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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.4 by root, Tue Nov 15 21:13:42 2011 UTC vs.
Revision 1.82 by root, Thu Oct 11 05:02:54 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"
4 6
7#include "ecb.h"//D
8
9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
10#include "xthread.h"
11#include "schmorp.h"
12
13#ifdef I_DLFCN
14 #include <dlfcn.h>
15#endif
16
17// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
18// but then fail to define the api functions unless you ALSO define
19// this. This breaks 100% of the opencl 1.1 apps, for what reason?
20// after all, the functions are deprecated, not removed.
21// in addition, you cannot test for this in any future-proof way.
22// each time a new opencl version comes out, you need to make a new
23// release.
24#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
25
26#ifndef PREFER_1_1
27 #define PREFER_1_1 1
28#endif
29
30#if PREFER_1_1
31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
32#endif
33
34#ifdef __APPLE__
35 #define CLHDR(name) <OpenCL/name>
36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
5#include <CL/opencl.h> 40#include CLHDR(opencl.h)
6 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>
54#endif
55
56#ifndef CL_VERSION_1_2
57 #undef PREFER_1_1
58 #define PREFER_1_1 1
59#endif
60
61// make sure all constants we might use are actually defined
62#include "default.h"
63
7typedef cl_platform_id OpenCL__Platform; 64typedef cl_platform_id OpenCL__Platform;
8typedef cl_device_id OpenCL__Device; 65typedef cl_device_id OpenCL__Device;
66typedef cl_device_id OpenCL__SubDevice;
9typedef cl_context OpenCL__Context; 67typedef cl_context OpenCL__Context;
10typedef cl_command_queue OpenCL__Queue; 68typedef cl_command_queue OpenCL__Queue;
11typedef cl_mem OpenCL__Memory; 69typedef cl_mem OpenCL__Memory;
12typedef cl_mem OpenCL__Buffer; 70typedef cl_mem OpenCL__Buffer;
71typedef cl_mem OpenCL__BufferObj;
13typedef cl_mem OpenCL__Image; 72typedef cl_mem OpenCL__Image;
14typedef cl_mem OpenCL__Image2D;
15typedef cl_mem OpenCL__Image3D;
16typedef cl_mem OpenCL__Memory_ornull; 73typedef cl_mem OpenCL__Memory_ornull;
17typedef cl_mem OpenCL__Buffer_ornull; 74typedef cl_mem OpenCL__Buffer_ornull;
18typedef cl_mem OpenCL__Image_ornull; 75typedef cl_mem OpenCL__Image_ornull;
19typedef cl_mem OpenCL__Image2D_ornull;
20typedef cl_mem OpenCL__Image3D_ornull;
21typedef cl_sampler OpenCL__Sampler; 76typedef cl_sampler OpenCL__Sampler;
22typedef cl_program OpenCL__Program; 77typedef cl_program OpenCL__Program;
23typedef cl_kernel OpenCL__Kernel; 78typedef cl_kernel OpenCL__Kernel;
24typedef cl_event OpenCL__Event; 79typedef cl_event OpenCL__Event;
80typedef cl_event OpenCL__UserEvent;
81
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;
108
109/*****************************************************************************/
110
111// name must include a leading underscore
112// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
113// for perl. doh.
114static void *
115glsym (const char *name)
116{
117 void *fun = 0;
118
119 #if defined I_DLFCN && defined RTLD_DEFAULT
120 fun = dlsym (RTLD_DEFAULT, name + 1);
121 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
122
123 if (!fun)
124 {
125 static void *libgl;
126 static const char *glso[] = {
127 "libGL.so.1",
128 "libGL.so.3",
129 "libGL.so.4.0",
130 "libGL.so",
131 "/usr/lib/libGL.so",
132 "/usr/X11R6/lib/libGL.1.dylib"
133 };
134 int i;
135
136 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
137 {
138 libgl = dlopen (glso [i], RTLD_LAZY);
139 if (libgl)
140 break;
141 }
142
143 if (libgl)
144 {
145 fun = dlsym (libgl, name + 1);
146 if (!fun) fun = dlsym (libgl, name);
147 }
148 }
149 #endif
150
151 return fun;
152}
153
154/*****************************************************************************/
155
156/* up to two temporary buffers */
157static void * ecb_noinline
158tmpbuf (size_t size)
159{
160 enum { buffers = 4 };
161 static int idx;
162 static void *buf [buffers];
163 static size_t len [buffers];
164
165 idx = (idx + 1) % buffers;
166
167 if (len [idx] < size)
168 {
169 free (buf [idx]);
170 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
171 buf [idx] = malloc (len [idx]);
172 }
173
174 return buf [idx];
175}
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
192/*****************************************************************************/
25 193
26typedef struct 194typedef struct
27{ 195{
28 IV iv; 196 IV iv;
29 const char *name; 197 const char *name;
30 #define const_iv(name) { (IV)CL_ ## name, # name }, 198 #define const_iv(name) { (IV)CL_ ## name, # name },
31} ivstr; 199} ivstr;
32 200
201typedef struct
202{
203 NV nv;
204 const char *name;
205 #define const_nv(name) { (NV)CL_ ## name, # name },
206} nvstr;
207
33static const char * 208static const char *
34iv2str (IV value, const ivstr *base, int count, const char *fallback) 209iv2str (IV value, const ivstr *base, int count, const char *fallback)
35{ 210{
36 int i; 211 int i;
37 static char strbuf [32]; 212 static char strbuf [32];
63 }; 238 };
64 239
65 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)"); 240 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)");
66} 241}
67 242
243/*****************************************************************************/
244
245static cl_int res;
246
68#define FAIL(name,res) \ 247#define FAIL(name) \
69 croak ("cl" # name ": %s", err2str (res)); 248 croak ("cl" # name ": %s", err2str (res));
70 249
71#define NEED_SUCCESS(name,args) \ 250#define NEED_SUCCESS(name,args) \
72 do { \ 251 do { \
73 cl_int res = cl ## name args; \ 252 res = cl ## name args; \
74 \ 253 \
75 if (res) \ 254 if (res) \
76 FAIL (name, res); \ 255 FAIL (name); \
77 } while (0) 256 } while (0)
78 257
79#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 258#define NEED_SUCCESS_ARG(retdecl, name, args) \
80#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr)) 259 retdecl = cl ## name args; \
260 if (res) \
261 FAIL (name);
81 262
82/*TODO*/ 263/*****************************************************************************/
83#define EVENT_LIST(items,count) cl_uint event_list_count = 0; cl_event *event_list_ptr = 0 264
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
298static cl_context_properties * ecb_noinline
299SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
300{
301 if (!sv || !SvOK (sv))
302 if (extra)
303 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
304 else
305 return 0;
306
307 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
308 {
309 AV *av = (AV *)SvRV (sv);
310 int i, len = av_len (av) + 1;
311 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
312 cl_context_properties *l = p;
313
314 if (len & 1)
315 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
316
317 while (extracount--)
318 *l++ = *extra++;
319
320 for (i = 0; i < len; i += 2)
321 {
322 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
323 SV *p_sv = *av_fetch (av, i + 1, 0);
324 cl_context_properties v = SvIV (p_sv); // code below can override
325
326 switch (t)
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
333 case CL_GLX_DISPLAY_KHR:
334 if (!SvOK (p_sv))
335 {
336 void *func = glsym ("_glXGetCurrentDisplay");
337 if (func)
338 v = (cl_context_properties)((void *(*)(void))func)();
339 }
340 break;
341
342 case CL_GL_CONTEXT_KHR:
343 if (!SvOK (p_sv))
344 {
345 void *func = glsym ("_glXGetCurrentContext");
346 if (func)
347 v = (cl_context_properties)((void *(*)(void))func)();
348 }
349 break;
350
351 default:
352 /* unknown property, treat as int */
353 break;
354 }
355
356 *l++ = t;
357 *l++ = v;
358 }
359
360 *l = 0;
361
362 return p;
363 }
364
365 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
366}
367
368// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
369// is a pointer (and only on well-behaved systems).
