
624void 918void
632 list = tmpbuf (sizeof (*list) * count); 926 list = tmpbuf (sizeof (*list) * count);
633 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 927 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
634 928
635 EXTEND (SP, count); 929 EXTEND (SP, count);
636 for (i = 0; i < count; ++i) 930 for (i = 0; i < count; ++i)
637 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 931 PUSH_CLOBJ (stash_platform, list [i]);
638 932
639void 933void
640context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 934context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
641 PPCODE: 935 PPCODE:
936 CONTEXT_NOTIFY_CALLBACK;
642 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 937 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
643 XPUSH_CLOBJ ("OpenCL::Context", ctx); 938 XPUSH_CLOBJ_CONTEXT;
644 939
645#if 0
646
647void 940void
648context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 941context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
649 PPCODE: 942 PPCODE:
650 /* der Gipfel der Kunst */ 943 cl_uint device_count;
944 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
651 945
652#endif 946 CONTEXT_NOTIFY_CALLBACK;
947 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
948 XPUSH_CLOBJ_CONTEXT;
653 949
654void 950void
655wait_for_events (...) 951wait_for_events (...)
656 CODE: 952 CODE:
657 EVENT_LIST (0, items); 953 EVENT_LIST (0);
658 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 954 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
659 955
660PROTOTYPES: DISABLE 956PROTOTYPES: DISABLE
661 957
662MODULE = OpenCL PACKAGE = OpenCL::Platform 958MODULE = OpenCL PACKAGE = OpenCL::Platform
705 list = tmpbuf (sizeof (*list) * count); 1001 list = tmpbuf (sizeof (*list) * count);
706 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 1002 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
707 1003
708 EXTEND (SP, count); 1004 EXTEND (SP, count);
709 for (i = 0; i < count; ++i) 1005 for (i = 0; i < count; ++i)
710 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1006 PUSH_CLOBJ (stash_device, list [i]);
711 1007
712void 1008void
713context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = 0) 1009context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
714 PPCODE: 1010 PPCODE:
1011 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1012 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1013
1014 cl_uint device_count;
1015 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1016
1017 CONTEXT_NOTIFY_CALLBACK;
1018 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1019 XPUSH_CLOBJ_CONTEXT;
1020
1021void
1022context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1023 PPCODE:
1024 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1025 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1026
1027 CONTEXT_NOTIFY_CALLBACK;
1028 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1029 XPUSH_CLOBJ_CONTEXT;
1030
1031MODULE = OpenCL PACKAGE = OpenCL::Device
1032
1033void
1034info (OpenCL::Device self, cl_device_info name)
1035 PPCODE:
1036 INFO (Device)
1037
1038#if CL_VERSION_1_2
1039
1040void
1041sub_devices (OpenCL::Device self, SV *properties)
1042 PPCODE:
715 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1043 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
716 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1044 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
717 1045
718 AV *av = (AV *)SvRV (devices); 1046 properties = SvRV (properties);
719 cl_uint num_devices = av_len (av) + 1; 1047
720 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1048 cl_uint count = av_len ((AV *)properties) + 1;
1049 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
721 1050
722 int i; 1051 int i;
723 for (i = num_devices; i--; ) 1052 for (i = 0; i < count; ++i)
724 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1053 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
725 1054
726 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; 1055 props [count] = 0;
727 void *user_data = 0;
728
729 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
730 XPUSH_CLOBJ ("OpenCL::Context", ctx);
731 1056
732void 1057 cl_uint num_devices;
733context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1058 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
734 PPCODE: 1059 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
735 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1060 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
736 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
737 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
738 XPUSH_CLOBJ ("OpenCL::Context", ctx);
739 1061
740MODULE = OpenCL PACKAGE = OpenCL::Device 1062 EXTEND (SP, num_devices);
1063 for (i = 0; i < count; ++i)
1064 PUSH_CLOBJ (stash_subdevice, list [i]);
741 1065
742void 1066#endif
743info (OpenCL::Device self, cl_device_info name)
744 PPCODE:
745 INFO (Device)
746 1067
747#BEGIN:device 1068#BEGIN:device
748 1069
749void 1070void
750type (OpenCL::Device self) 1071type (OpenCL::Device self)
1072 ALIAS:
1073 type = CL_DEVICE_TYPE
1074 address_bits = CL_DEVICE_ADDRESS_BITS
1075 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1076 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1077 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1078 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1079 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1080 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1081 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1082 properties = CL_DEVICE_QUEUE_PROPERTIES
1083 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1084 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
751 PPCODE: 1085 PPCODE:
752 cl_device_type value [1]; 1086 cl_ulong value [1];
753 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
754 EXTEND (SP, 1); 1088 EXTEND (SP, 1);
755 const int i = 0; 1089 const int i = 0;
756 PUSHs (sv_2mortal (newSViv (value [i]))); 1090 PUSHs (sv_2mortal (newSVuv (value [i])));
757 1091
758void 1092void
759vendor_id (OpenCL::Device self) 1093vendor_id (OpenCL::Device self)
760 ALIAS: 1094 ALIAS:
761 vendor_id = CL_DEVICE_VENDOR_ID 1095 vendor_id = CL_DEVICE_VENDOR_ID
772 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1106 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
773 image_support = CL_DEVICE_IMAGE_SUPPORT 1107 image_support = CL_DEVICE_IMAGE_SUPPORT
774 max_samplers = CL_DEVICE_MAX_SAMPLERS 1108 max_samplers = CL_DEVICE_MAX_SAMPLERS
775 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1109 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
776 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1110 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1111 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
777 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1112 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
778 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1113 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1114 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
779 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1115 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
780 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1116 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
781 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1117 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
782 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1118 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
783 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1119 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
821 EXTEND (SP, n); 1157 EXTEND (SP, n);
822 for (i = 0; i < n; ++i) 1158 for (i = 0; i < n; ++i)
823 PUSHs (sv_2mortal (newSVuv (value [i]))); 1159 PUSHs (sv_2mortal (newSVuv (value [i])));
824 1160
825void 1161void
826address_bits (OpenCL::Device self)
827 PPCODE:
828 cl_bitfield value [1];
829 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
830 EXTEND (SP, 1);
831 const int i = 0;
832 PUSHs (sv_2mortal (newSVuv (value [i])));
833
834void
835max_mem_alloc_size (OpenCL::Device self)
836 ALIAS:
837 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
838 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
839 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
840 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
841 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
842 PPCODE:
843 cl_ulong value [1];
844 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
845 EXTEND (SP, 1);
846 const int i = 0;
847 PUSHs (sv_2mortal (newSVuv (value [i])));
848
849void
850single_fp_config (OpenCL::Device self)
851 ALIAS:
852 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
853 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
854 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
855 PPCODE:
856 cl_device_fp_config value [1];
857 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
858 EXTEND (SP, 1);
859 const int i = 0;
860 PUSHs (sv_2mortal (newSVuv (value [i])));
861
862void
863global_mem_cache_type (OpenCL::Device self)
864 PPCODE:
865 cl_device_mem_cache_type value [1];
866 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
867 EXTEND (SP, 1);
868 const int i = 0;
869 PUSHs (sv_2mortal (newSVuv (value [i])));
870
871void
872local_mem_type (OpenCL::Device self)
873 PPCODE:
874 cl_device_local_mem_type value [1];
875 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
876 EXTEND (SP, 1);
877 const int i = 0;
878 PUSHs (sv_2mortal (newSVuv (value [i])));
879
880void
881error_correction_support (OpenCL::Device self) 1162error_correction_support (OpenCL::Device self)
882 ALIAS: 1163 ALIAS:
883 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1164 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
884 endian_little = CL_DEVICE_ENDIAN_LITTLE 1165 endian_little = CL_DEVICE_ENDIAN_LITTLE
885 available = CL_DEVICE_AVAILABLE 1166 available = CL_DEVICE_AVAILABLE
886 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1167 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
887 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1168 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
888 PPCODE: 1169 PPCODE:
889 cl_bool value [1]; 1170 cl_uint value [1];
890 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1171 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
891 EXTEND (SP, 1); 1172 EXTEND (SP, 1);
892 const int i = 0; 1173 const int i = 0;
893 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1174 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
894
895void
896execution_capabilities (OpenCL::Device self)
897 PPCODE:
898 cl_device_exec_capabilities value [1];
899 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
900 EXTEND (SP, 1);
901 const int i = 0;
902 PUSHs (sv_2mortal (newSVuv (value [i])));
903
904void
905properties (OpenCL::Device self)
906 PPCODE:
907 cl_command_queue_properties value [1];
908 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
909 EXTEND (SP, 1);
910 const int i = 0;
911 PUSHs (sv_2mortal (newSViv (value [i])));
912 1175
913void 1176void
914platform (OpenCL::Device self) 1177platform (OpenCL::Device self)
915 PPCODE: 1178 PPCODE:
916 cl_platform_id value [1]; 1179 cl_platform_id value [1];
917 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1180 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
918 EXTEND (SP, 1); 1181 EXTEND (SP, 1);
919 const int i = 0; 1182 const int i = 0;
920 {
921 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1183 PUSH_CLOBJ (stash_platform, value [i]);
922 }
923 1184
924void 1185void
925name (OpenCL::Device self) 1186name (OpenCL::Device self)
926 ALIAS: 1187 ALIAS:
927 name = CL_DEVICE_NAME 1188 name = CL_DEVICE_NAME
944 PPCODE: 1205 PPCODE:
945 cl_device_id value [1]; 1206 cl_device_id value [1];
946 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1207 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
947 EXTEND (SP, 1); 1208 EXTEND (SP, 1);
948 const int i = 0; 1209 const int i = 0;
949 {
950 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1210 PUSH_CLOBJ (stash_device, value [i]);
951 }
952 1211
953void 1212void
954partition_types_ext (OpenCL::Device self) 1213partition_types_ext (OpenCL::Device self)
955 ALIAS: 1214 ALIAS:
956 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1215 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
966 for (i = 0; i < n; ++i) 1225 for (i = 0; i < n; ++i)
967 PUSHs (sv_2mortal (newSVuv (value [i]))); 1226 PUSHs (sv_2mortal (newSVuv (value [i])));
968 1227
969#END:device 1228#END:device
970 1229
1230MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1231
1232#if CL_VERSION_1_2
1233
1234void
1235DESTROY (OpenCL::SubDevice self)
1236 CODE:
1237 clReleaseDevice (self);
1238
1239#endif
1240
971MODULE = OpenCL PACKAGE = OpenCL::Context 1241MODULE = OpenCL PACKAGE = OpenCL::Context
972 1242
973void 1243void
974DESTROY (OpenCL::Context context) 1244DESTROY (OpenCL::Context self)
975 CODE: 1245 CODE:
976 clReleaseContext (context); 1246 clReleaseContext (self);
977 1247
978void 1248void
979info (OpenCL::Context self, cl_context_info name) 1249info (OpenCL::Context self, cl_context_info name)
980 PPCODE: 1250 PPCODE:
981 INFO (Context) 1251 INFO (Context)
982 1252
983void 1253void
984queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1254queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
985 PPCODE: 1255 PPCODE:
986 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1256 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
987 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1257 XPUSH_CLOBJ (stash_queue, queue);
988 1258
989void 1259void
990user_event (OpenCL::Context self) 1260user_event (OpenCL::Context self)
991 PPCODE: 1261 PPCODE:
992 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1262 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
993 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1263 XPUSH_CLOBJ (stash_userevent, ev);
994 1264
995void 1265void
996buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1266buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
997 PPCODE: 1267 PPCODE:
998 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1268 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
999 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1269 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
1000 1270
1001 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1002 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1272 XPUSH_CLOBJ (stash_bufferobj, mem);
1003 1273
1004void 1274void
1005buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1275buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
1006 PPCODE: 1276 PPCODE:
1007 STRLEN len; 1277 STRLEN len;
1008 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1278 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1009 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1279 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
1010 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1280 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
1011 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
1012 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1282 XPUSH_CLOBJ (stash_bufferobj, mem);
1013 1283
1014#if CL_VERSION_1_2 1284#if CL_VERSION_1_2
1015 1285
1016void 1286void
1017image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1287image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
1018 PPCODE: 1288 PPCODE:
1019 STRLEN len; 1289 STRLEN len;
1020 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1290 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1021 const cl_image_format format = { channel_order, channel_type }; 1291 const cl_image_format format = { channel_order, channel_type };
1022 const cl_image_desc desc = { 1292 const cl_image_desc desc = {
1023 type, 1293 type,
1024 width, height, depth, 1294 width, height, depth,
1025 array_size, row_pitch, slice_pitch, 1295 array_size, row_pitch, slice_pitch,
1026 num_mip_level, num_samples, 1296 num_mip_level, num_samples,
1027 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1297 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1028 }; 1298 };
1029 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1299 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1030 char *klass = "OpenCL::Image"; 1300 HV *stash = stash_image;
1031 switch (type) 1301 switch (type)
1032 { 1302 {
1033 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1303 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1034 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1304 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1035 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1305 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1036 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1306 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1037 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1307 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1038 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1308 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1039 } 1309 }
1040 XPUSH_CLOBJ (klass, mem); 1310 XPUSH_CLOBJ (stash, mem);
1041 1311
1042#endif 1312#endif
1043 1313
1044void 1314void
1045image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1315image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
1051 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1052#else 1322#else
1053 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1323 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1054 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1055#endif 1325#endif
1056 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1326 XPUSH_CLOBJ (stash_image2d, mem);
1057 1327
1058void 1328void
1059image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1329image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1060 PPCODE: 1330 PPCODE:
1061 STRLEN len; 1331 STRLEN len;
1065 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1066#else 1336#else
1067 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1337 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1068 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1069#endif 1339#endif
1070 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1340 XPUSH_CLOBJ (stash_image3d, mem);
1071 1341
1072#if cl_apple_gl_sharing || cl_khr_gl_sharing 1342#if cl_apple_gl_sharing || cl_khr_gl_sharing
1073 1343
1074void 1344void
1075gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1345gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1076 PPCODE: 1346 PPCODE:
1077 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1078 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1348 XPUSH_CLOBJ (stash_bufferobj, mem);
1079 1349
1080void 1350void
1081gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1351gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1082 PPCODE: 1352 PPCODE:
1083 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1353 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1084 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1354 XPUSH_CLOBJ (stash_image2d, mem);
1085 1355
1086#if CL_VERSION_1_2 1356#if CL_VERSION_1_2
1087 1357
1088void 1358void
1089gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1359gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1090 ALIAS: 1360 ALIAS:
1091 PPCODE: 1361 PPCODE:
1092 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1362 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1093 cl_gl_object_type type; 1363 cl_gl_object_type type;
1094 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1364 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1095 char *klass = "OpenCL::Memory"; 1365 HV *stash = stash_memory;
1096 switch (type) 1366 switch (type)
1097 { 1367 {
1098 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1368 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1099 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1369 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1100 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1370 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1101 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1371 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1102 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1372 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1103 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1373 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1104 } 1374 }
1105 XPUSH_CLOBJ (klass, mem); 1375 XPUSH_CLOBJ (stash, mem);
1106 1376
1107#endif 1377#endif
1108 1378
1109void 1379void
1110gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1380gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1112#if PREFER_1_1 1382#if PREFER_1_1
1113 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1114#else 1384#else
1115 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1385 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1116#endif 1386#endif
1117 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1387 XPUSH_CLOBJ (stash_image2d, mem);
1118 1388
1119void 1389void
1120gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1390gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1121 PPCODE: 1391 PPCODE:
1122#if PREFER_1_1 1392#if PREFER_1_1
1123 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1124#else 1394#else
1125 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1395 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1126#endif 1396#endif
1127 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1397 XPUSH_CLOBJ (stash_image3d, mem);
1128 1398
1129#endif 1399#endif
1130 1400
1131void 1401void
1132supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1402supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1152 1422
1153void 1423void
1154sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1424sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1155 PPCODE: 1425 PPCODE:
1156 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1426 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1157 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1427 XPUSH_CLOBJ (stash_sampler, sampler);
1158 1428
1159void 1429void
1160program_with_source (OpenCL::Context self, SV *program) 1430program_with_source (OpenCL::Context self, SV *program)
1161 PPCODE: 1431 PPCODE:
1162 STRLEN len; 1432 STRLEN len;
1163 size_t len2; 1433 size_t len2;
1164 const char *ptr = SvPVbyte (program, len); 1434 const char *ptr = SvPVbyte (program, len);
1165 1435
1166 len2 = len; 1436 len2 = len;
1167 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1437 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1168 XPUSH_CLOBJ ("OpenCL::Program", prog); 1438 XPUSH_CLOBJ (stash_program, prog);
1439
1440void
1441program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1442 PPCODE:
1443 cl_uint device_count;
1444 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1445
1446 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1447 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1448
1449 binaries = SvRV (binaries);
1450
1451 if (device_count != av_len ((AV *)binaries) + 1)
1452 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1453
1454 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1455 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1456 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1457
1458 int i;
1459 for (i = 0; i < device_count; ++i)
1460 {
1461 STRLEN len;
1462 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1463 length_list [i] = len;
1464 }
1465
1466 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1467 length_list, binary_list,
1468 GIMME_V == G_ARRAY ? status_list : 0, &res));
1469
1470 EXTEND (SP, 2);
1471 PUSH_CLOBJ (stash_program, prog);
1472
1473 if (GIMME_V == G_ARRAY)
1474 {
1475 AV *av = newAV ();
1476 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1477
1478 for (i = device_count; i--; )
1479 av_store (av, i, newSViv (status_list [i]));
1480 }
1481
1482#if CL_VERSION_1_2
1483
1484void
1485program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1486 PPCODE:
1487 cl_uint device_count;
1488 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1489
1490 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1491
1492 XPUSH_CLOBJ (stash_program, prog);
1493
1494void
1495link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1496 CODE:
1497 cl_uint device_count = 0;
1498 cl_device_id *device_list = 0;
1499
1500 if (SvOK (devices))
1501 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1502
1503 cl_uint program_count;
1504 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1505
1506 void *user_data;
1507 program_callback pfn_notify = make_program_callback (notify, &user_data);
1508
1509 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1510 program_count, program_list, pfn_notify, user_data, &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514#endif
1169 1515
1170#BEGIN:context 1516#BEGIN:context
1171 1517
1172void 1518void
1173reference_count (OpenCL::Context self) 1519reference_count (OpenCL::Context self)
1189 cl_device_id *value = tmpbuf (size); 1535 cl_device_id *value = tmpbuf (size);
1190 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1536 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1191 int i, n = size / sizeof (*value); 1537 int i, n = size / sizeof (*value);
1192 EXTEND (SP, n); 1538 EXTEND (SP, n);
1193 for (i = 0; i < n; ++i) 1539 for (i = 0; i < n; ++i)
1194 {
1195 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1540 PUSH_CLOBJ (stash_device, value [i]);
1196 }
1197 1541
1198void 1542void
1199properties (OpenCL::Context self) 1543properties (OpenCL::Context self)
1200 PPCODE: 1544 PPCODE:
1201 size_t size; 1545 size_t size;
1215DESTROY (OpenCL::Queue self) 1559DESTROY (OpenCL::Queue self)
1216 CODE: 1560 CODE:
1217 clReleaseCommandQueue (self); 1561 clReleaseCommandQueue (self);
1218 1562
1219void 1563void
1220enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1564read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1565 ALIAS:
1566 enqueue_read_buffer = 0
1221 PPCODE: 1567 PPCODE:
1222 cl_event ev = 0;
1223 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
1224 1569
1225 SvUPGRADE (data, SVt_PV); 1570 SvUPGRADE (data, SVt_PV);
1226 SvGROW (data, len); 1571 SvGROW (data, len);
1227 SvPOK_only (data); 1572 SvPOK_only (data);
1228 SvCUR_set (data, len); 1573 SvCUR_set (data, len);
1574
1575 cl_event ev = 0;
1229 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1230 1577
1231 if (ev) 1578 if (ev)
1232 XPUSH_CLOBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
1233 1580
1234void 1581void
1235enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1582write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1583 ALIAS:
1584 enqueue_write_buffer = 0
1236 PPCODE: 1585 PPCODE:
1237 cl_event ev = 0; 1586 EVENT_LIST (5);
1587
1238 STRLEN len; 1588 STRLEN len;
1239 char *ptr = SvPVbyte (data, len); 1589 char *ptr = SvPVbyte (data, len);
1240 EVENT_LIST (5, items - 5);
1241 1590
1591 cl_event ev = 0;
1242 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1592 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1243 1593
1244 if (ev) 1594 if (ev)
1245 XPUSH_CLOBJ ("OpenCL::Event", ev); 1595 XPUSH_CLOBJ (stash_event, ev);
1246 1596
1247#if CL_VERSION_1_2 1597#if CL_VERSION_1_2
1248 1598
1249void 1599void
1250enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1600fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1601 ALIAS:
1602 enqueue_fill_buffer = 0
1251 PPCODE: 1603 PPCODE:
1252 cl_event ev = 0; 1604 EVENT_LIST (5);
1605
1253 STRLEN len; 1606 STRLEN len;
1254 char *ptr = SvPVbyte (data, len); 1607 char *ptr = SvPVbyte (data, len);
1255 EVENT_LIST (5, items - 5);
1256 1608
1609 cl_event ev = 0;
1257 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1610 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1258 1611
1259 if (ev) 1612 if (ev)
1260 XPUSH_CLOBJ ("OpenCL::Event", ev); 1613 XPUSH_CLOBJ (stash_event, ev);
1261 1614
1262void 1615void
1263enqueue_fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 1616fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1617 ALIAS:
1618 enqueue_fill_image = 0
1264 PPCODE: 1619 PPCODE:
1265 cl_event ev = 0; 1620 EVENT_LIST (12);
1266 STRLEN len; 1621
1267 const size_t origin [3] = { x, y, z }; 1622 const size_t origin [3] = { x, y, z };
1268 const size_t region [3] = { width, height, depth }; 1623 const size_t region [3] = { width, height, depth };