370static void * ecb_noinline
371object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
372{
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)
382 {
383 list = tmpbuf (sizeof (*list) * count);
384 int i;
385 for (i = 0; i < count; ++i)
386 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
387 }
388
389 if (!count && !or_undef)
390 croak ("%s: '%s' must contain at least one %s object",
391 cv_get_name (cv), argname, klass);
392
393 *rcount = count;
394 return (void *)list;
395}
396
397/*****************************************************************************/
398/* callback stuff */
399
400/* default context callback, log to stderr */
401static void CL_CALLBACK
402context_default_notify (const char *msg, const void *info, size_t cb, void *data)
403{
404 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
405}
406
407typedef struct
408{
409 int free_cb;
410 void (*push)(void *data1, void *data2, void *data3);
411} eq_vtbl;
412
413typedef struct eq_item
414{
415 struct eq_item *next;
416 eq_vtbl *vtbl;
417 SV *cb;
418 void *data1, *data2, *data3;
419} eq_item;
420
421static void (*eq_signal_func)(void *signal_arg, int value);
422static void *eq_signal_arg;
423static xmutex_t eq_lock = X_MUTEX_INIT;
424static eq_item *eq_head, *eq_tail;
425
426static void ecb_noinline
427eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
428{
429 eq_item *item = malloc (sizeof (eq_item));
430
431 item->next = 0;
432 item->vtbl = vtbl;
433 item->cb = cb;
434 item->data1 = data1;
435 item->data2 = data2;
436 item->data3 = data3;
437
438 X_LOCK (eq_lock);
439
440 *(eq_head ? &eq_tail->next : &eq_head) = item;
441 eq_tail = item;
442
443 X_UNLOCK (eq_lock);
444
445 eq_signal_func (eq_signal_arg, 0);
446}
447
448static eq_item *
449eq_dec (void)
450{
451 eq_item *res;
452
453 X_LOCK (eq_lock);
454
455 res = eq_head;
456 if (res)
457 eq_head = res->next;
458
459 X_UNLOCK (eq_lock);
460
461 return res;
462}
463
464static void
465eq_poll (void)
466{
467 eq_item *item;
468
469 while ((item = eq_dec ()))
470 {
471 ENTER;
472 SAVETMPS;
473
474 dSP;
475 PUSHMARK (SP);
476 EXTEND (SP, 2);
477
478 if (item->vtbl->free_cb)
479 sv_2mortal (item->cb);
480
481 PUTBACK;
482 item->vtbl->push (item->data1, item->data2, item->data3);
483
484 SV *cb = item->cb;
485 free (item);
486
487 call_sv (cb, G_DISCARD | G_VOID);
488
489 FREETMPS;
490 LEAVE;
491 }
492}
493
494static void
495eq_poll_interrupt (pTHX_ void *c_arg, int value)
496{
497 eq_poll ();
498}
499
500/*****************************************************************************/
501/* context notify */
502
503static void ecb_noinline
504eq_context_push (void *data1, void *data2, void *data3)
505{
506 dSP;
507 PUSHs (sv_2mortal (newSVpv (data1, 0)));
508 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
509 PUTBACK;
510
511 free (data1);
512 free (data2);
513}
514
515static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
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/*****************************************************************************/
549/* build/compile/link notify */
550
551static void
552eq_program_push (void *data1, void *data2, void *data3)
553{
554 dSP;
555 PUSH_CLOBJ (stash_program, data1);
556 PUTBACK;
557}
558
559static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
560
561static void CL_CALLBACK
562eq_program_notify (cl_program program, void *user_data)
563{
564 clRetainProgram (program);
565
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 }
584}
585
586struct build_args
587{
588 cl_program program;
589 char *options;
590 void *user_data;
591 cl_uint num_devices;
592};
593
594X_THREAD_PROC (build_program_thread)
595{
596 struct build_args *arg = thr_arg;
597
598 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
599
600 if (arg->user_data)
601 eq_program_notify (arg->program, arg->user_data);
602 else
603 clReleaseProgram (arg->program);
604
605 free (arg->options);
606 free (arg);
607
608 return 0;
609}
610
611static void
612build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
613{
614 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
615
616 arg->program = program;
617 arg->options = strdup (options);
618 arg->user_data = user_data;
619 arg->num_devices = num_devices;
620 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
621
622 xthread_t id;
623 xthread_create (&id, build_program_thread, arg);
624}
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/*****************************************************************************/
643/* event objects */
644
645static void
646eq_event_push (void *data1, void *data2, void *data3)
647{
648 dSP;
649 PUSH_CLOBJ (stash_event, data1);
650 PUSHs (sv_2mortal (newSViv ((IV)data2)));
651 PUTBACK;
652}
653
654static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
655
656static void CL_CALLBACK
657eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
658{
659 clRetainEvent (event);
660 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
661}
662
663/*****************************************************************************/
664/* utilities for XS code */
665
666static size_t
667img_row_pitch (cl_mem img)
668{
669 size_t res;
670 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
671 return res;
672}
673
674static cl_event * ecb_noinline
675event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
676{
677 cl_uint count = *rcount;
678
679 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
680 *rcount = count = 0;
681
682 if (!count && !extra)
683 return 0;
684
685 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
686 int i = 0;
687
688 while (count--)
689 if (SvOK (items [count]))
690 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
691
692 if (extra)
693 list [i++] = extra;
694
695 *rcount = i;
696
697 return i ? list : 0;
698}
699
700#define EVENT_LIST(skip) \
701 cl_uint event_list_count = items - (skip); \
702 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
84 703
85#define INFO(class) \ 704#define INFO(class) \
86{ \ 705{ \
87 size_t size; \ 706 size_t size; \
88 SV *sv; \
89 \
90 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 707 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
91 sv = sv_2mortal (newSV (size)); \ 708 SV *sv = sv_2mortal (newSV (size)); \
92 SvUPGRADE (sv, SVt_PV); \ 709 SvUPGRADE (sv, SVt_PV); \
93 SvPOK_only (sv); \ 710 SvPOK_only (sv); \
94 SvCUR_set (sv, size); \ 711 SvCUR_set (sv, size); \
95 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 712 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
96 XPUSHs (sv); \ 713 XPUSHs (sv); \
97} 714}
98 715
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
99static void * 787static void
100SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 788mapped_detach (SV *sv, OpenCL__Mapped mapped)
101{ 789{
102 if (SvROK (sv) && sv_derived_from (sv, pkg)) 790 SV *data = SvRV (sv);
103 return (void *)SvIV (SvRV (sv));
104 791
105 croak ("%s: %s is not of type %s", func, svname, pkg); 792 // the next check checks both before AND after detach, where SvPVX should STILL be 0
106} 793 if (SvPVX (data) != (char *)mapped->ptr)
107 794 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
108static void * 795 else
109tmpbuf (size_t size)
110{
111 static void *buf;
112 static size_t len;
113
114 if (len < size)
115 { 796 {
116 free (buf); 797 SvREADONLY_off (data);
117 len = ((size + 31) & ~4095) + 4096 - 32; 798 SvCUR_set (data, 0);
118 buf = malloc (len); 799 SvPVX (data) = 0;
800 SvOK_off (data);
119 } 801 }
120 802
121 return buf; 803 mapped->ptr = 0;
122} 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/*****************************************************************************/
123 830
124MODULE = OpenCL PACKAGE = OpenCL 831MODULE = OpenCL PACKAGE = OpenCL
125 832
126PROTOTYPES: ENABLE 833PROTOTYPES: ENABLE
127 834
835void
836poll ()
837 CODE:
838 eq_poll ();
839
840void
841_eq_initialise (IV func, IV arg)
842 CODE:
843 eq_signal_func = (void (*)(void *, int))func;
844 eq_signal_arg = (void*)arg;
845
128BOOT: 846BOOT:
129{ 847{
130 HV *stash = gv_stashpv ("OpenCL", 1); 848 HV *stash = gv_stashpv ("OpenCL", 1);
849
131 static const ivstr *civ, const_iv[] = { 850 static const ivstr *civ, const_iv[] = {
132 { sizeof (cl_char ), "SIZEOF_CHAR" }, 851 { sizeof (cl_char ), "SIZEOF_CHAR" },
133 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 852 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
134 { sizeof (cl_short ), "SIZEOF_SHORT" }, 853 { sizeof (cl_short ), "SIZEOF_SHORT" },
135 { sizeof (cl_ushort), "SIZEOF_USHORT"}, 854 { sizeof (cl_ushort), "SIZEOF_USHORT" },
136 { sizeof (cl_int ), "SIZEOF_INT" }, 855 { sizeof (cl_int ), "SIZEOF_INT" },
137 { sizeof (cl_uint ), "SIZEOF_UINT" }, 856 { sizeof (cl_uint ), "SIZEOF_UINT" },
138 { sizeof (cl_long ), "SIZEOF_LONG" }, 857 { sizeof (cl_long ), "SIZEOF_LONG" },
139 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 858 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
140 { sizeof (cl_half ), "SIZEOF_HALF" }, 859 { sizeof (cl_half ), "SIZEOF_HALF" },
141 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 860 { sizeof (cl_float ), "SIZEOF_FLOAT" },
861 { sizeof (cl_double), "SIZEOF_DOUBLE" },
862 { PREFER_1_1 , "PREFER_1_1" },
142#include "constiv.h" 863#include "constiv.h"
143 }; 864 };
865
144 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--)
145 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);
901
902 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
146} 903}
904
905cl_int
906errno ()
907 CODE:
908 RETVAL = res;
909 OUTPUT:
910 RETVAL
147 911
148const char * 912const char *
149err2str (cl_int err) 913err2str (cl_int err = res)
150 914
151const char * 915const char *
152enum2str (cl_uint value) 916enum2str (cl_uint value)
153 917
154void 918void
155platforms () 919platforms ()
156 PPCODE: 920 PPCODE:
157{
158 cl_platform_id *list; 921 cl_platform_id *list;
159 cl_uint count; 922 cl_uint count;
160 int i; 923 int i;
161 924
162 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count)); 925 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count));
163 list = tmpbuf (sizeof (*list) * count); 926 list = tmpbuf (sizeof (*list) * count);
164 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 927 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
165 928
166 EXTEND (SP, count); 929 EXTEND (SP, count);
167 for (i = 0; i < count; ++i) 930 for (i = 0; i < count; ++i)
168 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 931 PUSH_CLOBJ (stash_platform, list [i]);
169}
170 932
171void 933void
172context_from_type_simple (cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 934context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
173 PPCODE: 935 PPCODE:
174{ 936 CONTEXT_NOTIFY_CALLBACK;
175 cl_int res; 937 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
176 cl_context ctx = clCreateContextFromType (0, type, 0, 0, &res); 938 XPUSH_CLOBJ_CONTEXT;
177 939
178 if (res) 940void
179 FAIL (CreateContextFromType, res); 941context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
942 PPCODE:
943 cl_uint device_count;
944 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
180 945
181 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 946 CONTEXT_NOTIFY_CALLBACK;
182} 947 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
948 XPUSH_CLOBJ_CONTEXT;
183 949
184void 950void
185wait_for_events (...) 951wait_for_events (...)