1269 EVENT_LIST (12, items - 12);
1270 1624
1271 const cl_float c_f [4] = { r, g, b, a }; 1625 const cl_float c_f [4] = { r, g, b, a };
1272 const cl_uint c_u [4] = { r, g, b, a }; 1626 const cl_uint c_u [4] = { r, g, b, a };
1273 const cl_int c_s [4] = { r, g, b, a }; 1627 const cl_int c_s [4] = { r, g, b, a };
1274 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1628 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1277 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1631 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1278 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1632 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1279 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1633 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1280 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1634 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1281 1635
1636 cl_event ev = 0;
1282 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1637 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1283 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1638 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1284 1639
1285 if (ev) 1640 if (ev)
1286 XPUSH_CLOBJ ("OpenCL::Event", ev); 1641 XPUSH_CLOBJ (stash_event, ev);
1287 1642
1288#endif 1643#endif
1289 1644
1290void 1645void
1291enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1646copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1647 ALIAS:
1648 enqueue_copy_buffer = 0
1292 PPCODE: 1649 PPCODE:
1650 EVENT_LIST (6);
1651
1293 cl_event ev = 0; 1652 cl_event ev = 0;
1294 EVENT_LIST (6, items - 6);
1295
1296 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1653 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1297 1654
1298 if (ev) 1655 if (ev)
1299 XPUSH_CLOBJ ("OpenCL::Event", ev); 1656 XPUSH_CLOBJ (stash_event, ev);
1300 1657
1301void 1658void
1302enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1659read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1660 ALIAS:
1661 enqueue_read_buffer_rect = 0
1303 PPCODE: 1662 PPCODE:
1304 cl_event ev = 0; 1663 EVENT_LIST (17);
1664
1305 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1665 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1306 const size_t host_origin[3] = { host_x, host_y, host_z }; 1666 const size_t host_origin[3] = { host_x, host_y, host_z };
1307 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1308 EVENT_LIST (17, items - 17);
1309 1668
1310 if (!buf_row_pitch) 1669 if (!buf_row_pitch)
1311 buf_row_pitch = region [0]; 1670 buf_row_pitch = region [0];
1312 1671
1313 if (!buf_slice_pitch) 1672 if (!buf_slice_pitch)
1323 1682
1324 SvUPGRADE (data, SVt_PV); 1683 SvUPGRADE (data, SVt_PV);
1325 SvGROW (data, len); 1684 SvGROW (data, len);
1326 SvPOK_only (data); 1685 SvPOK_only (data);
1327 SvCUR_set (data, len); 1686 SvCUR_set (data, len);
1687
1688 cl_event ev = 0;
1328 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1689 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1329 1690
1330 if (ev) 1691 if (ev)
1331 XPUSH_CLOBJ ("OpenCL::Event", ev); 1692 XPUSH_CLOBJ (stash_event, ev);
1332 1693
1333void 1694void
1334enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1695write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1696 ALIAS:
1697 enqueue_write_buffer_rect = 0
1335 PPCODE: 1698 PPCODE:
1336 cl_event ev = 0; 1699 EVENT_LIST (17);
1700
1337 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1701 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1338 const size_t host_origin[3] = { host_x, host_y, host_z }; 1702 const size_t host_origin[3] = { host_x, host_y, host_z };
1339 const size_t region[3] = { width, height, depth }; 1703 const size_t region[3] = { width, height, depth };
1340 STRLEN len; 1704 STRLEN len;
1341 char *ptr = SvPVbyte (data, len); 1705 char *ptr = SvPVbyte (data, len);
1342 EVENT_LIST (17, items - 17);
1343 1706
1344 if (!buf_row_pitch) 1707 if (!buf_row_pitch)
1345 buf_row_pitch = region [0]; 1708 buf_row_pitch = region [0];
1346 1709
1347 if (!buf_slice_pitch) 1710 if (!buf_slice_pitch)
1356 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1719 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1357 1720
1358 if (len < min_len) 1721 if (len < min_len)
1359 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1722 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1360 1723
1724 cl_event ev = 0;
1361 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1725 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1362 1726
1363 if (ev) 1727 if (ev)
1364 XPUSH_CLOBJ ("OpenCL::Event", ev); 1728 XPUSH_CLOBJ (stash_event, ev);
1365 1729
1366void 1730void
1367enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1731copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1732 ALIAS:
1733 enqueue_copy_buffer_rect = 0
1368 PPCODE: 1734 PPCODE:
1369 cl_event ev = 0; 1735 EVENT_LIST (16);
1736
1370 const size_t src_origin[3] = { src_x, src_y, src_z }; 1737 const size_t src_origin[3] = { src_x, src_y, src_z };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1738 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16);
1374 1740
1741 cl_event ev = 0;
1375 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1742 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1376 1743
1377 if (ev) 1744 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev); 1745 XPUSH_CLOBJ (stash_event, ev);
1379 1746
1380void 1747void
1381enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1748read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1749 ALIAS:
1750 enqueue_read_image = 0
1382 PPCODE: 1751 PPCODE:
1383 cl_event ev = 0; 1752 EVENT_LIST (12);
1753
1384 const size_t src_origin[3] = { src_x, src_y, src_z }; 1754 const size_t src_origin[3] = { src_x, src_y, src_z };
1385 const size_t region[3] = { width, height, depth }; 1755 const size_t region[3] = { width, height, depth };
1386 EVENT_LIST (12, items - 12);
1387 1756
1388 if (!row_pitch) 1757 if (!row_pitch)
1389 row_pitch = img_row_pitch (src); 1758 row_pitch = img_row_pitch (src);
1390 1759
1391 if (depth > 1 && !slice_pitch) 1760 if (depth > 1 && !slice_pitch)
1395 1764
1396 SvUPGRADE (data, SVt_PV); 1765 SvUPGRADE (data, SVt_PV);
1397 SvGROW (data, len); 1766 SvGROW (data, len);
1398 SvPOK_only (data); 1767 SvPOK_only (data);
1399 SvCUR_set (data, len); 1768 SvCUR_set (data, len);
1769
1770 cl_event ev = 0;
1400 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1771 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1401 1772
1402 if (ev) 1773 if (ev)
1403 XPUSH_CLOBJ ("OpenCL::Event", ev); 1774 XPUSH_CLOBJ (stash_event, ev);
1404 1775
1405void 1776void
1406enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1777write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1778 ALIAS:
1779 enqueue_write_image = 0
1407 PPCODE: 1780 PPCODE:
1408 cl_event ev = 0; 1781 EVENT_LIST (12);
1782
1409 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1783 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1410 const size_t region[3] = { width, height, depth }; 1784 const size_t region[3] = { width, height, depth };
1411 STRLEN len; 1785 STRLEN len;
1412 char *ptr = SvPVbyte (data, len); 1786 char *ptr = SvPVbyte (data, len);
1413 EVENT_LIST (12, items - 12);
1414 1787
1415 if (!row_pitch) 1788 if (!row_pitch)
1416 row_pitch = img_row_pitch (dst); 1789 row_pitch = img_row_pitch (dst);
1417 1790
1418 if (depth > 1 && !slice_pitch) 1791 if (depth > 1 && !slice_pitch)
1421 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1794 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1422 1795
1423 if (len < min_len) 1796 if (len < min_len)
1424 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1797 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1425 1798
1799 cl_event ev = 0;
1426 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1800 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1427 1801
1428 if (ev) 1802 if (ev)
1429 XPUSH_CLOBJ ("OpenCL::Event", ev); 1803 XPUSH_CLOBJ (stash_event, ev);
1430 1804
1431void 1805void
1432enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1806copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1807 ALIAS:
1808 enqueue_copy_image = 0
1433 PPCODE: 1809 PPCODE:
1434 cl_event ev = 0; 1810 EVENT_LIST (12);
1811
1435 const size_t src_origin[3] = { src_x, src_y, src_z }; 1812 const size_t src_origin[3] = { src_x, src_y, src_z };
1436 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1813 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1437 const size_t region[3] = { width, height, depth }; 1814 const size_t region[3] = { width, height, depth };
1438 EVENT_LIST (12, items - 12);
1439 1815
1816 cl_event ev = 0;
1440 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1817 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1818
1442 if (ev) 1819 if (ev)
1443 XPUSH_CLOBJ ("OpenCL::Event", ev); 1820 XPUSH_CLOBJ (stash_event, ev);
1444 1821
1445void 1822void
1446enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1823copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1824 ALIAS:
1825 enqueue_copy_image_to_buffer = 0
1447 PPCODE: 1826 PPCODE:
1827 EVENT_LIST (10);
1828
1829 const size_t src_origin[3] = { src_x, src_y, src_z };
1830 const size_t region [3] = { width, height, depth };
1831
1448 cl_event ev = 0; 1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1840 ALIAS:
1841 enqueue_copy_buffer_to_image = 0
1842 PPCODE:
1843 EVENT_LIST (10);
1844
1449 const size_t src_origin[3] = { src_x, src_y, src_z }; 1845 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1846 const size_t region [3] = { width, height, depth };
1847
1848 cl_event ev = 0;
1849 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1850
1851 if (ev)
1852 XPUSH_CLOBJ (stash_event, ev);
1853
1854void
1855map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1856 ALIAS:
1857 enqueue_map_buffer = 0
1858 PPCODE:
1859 EVENT_LIST (6);
1860
1861 size_t cb = SvIV (cb_);
1862
1863 if (!SvOK (cb_))
1864 {
1865 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1866 cb -= offset;
1867 }
1868
1869 cl_event ev;
1870 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1871 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1872
1873void
1874map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1875 ALIAS:
1876 enqueue_map_image = 0
1877 PPCODE:
1878 size_t width = SvIV (width_);
1879 if (!SvOK (width_))
1880 {
1881 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1882 width -= x;
1883 }
1884
1885 size_t height = SvIV (width_);
1886 if (!SvOK (height_))
1887 {
1888 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1889 height -= y;
1890
1891 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1892 if (!height)
1893 height = 1;
1894 }
1895
1896 size_t depth = SvIV (width_);
1897 if (!SvOK (depth_))
1898 {
1899 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1900 depth -= z;
1901
1902 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1903 if (!depth)
1904 depth = 1;
1905 }
1906
1907 const size_t origin[3] = { x, y, z };
1450 const size_t region[3] = { width, height, depth }; 1908 const size_t region[3] = { width, height, depth };
1909 size_t row_pitch, slice_pitch;
1451 EVENT_LIST (10, items - 10); 1910 EVENT_LIST (10);
1452 1911
1912 cl_event ev;
1913 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1914
1915 size_t cb = slice_pitch ? slice_pitch * region [2]
1916 : row_pitch ? row_pitch * region [1]
1917 : region [0];
1918
1919 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1920
1921void
1922unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1923 PPCODE:
1924 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1925 if (GIMME_V != G_VOID)
1926 {
1927 clRetainEvent (mapped->event);
1928 XPUSH_CLOBJ (stash_event, mapped->event);
1929 }
1930
1931void
1932task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1933 ALIAS:
1934 enqueue_task = 0
1935 PPCODE:
1936 EVENT_LIST (2);
1937
1938 cl_event ev = 0;
1453 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1939 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1454 1940
1455 if (ev) 1941 if (ev)
1456 XPUSH_CLOBJ ("OpenCL::Event", ev); 1942 XPUSH_CLOBJ (stash_event, ev);
1457 1943
1458void 1944void
1459enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1460 PPCODE:
1461 cl_event ev = 0;
1462 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1463 const size_t region[3] = { width, height, depth };
1464 EVENT_LIST (10, items - 10);
1465
1466 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1467
1468 if (ev)
1469 XPUSH_CLOBJ ("OpenCL::Event", ev);
1470
1471void
1472enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1473 PPCODE:
1474 cl_event ev = 0;
1475 EVENT_LIST (2, items - 2);
1476
1477 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1478
1479 if (ev)
1480 XPUSH_CLOBJ ("OpenCL::Event", ev);
1481
1482void
1483enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1945nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1946 ALIAS:
1947 enqueue_nd_range_kernel = 0
1484 PPCODE: 1948 PPCODE:
1485 cl_event ev = 0; 1949 EVENT_LIST (5);
1950
1486 size_t *gwo = 0, *gws, *lws = 0; 1951 size_t *gwo = 0, *gws, *lws = 0;
1487 int gws_len; 1952 int gws_len;
1488 size_t *lists; 1953 size_t *lists;
1489 int i; 1954 int i;
1490 EVENT_LIST (5, items - 5);
1491 1955
1492 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1956 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1493 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1957 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1494 1958
1495 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1959 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1496 1960
1497 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1961 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1498 1962
1499 gws = lists + gws_len * 0; 1963 gws = lists + gws_len * 0;
1500 for (i = 0; i < gws_len; ++i) 1964 for (i = 0; i < gws_len; ++i)
1965 {
1501 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1966 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1967 // at least nvidia crashes for 0-sized work group sizes, work around
1968 if (!gws [i])
1969 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1970 }
1502 1971
1503 if (SvOK (global_work_offset)) 1972 if (SvOK (global_work_offset))
1504 { 1973 {
1505 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1974 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1506 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1514 } 1983 }
1515 1984
1516 if (SvOK (local_work_size)) 1985 if (SvOK (local_work_size))
1517 { 1986 {
1518 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1987 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1519 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1988 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1520 1989
1521 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1990 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1522 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1991 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1523 1992
1524 lws = lists + gws_len * 2; 1993 lws = lists + gws_len * 2;
1525 for (i = 0; i < gws_len; ++i) 1994 for (i = 0; i < gws_len; ++i)
1995 {
1526 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1996 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1997 // at least nvidia crashes for 0-sized work group sizes, work around
1998 if (!lws [i])
1999 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2000 }
1527 } 2001 }
1528 2002
2003 cl_event ev = 0;
1529 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2004 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1530 2005
1531 if (ev) 2006 if (ev)
1532 XPUSH_CLOBJ ("OpenCL::Event", ev); 2007 XPUSH_CLOBJ (stash_event, ev);
2008
2009#if CL_VERSION_1_2
2010
2011void
2012migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2013 ALIAS:
2014 enqueue_migrate_mem_objects = 0
2015 PPCODE:
2016 EVENT_LIST (3);
2017
2018 cl_uint obj_count;
2019 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2020
2021 cl_event ev = 0;
2022 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2023
2024 if (ev)
2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#endif
1533 2028
1534#if cl_apple_gl_sharing || cl_khr_gl_sharing 2029#if cl_apple_gl_sharing || cl_khr_gl_sharing
1535 2030
1536void 2031void
1537enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2032acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1538 ALIAS: 2033 ALIAS:
2034 release_gl_objects = 1
2035 enqueue_acquire_gl_objects = 0
1539 enqueue_release_gl_objects = 1 2036 enqueue_release_gl_objects = 1
1540 PPCODE: 2037 PPCODE:
1541 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2038 EVENT_LIST (2);
1542 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2039
2040 cl_uint obj_count;
2041 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1543 2042
1544 cl_event ev = 0; 2043 cl_event ev = 0;
1545 EVENT_LIST (2, items - 2);
1546 AV *av = (AV *)SvRV (objects);
1547 cl_uint num_objects = av_len (av) + 1;
1548 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1549 int i;
1550
1551 for (i = num_objects; i--; )
1552 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1553 2044
1554 if (ix) 2045 if (ix)
1555 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1556 else 2047 else
1557 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2048 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1558 2049
1559 if (ev) 2050 if (ev)
1560 XPUSH_CLOBJ ("OpenCL::Event", ev); 2051 XPUSH_CLOBJ (stash_event, ev);
1561 2052
1562#endif 2053#endif
1563 2054
1564void 2055void
1565enqueue_wait_for_events (OpenCL::Queue self, ...) 2056wait_for_events (OpenCL::Queue self, ...)
2057 ALIAS:
2058 enqueue_wait_for_events = 0
1566 CODE: 2059 CODE:
1567 EVENT_LIST (1, items - 1); 2060 EVENT_LIST (1);
1568#if PREFER_1_1 2061#if PREFER_1_1
1569 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1570#else 2063#else
1571 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2064 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1572#endif 2065#endif
1573 2066
1574void 2067void
1575enqueue_marker (OpenCL::Queue self, ...) 2068marker (OpenCL::Queue self, ...)
2069 ALIAS:
2070 enqueue_marker = 0
1576 PPCODE: 2071 PPCODE:
2072 EVENT_LIST (1);
1577 cl_event ev = 0; 2073 cl_event ev = 0;
1578 EVENT_LIST (1, items - 1);
1579#if PREFER_1_1 2074#if PREFER_1_1
1580 if (!event_list_count) 2075 if (!event_list_count)
1581 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2076 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1582 else 2077 else
1583#if CL_VERSION_1_2 2078#if CL_VERSION_1_2
1590#endif 2085#endif
1591#else 2086#else
1592 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2087 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1593#endif 2088#endif
1594 if (ev) 2089 if (ev)
1595 XPUSH_CLOBJ ("OpenCL::Event", ev); 2090 XPUSH_CLOBJ (stash_event, ev);
1596 2091
1597void 2092void
1598enqueue_barrier (OpenCL::Queue self, ...) 2093barrier (OpenCL::Queue self, ...)