186 CODE: 952 CODE:
187{
188 EVENT_LIST (0, items); 953 EVENT_LIST (0);
189 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 954 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
190}
191 955
192PROTOTYPES: DISABLE 956PROTOTYPES: DISABLE
193 957
194MODULE = OpenCL PACKAGE = OpenCL::Platform 958MODULE = OpenCL PACKAGE = OpenCL::Platform
195 959
196void 960void
197info (OpenCL::Platform this, cl_platform_info name) 961info (OpenCL::Platform self, cl_platform_info name)
198 PPCODE: 962 PPCODE:
199 INFO (Platform) 963 INFO (Platform)
200 964
201void 965void
966unload_compiler (OpenCL::Platform self)
967 CODE:
968#if CL_VERSION_1_2
969 clUnloadPlatformCompiler (self);
970#endif
971
972#BEGIN:platform
973
974void
975profile (OpenCL::Platform self)
976 ALIAS:
977 profile = CL_PLATFORM_PROFILE
978 version = CL_PLATFORM_VERSION
979 name = CL_PLATFORM_NAME
980 vendor = CL_PLATFORM_VENDOR
981 extensions = CL_PLATFORM_EXTENSIONS
982 PPCODE:
983 size_t size;
984 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
985 char *value = tmpbuf (size);
986 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
987 EXTEND (SP, 1);
988 const int i = 0;
989 PUSHs (sv_2mortal (newSVpv (value, 0)));
990
991#END:platform
992
993void
202devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 994devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
203 PPCODE: 995 PPCODE:
204{
205 cl_device_id *list; 996 cl_device_id *list;
206 cl_uint count; 997 cl_uint count;
207 int i; 998 int i;
208 999
209 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 1000 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
210 list = tmpbuf (sizeof (*list) * count); 1001 list = tmpbuf (sizeof (*list) * count);
211 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 1002 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
212 1003
213 EXTEND (SP, count); 1004 EXTEND (SP, count);
214 for (i = 0; i < count; ++i) 1005 for (i = 0; i < count; ++i)
215 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1006 PUSH_CLOBJ (stash_device, list [i]);
216}
217 1007
218void 1008void
219context_from_type_simple (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 1009context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
220 PPCODE: 1010 PPCODE:
221{
222 cl_int res;
223 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 1011 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
224 cl_context ctx = clCreateContextFromType (props, type, 0, 0, &res); 1012 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
225 1013
226 if (res) 1014 cl_uint device_count;
227 FAIL (CreateContextFromType, res); 1015 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
228 1016
229 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1017 CONTEXT_NOTIFY_CALLBACK;
230} 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;
231 1030
232MODULE = OpenCL PACKAGE = OpenCL::Device 1031MODULE = OpenCL PACKAGE = OpenCL::Device
233 1032
234void 1033void
235info (OpenCL::Device this, cl_device_info name) 1034info (OpenCL::Device self, cl_device_info name)
236 PPCODE: 1035 PPCODE:
237 INFO (Device) 1036 INFO (Device)
238 1037
1038#if CL_VERSION_1_2
1039
239void 1040void
240context_simple (OpenCL::Device this) 1041sub_devices (OpenCL::Device self, SV *properties)
1042 PPCODE:
1043 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
1044 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
1045
1046 properties = SvRV (properties);
1047
1048 cl_uint count = av_len ((AV *)properties) + 1;
1049 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1050
1051 int i;
1052 for (i = 0; i < count; ++i)
1053 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
1054
1055 props [count] = 0;
1056
1057 cl_uint num_devices;
1058 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1059 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1060 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
1061
1062 EXTEND (SP, num_devices);
1063 for (i = 0; i < count; ++i)
1064 PUSH_CLOBJ (stash_subdevice, list [i]);
1065
1066#endif
1067
1068#BEGIN:device
1069
1070void
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
241 PPCODE: 1085 PPCODE:
242{ 1086 cl_ulong value [1];
243 cl_int res; 1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
244 cl_context ctx = clCreateContext (0, 1, &this, 0, 0, &res); 1088 EXTEND (SP, 1);
1089 const int i = 0;
1090 PUSHs (sv_2mortal (newSVuv (value [i])));
245 1091
246 if (res) 1092void
247 FAIL (CreateContext, res); 1093vendor_id (OpenCL::Device self)
1094 ALIAS:
1095 vendor_id = CL_DEVICE_VENDOR_ID
1096 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
1097 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
1098 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
1099 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
1100 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
1101 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
1102 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
1103 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
1104 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
1105 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
1106 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
1107 image_support = CL_DEVICE_IMAGE_SUPPORT
1108 max_samplers = CL_DEVICE_MAX_SAMPLERS
1109 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
1110 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1111 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
1112 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
1113 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1114 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
1115 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
1116 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
1117 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
1118 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
1119 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
1120 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
1121 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
1122 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
1123 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
1124 PPCODE:
1125 cl_uint value [1];
1126 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1127 EXTEND (SP, 1);
1128 const int i = 0;
1129 PUSHs (sv_2mortal (newSVuv (value [i])));
248 1130
249 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1131void
250} 1132max_work_group_size (OpenCL::Device self)
1133 ALIAS:
1134 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
1135 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
1136 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
1137 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
1138 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
1139 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
1140 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
1141 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
1142 PPCODE:
1143 size_t value [1];
1144 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1145 EXTEND (SP, 1);
1146 const int i = 0;
1147 PUSHs (sv_2mortal (newSVuv (value [i])));
1148
1149void
1150max_work_item_sizes (OpenCL::Device self)
1151 PPCODE:
1152 size_t size;
1153 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
1154 size_t *value = tmpbuf (size);
1155 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
1156 int i, n = size / sizeof (*value);
1157 EXTEND (SP, n);
1158 for (i = 0; i < n; ++i)
1159 PUSHs (sv_2mortal (newSVuv (value [i])));
1160
1161void
1162error_correction_support (OpenCL::Device self)
1163 ALIAS:
1164 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1165 endian_little = CL_DEVICE_ENDIAN_LITTLE
1166 available = CL_DEVICE_AVAILABLE
1167 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1168 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1169 PPCODE:
1170 cl_uint value [1];
1171 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1172 EXTEND (SP, 1);
1173 const int i = 0;
1174 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1175
1176void
1177platform (OpenCL::Device self)
1178 PPCODE:
1179 cl_platform_id value [1];
1180 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1181 EXTEND (SP, 1);
1182 const int i = 0;
1183 PUSH_CLOBJ (stash_platform, value [i]);
1184
1185void
1186name (OpenCL::Device self)
1187 ALIAS:
1188 name = CL_DEVICE_NAME
1189 vendor = CL_DEVICE_VENDOR
1190 driver_version = CL_DRIVER_VERSION
1191 profile = CL_DEVICE_PROFILE
1192 version = CL_DEVICE_VERSION
1193 extensions = CL_DEVICE_EXTENSIONS
1194 PPCODE:
1195 size_t size;
1196 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1197 char *value = tmpbuf (size);
1198 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1199 EXTEND (SP, 1);
1200 const int i = 0;
1201 PUSHs (sv_2mortal (newSVpv (value, 0)));
1202
1203void
1204parent_device_ext (OpenCL::Device self)
1205 PPCODE:
1206 cl_device_id value [1];
1207 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1208 EXTEND (SP, 1);
1209 const int i = 0;
1210 PUSH_CLOBJ (stash_device, value [i]);
1211
1212void
1213partition_types_ext (OpenCL::Device self)
1214 ALIAS:
1215 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1216 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
1217 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
1218 PPCODE:
1219 size_t size;
1220 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1221 cl_device_partition_property_ext *value = tmpbuf (size);
1222 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1223 int i, n = size / sizeof (*value);
1224 EXTEND (SP, n);
1225 for (i = 0; i < n; ++i)
1226 PUSHs (sv_2mortal (newSVuv (value [i])));
1227
1228#END:device
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
251 1240
252MODULE = OpenCL PACKAGE = OpenCL::Context 1241MODULE = OpenCL PACKAGE = OpenCL::Context
253 1242
254void 1243void
255DESTROY (OpenCL::Context context) 1244DESTROY (OpenCL::Context self)
256 CODE: 1245 CODE:
257 clReleaseContext (context); 1246 clReleaseContext (self);
258 1247
259void 1248void
260info (OpenCL::Context this, cl_context_info name) 1249info (OpenCL::Context self, cl_context_info name)
261 PPCODE: 1250 PPCODE:
262 INFO (Context) 1251 INFO (Context)
263 1252
264void 1253void
265command_queue_simple (OpenCL::Context this, OpenCL::Device device) 1254queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
266 PPCODE: 1255 PPCODE:
267{ 1256 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
268 cl_int res; 1257 XPUSH_CLOBJ (stash_queue, queue);
269 cl_command_queue queue = clCreateCommandQueue (this, device, 0, &res);
270 1258
271 if (res)
272 FAIL (CreateCommandQueue, res);
273
274 XPUSH_NEW_OBJ ("OpenCL::Queue", queue);
275}
276
277void 1259void
1260user_event (OpenCL::Context self)
1261 PPCODE:
1262 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
1263 XPUSH_CLOBJ (stash_userevent, ev);
1264
1265void
278buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1266buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
279 PPCODE: 1267 PPCODE:
280{
281 cl_int res;
282 cl_mem mem;
283
284 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))
285 croak ("clCreateBuffer: 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?");
286 1270
287 mem = clCreateBuffer (this, flags, len, 0, &res); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1272 XPUSH_CLOBJ (stash_bufferobj, mem);
288 1273
289 if (res)
290 FAIL (CreateBuffer, res);
291
292 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem);
293}
294
295void 1274void
296buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1275buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
297 PPCODE: 1276 PPCODE:
298{
299 STRLEN len; 1277 STRLEN len;
300 char *ptr = SvPVbyte (data, len); 1278 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
301 cl_int res;
302 cl_mem mem;
303
304 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)))
305 croak ("clCreateBuffer: 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?");
306 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
307 mem = clCreateBuffer (this, flags, len, ptr, &res); 1282 XPUSH_CLOBJ (stash_bufferobj, mem);
308 1283
309 if (res) 1284#if CL_VERSION_1_2
310 FAIL (CreateBuffer, res);
311 1285
312 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem);
313}
314
315void 1286void
316image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, SV *data) 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)
317 PPCODE: 1288 PPCODE:
318{
319 STRLEN len; 1289 STRLEN len;
320 char *ptr = SvPVbyte (data, len); 1290 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
321 const cl_image_format format = { channel_order, channel_type }; 1291 const cl_image_format format = { channel_order, channel_type };
322 cl_int res; 1292 const cl_image_desc desc = {
323 cl_mem mem = clCreateImage2D (this, flags, &format, width, height, len / height, ptr, &res); 1293 type,
1294 width, height, depth,
1295 array_size, row_pitch, slice_pitch,
1296 num_mip_level, num_samples,
1297 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1298 };
1299 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1300 HV *stash = stash_image;
1301 switch (type)
1302 {
1303 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1304 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1305 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1306 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1307 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1308 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1309 }
1310 XPUSH_CLOBJ (stash, mem);
324 1311
325 if (res) 1312#endif
326 FAIL (CreateImage2D, res);
327 1313
328 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
329}
330
331void 1314void
332image3d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t slice_pitch, SV *data) 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)
333 PPCODE: 1316 PPCODE:
334{
335 STRLEN len; 1317 STRLEN len;
336 char *ptr = SvPVbyte (data, len); 1318 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
337 const cl_image_format format = { channel_order, channel_type }; 1319 const cl_image_format format = { channel_order, channel_type };
338 cl_int res; 1320#if PREFER_1_1
339 cl_mem mem = clCreateImage3D (this, flags, &format, width, height, 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
340 depth, len / (height * slice_pitch), slice_pitch, ptr, &res); 1322#else
1323 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1325#endif
1326 XPUSH_CLOBJ (stash_image2d, mem);
341 1327
342 if (res)
343 FAIL (CreateImage3D, res);
344
345 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem);
346}
347
348void 1328void
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)
1330 PPCODE:
1331 STRLEN len;
1332 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1333 const cl_image_format format = { channel_order, channel_type };
1334#if PREFER_1_1
1335 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1336#else
1337 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1339#endif
1340 XPUSH_CLOBJ (stash_image3d, mem);
1341
1342#if cl_apple_gl_sharing || cl_khr_gl_sharing
1343
1344void
1345gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1346 PPCODE:
1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1348 XPUSH_CLOBJ (stash_bufferobj, mem);
1349
1350void
1351gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1352 PPCODE:
1353 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1354 XPUSH_CLOBJ (stash_image2d, mem);
1355
1356#if CL_VERSION_1_2
1357
1358void
1359gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1360 ALIAS:
1361 PPCODE:
1362 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1363 cl_gl_object_type type;
1364 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1365 HV *stash = stash_memory;
1366 switch (type)
1367 {
1368 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1369 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1370 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1371 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1372 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1373 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1374 }
1375 XPUSH_CLOBJ (stash, mem);
1376
1377#endif
1378
1379void
1380gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1381 PPCODE:
1382#if PREFER_1_1
1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1384#else
1385 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1386#endif
1387 XPUSH_CLOBJ (stash_image2d, mem);
1388
1389void
1390gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1391 PPCODE:
1392#if PREFER_1_1
1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1394#else
1395 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1396#endif
1397 XPUSH_CLOBJ (stash_image3d, mem);
1398
1399#endif
1400
1401void
349supported_image_formats (OpenCL::Context this, 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)
350 PPCODE: 1403 PPCODE:
351{ 1404{
352 cl_uint count; 1405 cl_uint count;
353 cl_image_format *list; 1406 cl_image_format *list;
354 int i; 1407 int i;
355 1408
356 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1409 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
357 Newx (list, count, cl_image_format); 1410 Newx (list, count, cl_image_format);
358 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1411 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
359 1412
360 EXTEND (SP, count); 1413 EXTEND (SP, count);
361 for (i = 0; i < count; ++i) 1414 for (i = 0; i < count; ++i)
362 { 1415 {
363 AV *av = newAV (); 1416 AV *av = newAV ();
366 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1419 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
367 } 1420 }
368} 1421}
369 1422
370void 1423void
371sampler (OpenCL::Context this, 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)
372 PPCODE: 1425 PPCODE:
373{
374 cl_int res;
375 cl_sampler sampler = clCreateSampler (this, normalized_coords, addressing_mode, filter_mode, &res); 1426 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
376
377 if (res)
378 FAIL (CreateSampler, res);
379
380 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1427 XPUSH_CLOBJ (stash_sampler, sampler);
381}
382 1428
383void 1429void
384program_with_source (OpenCL::Context this, SV *program) 1430program_with_source (OpenCL::Context self, SV *program)
385 PPCODE: 1431 PPCODE:
386{
387 STRLEN len; 1432 STRLEN len;
388 size_t len2; 1433 size_t len2;
389 const char *ptr = SvPVbyte (program, len); 1434 const char *ptr = SvPVbyte (program, len);
390 cl_int res;
391 cl_program prog;
392 1435
393 len2 = len; 1436 len2 = len;
394 prog = clCreateProgramWithSource (this, 1, &ptr, &len2, &res); 1437 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
395
396 if (res)
397 FAIL (CreateProgramWithSource, res);
398
399 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1438 XPUSH_CLOBJ (stash_program, prog);
400} 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
1515
1516#BEGIN:context
1517
1518void
1519reference_count (OpenCL::Context self)
1520 ALIAS:
1521 reference_count = CL_CONTEXT_REFERENCE_COUNT
1522 num_devices = CL_CONTEXT_NUM_DEVICES
1523 PPCODE:
1524 cl_uint value [1];
1525 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1526 EXTEND (SP, 1);
1527 const int i = 0;
1528 PUSHs (sv_2mortal (newSVuv (value [i])));
1529
1530void
1531devices (OpenCL::Context self)
1532 PPCODE:
1533 size_t size;
1534 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1535 cl_device_id *value = tmpbuf (size);
1536 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1537 int i, n = size / sizeof (*value);
1538 EXTEND (SP, n);
1539 for (i = 0; i < n; ++i)
1540 PUSH_CLOBJ (stash_device, value [i]);
1541
1542void
1543properties (OpenCL::Context self)
1544 PPCODE:
1545 size_t size;
1546 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1547 cl_context_properties *value = tmpbuf (size);
1548 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1549 int i, n = size / sizeof (*value);
1550 EXTEND (SP, n);
1551 for (i = 0; i < n; ++i)
1552 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1553
1554#END:context
401 1555
402MODULE = OpenCL PACKAGE = OpenCL::Queue 1556MODULE = OpenCL PACKAGE = OpenCL::Queue
403 1557
404void 1558void
405DESTROY (OpenCL::Queue this) 1559DESTROY (OpenCL::Queue self)
406 CODE: 1560 CODE:
407 clReleaseCommandQueue (this); 1561 clReleaseCommandQueue (self);
408 1562
409void 1563void
410info (OpenCL::Queue this, cl_command_queue_info name)
411 PPCODE:
412 INFO (CommandQueue)
413
414void
415enqueue_read_buffer (OpenCL::Queue this, 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
416 PPCODE: 1567 PPCODE:
417{
418 cl_event ev = 0;
419 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
420 1569
421 SvUPGRADE (data, SVt_PV); 1570 SvUPGRADE (data, SVt_PV);
422 SvGROW (data, len); 1571 SvGROW (data, len);
423 SvPOK_only (data); 1572 SvPOK_only (data);
424 SvCUR_set (data, len); 1573 SvCUR_set (data, len);
1574
1575 cl_event ev = 0;
425 NEED_SUCCESS (EnqueueReadBuffer, (this, 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));
426 1577
427 if (ev) 1578 if (ev)
428 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
429}
430 1580
431void 1581void
432enqueue_write_buffer (OpenCL::Queue this, 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
433 PPCODE: 1585 PPCODE:
434{ 1586 EVENT_LIST (5);
435 cl_event ev = 0; 1587
436 STRLEN len; 1588 STRLEN len;
437 char *ptr = SvPVbyte (data, len); 1589 char *ptr = SvPVbyte (data, len);
438 EVENT_LIST (5, items - 5);
439 1590
1591 cl_event ev = 0;
440 NEED_SUCCESS (EnqueueReadBuffer, (this, 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));
441 1593
442 if (ev) 1594 if (ev)
443 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1595 XPUSH_CLOBJ (stash_event, ev);
444}
445 1596
1597#if CL_VERSION_1_2
1598
446void 1599void
447enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1600fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1601 ALIAS:
1602 enqueue_fill_buffer = 0
448 PPCODE: 1603 PPCODE:
449{ 1604 EVENT_LIST (5);
1605
1606 STRLEN len;
1607 char *ptr = SvPVbyte (data, len);
1608
450 cl_event ev = 0; 1609 cl_event ev = 0;
451 EVENT_LIST (6, items - 6);
452
453 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, 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));
454 1611
455 if (ev) 1612 if (ev)
456 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1613 XPUSH_CLOBJ (stash_event, ev);
457}
458 1614
459 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html */
460 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html */
461
462void 1615void
463enqueue_read_image (OpenCL::Queue this, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 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
464 PPCODE: 1619 PPCODE:
465{ 1620 EVENT_LIST (12);
1621
1622 const size_t origin [3] = { x, y, z };
1623 const size_t region [3] = { width, height, depth };
1624
1625 const cl_float c_f [4] = { r, g, b, a };
1626 const cl_uint c_u [4] = { r, g, b, a };
1627 const cl_int c_s [4] = { r, g, b, a };
1628 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1629 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1630 cl_image_format format;
1631 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1632 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1633 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1634 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1635
466 cl_event ev = 0; 1636 cl_event ev = 0;
1637 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1638 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1639
1640 if (ev)
1641 XPUSH_CLOBJ (stash_event, ev);
1642
1643#endif
1644
1645void
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
1649 PPCODE:
1650 EVENT_LIST (6);
1651
1652 cl_event 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));
1654
1655 if (ev)
1656 XPUSH_CLOBJ (stash_event, ev);
1657
1658void
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
1662 PPCODE:
1663 EVENT_LIST (17);
1664
1665 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
467 const size_t src_origin[3] = { src_x, src_y, src_z }; 1666 const size_t host_origin[3] = { host_x, host_y, host_z };
468 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1668
1669 if (!buf_row_pitch)
1670 buf_row_pitch = region [0];
1671
1672 if (!buf_slice_pitch)
1673 buf_slice_pitch = region [1] * buf_row_pitch;
1674
1675 if (!host_row_pitch)
1676 host_row_pitch = region [0];
1677
1678 if (!host_slice_pitch)
1679 host_slice_pitch = region [1] * host_row_pitch;
1680
469 size_t len = row_pitch * slice_pitch * depth; 1681 size_t len = host_row_pitch * host_slice_pitch * region [2];
470 EVENT_LIST (11, items - 11);
471 1682
472 SvUPGRADE (data, SVt_PV); 1683 SvUPGRADE (data, SVt_PV);
473 SvGROW (data, len); 1684 SvGROW (data, len);
474 SvPOK_only (data); 1685 SvPOK_only (data);
475 SvCUR_set (data, len); 1686 SvCUR_set (data, len);
476 NEED_SUCCESS (EnqueueReadImage, (this, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1687
1688 cl_event 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));
477 1690
478 if (ev) 1691 if (ev)
479 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1692 XPUSH_CLOBJ (stash_event, ev);
480}
481 1693
482void 1694void
483enqueue_write_image (OpenCL::Queue this, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, SV *data, ...) 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
484 PPCODE: 1698 PPCODE:
485{ 1699 EVENT_LIST (17);
1700
1701 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1702 const size_t host_origin[3] = { host_x, host_y, host_z };
1703 const size_t region[3] = { width, height, depth };
1704 STRLEN len;
1705 char *ptr = SvPVbyte (data, len);
1706
1707 if (!buf_row_pitch)
1708 buf_row_pitch = region [0];
1709
1710 if (!buf_slice_pitch)
1711 buf_slice_pitch = region [1] * buf_row_pitch;
1712
1713 if (!host_row_pitch)
1714 host_row_pitch = region [0];
1715
1716 if (!host_slice_pitch)
1717 host_slice_pitch = region [1] * host_row_pitch;
1718
1719 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1720
1721 if (len < min_len)
1722 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1723
486 cl_event ev = 0; 1724 cl_event 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));
1726
1727 if (ev)
1728 XPUSH_CLOBJ (stash_event, ev);
1729
1730void
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
1734 PPCODE:
1735 EVENT_LIST (16);
1736
1737 const size_t src_origin[3] = { src_x, src_y, src_z };
1738 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1739 const size_t region[3] = { width, height, depth };
1740
1741 cl_event 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));
1743
1744 if (ev)
1745 XPUSH_CLOBJ (stash_event, ev);
1746
1747void
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
1751 PPCODE:
1752 EVENT_LIST (12);
1753
1754 const size_t src_origin[3] = { src_x, src_y, src_z };
1755 const size_t region[3] = { width, height, depth };
1756
1757 if (!row_pitch)
1758 row_pitch = img_row_pitch (src);
1759
1760 if (depth > 1 && !slice_pitch)
1761 slice_pitch = row_pitch * height;
1762
1763 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1764
1765 SvUPGRADE (data, SVt_PV);
1766 SvGROW (data, len);
1767 SvPOK_only (data);
1768 SvCUR_set (data, len);
1769
1770 cl_event 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));
1772
1773 if (ev)
1774 XPUSH_CLOBJ (stash_event, ev);
1775
1776void
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
1780 PPCODE:
1781 EVENT_LIST (12);
1782
487 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 };
488 const size_t region[3] = { width, height, depth }; 1784 const size_t region[3] = { width, height, depth };
489 STRLEN len; 1785 STRLEN len;
490 char *ptr = SvPVbyte (data, len); 1786 char *ptr = SvPVbyte (data, len);
1787
1788 if (!row_pitch)
1789 row_pitch = img_row_pitch (dst);
1790
1791 if (depth > 1 && !slice_pitch)
491 size_t slice_pitch = len / (row_pitch * height); 1792 slice_pitch = row_pitch * height;
492 EVENT_LIST (11, items - 11);
493 1793
1794 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1795
1796 if (len < min_len)
1797 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1798
1799 cl_event ev = 0;
494 NEED_SUCCESS (EnqueueWriteImage, (this, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 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));
495 1801
496 if (ev) 1802 if (ev)
497 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1803 XPUSH_CLOBJ (stash_event, ev);
498}
499 1804
500void 1805void
501enqueue_copy_buffer_rect (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 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
502 PPCODE: 1809 PPCODE:
503{ 1810 EVENT_LIST (12);
504 cl_event ev = 0; 1811
505 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 };
506 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 };
507 const size_t region[3] = { width, height, depth }; 1814 const size_t region[3] = { width, height, depth };
508 EVENT_LIST (16, items - 16);
509 1815
1816 cl_event ev = 0;
510 NEED_SUCCESS (EnqueueCopyBufferRect, (this, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1817 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
511 1818
512 if (ev) 1819 if (ev)
513 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1820 XPUSH_CLOBJ (stash_event, ev);
514}
515 1821
516void 1822void
517enqueue_copy_buffer_to_image (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 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
518 PPCODE: 1826 PPCODE:
519{ 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
520 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
521 const size_t dst_origin[3] = { dst_x, dst_y, dst_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 };
522 const size_t region[3] = { width, height, depth }; 1908 const size_t region[3] = { width, height, depth };
1909 size_t row_pitch, slice_pitch;
523 EVENT_LIST (10, items - 10); 1910 EVENT_LIST (10);
524 1911
525 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 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;
1939 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
526 1940
527 if (ev) 1941 if (ev)
528 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1942 XPUSH_CLOBJ (stash_event, ev);
529}
530 1943
531void 1944void
532enqueue_copy_image (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
533 PPCODE:
534{
535 cl_event ev = 0;
536 const size_t src_origin[3] = { src_x, src_y, src_z };
537 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
538 const size_t region[3] = { width, height, depth };
539 EVENT_LIST (12, items - 12);
540
541 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
542
543 if (ev)
544 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
545}
546
547void
548enqueue_copy_image_to_buffer (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
549 PPCODE:
550{
551 cl_event ev = 0;
552 const size_t src_origin[3] = { src_x, src_y, src_z };
553 const size_t region[3] = { width, height, depth };
554 EVENT_LIST (10, items - 10);
555
556 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
557
558 if (ev)
559 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
560}
561
562void
563enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
564 PPCODE:
565{
566 cl_event ev = 0;
567 EVENT_LIST (2, items - 2);
568
569 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
570
571 if (ev)
572 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
573}
574
575void
576enqueue_nd_range_kernel (OpenCL::Queue this, 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
577 PPCODE: 1948 PPCODE:
578{ 1949 EVENT_LIST (5);
579 cl_event ev = 0; 1950
580 size_t *gwo = 0, *gws, *lws = 0; 1951 size_t *gwo = 0, *gws, *lws = 0;
581 int gws_len; 1952 int gws_len;
582 size_t *lists; 1953 size_t *lists;
583 int i; 1954 int i;
584 EVENT_LIST (2, items - 2);
585 1955
586 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)
587 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1957 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
588 1958
589 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1959 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
590 1960
591 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1961 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
592 1962
593 gws = lists + gws_len * 0; 1963 gws = lists + gws_len * 0;
594 for (i = 0; i < gws_len; ++i) 1964 for (i = 0; i < gws_len; ++i)
1965 {
595 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 }
596 1971
597 if (SvOK (global_work_offset)) 1972 if (SvOK (global_work_offset))
598 { 1973 {
599 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)
600 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");
607 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1982 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
608 } 1983 }
609 1984
610 if (SvOK (local_work_size)) 1985 if (SvOK (local_work_size))
611 { 1986 {
612 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)
613 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");
614 1989
615 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1990 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
616 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");
617 1992
618 lws = lists + gws_len * 2; 1993 lws = lists + gws_len * 2;
619 for (i = 0; i < gws_len; ++i) 1994 for (i = 0; i < gws_len; ++i)
1995 {
620 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 }
621 } 2001 }
622 2002
2003 cl_event ev = 0;
623 NEED_SUCCESS (EnqueueNDRangeKernel, (this, 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));
624 2005
625 if (ev) 2006 if (ev)
626 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2007 XPUSH_CLOBJ (stash_event, ev);
627}
628 2008
2009#if CL_VERSION_1_2
2010
629void 2011void
630enqueue_marker (OpenCL::Queue this) 2012migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2013 ALIAS:
2014 enqueue_migrate_mem_objects = 0
631 PPCODE: 2015 PPCODE:
632{ 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
2028
2029#if cl_apple_gl_sharing || cl_khr_gl_sharing
2030
2031void
2032acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2033 ALIAS:
2034 release_gl_objects = 1
2035 enqueue_acquire_gl_objects = 0
2036 enqueue_release_gl_objects = 1
2037 PPCODE:
2038 EVENT_LIST (2);
2039
2040 cl_uint obj_count;
2041 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2042
633 cl_event ev; 2043 cl_event ev = 0;
634 NEED_SUCCESS (EnqueueMarker, (this, &ev));
635 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
636}
637 2044
2045 if (ix)
2046 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2047 else
2048 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2049
2050 if (ev)
2051 XPUSH_CLOBJ (stash_event, ev);
2052
2053#endif
2054
638void 2055void
639enqueue_wait_for_events (OpenCL::Queue this, ...) 2056wait_for_events (OpenCL::Queue self, ...)
2057 ALIAS:
2058 enqueue_wait_for_events = 0
640 CODE: 2059 CODE:
641{
642 EVENT_LIST (1, items - 1); 2060 EVENT_LIST (1);
2061#if PREFER_1_1
643 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
644} 2063#else
2064 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2065#endif
645 2066
646void 2067void
647enqueue_barrier (OpenCL::Queue this) 2068marker (OpenCL::Queue self, ...)
2069 ALIAS:
2070 enqueue_marker = 0
648 CODE: 2071 PPCODE:
2072 EVENT_LIST (1);
2073 cl_event ev = 0;
2074#if PREFER_1_1
2075 if (!event_list_count)
2076 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2077 else
2078#if CL_VERSION_1_2
2079 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2080#else
2081 {
2082 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2083 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2084 }
2085#endif
2086#else
2087 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2088#endif
2089 if (ev)
2090 XPUSH_CLOBJ (stash_event, ev);
2091
2092void
2093barrier (OpenCL::Queue self, ...)
2094 ALIAS:
2095 enqueue_barrier = 0
2096 PPCODE:
2097 EVENT_LIST (1);
2098 cl_event ev = 0;
2099#if PREFER_1_1
2100 if (!event_list_count && GIMME_V == G_VOID)
649 NEED_SUCCESS (EnqueueBarrier, (this)); 2101 NEED_SUCCESS (EnqueueBarrier, (self));
2102 else
2103#if CL_VERSION_1_2
2104 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2105#else
2106 {
2107 if (event_list_count)
2108 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
650 2109
2110 if (GIMME_V != G_VOID)
2111 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2112 }
2113#endif
2114#else
2115 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2116#endif
2117 if (ev)
2118 XPUSH_CLOBJ (stash_event, ev);
2119
651void 2120void
652flush (OpenCL::Queue this) 2121flush (OpenCL::Queue self)
653 CODE: 2122 CODE:
654 NEED_SUCCESS (Flush, (this)); 2123 NEED_SUCCESS (Flush, (self));
655 2124
656void 2125void
657finish (OpenCL::Queue this) 2126finish (OpenCL::Queue self)
658 CODE: 2127 CODE:
659 NEED_SUCCESS (Finish, (this)); 2128 NEED_SUCCESS (Finish, (self));
2129
2130void
2131info (OpenCL::Queue self, cl_command_queue_info name)
2132 PPCODE:
2133 INFO (CommandQueue)
2134
2135#BEGIN:command_queue
2136
2137void
2138context (OpenCL::Queue self)
2139 PPCODE:
2140 cl_context value [1];
2141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
2142 EXTEND (SP, 1);
2143 const int i = 0;
2144 NEED_SUCCESS (RetainContext, (value [i]));
2145 PUSH_CLOBJ (stash_context, value [i]);
2146
2147void
2148device (OpenCL::Queue self)
2149 PPCODE:
2150 cl_device_id value [1];
2151 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
2152 EXTEND (SP, 1);
2153 const int i = 0;
2154 PUSH_CLOBJ (stash_device, value [i]);
2155
2156void
2157reference_count (OpenCL::Queue self)
2158 PPCODE:
2159 cl_uint value [1];
2160 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
2161 EXTEND (SP, 1);
2162 const int i = 0;
2163 PUSHs (sv_2mortal (newSVuv (value [i])));
2164
2165void
2166properties (OpenCL::Queue self)
2167 PPCODE:
2168 cl_ulong value [1];
2169 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
2170 EXTEND (SP, 1);
2171 const int i = 0;
2172 PUSHs (sv_2mortal (newSVuv (value [i])));
2173
2174#END:command_queue
660 2175
661MODULE = OpenCL PACKAGE = OpenCL::Memory 2176MODULE = OpenCL PACKAGE = OpenCL::Memory
662 2177
663void 2178void
664DESTROY (OpenCL::Memory this) 2179DESTROY (OpenCL::Memory self)
665 CODE: 2180 CODE:
666 clReleaseMemObject (this); 2181 clReleaseMemObject (self);
667 2182
668void 2183void
669info (OpenCL::Memory this, cl_mem_info name) 2184info (OpenCL::Memory self, cl_mem_info name)
670 PPCODE: 2185 PPCODE:
671 INFO (MemObject) 2186 INFO (MemObject)
672 2187
2188void
2189destructor_callback (OpenCL::Memory self, SV *notify)
2190 PPCODE:
2191 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2192
2193#BEGIN:mem
2194
2195void
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
2201 PPCODE:
2202 cl_uint value [1];
2203 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2204 EXTEND (SP, 1);
2205 const int i = 0;
2206 PUSHs (sv_2mortal (newSVuv (value [i])));
2207
2208void
2209flags (OpenCL::Memory self)
2210 PPCODE:
2211 cl_ulong value [1];
2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
2213 EXTEND (SP, 1);
2214 const int i = 0;
2215 PUSHs (sv_2mortal (newSVuv (value [i])));
2216
2217void
2218size (OpenCL::Memory self)
2219 ALIAS:
2220 size = CL_MEM_SIZE
2221 offset = CL_MEM_OFFSET
2222 PPCODE:
2223 size_t value [1];
2224 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2225 EXTEND (SP, 1);
2226 const int i = 0;
2227 PUSHs (sv_2mortal (newSVuv (value [i])));
2228
2229void
2230host_ptr (OpenCL::Memory self)
2231 PPCODE:
2232 void * value [1];
2233 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
2234 EXTEND (SP, 1);
2235 const int i = 0;
2236 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
2237
2238void
2239context (OpenCL::Memory self)
2240 PPCODE:
2241 cl_context value [1];
2242 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
2243 EXTEND (SP, 1);
2244 const int i = 0;
2245 NEED_SUCCESS (RetainContext, (value [i]));
2246 PUSH_CLOBJ (stash_context, value [i]);
2247
2248void
2249associated_memobject (OpenCL::Memory self)
2250 PPCODE:
2251 cl_mem value [1];
2252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
2253 EXTEND (SP, 1);
2254 const int i = 0;
2255 NEED_SUCCESS (RetainMemObject, (value [i]));
2256 PUSH_CLOBJ (stash_memory, value [i]);
2257
2258#END:mem
2259
2260#if cl_apple_gl_sharing || cl_khr_gl_sharing
2261
2262void
2263gl_object_info (OpenCL::Memory self)
2264 PPCODE:
2265 cl_gl_object_type type;
2266 cl_GLuint name;
2267 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2268 EXTEND (SP, 2);
2269 PUSHs (sv_2mortal (newSVuv (type)));
2270 PUSHs (sv_2mortal (newSVuv (name)));
2271
2272#endif
2273
2274MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2275
2276void
2277sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2278 PPCODE:
2279 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2280 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2281
2282 cl_buffer_region crdata = { origin, size };
2283
2284 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2285 XPUSH_CLOBJ (stash_buffer, mem);
2286
2287MODULE = OpenCL PACKAGE = OpenCL::Image
2288
2289void
2290image_info (OpenCL::Image self, cl_image_info name)
2291 PPCODE:
2292 INFO (Image)
2293
2294void
2295format (OpenCL::Image self)
2296 PPCODE:
2297 cl_image_format format;
2298 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2299 EXTEND (SP, 2);
2300 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2301 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2302
2303#BEGIN:image
2304
2305void
2306element_size (OpenCL::Image self)
2307 ALIAS:
2308 element_size = CL_IMAGE_ELEMENT_SIZE
2309 row_pitch = CL_IMAGE_ROW_PITCH
2310 slice_pitch = CL_IMAGE_SLICE_PITCH
2311 width = CL_IMAGE_WIDTH
2312 height = CL_IMAGE_HEIGHT
2313 depth = CL_IMAGE_DEPTH
2314 PPCODE:
2315 size_t value [1];
2316 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
2317 EXTEND (SP, 1);
2318 const int i = 0;
2319 PUSHs (sv_2mortal (newSVuv (value [i])));
2320
2321#END:image
2322
2323#if cl_apple_gl_sharing || cl_khr_gl_sharing
2324
2325#BEGIN:gl_texture
2326
2327void
2328target (OpenCL::Image self)
2329 PPCODE:
2330 cl_GLenum value [1];
2331 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2332 EXTEND (SP, 1);
2333 const int i = 0;
2334 PUSHs (sv_2mortal (newSVuv (value [i])));
2335
2336void
2337gl_mipmap_level (OpenCL::Image self)
2338 PPCODE:
2339 cl_GLint value [1];
2340 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2341 EXTEND (SP, 1);
2342 const int i = 0;
2343 PUSHs (sv_2mortal (newSViv (value [i])));
2344
2345#END:gl_texture
2346
2347#endif
2348
673MODULE = OpenCL PACKAGE = OpenCL::Sampler 2349MODULE = OpenCL PACKAGE = OpenCL::Sampler
674 2350
675void 2351void
676DESTROY (OpenCL::Sampler this) 2352DESTROY (OpenCL::Sampler self)
677 CODE: 2353 CODE:
678 clReleaseSampler (this); 2354 clReleaseSampler (self);
679 2355
680void 2356void
681info (OpenCL::Sampler this, cl_sampler_info name) 2357info (OpenCL::Sampler self, cl_sampler_info name)
682 PPCODE: 2358 PPCODE:
683 INFO (Sampler) 2359 INFO (Sampler)
684 2360
2361#BEGIN:sampler
2362
2363void
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
2369 PPCODE:
2370 cl_uint value [1];
2371 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
2372 EXTEND (SP, 1);
2373 const int i = 0;
2374 PUSHs (sv_2mortal (newSVuv (value [i])));
2375
2376void
2377context (OpenCL::Sampler self)
2378 PPCODE:
2379 cl_context value [1];
2380 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2381 EXTEND (SP, 1);
2382 const int i = 0;
2383 NEED_SUCCESS (RetainContext, (value [i]));
2384 PUSH_CLOBJ (stash_context, value [i]);
2385
2386void
2387filter_mode (OpenCL::Sampler self)
2388 PPCODE:
2389 cl_uint value [1];
2390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2391 EXTEND (SP, 1);
2392 const int i = 0;
2393 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2394
2395#END:sampler
2396
685MODULE = OpenCL PACKAGE = OpenCL::Program 2397MODULE = OpenCL PACKAGE = OpenCL::Program
686 2398
687void 2399void
688DESTROY (OpenCL::Program this) 2400DESTROY (OpenCL::Program self)
689 CODE: 2401 CODE:
690 clReleaseProgram (this); 2402 clReleaseProgram (self);
691 2403
692void 2404void
693info (OpenCL::Program this, cl_program_info name) 2405build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2406 ALIAS:
2407 build_async = 1
694 PPCODE: 2408 CODE:
695 INFO (Program) 2409 cl_uint device_count = 0;
2410 cl_device_id *device_list = 0;
696 2411
2412 if (SvOK (devices))
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
697void 2425void
698build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2426compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
699 CODE: 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 {
2440 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2441 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2442
2443 HV *hv = (HV *)SvRV (devices);
2444
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))
2452 {
2453 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2454 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2455 ++i;
2456 }
2457 }
2458
2459 void *user_data;
2460 program_callback pfn_notify = make_program_callback (notify, &user_data);
2461
700 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2462 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2463 header_count, header_list, header_name, pfn_notify, user_data));
701 2464
2465#endif
2466
702void 2467void
703build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2468build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
704 PPCODE: 2469 PPCODE:
705{
706 size_t size; 2470 size_t size;
707 SV *sv;
708
709 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2471 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
710 sv = sv_2mortal (newSV (size)); 2472 SV *sv = sv_2mortal (newSV (size));
711 SvUPGRADE (sv, SVt_PV); 2473 SvUPGRADE (sv, SVt_PV);
712 SvPOK_only (sv); 2474 SvPOK_only (sv);
713 SvCUR_set (sv, size); 2475 SvCUR_set (sv, size);
714 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2476 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
715 XPUSHs (sv); 2477 XPUSHs (sv);
716} 2478
2479#BEGIN:program_build
2480
2481void
2482build_status (OpenCL::Program self, OpenCL::Device device)
2483 PPCODE:
2484 cl_int value [1];
2485 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2486 EXTEND (SP, 1);
2487 const int i = 0;
2488 PUSHs (sv_2mortal (newSViv (value [i])));
2489
2490void
2491build_options (OpenCL::Program self, OpenCL::Device device)
2492 ALIAS:
2493 build_options = CL_PROGRAM_BUILD_OPTIONS
2494 build_log = CL_PROGRAM_BUILD_LOG
2495 PPCODE:
2496 size_t size;
2497 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2498 char *value = tmpbuf (size);
2499 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2500 EXTEND (SP, 1);
2501 const int i = 0;
2502 PUSHs (sv_2mortal (newSVpv (value, 0)));
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
2513#END:program_build
717 2514
718void 2515void
719kernel (OpenCL::Program program, SV *function) 2516kernel (OpenCL::Program program, SV *function)
720 PPCODE: 2517 PPCODE:
721{
722 cl_int res;
723 cl_kernel kernel = clCreateKernel (program, SvPVbyte_nolen (function), &res); 2518 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2519 XPUSH_CLOBJ (stash_kernel, kernel);
724 2520
725 if (res) 2521void
726 FAIL (CreateKernel, res); 2522kernels_in_program (OpenCL::Program program)
2523 PPCODE:
2524 cl_uint num_kernels;
2525 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2526 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2527 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
727 2528
728 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2529 int i;
729} 2530 EXTEND (SP, num_kernels);
2531 for (i = 0; i < num_kernels; ++i)
2532 PUSH_CLOBJ (stash_kernel, kernels [i]);
2533
2534void
2535info (OpenCL::Program self, cl_program_info name)
2536 PPCODE:
2537 INFO (Program)
2538
2539void
2540binaries (OpenCL::Program self)
2541 PPCODE:
2542 cl_uint n, i;
2543 size_t size;
2544
2545 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2546 if (!n) XSRETURN_EMPTY;
2547
2548 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2549 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2550 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2551 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2552
2553 EXTEND (SP, n);
2554 for (i = 0; i < n; ++i)
2555 {
2556 SV *sv = sv_2mortal (newSV (sizes [i]));
2557 SvUPGRADE (sv, SVt_PV);
2558 SvPOK_only (sv);
2559 SvCUR_set (sv, sizes [i]);
2560 ptrs [i] = (void *)SvPVX (sv);
2561 PUSHs (sv);
2562 }
2563
2564 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2565 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2566
2567#BEGIN:program
2568
2569void
2570reference_count (OpenCL::Program self)
2571 ALIAS:
2572 reference_count = CL_PROGRAM_REFERENCE_COUNT
2573 num_devices = CL_PROGRAM_NUM_DEVICES
2574 PPCODE:
2575 cl_uint value [1];
2576 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2577 EXTEND (SP, 1);
2578 const int i = 0;
2579 PUSHs (sv_2mortal (newSVuv (value [i])));
2580
2581void
2582context (OpenCL::Program self)
2583 PPCODE:
2584 cl_context value [1];
2585 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2586 EXTEND (SP, 1);
2587 const int i = 0;
2588 NEED_SUCCESS (RetainContext, (value [i]));
2589 PUSH_CLOBJ (stash_context, value [i]);
2590
2591void
2592devices (OpenCL::Program self)
2593 PPCODE:
2594 size_t size;
2595 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2596 cl_device_id *value = tmpbuf (size);
2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2598 int i, n = size / sizeof (*value);
2599 EXTEND (SP, n);
2600 for (i = 0; i < n; ++i)
2601 PUSH_CLOBJ (stash_device, value [i]);
2602
2603void
2604source (OpenCL::Program self)
2605 PPCODE:
2606 size_t size;
2607 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2608 char *value = tmpbuf (size);
2609 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2610 EXTEND (SP, 1);
2611 const int i = 0;
2612 PUSHs (sv_2mortal (newSVpv (value, 0)));
2613
2614void
2615binary_sizes (OpenCL::Program self)
2616 PPCODE:
2617 size_t size;
2618 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2619 size_t *value = tmpbuf (size);
2620 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2621 int i, n = size / sizeof (*value);
2622 EXTEND (SP, n);
2623 for (i = 0; i < n; ++i)
2624 PUSHs (sv_2mortal (newSVuv (value [i])));
2625
2626#END:program
730 2627
731MODULE = OpenCL PACKAGE = OpenCL::Kernel 2628MODULE = OpenCL PACKAGE = OpenCL::Kernel
732 2629
733void 2630void
734DESTROY (OpenCL::Kernel this) 2631DESTROY (OpenCL::Kernel self)
735 CODE: 2632 CODE:
736 clReleaseKernel (this); 2633 clReleaseKernel (self);
737 2634
738void 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
2701set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2702 CODE:
2703 clSetKernelArg (self, idx, sizeof (value), &value);
2704
2705void
2706set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2707 CODE:
2708 clSetKernelArg (self, idx, sizeof (value), &value);
2709
2710void
2711set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2712 CODE:
2713 clSetKernelArg (self, idx, sizeof (value), &value);
2714
2715void
2716set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2717 CODE:
2718 clSetKernelArg (self, idx, sizeof (value), &value);
2719
2720void
2721set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2722 CODE:
2723 clSetKernelArg (self, idx, sizeof (value), &value);
2724
2725void
2726set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2727 CODE:
2728 clSetKernelArg (self, idx, sizeof (value), &value);
2729
2730void
2731set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2732 CODE:
2733 clSetKernelArg (self, idx, sizeof (value), &value);
2734
2735void
2736set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2737 CODE:
2738 clSetKernelArg (self, idx, sizeof (value), &value);
2739
2740void
2741set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2742 CODE:
2743 clSetKernelArg (self, idx, sizeof (value), &value);
2744
2745void
2746set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2747 CODE:
2748 clSetKernelArg (self, idx, sizeof (value), &value);
2749
2750void
2751set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2752 CODE:
2753 clSetKernelArg (self, idx, sizeof (value), &value);
2754
2755void
2756set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2757 CODE:
2758 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2759
2760void
2761set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2762 CODE:
2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2764
2765void
2766set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2767 CODE:
2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2769
2770void
2771set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2772 CODE:
2773 clSetKernelArg (self, idx, sizeof (value), &value);
2774
2775void
2776set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2777 CODE:
2778 clSetKernelArg (self, idx, size, 0);
2779
2780void
2781set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2782 CODE:
2783 clSetKernelArg (self, idx, sizeof (value), &value);
2784
2785void
739info (OpenCL::Kernel this, cl_kernel_info name) 2786info (OpenCL::Kernel self, cl_kernel_info name)
740 PPCODE: 2787 PPCODE:
741 INFO (Kernel) 2788 INFO (Kernel)
742 2789
2790#BEGIN:kernel
2791
743void 2792void
744set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2793function_name (OpenCL::Kernel self)
2794 PPCODE:
2795 size_t size;
2796 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2797 char *value = tmpbuf (size);
2798 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2799 EXTEND (SP, 1);
2800 const int i = 0;
2801 PUSHs (sv_2mortal (newSVpv (value, 0)));
2802
2803void
2804num_args (OpenCL::Kernel self)
2805 ALIAS:
2806 num_args = CL_KERNEL_NUM_ARGS
2807 reference_count = CL_KERNEL_REFERENCE_COUNT
2808 PPCODE:
2809 cl_uint value [1];
2810 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2811 EXTEND (SP, 1);
2812 const int i = 0;
2813 PUSHs (sv_2mortal (newSVuv (value [i])));
2814
2815void
2816context (OpenCL::Kernel self)
2817 PPCODE:
2818 cl_context value [1];
2819 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2820 EXTEND (SP, 1);
2821 const int i = 0;
2822 NEED_SUCCESS (RetainContext, (value [i]));
2823 PUSH_CLOBJ (stash_context, value [i]);
2824
2825void
2826program (OpenCL::Kernel self)
2827 PPCODE:
2828 cl_program value [1];
2829 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2830 EXTEND (SP, 1);
2831 const int i = 0;
2832 NEED_SUCCESS (RetainProgram, (value [i]));
2833 PUSH_CLOBJ (stash_program, value [i]);
2834
2835#END:kernel
2836
2837void
2838work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
745 CODE: 2839 PPCODE:
746 clSetKernelArg (this, idx, sizeof (value), &value); 2840 size_t size;
2841 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2842 SV *sv = sv_2mortal (newSV (size));
2843 SvUPGRADE (sv, SVt_PV);
2844 SvPOK_only (sv);
2845 SvCUR_set (sv, size);
2846 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2847 XPUSHs (sv);
747 2848
2849#BEGIN:kernel_work_group
2850
748void 2851void
749set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value) 2852work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2853 ALIAS:
2854 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2855 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2856 PPCODE:
2857 size_t value [1];
2858 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2859 EXTEND (SP, 1);
2860 const int i = 0;
2861 PUSHs (sv_2mortal (newSVuv (value [i])));
2862
2863void
2864compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2865 PPCODE:
2866 size_t size;
2867 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2868 size_t *value = tmpbuf (size);
2869 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2870 int i, n = size / sizeof (*value);
2871 EXTEND (SP, n);
2872 for (i = 0; i < n; ++i)
2873 PUSHs (sv_2mortal (newSVuv (value [i])));
2874
2875void
2876local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
2877 ALIAS:
2878 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2879 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2880 PPCODE:
2881 cl_ulong value [1];
2882 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2883 EXTEND (SP, 1);
2884 const int i = 0;
2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2886
2887#END:kernel_work_group
2888
2889#if CL_VERSION_1_2
2890
2891void
2892arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
750 CODE: 2893 PPCODE:
751 clSetKernelArg (this, idx, sizeof (value), &value); 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);
752 2902
753void 2903#BEGIN:kernel_arg
754set_short (OpenCL::Kernel this, cl_uint idx, cl_short value)
755 CODE:
756 clSetKernelArg (this, idx, sizeof (value), &value);
757 2904
758void 2905void
759set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2906arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
760 CODE: 2907 ALIAS:
761 clSetKernelArg (this, idx, sizeof (value), &value); 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])));
762 2916
763void 2917void
764set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2918arg_type_name (OpenCL::Kernel self, cl_uint idx)
765 CODE: 2919 ALIAS:
766 clSetKernelArg (this, idx, sizeof (value), &value); 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)));
767 2930
768void 2931void
769set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value) 2932arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
770 CODE: 2933 PPCODE:
771 clSetKernelArg (this, idx, sizeof (value), &value); 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])));
772 2939
773void 2940#END:kernel_arg
774set_long (OpenCL::Kernel this, cl_uint idx, cl_long value)
775 CODE:
776 clSetKernelArg (this, idx, sizeof (value), &value);
777 2941
778void 2942#endif
779set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
780 CODE:
781 clSetKernelArg (this, idx, sizeof (value), &value);
782
783void
784set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
785 CODE:
786 clSetKernelArg (this, idx, sizeof (value), &value);
787
788void
789set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
790 CODE:
791 clSetKernelArg (this, idx, sizeof (value), &value);
792
793void
794set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
795 CODE:
796 clSetKernelArg (this, idx, sizeof (value), &value);
797
798void
799set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
800 CODE:
801 clSetKernelArg (this, idx, sizeof (value), &value);
802
803void
804set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
805 CODE:
806 clSetKernelArg (this, idx, sizeof (value), &value);
807
808void
809set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
810 CODE:
811 clSetKernelArg (this, idx, sizeof (value), &value);
812
813void
814set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
815 CODE:
816 clSetKernelArg (this, idx, sizeof (value), &value);
817
818void
819set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
820 CODE:
821 clSetKernelArg (this, idx, sizeof (value), &value);
822 2943
823MODULE = OpenCL PACKAGE = OpenCL::Event 2944MODULE = OpenCL PACKAGE = OpenCL::Event
824 2945
825void 2946void
826DESTROY (OpenCL::Event this) 2947DESTROY (OpenCL::Event self)
827 CODE: 2948 CODE:
828 clReleaseEvent (this); 2949 clReleaseEvent (self);
829 2950
830void 2951void
2952wait (OpenCL::Event self)
2953 CODE:
2954 clWaitForEvents (1, &self);
2955
2956void
2957cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2958 CODE:
2959 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2960
2961void
831info (OpenCL::Event this, cl_event_info name) 2962info (OpenCL::Event self, cl_event_info name)
832 PPCODE: 2963 PPCODE:
833 INFO (Event) 2964 INFO (Event)
834 2965
2966#BEGIN:event
2967
835void 2968void
836wait (OpenCL::Event this) 2969command_queue (OpenCL::Event self)
2970 PPCODE:
2971 cl_command_queue value [1];
2972 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2973 EXTEND (SP, 1);
2974 const int i = 0;
2975 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2976 PUSH_CLOBJ (stash_queue, value [i]);
2977
2978void
2979command_type (OpenCL::Event self)
2980 ALIAS:
2981 command_type = CL_EVENT_COMMAND_TYPE
2982 reference_count = CL_EVENT_REFERENCE_COUNT
2983 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2984 PPCODE:
2985 cl_uint value [1];
2986 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2987 EXTEND (SP, 1);
2988 const int i = 0;
2989 PUSHs (sv_2mortal (newSVuv (value [i])));
2990
2991void
2992context (OpenCL::Event self)
2993 PPCODE:
2994 cl_context value [1];
2995 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2996 EXTEND (SP, 1);
2997 const int i = 0;
2998 NEED_SUCCESS (RetainContext, (value [i]));
2999 PUSH_CLOBJ (stash_context, value [i]);
3000
3001#END:event
3002
3003void
3004profiling_info (OpenCL::Event self, cl_profiling_info name)
837 CODE: 3005 PPCODE:
838 clWaitForEvents (1, &this); 3006 INFO (EventProfiling)
839 3007
3008#BEGIN:profiling
3009
3010void
3011profiling_command_queued (OpenCL::Event self)
3012 ALIAS:
3013 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
3014 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
3015 profiling_command_start = CL_PROFILING_COMMAND_START
3016 profiling_command_end = CL_PROFILING_COMMAND_END
3017 PPCODE:
3018 cl_ulong value [1];
3019 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
3020 EXTEND (SP, 1);
3021 const int i = 0;
3022 PUSHs (sv_2mortal (newSVuv (value [i])));
3023
3024#END:profiling
3025
3026MODULE = OpenCL PACKAGE = OpenCL::UserEvent
3027
3028void
3029set_status (OpenCL::UserEvent self, cl_int execution_status)
3030 CODE:
3031 clSetUserEventStatus (self, execution_status);
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