2094 ALIAS:
2095 enqueue_barrier = 0
1599 PPCODE: 2096 PPCODE:
2097 EVENT_LIST (1);
1600 cl_event ev = 0; 2098 cl_event ev = 0;
1601 EVENT_LIST (1, items - 1);
1602#if PREFER_1_1 2099#if PREFER_1_1
1603 if (!event_list_count && GIMME_V == G_VOID) 2100 if (!event_list_count && GIMME_V == G_VOID)
1604 NEED_SUCCESS (EnqueueBarrier, (self)); 2101 NEED_SUCCESS (EnqueueBarrier, (self));
1605 else 2102 else
1606#if CL_VERSION_1_2 2103#if CL_VERSION_1_2
1616#endif 2113#endif
1617#else 2114#else
1618 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2115 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1619#endif 2116#endif
1620 if (ev) 2117 if (ev)
1621 XPUSH_CLOBJ ("OpenCL::Event", ev); 2118 XPUSH_CLOBJ (stash_event, ev);
1622 2119
1623void 2120void
1624flush (OpenCL::Queue self) 2121flush (OpenCL::Queue self)
1625 CODE: 2122 CODE:
1626 NEED_SUCCESS (Flush, (self)); 2123 NEED_SUCCESS (Flush, (self));
1642 PPCODE: 2139 PPCODE:
1643 cl_context value [1]; 2140 cl_context value [1];
1644 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1645 EXTEND (SP, 1); 2142 EXTEND (SP, 1);
1646 const int i = 0; 2143 const int i = 0;
1647 {
1648 NEED_SUCCESS (RetainContext, (value [i])); 2144 NEED_SUCCESS (RetainContext, (value [i]));
1649 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2145 PUSH_CLOBJ (stash_context, value [i]);
1650 }
1651 2146
1652void 2147void
1653device (OpenCL::Queue self) 2148device (OpenCL::Queue self)
1654 PPCODE: 2149 PPCODE:
1655 cl_device_id value [1]; 2150 cl_device_id value [1];
1656 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2151 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1657 EXTEND (SP, 1); 2152 EXTEND (SP, 1);
1658 const int i = 0; 2153 const int i = 0;
1659 {
1660 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2154 PUSH_CLOBJ (stash_device, value [i]);
1661 }
1662 2155
1663void 2156void
1664reference_count (OpenCL::Queue self) 2157reference_count (OpenCL::Queue self)
1665 PPCODE: 2158 PPCODE:
1666 cl_uint value [1]; 2159 cl_uint value [1];
1670 PUSHs (sv_2mortal (newSVuv (value [i]))); 2163 PUSHs (sv_2mortal (newSVuv (value [i])));
1671 2164
1672void 2165void
1673properties (OpenCL::Queue self) 2166properties (OpenCL::Queue self)
1674 PPCODE: 2167 PPCODE:
1675 cl_command_queue_properties value [1]; 2168 cl_ulong value [1];
1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2169 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1677 EXTEND (SP, 1); 2170 EXTEND (SP, 1);
1678 const int i = 0; 2171 const int i = 0;
1679 PUSHs (sv_2mortal (newSViv (value [i]))); 2172 PUSHs (sv_2mortal (newSVuv (value [i])));
1680 2173
1681#END:command_queue 2174#END:command_queue
1682 2175
1683MODULE = OpenCL PACKAGE = OpenCL::Memory 2176MODULE = OpenCL PACKAGE = OpenCL::Memory
1684 2177
1690void 2183void
1691info (OpenCL::Memory self, cl_mem_info name) 2184info (OpenCL::Memory self, cl_mem_info name)
1692 PPCODE: 2185 PPCODE:
1693 INFO (MemObject) 2186 INFO (MemObject)
1694 2187
2188void
2189destructor_callback (OpenCL::Memory self, SV *notify)
2190 PPCODE:
2191 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2192
1695#BEGIN:mem 2193#BEGIN:mem
1696 2194
1697void 2195void
1698type (OpenCL::Memory self) 2196type (OpenCL::Memory self)
2197 ALIAS:
2198 type = CL_MEM_TYPE
2199 map_count = CL_MEM_MAP_COUNT
2200 reference_count = CL_MEM_REFERENCE_COUNT
1699 PPCODE: 2201 PPCODE:
1700 cl_mem_object_type value [1]; 2202 cl_uint value [1];
1701 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2203 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1702 EXTEND (SP, 1); 2204 EXTEND (SP, 1);
1703 const int i = 0; 2205 const int i = 0;
1704 PUSHs (sv_2mortal (newSViv (value [i]))); 2206 PUSHs (sv_2mortal (newSVuv (value [i])));
1705 2207
1706void 2208void
1707flags (OpenCL::Memory self) 2209flags (OpenCL::Memory self)
1708 PPCODE: 2210 PPCODE:
1709 cl_mem_flags value [1]; 2211 cl_ulong value [1];
1710 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1711 EXTEND (SP, 1); 2213 EXTEND (SP, 1);
1712 const int i = 0; 2214 const int i = 0;
1713 PUSHs (sv_2mortal (newSViv (value [i]))); 2215 PUSHs (sv_2mortal (newSVuv (value [i])));
1714 2216
1715void 2217void
1716size (OpenCL::Memory self) 2218size (OpenCL::Memory self)
1717 ALIAS: 2219 ALIAS:
1718 size = CL_MEM_SIZE 2220 size = CL_MEM_SIZE
1732 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1733 const int i = 0; 2235 const int i = 0;
1734 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2236 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1735 2237
1736void 2238void
1737map_count (OpenCL::Memory self)
1738 ALIAS:
1739 map_count = CL_MEM_MAP_COUNT
1740 reference_count = CL_MEM_REFERENCE_COUNT
1741 PPCODE:
1742 cl_uint value [1];
1743 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1744 EXTEND (SP, 1);
1745 const int i = 0;
1746 PUSHs (sv_2mortal (newSVuv (value [i])));
1747
1748void
1749context (OpenCL::Memory self) 2239context (OpenCL::Memory self)
1750 PPCODE: 2240 PPCODE:
1751 cl_context value [1]; 2241 cl_context value [1];
1752 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2242 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1753 EXTEND (SP, 1); 2243 EXTEND (SP, 1);
1754 const int i = 0; 2244 const int i = 0;
1755 {
1756 NEED_SUCCESS (RetainContext, (value [i])); 2245 NEED_SUCCESS (RetainContext, (value [i]));
1757 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2246 PUSH_CLOBJ (stash_context, value [i]);
1758 }
1759 2247
1760void 2248void
1761associated_memobject (OpenCL::Memory self) 2249associated_memobject (OpenCL::Memory self)
1762 PPCODE: 2250 PPCODE:
1763 cl_mem value [1]; 2251 cl_mem value [1];
1764 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1765 EXTEND (SP, 1); 2253 EXTEND (SP, 1);
1766 const int i = 0; 2254 const int i = 0;
1767 {
1768 NEED_SUCCESS (RetainMemObject, (value [i])); 2255 NEED_SUCCESS (RetainMemObject, (value [i]));
1769 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2256 PUSH_CLOBJ (stash_memory, value [i]);
1770 }
1771 2257
1772#END:mem 2258#END:mem
1773 2259
1774#if cl_apple_gl_sharing || cl_khr_gl_sharing 2260#if cl_apple_gl_sharing || cl_khr_gl_sharing
1775 2261
1794 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2280 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1795 2281
1796 cl_buffer_region crdata = { origin, size }; 2282 cl_buffer_region crdata = { origin, size };
1797 2283
1798 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2284 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1799 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2285 XPUSH_CLOBJ (stash_buffer, mem);
1800 2286
1801MODULE = OpenCL PACKAGE = OpenCL::Image 2287MODULE = OpenCL PACKAGE = OpenCL::Image
1802 2288
1803void 2289void
1804image_info (OpenCL::Image self, cl_image_info name) 2290image_info (OpenCL::Image self, cl_image_info name)
1874 2360
1875#BEGIN:sampler 2361#BEGIN:sampler
1876 2362
1877void 2363void
1878reference_count (OpenCL::Sampler self) 2364reference_count (OpenCL::Sampler self)
2365 ALIAS:
2366 reference_count = CL_SAMPLER_REFERENCE_COUNT
2367 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2368 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1879 PPCODE: 2369 PPCODE:
1880 cl_uint value [1]; 2370 cl_uint value [1];
1881 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1882 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
1883 const int i = 0; 2373 const int i = 0;
1884 PUSHs (sv_2mortal (newSVuv (value [i]))); 2374 PUSHs (sv_2mortal (newSVuv (value [i])));
1885 2375
1886void 2376void
1888 PPCODE: 2378 PPCODE:
1889 cl_context value [1]; 2379 cl_context value [1];
1890 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2380 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1891 EXTEND (SP, 1); 2381 EXTEND (SP, 1);
1892 const int i = 0; 2382 const int i = 0;
1893 {
1894 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
1895 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2384 PUSH_CLOBJ (stash_context, value [i]);
1896 }
1897
1898void
1899normalized_coords (OpenCL::Sampler self)
1900 PPCODE:
1901 cl_addressing_mode value [1];
1902 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1903 EXTEND (SP, 1);
1904 const int i = 0;
1905 PUSHs (sv_2mortal (newSViv (value [i])));
1906
1907void
1908addressing_mode (OpenCL::Sampler self)
1909 PPCODE:
1910 cl_filter_mode value [1];
1911 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1912 EXTEND (SP, 1);
1913 const int i = 0;
1914 PUSHs (sv_2mortal (newSViv (value [i])));
1915 2385
1916void 2386void
1917filter_mode (OpenCL::Sampler self) 2387filter_mode (OpenCL::Sampler self)
1918 PPCODE: 2388 PPCODE:
1919 cl_bool value [1]; 2389 cl_uint value [1];
1920 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1921 EXTEND (SP, 1); 2391 EXTEND (SP, 1);
1922 const int i = 0; 2392 const int i = 0;
1923 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2393 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1924 2394
1934void 2404void
1935build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2405build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1936 ALIAS: 2406 ALIAS:
1937 build_async = 1 2407 build_async = 1
1938 CODE: 2408 CODE:
1939 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1940 void *user_data = 0;
1941 cl_uint num_devices = 0; 2409 cl_uint device_count = 0;
1942 cl_device_id *device_list = 0; 2410 cl_device_id *device_list = 0;
1943 2411
1944 if (SvOK (devices)) 2412 if (SvOK (devices))
1945 { 2413 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2414
2415 void *user_data;
2416 program_callback pfn_notify = make_program_callback (notify, &user_data);
2417
2418 if (ix)
2419 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2420 else
2421 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2422
2423#if CL_VERSION_1_2
2424
2425void
2426compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2427 CODE:
2428 cl_uint device_count = 0;
2429 cl_device_id *device_list = 0;
2430
2431 if (SvOK (devices))
2432 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2433
2434 cl_uint header_count = 0;
2435 cl_program *header_list = 0;
2436 const char **header_name = 0;
2437
2438 if (SvOK (headers))
2439 {
1946 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2440 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1947 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2441 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1948 2442
1949 AV *av = (AV *)SvRV (devices); 2443 HV *hv = (HV *)SvRV (devices);
1950 num_devices = av_len (av) + 1;
1951 2444
1952 if (num_devices) 2445 header_count = hv_iterinit (hv);
2446 header_list = tmpbuf (sizeof (*header_list) * header_count);
2447 header_name = tmpbuf (sizeof (*header_name) * header_count);
2448
2449 HE *he;
2450 int i = 0;
2451 while (he = hv_iternext (hv))
1953 { 2452 {
1954 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2453 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2454 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1955 int count; 2455 ++i;
1956 for (count = 0; count < num_devices; ++count)
1957 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1958 } 2456 }
1959 } 2457 }
1960 2458
1961 if (SvOK (notify)) 2459 void *user_data;
1962 { 2460 program_callback pfn_notify = make_program_callback (notify, &user_data);
1963 NEED_SUCCESS (RetainProgram, (self));
1964 pfn_notify = eq_program_notify;
1965 user_data = SvREFCNT_inc (s_get_cv (notify));
1966 }
1967 2461
1968 if (ix)
1969 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1970 else
1971 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2462 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2463 header_count, header_list, header_name, pfn_notify, user_data));
2464
2465#endif
1972 2466
1973void 2467void
1974build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2468build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1975 PPCODE: 2469 PPCODE:
1976 size_t size; 2470 size_t size;
1985#BEGIN:program_build 2479#BEGIN:program_build
1986 2480
1987void 2481void
1988build_status (OpenCL::Program self, OpenCL::Device device) 2482build_status (OpenCL::Program self, OpenCL::Device device)
1989 PPCODE: 2483 PPCODE:
1990 cl_build_status value [1]; 2484 cl_int value [1];
1991 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2485 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1992 EXTEND (SP, 1); 2486 EXTEND (SP, 1);
1993 const int i = 0; 2487 const int i = 0;
1994 PUSHs (sv_2mortal (newSViv (value [i]))); 2488 PUSHs (sv_2mortal (newSViv (value [i])));
1995 2489
2005 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2499 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2006 EXTEND (SP, 1); 2500 EXTEND (SP, 1);
2007 const int i = 0; 2501 const int i = 0;
2008 PUSHs (sv_2mortal (newSVpv (value, 0))); 2502 PUSHs (sv_2mortal (newSVpv (value, 0)));
2009 2503
2504void
2505binary_type (OpenCL::Program self, OpenCL::Device device)
2506 PPCODE:
2507 cl_uint value [1];
2508 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2509 EXTEND (SP, 1);
2510 const int i = 0;
2511 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2512
2010#END:program_build 2513#END:program_build
2011 2514
2012void 2515void
2013kernel (OpenCL::Program program, SV *function) 2516kernel (OpenCL::Program program, SV *function)
2014 PPCODE: 2517 PPCODE:
2015 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2518 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2016 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2519 XPUSH_CLOBJ (stash_kernel, kernel);
2017 2520
2018void 2521void
2019kernels_in_program (OpenCL::Program program) 2522kernels_in_program (OpenCL::Program program)
2020 PPCODE: 2523 PPCODE:
2021 cl_uint num_kernels; 2524 cl_uint num_kernels;
2024 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2527 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2025 2528
2026 int i; 2529 int i;
2027 EXTEND (SP, num_kernels); 2530 EXTEND (SP, num_kernels);
2028 for (i = 0; i < num_kernels; ++i) 2531 for (i = 0; i < num_kernels; ++i)
2029 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2532 PUSH_CLOBJ (stash_kernel, kernels [i]);
2030 2533
2031void 2534void
2032info (OpenCL::Program self, cl_program_info name) 2535info (OpenCL::Program self, cl_program_info name)
2033 PPCODE: 2536 PPCODE:
2034 INFO (Program) 2537 INFO (Program)
2080 PPCODE: 2583 PPCODE:
2081 cl_context value [1]; 2584 cl_context value [1];
2082 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2585 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2083 EXTEND (SP, 1); 2586 EXTEND (SP, 1);
2084 const int i = 0; 2587 const int i = 0;
2085 {
2086 NEED_SUCCESS (RetainContext, (value [i])); 2588 NEED_SUCCESS (RetainContext, (value [i]));
2087 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2589 PUSH_CLOBJ (stash_context, value [i]);
2088 }
2089 2590
2090void 2591void
2091devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
2092 PPCODE: 2593 PPCODE:
2093 size_t size; 2594 size_t size;
2095 cl_device_id *value = tmpbuf (size); 2596 cl_device_id *value = tmpbuf (size);
2096 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2097 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
2098 EXTEND (SP, n); 2599 EXTEND (SP, n);
2099 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
2100 {
2101 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2601 PUSH_CLOBJ (stash_device, value [i]);
2102 }
2103 2602
2104void 2603void
2105source (OpenCL::Program self) 2604source (OpenCL::Program self)
2106 PPCODE: 2605 PPCODE:
2107 size_t size; 2606 size_t size;
2132DESTROY (OpenCL::Kernel self) 2631DESTROY (OpenCL::Kernel self)
2133 CODE: 2632 CODE:
2134 clReleaseKernel (self); 2633 clReleaseKernel (self);
2135 2634
2136void 2635void
2636setf (OpenCL::Kernel self, const char *format, ...)
2637 CODE:
2638 int i;
2639 for (i = 2; ; ++i)
2640 {
2641 while (*format == ' ')
2642 ++format;
2643
2644 char type = *format++;
2645
2646 if (!type)
2647 break;
2648
2649 if (i >= items)
2650 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2651
2652 SV *sv = ST (i);
2653
2654 union
2655 {
2656 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2657 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2658 cl_half ch; cl_float cf; cl_double cd;
2659 cl_mem cm;
2660 cl_sampler ca;
2661 size_t cz;
2662 cl_event ce;
2663 } arg;
2664 size_t size;
2665 int nullarg = 0;
2666
2667 switch (type)
2668 {
2669 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2670 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2671 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2672 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2673 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2674 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2675 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2676 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2677
2678 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2679 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2680 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2681
2682 case 'z': nullarg = 1; size = SvIV (sv); break;
2683
2684 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2685 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2686 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2687
2688 default:
2689 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2690 }
2691
2692 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2693 if (res)
2694 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2695 }
2696
2697 if (i != items)
2698 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2699
2700void
2137set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2701set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2138 CODE: 2702 CODE:
2139 clSetKernelArg (self, idx, sizeof (value), &value); 2703 clSetKernelArg (self, idx, sizeof (value), &value);
2140 2704
2141void 2705void
2189 clSetKernelArg (self, idx, sizeof (value), &value); 2753 clSetKernelArg (self, idx, sizeof (value), &value);
2190 2754
2191void 2755void
2192set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2756set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2193 CODE: 2757 CODE:
2194 clSetKernelArg (self, idx, sizeof (value), &value); 2758 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2195 2759
2196void 2760void
2197set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2761set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2198 CODE: 2762 CODE:
2199 clSetKernelArg (self, idx, sizeof (value), &value); 2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2200 2764
2201void 2765void
2202set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2766set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2203 CODE: 2767 CODE:
2204 clSetKernelArg (self, idx, sizeof (value), &value); 2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2205
2206void
2207set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
2208 CODE:
2209 clSetKernelArg (self, idx, sizeof (value), &value);
2210 2769
2211void 2770void
2212set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2771set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2213 CODE: 2772 CODE:
2214 clSetKernelArg (self, idx, sizeof (value), &value); 2773 clSetKernelArg (self, idx, sizeof (value), &value);
2258 PPCODE: 2817 PPCODE:
2259 cl_context value [1]; 2818 cl_context value [1];
2260 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2819 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2261 EXTEND (SP, 1); 2820 EXTEND (SP, 1);
2262 const int i = 0; 2821 const int i = 0;
2263 {
2264 NEED_SUCCESS (RetainContext, (value [i])); 2822 NEED_SUCCESS (RetainContext, (value [i]));
2265 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2823 PUSH_CLOBJ (stash_context, value [i]);
2266 }
2267 2824
2268void 2825void
2269program (OpenCL::Kernel self) 2826program (OpenCL::Kernel self)
2270 PPCODE: 2827 PPCODE:
2271 cl_program value [1]; 2828 cl_program value [1];
2272 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2829 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2273 EXTEND (SP, 1); 2830 EXTEND (SP, 1);
2274 const int i = 0; 2831 const int i = 0;
2275 {
2276 NEED_SUCCESS (RetainProgram, (value [i])); 2832 NEED_SUCCESS (RetainProgram, (value [i]));
2277 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2833 PUSH_CLOBJ (stash_program, value [i]);
2278 }
2279 2834
2280#END:kernel 2835#END:kernel
2281 2836
2282void 2837void
2283work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2838work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2329 const int i = 0; 2884 const int i = 0;
2330 PUSHs (sv_2mortal (newSVuv (value [i]))); 2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2331 2886
2332#END:kernel_work_group 2887#END:kernel_work_group
2333 2888
2889#if CL_VERSION_1_2
2890
2891void
2892arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2893 PPCODE:
2894 size_t size;
2895 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2896 SV *sv = sv_2mortal (newSV (size));
2897 SvUPGRADE (sv, SVt_PV);
2898 SvPOK_only (sv);
2899 SvCUR_set (sv, size);
2900 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2901 XPUSHs (sv);
2902
2903#BEGIN:kernel_arg
2904
2905void
2906arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2907 ALIAS:
2908 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2909 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2910 PPCODE:
2911 cl_uint value [1];
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2913 EXTEND (SP, 1);
2914 const int i = 0;
2915 PUSHs (sv_2mortal (newSVuv (value [i])));
2916
2917void
2918arg_type_name (OpenCL::Kernel self, cl_uint idx)
2919 ALIAS:
2920 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2921 arg_name = CL_KERNEL_ARG_NAME
2922 PPCODE:
2923 size_t size;
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2925 char *value = tmpbuf (size);
2926 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2927 EXTEND (SP, 1);
2928 const int i = 0;
2929 PUSHs (sv_2mortal (newSVpv (value, 0)));
2930
2931void
2932arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2933 PPCODE:
2934 cl_ulong value [1];
2935 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2936 EXTEND (SP, 1);
2937 const int i = 0;
2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2939
2940#END:kernel_arg
2941
2942#endif
2943
2334MODULE = OpenCL PACKAGE = OpenCL::Event 2944MODULE = OpenCL PACKAGE = OpenCL::Event
2335 2945
2336void 2946void
2337DESTROY (OpenCL::Event self) 2947DESTROY (OpenCL::Event self)
2338 CODE: 2948 CODE:
2360 PPCODE: 2970 PPCODE:
2361 cl_command_queue value [1]; 2971 cl_command_queue value [1];
2362 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2972 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2363 EXTEND (SP, 1); 2973 EXTEND (SP, 1);
2364 const int i = 0; 2974 const int i = 0;
2365 {
2366 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2975 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2367 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2976 PUSH_CLOBJ (stash_queue, value [i]);
2368 }
2369 2977
2370void 2978void
2371command_type (OpenCL::Event self) 2979command_type (OpenCL::Event self)
2372 PPCODE:
2373 cl_command_type value [1];
2374 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2375 EXTEND (SP, 1);
2376 const int i = 0;
2377 PUSHs (sv_2mortal (newSVuv (value [i])));
2378
2379void
2380reference_count (OpenCL::Event self)
2381 ALIAS: 2980 ALIAS:
2981 command_type = CL_EVENT_COMMAND_TYPE
2382 reference_count = CL_EVENT_REFERENCE_COUNT 2982 reference_count = CL_EVENT_REFERENCE_COUNT
2383 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2983 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2384 PPCODE: 2984 PPCODE:
2385 cl_uint value [1]; 2985 cl_uint value [1];
2386 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2986 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2393 PPCODE: 2993 PPCODE:
2394 cl_context value [1]; 2994 cl_context value [1];
2395 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2995 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2396 EXTEND (SP, 1); 2996 EXTEND (SP, 1);
2397 const int i = 0; 2997 const int i = 0;
2398 {
2399 NEED_SUCCESS (RetainContext, (value [i])); 2998 NEED_SUCCESS (RetainContext, (value [i]));
2400 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2999 PUSH_CLOBJ (stash_context, value [i]);
2401 }
2402 3000
2403#END:event 3001#END:event
2404 3002
2405void 3003void
2406profiling_info (OpenCL::Event self, cl_profiling_info name) 3004profiling_info (OpenCL::Event self, cl_profiling_info name)
2430void 3028void
2431set_status (OpenCL::UserEvent self, cl_int execution_status) 3029set_status (OpenCL::UserEvent self, cl_int execution_status)
2432 CODE: 3030 CODE:
2433 clSetUserEventStatus (self, execution_status); 3031 clSetUserEventStatus (self, execution_status);
2434 3032
3033MODULE = OpenCL PACKAGE = OpenCL::Mapped
3034
3035void
3036DESTROY (SV *self)
3037 CODE:
3038 OpenCL__Mapped mapped = SvMAPPED (self);
3039
3040 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3041 mapped_detach (self, mapped);
3042
3043 clReleaseCommandQueue (mapped->queue);
3044 clReleaseEvent (mapped->event);
3045 Safefree (mapped);
3046
3047void
3048unmap (OpenCL::Mapped self, ...)
3049 CODE:
3050 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3051
3052bool
3053mapped (OpenCL::Mapped self)
3054 CODE:
3055 RETVAL = !!self->ptr;
3056 OUTPUT:
3057 RETVAL
3058
3059void
3060wait (OpenCL::Mapped self)
3061 PPCODE:
3062 if (self->event)
3063 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3064
3065void
3066event (OpenCL::Mapped self)
3067 PPCODE:
3068 if (!self->event)
3069 XSRETURN_UNDEF;
3070
3071 clRetainEvent (self->event);
3072 XPUSH_CLOBJ (stash_event, self->event);
3073
3074#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3075#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3076#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3077#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3078#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3079#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3080
3081IV
3082size (OpenCL::Mapped self)
3083 ALIAS:
3084 size = MAPPED_OFFSET_CB
3085 row_pitch = MAPPED_OFFSET_ROW_PITCH
3086 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3087 width = MAPPED_OFFSET_WIDTH
3088 height = MAPPED_OFFSET_HEIGHT
3089 depth = MAPPED_OFFSET_DEPTH
3090 CODE:
3091 RETVAL = *(size_t *)((char *)self + ix);
3092 OUTPUT:
3093 RETVAL
3094
3095IV
3096ptr (OpenCL::Mapped self)
3097 CODE:
3098 RETVAL = PTR2IV (self->ptr);
3099 OUTPUT:
3100 RETVAL
3101
3102void
3103set (OpenCL::Mapped self, size_t offset, SV *data)
3104 CODE:
3105 STRLEN len;
3106 const char *ptr = SvPVbyte (data, len);
3107
3108 if (offset + len > self->cb)
3109 croak ("OpenCL::Mapped::set out of bound condition detected");
3110
3111 memcpy (offset + (char *)self->ptr, ptr, len);
3112
3113void
3114get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3115 PPCODE:
3116 if (!SvOK (ST (1)))
3117 count = self->width - x;
3118
3119 if (x + count > self->width)
3120 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3121
3122 if (y >= self->height)
3123 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3124
3125 if (z >= self->depth)
3126 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3127
3128 size_t element = mapped_element_size (self);
3129
3130 count *= element;
3131 x *= element;
3132
3133 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3134 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3135
3136void
3137set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3138 PPCODE:
3139 STRLEN count;
3140 char *dataptr = SvPVbyte (data, count);
3141 size_t element = mapped_element_size (self);
3142
3143 x *= element;
3144
3145 if (x + count > self->width * element)
3146 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3147
3148 if (y >= self->height)
3149 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3150
3151 if (z >= self->depth)
3152 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3153
3154 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3155 memcpy (ptr, dataptr, count);
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3158
3159MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3160
3161IV
3162element_size (OpenCL::Mapped self)
3163 CODE:
3164 RETVAL = mapped_element_size (self);
3165 OUTPUT:
3166 RETVAL
3167

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines