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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.5 by root, Wed Nov 16 00:35:30 2011 UTC vs.
Revision 1.67 by root, Fri May 4 14:49:42 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
10
11#ifdef I_DLFCN
12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
30#endif
31
32#ifdef __APPLE__
33 #include <OpenCL/opencl.h>
34#else
5#include <CL/opencl.h> 35 #include <CL/opencl.h>
36#endif
6 37
38#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1
40 #define PREFER_1_1 1
41#endif
42
7typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
8typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
9typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
10typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
11typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
12typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
50typedef cl_mem OpenCL__BufferObj;
13typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
14typedef cl_mem OpenCL__Image2D;
15typedef cl_mem OpenCL__Image3D;
16typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
17typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
18typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
19typedef cl_mem OpenCL__Image2D_ornull;
20typedef cl_mem OpenCL__Image3D_ornull;
21typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
22typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
23typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
24typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
25typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
60
61typedef struct mapped * OpenCL__Mapped;
62
63typedef SV *FUTURE;
64
65static HV
66 *stash_platform,
67 *stash_device,
68 *stash_subdevice,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
26 89
27/*****************************************************************************/ 90/*****************************************************************************/
28 91
92// name must include a leading underscore
93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
94// for perl. doh.
95static void *
96glsym (const char *name)
97{
98 void *fun = 0;
99
100 #if defined I_DLFCN && defined RTLD_DEFAULT
101 fun = dlsym (RTLD_DEFAULT, name + 1);
102 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
103
104 if (!fun)
105 {
106 static void *libgl;
107 static const char *glso[] = {
108 "libGL.so.1",
109 "libGL.so.3",
110 "libGL.so.4.0",
111 "libGL.so",
112 "/usr/lib/libGL.so",
113 "/usr/X11R6/lib/libGL.1.dylib"
114 };
115 int i;
116
117 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
118 {
119 libgl = dlopen (glso [i], RTLD_LAZY);
120 if (libgl)
121 break;
122 }
123
124 if (libgl)
125 {
126 fun = dlsym (libgl, name + 1);
127 if (!fun) fun = dlsym (libgl, name);
128 }
129 }
130 #endif
131
132 return fun;
133}
134
135/*****************************************************************************/
136
29/* up to two temporary buffers */ 137/* up to two temporary buffers */
30static void * 138static void * ecb_noinline
31tmpbuf (size_t size) 139tmpbuf (size_t size)
32{ 140{
141 enum { buffers = 4 };
33 static int idx; 142 static int idx;
34 static void *buf [2]; 143 static void *buf [buffers];
35 static size_t len [2]; 144 static size_t len [buffers];
36 145
37 idx ^= 1; 146 idx = (idx + 1) % buffers;
38 147
39 if (len [idx] < size) 148 if (len [idx] < size)
40 { 149 {
41 free (buf [idx]); 150 free (buf [idx]);
42 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
90 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)"); 199 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)");
91} 200}
92 201
93/*****************************************************************************/ 202/*****************************************************************************/
94 203
95static cl_int last_error; 204static cl_int res;
96 205
97#define FAIL(name,err) \ 206#define FAIL(name) \
98 croak ("cl" # name ": %s", err2str (last_error = err)); 207 croak ("cl" # name ": %s", err2str (res));
99 208
100#define NEED_SUCCESS(name,args) \ 209#define NEED_SUCCESS(name,args) \
101 do { \ 210 do { \
102 cl_int res = cl ## name args; \ 211 res = cl ## name args; \
103 \ 212 \
104 if (res) \ 213 if (res) \
105 FAIL (name, res); \ 214 FAIL (name); \
106 } while (0) 215 } while (0)
107 216
217#define NEED_SUCCESS_ARG(retdecl, name, args) \
218 retdecl = cl ## name args; \
219 if (res) \
220 FAIL (name);
221
108/*****************************************************************************/ 222/*****************************************************************************/
109 223
110#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
111#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
112
113static void * 224static SV *
225new_clobj (HV *stash, IV id)
226{
227 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
228}
229
230#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
231#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
232
233/* cl objects are either \$iv, or [$iv, ...] */
234/* they can be upgraded at runtime to the array form */
235static void * ecb_noinline
114SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 236SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
115{ 237{
238 // sv_derived_from is quite slow :(
116 if (SvROK (sv) && sv_derived_from (sv, pkg)) 239 if (SvROK (sv) && sv_derived_from (sv, pkg))
117 return (void *)SvIV (SvRV (sv)); 240 return (void *)SvIV (SvRV (sv));
118 241
119 croak ("%s: %s is not of type %s", func, svname, pkg); 242 croak ("%s: %s is not of type %s", func, svname, pkg);
243}
244
245// the "no-inherit" version of the above
246static void * ecb_noinline
247SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
248{
249 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
250 return (void *)SvIV (SvRV (sv));
251
252 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
120} 253}
121 254
122/*****************************************************************************/ 255/*****************************************************************************/
123 256
124static cl_event * 257static cl_context_properties * ecb_noinline
258SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
259{
260 if (!sv || !SvOK (sv))
261 if (extra)
262 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
263 else
264 return 0;
265
266 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
267 {
268 AV *av = (AV *)SvRV (sv);
269 int i, len = av_len (av) + 1;
270 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
271 cl_context_properties *l = p;
272
273 if (len & 1)
274 croak ("%s: %s is not a property list (must contain an even number of elements)", func, svname);
275
276 while (extracount--)
277 *l++ = *extra++;
278
279 for (i = 0; i < len; i += 2)
280 {
281 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
282 SV *p_sv = *av_fetch (av, i + 1, 0);
283 cl_context_properties v = SvIV (p_sv); // code below can override
284
285 switch (t)
286 {
287 case CL_CONTEXT_PLATFORM:
288 if (SvROK (p_sv))
289 v = (cl_context_properties)SvCLOBJ (func, svname, p_sv, "OpenCL::Platform");
290 break;
291
292 case CL_GLX_DISPLAY_KHR:
293 if (!SvOK (p_sv))
294 {
295 void *func = glsym ("_glXGetCurrentDisplay");
296 if (func)
297 v = (cl_context_properties)((void *(*)(void))func)();
298 }
299 break;
300
301 case CL_GL_CONTEXT_KHR:
302 if (!SvOK (p_sv))
303 {
304 void *func = glsym ("_glXGetCurrentContext");
305 if (func)
306 v = (cl_context_properties)((void *(*)(void))func)();
307 }
308 break;
309
310 default:
311 /* unknown property, treat as int */
312 break;
313 }
314
315 *l++ = t;
316 *l++ = v;
317 }
318
319 *l = 0;
320
321 return p;
322 }
323
324 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
325}
326
327/*****************************************************************************/
328/* callback stuff */
329
330/* default context callback, log to stderr */
331static void CL_CALLBACK
332context_default_notify (const char *msg, const void *info, size_t cb, void *data)
333{
334 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
335}
336
337typedef struct
338{
339 int free_cb;
340 void (*push)(void *data1, void *data2, void *data3);
341} eq_vtbl;
342
343typedef struct eq_item
344{
345 struct eq_item *next;
346 eq_vtbl *vtbl;
347 SV *cb;
348 void *data1, *data2, *data3;
349} eq_item;
350
351static void (*eq_signal_func)(void *signal_arg, int value);
352static void *eq_signal_arg;
353static xmutex_t eq_lock = X_MUTEX_INIT;
354static eq_item *eq_head, *eq_tail;
355
356static void ecb_noinline
357eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
358{
359 eq_item *item = malloc (sizeof (eq_item));
360
361 item->next = 0;
362 item->vtbl = vtbl;
363 item->cb = cb;
364 item->data1 = data1;
365 item->data2 = data2;
366 item->data3 = data3;
367
368 X_LOCK (eq_lock);
369
370 *(eq_head ? &eq_tail->next : &eq_head) = item;
371 eq_tail = item;
372
373 X_UNLOCK (eq_lock);
374
375 eq_signal_func (eq_signal_arg, 0);
376}
377
378static eq_item *
379eq_dec (void)
380{
381 eq_item *res;
382
383 X_LOCK (eq_lock);
384
385 res = eq_head;
386 if (res)
387 eq_head = res->next;
388
389 X_UNLOCK (eq_lock);
390
391 return res;
392}
393
394static void
395eq_poll (void)
396{
397 eq_item *item;
398
399 while ((item = eq_dec ()))
400 {
401 ENTER;
402 SAVETMPS;
403
404 dSP;
405 PUSHMARK (SP);
406 EXTEND (SP, 2);
407
408 if (item->vtbl->free_cb)
409 sv_2mortal (item->cb);
410
411 PUTBACK;
412 item->vtbl->push (item->data1, item->data2, item->data3);
413
414 SV *cb = item->cb;
415 free (item);
416
417 call_sv (cb, G_DISCARD | G_VOID);
418
419 FREETMPS;
420 LEAVE;
421 }
422}
423
424static void
425eq_poll_interrupt (pTHX_ void *c_arg, int value)
426{
427 eq_poll ();
428}
429
430/*****************************************************************************/
431/* context notify */
432
433static void ecb_noinline
434eq_context_push (void *data1, void *data2, void *data3)
435{
436 dSP;
437 PUSHs (sv_2mortal (newSVpv (data1, 0)));
438 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
439 PUTBACK;
440
441 free (data1);
442 free (data2);
443}
444
445static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
446
447static void CL_CALLBACK
448eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
449{
450 void *pvt_copy = malloc (cb);
451 memcpy (pvt_copy, pvt, cb);
452 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
453}
454
455#define CONTEXT_NOTIFY_CALLBACK \
456 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
457 void *user_data = 0; \
458 \
459 if (SvOK (notify)) \
460 { \
461 pfn_notify = eq_context_notify; \
462 user_data = s_get_cv (notify); \
463 }
464
465static SV * ecb_noinline
466new_clobj_context (cl_context ctx, void *user_data)
467{
468 SV *sv = new_clobj (stash_context, (IV)ctx);
469
470 if (user_data)
471 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
472
473 return sv;
474}
475
476#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
477
478/*****************************************************************************/
479/* build/compile/link notify */
480
481static void
482eq_program_push (void *data1, void *data2, void *data3)
483{
484 dSP;
485 PUSH_CLOBJ (stash_program, data1);
486 PUTBACK;
487}
488
489static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
490
491static void CL_CALLBACK
492eq_program_notify (cl_program program, void *user_data)
493{
494 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
495}
496
497struct build_args
498{
499 cl_program program;
500 char *options;
501 void *user_data;
502 cl_uint num_devices;
503};
504
505X_THREAD_PROC (build_program_thread)
506{
507 struct build_args *arg = thr_arg;
508
509 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
510
511 if (arg->user_data)
512 eq_program_notify (arg->program, arg->user_data);
513 else
514 clReleaseProgram (arg->program);
515
516 free (arg->options);
517 free (arg);
518
519 return 0;
520}
521
522static void
523build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
524{
525 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
526
527 arg->program = program;
528 arg->options = strdup (options);
529 arg->user_data = user_data;
530 arg->num_devices = num_devices;
531 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
532
533 xthread_t id;
534 thread_create (&id, build_program_thread, arg);
535}
536
537/*****************************************************************************/
538/* event objects */
539
540static void
541eq_event_push (void *data1, void *data2, void *data3)
542{
543 dSP;
544 PUSH_CLOBJ (stash_event, data1);
545 PUSHs (sv_2mortal (newSViv ((IV)data2)));
546 PUTBACK;
547}
548
549static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
550
551static void CL_CALLBACK
552eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
553{
554 clRetainEvent (event);
555 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
556}
557
558/*****************************************************************************/
559/* utilities for XS code */
560
561static size_t
562img_row_pitch (cl_mem img)
563{
564 size_t res;
565 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
566 return res;
567}
568
569static cl_event * ecb_noinline
125event_list (SV **items, int count) 570event_list (SV **items, cl_uint *rcount, cl_event extra)
126{ 571{
572 cl_uint count = *rcount;
573
574 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
575 *rcount = count = 0;
576
577 if (!count && !extra)
578 return 0;
579
127 cl_event *list = tmpbuf (sizeof (cl_event) * count); 580 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
581 int i = 0;
128 582
129 while (count--) 583 while (count--)
584 if (SvOK (items [count]))
130 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 585 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
131 586
587 if (extra)
588 list [i++] = extra;
589
590 *rcount = i;
591
132 return list; 592 return i ? list : 0;
133} 593}
134 594
135#define EVENT_LIST(items,count) \ 595#define EVENT_LIST(skip) \
136 cl_uint event_list_count = (count); \ 596 cl_uint event_list_count = items - (skip); \
137 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 597 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
138 598
139#define INFO(class) \ 599#define INFO(class) \
140{ \ 600{ \
141 size_t size; \ 601 size_t size; \
142 SV *sv; \
143 \
144 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 602 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
145 sv = sv_2mortal (newSV (size)); \ 603 SV *sv = sv_2mortal (newSV (size)); \
146 SvUPGRADE (sv, SVt_PV); \ 604 SvUPGRADE (sv, SVt_PV); \
147 SvPOK_only (sv); \ 605 SvPOK_only (sv); \
148 SvCUR_set (sv, size); \ 606 SvCUR_set (sv, size); \
149 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 607 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
150 XPUSHs (sv); \ 608 XPUSHs (sv); \
151} 609}
152 610
611/*****************************************************************************/
612/* mapped_xxx */
613
614static OpenCL__Mapped
615SvMAPPED (SV *sv)
616{
617 // no typechecking atm., keep your fingers crossed
618 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
619}
620
621struct mapped
622{
623 cl_command_queue queue;
624 cl_mem memobj;
625 void *ptr;
626 size_t cb;
627 cl_event event;
628 size_t row_pitch;
629 size_t slice_pitch;
630};
631
632static SV *
633mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
634{
635 SV *data = newSV (0);
636 SvUPGRADE (data, SVt_PVMG);
637
638 OpenCL__Mapped mapped;
639 New (0, mapped, 1, struct mapped);
640
641 clRetainCommandQueue (queue);
642
643 mapped->queue = queue;
644 mapped->memobj = memobj;
645 mapped->ptr = ptr;
646 mapped->cb = cb;
647 mapped->event = ev;
648 mapped->row_pitch = row_pitch;
649 mapped->slice_pitch = slice_pitch;
650
651 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
652
653 if (SvLEN (data))
654 Safefree (data);
655
656 SvPVX (data) = (char *)ptr;
657 SvCUR_set (data, cb);
658 SvLEN_set (data, 0);
659 SvPOK_only (data);
660
661 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
662
663 if (!(flags & CL_MAP_WRITE))
664 SvREADONLY_on (data);
665
666 return obj;
667}
668
669static void
670mapped_detach (SV *sv, OpenCL__Mapped mapped)
671{
672 SV *data = SvRV (sv);
673
674 // the next check checks both before AND after detach, where SvPVX should STILL be 0
675 if (SvPVX (data) != (char *)mapped->ptr)
676 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
677 else
678 {
679 SvREADONLY_off (data);
680 SvCUR_set (data, 0);
681 SvPVX (data) = 0;
682 SvOK_off (data);
683 }
684
685 mapped->ptr = 0;
686}
687
688static void
689mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
690{
691 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
692 cl_event ev;
693
694 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
695
696 clReleaseEvent (mapped->event);
697 mapped->event = ev;
698
699 mapped_detach (self, mapped);
700}
701
702/*****************************************************************************/
703
153MODULE = OpenCL PACKAGE = OpenCL 704MODULE = OpenCL PACKAGE = OpenCL
154 705
155PROTOTYPES: ENABLE 706PROTOTYPES: ENABLE
156 707
708void
709poll ()
710 CODE:
711 eq_poll ();
712
713void
714_eq_initialise (IV func, IV arg)
715 CODE:
716 eq_signal_func = (void (*)(void *, int))func;
717 eq_signal_arg = (void*)arg;
718
157BOOT: 719BOOT:
158{ 720{
159 HV *stash = gv_stashpv ("OpenCL", 1); 721 HV *stash = gv_stashpv ("OpenCL", 1);
160 static const ivstr *civ, const_iv[] = { 722 static const ivstr *civ, const_iv[] = {
161 { sizeof (cl_char ), "SIZEOF_CHAR" }, 723 { sizeof (cl_char ), "SIZEOF_CHAR" },
162 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 724 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
163 { sizeof (cl_short ), "SIZEOF_SHORT" }, 725 { sizeof (cl_short ), "SIZEOF_SHORT" },
164 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 726 { sizeof (cl_ushort), "SIZEOF_USHORT" },
165 { sizeof (cl_int ), "SIZEOF_INT" }, 727 { sizeof (cl_int ), "SIZEOF_INT" },
166 { sizeof (cl_uint ), "SIZEOF_UINT" }, 728 { sizeof (cl_uint ), "SIZEOF_UINT" },
167 { sizeof (cl_long ), "SIZEOF_LONG" }, 729 { sizeof (cl_long ), "SIZEOF_LONG" },
168 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 730 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
169 { sizeof (cl_half ), "SIZEOF_HALF" }, 731 { sizeof (cl_half ), "SIZEOF_HALF" },
170 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 732 { sizeof (cl_float ), "SIZEOF_FLOAT" },
171 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 733 { sizeof (cl_double), "SIZEOF_DOUBLE" },
172#include "constiv.h" 734#include "constiv.h"
173 }; 735 };
736
174 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 737 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
175 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 738 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
739
740 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
741 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
742 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
743 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
744 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
745 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
746 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
747 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
748 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
749 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
750 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
751 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
752 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
753 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
754 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
755 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
756 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
757 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
758 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
759 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
760 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
761 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
762 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
763
764 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
176} 765}
177 766
178cl_int 767cl_int
179errno () 768errno ()
180 CODE: 769 CODE:
181 errno = last_error; 770 RETVAL = res;
771 OUTPUT:
772 RETVAL
182 773
183const char * 774const char *
184err2str (cl_int err) 775err2str (cl_int err = res)
185 776
186const char * 777const char *
187enum2str (cl_uint value) 778enum2str (cl_uint value)
188 779
189void 780void
190platforms () 781platforms ()
191 PPCODE: 782 PPCODE:
192{
193 cl_platform_id *list; 783 cl_platform_id *list;
194 cl_uint count; 784 cl_uint count;
195 int i; 785 int i;
196 786
197 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count)); 787 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count));
198 list = tmpbuf (sizeof (*list) * count); 788 list = tmpbuf (sizeof (*list) * count);
199 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 789 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
200 790
201 EXTEND (SP, count); 791 EXTEND (SP, count);
202 for (i = 0; i < count; ++i) 792 for (i = 0; i < count; ++i)
203 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 793 PUSH_CLOBJ (stash_platform, list [i]);
204}
205 794
206void 795void
207context_from_type_simple (cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 796context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
208 PPCODE: 797 PPCODE:
209{ 798 CONTEXT_NOTIFY_CALLBACK;
210 cl_int res; 799 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
211 cl_context ctx = clCreateContextFromType (0, type, 0, 0, &res); 800 XPUSH_CLOBJ_CONTEXT;
212 801
213 if (res) 802void
214 FAIL (CreateContextFromType, res); 803context (FUTURE properties, FUTURE devices, FUTURE notify)
215 804 PPCODE:
216 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 805 /* der Gipfel der Kunst */
217}
218 806
219void 807void
220wait_for_events (...) 808wait_for_events (...)
221 CODE: 809 CODE:
222{
223 EVENT_LIST (0, items); 810 EVENT_LIST (0);
224 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 811 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
225}
226 812
227PROTOTYPES: DISABLE 813PROTOTYPES: DISABLE
228 814
229MODULE = OpenCL PACKAGE = OpenCL::Platform 815MODULE = OpenCL PACKAGE = OpenCL::Platform
230 816
231void 817void
232info (OpenCL::Platform this, cl_platform_info name) 818info (OpenCL::Platform self, cl_platform_info name)
233 PPCODE: 819 PPCODE:
234 INFO (Platform) 820 INFO (Platform)
235 821
236void 822void
823unload_compiler (OpenCL::Platform self)
824 CODE:
825#if CL_VERSION_1_2
826 clUnloadPlatformCompiler (self);
827#endif
828
829#BEGIN:platform
830
831void
832profile (OpenCL::Platform self)
833 ALIAS:
834 profile = CL_PLATFORM_PROFILE
835 version = CL_PLATFORM_VERSION
836 name = CL_PLATFORM_NAME
837 vendor = CL_PLATFORM_VENDOR
838 extensions = CL_PLATFORM_EXTENSIONS
839 PPCODE:
840 size_t size;
841 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
842 char *value = tmpbuf (size);
843 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
844 EXTEND (SP, 1);
845 const int i = 0;
846 PUSHs (sv_2mortal (newSVpv (value, 0)));
847
848#END:platform
849
850void
237devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 851devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
238 PPCODE: 852 PPCODE:
239{
240 cl_device_id *list; 853 cl_device_id *list;
241 cl_uint count; 854 cl_uint count;
242 int i; 855 int i;
243 856
244 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 857 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
245 list = tmpbuf (sizeof (*list) * count); 858 list = tmpbuf (sizeof (*list) * count);
246 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 859 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
247 860
248 EXTEND (SP, count); 861 EXTEND (SP, count);
249 for (i = 0; i < count; ++i) 862 for (i = 0; i < count; ++i)
250 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 863 PUSH_CLOBJ (stash_device, list [i]);
251}
252 864
253void 865void
254context_from_type_simple (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 866context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
255 PPCODE: 867 PPCODE:
256{ 868 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
257 cl_int res; 869 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
870
258 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 871 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
259 cl_context ctx = clCreateContextFromType (props, type, 0, 0, &res); 872 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
260 873
261 if (res) 874 AV *av = (AV *)SvRV (devices);
262 FAIL (CreateContextFromType, res); 875 cl_uint num_devices = av_len (av) + 1;
876 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
263 877
264 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 878 int i;
265} 879 for (i = num_devices; i--; )
880 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
881
882 CONTEXT_NOTIFY_CALLBACK;
883 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res));
884 XPUSH_CLOBJ_CONTEXT;
885
886void
887context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
888 PPCODE:
889 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
890 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
891
892 CONTEXT_NOTIFY_CALLBACK;
893 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
894 XPUSH_CLOBJ_CONTEXT;
266 895
267MODULE = OpenCL PACKAGE = OpenCL::Device 896MODULE = OpenCL PACKAGE = OpenCL::Device
268 897
269void 898void
270info (OpenCL::Device this, cl_device_info name) 899info (OpenCL::Device self, cl_device_info name)
271 PPCODE: 900 PPCODE:
272 INFO (Device) 901 INFO (Device)
273 902
903#if CL_VERSION_1_2
904
274void 905void
275context_simple (OpenCL::Device this) 906sub_devices (OpenCL::Device self, SV *properties)
907 PPCODE:
908 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
909 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
910
911 properties = SvRV (properties);
912
913 cl_uint count = av_len ((AV *)properties) + 1;
914 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
915
916 int i;
917 for (i = 0; i < count; ++i)
918 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
919
920 props [count] = 0;
921
922 cl_uint num_devices;
923 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
924 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
925 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
926
927 EXTEND (SP, num_devices);
928 for (i = 0; i < count; ++i)
929 PUSH_CLOBJ (stash_subdevice, list [i]);
930
931#endif
932
933#BEGIN:device
934
935void
936type (OpenCL::Device self)
276 PPCODE: 937 PPCODE:
277{ 938 cl_device_type value [1];
278 cl_int res; 939 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
279 cl_context ctx = clCreateContext (0, 1, &this, 0, 0, &res); 940 EXTEND (SP, 1);
941 const int i = 0;
942 PUSHs (sv_2mortal (newSVuv (value [i])));
280 943
281 if (res) 944void
282 FAIL (CreateContext, res); 945vendor_id (OpenCL::Device self)
946 ALIAS:
947 vendor_id = CL_DEVICE_VENDOR_ID
948 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
949 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
950 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
951 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
952 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
953 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
954 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
955 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
956 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
957 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
958 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
959 image_support = CL_DEVICE_IMAGE_SUPPORT
960 max_samplers = CL_DEVICE_MAX_SAMPLERS
961 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
962 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
963 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
964 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
965 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
966 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
967 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
968 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
969 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
970 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
971 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
972 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
973 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
974 PPCODE:
975 cl_uint value [1];
976 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
977 EXTEND (SP, 1);
978 const int i = 0;
979 PUSHs (sv_2mortal (newSVuv (value [i])));
283 980
284 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 981void
285} 982max_work_group_size (OpenCL::Device self)
983 ALIAS:
984 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
985 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
986 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
987 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
988 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
989 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
990 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
991 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
992 PPCODE:
993 size_t value [1];
994 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
995 EXTEND (SP, 1);
996 const int i = 0;
997 PUSHs (sv_2mortal (newSVuv (value [i])));
998
999void
1000max_work_item_sizes (OpenCL::Device self)
1001 PPCODE:
1002 size_t size;
1003 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
1004 size_t *value = tmpbuf (size);
1005 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
1006 int i, n = size / sizeof (*value);
1007 EXTEND (SP, n);
1008 for (i = 0; i < n; ++i)
1009 PUSHs (sv_2mortal (newSVuv (value [i])));
1010
1011void
1012address_bits (OpenCL::Device self)
1013 PPCODE:
1014 cl_bitfield value [1];
1015 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
1016 EXTEND (SP, 1);
1017 const int i = 0;
1018 PUSHs (sv_2mortal (newSVuv (value [i])));
1019
1020void
1021max_mem_alloc_size (OpenCL::Device self)
1022 ALIAS:
1023 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1024 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1025 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1026 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1027 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1028 PPCODE:
1029 cl_ulong value [1];
1030 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1031 EXTEND (SP, 1);
1032 const int i = 0;
1033 PUSHs (sv_2mortal (newSVuv (value [i])));
1034
1035void
1036single_fp_config (OpenCL::Device self)
1037 ALIAS:
1038 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1039 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1040 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
1041 PPCODE:
1042 cl_device_fp_config value [1];
1043 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1044 EXTEND (SP, 1);
1045 const int i = 0;
1046 PUSHs (sv_2mortal (newSVuv (value [i])));
1047
1048void
1049global_mem_cache_type (OpenCL::Device self)
1050 PPCODE:
1051 cl_device_mem_cache_type value [1];
1052 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
1053 EXTEND (SP, 1);
1054 const int i = 0;
1055 PUSHs (sv_2mortal (newSVuv (value [i])));
1056
1057void
1058local_mem_type (OpenCL::Device self)
1059 PPCODE:
1060 cl_device_local_mem_type value [1];
1061 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
1062 EXTEND (SP, 1);
1063 const int i = 0;
1064 PUSHs (sv_2mortal (newSVuv (value [i])));
1065
1066void
1067error_correction_support (OpenCL::Device self)
1068 ALIAS:
1069 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1070 endian_little = CL_DEVICE_ENDIAN_LITTLE
1071 available = CL_DEVICE_AVAILABLE
1072 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1073 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1074 PPCODE:
1075 cl_bool value [1];
1076 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1077 EXTEND (SP, 1);
1078 const int i = 0;
1079 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1080
1081void
1082execution_capabilities (OpenCL::Device self)
1083 PPCODE:
1084 cl_device_exec_capabilities value [1];
1085 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1086 EXTEND (SP, 1);
1087 const int i = 0;
1088 PUSHs (sv_2mortal (newSVuv (value [i])));
1089
1090void
1091properties (OpenCL::Device self)
1092 PPCODE:
1093 cl_command_queue_properties value [1];
1094 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1095 EXTEND (SP, 1);
1096 const int i = 0;
1097 PUSHs (sv_2mortal (newSVuv (value [i])));
1098
1099void
1100platform (OpenCL::Device self)
1101 PPCODE:
1102 cl_platform_id value [1];
1103 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1104 EXTEND (SP, 1);
1105 const int i = 0;
1106 {
1107 PUSH_CLOBJ (stash_platform, value [i]);
1108 }
1109
1110void
1111name (OpenCL::Device self)
1112 ALIAS:
1113 name = CL_DEVICE_NAME
1114 vendor = CL_DEVICE_VENDOR
1115 driver_version = CL_DRIVER_VERSION
1116 profile = CL_DEVICE_PROFILE
1117 version = CL_DEVICE_VERSION
1118 extensions = CL_DEVICE_EXTENSIONS
1119 PPCODE:
1120 size_t size;
1121 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1122 char *value = tmpbuf (size);
1123 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1124 EXTEND (SP, 1);
1125 const int i = 0;
1126 PUSHs (sv_2mortal (newSVpv (value, 0)));
1127
1128void
1129parent_device_ext (OpenCL::Device self)
1130 PPCODE:
1131 cl_device_id value [1];
1132 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1133 EXTEND (SP, 1);
1134 const int i = 0;
1135 {
1136 PUSH_CLOBJ (stash_device, value [i]);
1137 }
1138
1139void
1140partition_types_ext (OpenCL::Device self)
1141 ALIAS:
1142 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1143 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
1144 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
1145 PPCODE:
1146 size_t size;
1147 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1148 cl_device_partition_property_ext *value = tmpbuf (size);
1149 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1150 int i, n = size / sizeof (*value);
1151 EXTEND (SP, n);
1152 for (i = 0; i < n; ++i)
1153 PUSHs (sv_2mortal (newSVuv (value [i])));
1154
1155#END:device
1156
1157MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1158
1159#if CL_VERSION_1_2
1160
1161void
1162DESTROY (OpenCL::SubDevice self)
1163 CODE:
1164 clReleaseDevice (self);
1165
1166#endif
286 1167
287MODULE = OpenCL PACKAGE = OpenCL::Context 1168MODULE = OpenCL PACKAGE = OpenCL::Context
288 1169
289void 1170void
290DESTROY (OpenCL::Context context) 1171DESTROY (OpenCL::Context self)
291 CODE: 1172 CODE:
292 clReleaseContext (context); 1173 clReleaseContext (self);
293 1174
294void 1175void
295info (OpenCL::Context this, cl_context_info name) 1176info (OpenCL::Context self, cl_context_info name)
296 PPCODE: 1177 PPCODE:
297 INFO (Context) 1178 INFO (Context)
298 1179
299void 1180void
300command_queue_simple (OpenCL::Context this, OpenCL::Device device) 1181queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
301 PPCODE: 1182 PPCODE:
302{ 1183 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
303 cl_int res; 1184 XPUSH_CLOBJ (stash_queue, queue);
304 cl_command_queue queue = clCreateCommandQueue (this, device, 0, &res);
305 1185
306 if (res)
307 FAIL (CreateCommandQueue, res);
308
309 XPUSH_NEW_OBJ ("OpenCL::Queue", queue);
310}
311
312void 1186void
313user_event (OpenCL::Context this) 1187user_event (OpenCL::Context self)
314 PPCODE: 1188 PPCODE:
315{ 1189 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
316 cl_int res; 1190 XPUSH_CLOBJ (stash_userevent, ev);
317 cl_event ev = clCreateUserEvent (this, &res);
318 1191
319 if (res)
320 FAIL (CreateUserevent, res);
321
322 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev);
323}
324
325void 1192void
326buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1193buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
327 PPCODE: 1194 PPCODE:
328{
329 cl_int res;
330 cl_mem mem;
331
332 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1195 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
333 croak ("clCreateBuffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1196 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
334 1197
335 mem = clCreateBuffer (this, flags, len, 0, &res); 1198 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1199 XPUSH_CLOBJ (stash_bufferobj, mem);
336 1200
337 if (res)
338 FAIL (CreateBuffer, res);
339
340 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem);
341}
342
343void 1201void
344buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1202buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
345 PPCODE: 1203 PPCODE:
346{
347 STRLEN len; 1204 STRLEN len;
348 char *ptr = SvPVbyte (data, len); 1205 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
349 cl_int res;
350 cl_mem mem;
351
352 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1206 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
353 croak ("clCreateBuffer: have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1207 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
354 1208 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
355 mem = clCreateBuffer (this, flags, len, ptr, &res); 1209 XPUSH_CLOBJ (stash_bufferobj, mem);
356 1210
357 if (res) 1211#if CL_VERSION_1_2
358 FAIL (CreateBuffer, res);
359 1212
360 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem);
361}
362
363void 1213void
364image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, SV *data) 1214image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
365 PPCODE: 1215 PPCODE:
366{
367 STRLEN len; 1216 STRLEN len;
368 char *ptr = SvPVbyte (data, len); 1217 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
369 const cl_image_format format = { channel_order, channel_type }; 1218 const cl_image_format format = { channel_order, channel_type };
370 cl_int res; 1219 const cl_image_desc desc = {
371 cl_mem mem = clCreateImage2D (this, flags, &format, width, height, len / height, ptr, &res); 1220 type,
1221 width, height, depth,
1222 array_size, row_pitch, slice_pitch,
1223 num_mip_level, num_samples,
1224 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1225 };
1226 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1227 HV *stash = stash_image;
1228 switch (type)
1229 {
1230 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1231 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1232 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1233 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1234 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1235 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1236 }
1237 XPUSH_CLOBJ (stash, mem);
372 1238
373 if (res) 1239#endif
374 FAIL (CreateImage2D, res);
375 1240
376 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
377}
378
379void 1241void
380image3d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t slice_pitch, SV *data) 1242image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
381 PPCODE: 1243 PPCODE:
382{
383 STRLEN len; 1244 STRLEN len;
384 char *ptr = SvPVbyte (data, len); 1245 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
385 const cl_image_format format = { channel_order, channel_type }; 1246 const cl_image_format format = { channel_order, channel_type };
386 cl_int res; 1247#if PREFER_1_1
387 cl_mem mem = clCreateImage3D (this, flags, &format, width, height, 1248 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
388 depth, len / (height * slice_pitch), slice_pitch, ptr, &res); 1249#else
1250 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1251 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1252#endif
1253 XPUSH_CLOBJ (stash_image2d, mem);
389 1254
390 if (res)
391 FAIL (CreateImage3D, res);
392
393 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem);
394}
395
396void 1255void
1256image3d (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)
1257 PPCODE:
1258 STRLEN len;
1259 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1260 const cl_image_format format = { channel_order, channel_type };
1261#if PREFER_1_1
1262 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1263#else
1264 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1265 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1266#endif
1267 XPUSH_CLOBJ (stash_image3d, mem);
1268
1269#if cl_apple_gl_sharing || cl_khr_gl_sharing
1270
1271void
1272gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1273 PPCODE:
1274 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1275 XPUSH_CLOBJ (stash_bufferobj, mem);
1276
1277void
1278gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1279 PPCODE:
1280 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1281 XPUSH_CLOBJ (stash_image2d, mem);
1282
1283#if CL_VERSION_1_2
1284
1285void
1286gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1287 ALIAS:
1288 PPCODE:
1289 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1290 cl_gl_object_type type;
1291 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1292 HV *stash = stash_memory;
1293 switch (type)
1294 {
1295 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1296 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1297 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1298 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1299 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1300 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1301 }
1302 XPUSH_CLOBJ (stash, mem);
1303
1304#endif
1305
1306void
1307gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1308 PPCODE:
1309#if PREFER_1_1
1310 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1311#else
1312 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1313#endif
1314 XPUSH_CLOBJ (stash_image2d, mem);
1315
1316void
1317gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1318 PPCODE:
1319#if PREFER_1_1
1320 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1321#else
1322 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1323#endif
1324 XPUSH_CLOBJ (stash_image3d, mem);
1325
1326#endif
1327
1328void
397supported_image_formats (OpenCL::Context this, cl_mem_flags flags, cl_mem_object_type image_type) 1329supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
398 PPCODE: 1330 PPCODE:
399{ 1331{
400 cl_uint count; 1332 cl_uint count;
401 cl_image_format *list; 1333 cl_image_format *list;
402 int i; 1334 int i;
403 1335
404 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1336 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
405 Newx (list, count, cl_image_format); 1337 Newx (list, count, cl_image_format);
406 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1338 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
407 1339
408 EXTEND (SP, count); 1340 EXTEND (SP, count);
409 for (i = 0; i < count; ++i) 1341 for (i = 0; i < count; ++i)
410 { 1342 {
411 AV *av = newAV (); 1343 AV *av = newAV ();
414 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1346 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
415 } 1347 }
416} 1348}
417 1349
418void 1350void
419sampler (OpenCL::Context this, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1351sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
420 PPCODE: 1352 PPCODE:
421{
422 cl_int res;
423 cl_sampler sampler = clCreateSampler (this, normalized_coords, addressing_mode, filter_mode, &res); 1353 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
424
425 if (res)
426 FAIL (CreateSampler, res);
427
428 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1354 XPUSH_CLOBJ (stash_sampler, sampler);
429}
430 1355
431void 1356void
432program_with_source (OpenCL::Context this, SV *program) 1357program_with_source (OpenCL::Context self, SV *program)
433 PPCODE: 1358 PPCODE:
434{
435 STRLEN len; 1359 STRLEN len;
436 size_t len2; 1360 size_t len2;
437 const char *ptr = SvPVbyte (program, len); 1361 const char *ptr = SvPVbyte (program, len);
438 cl_int res;
439 cl_program prog;
440 1362
441 len2 = len; 1363 len2 = len;
442 prog = clCreateProgramWithSource (this, 1, &ptr, &len2, &res); 1364 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
443
444 if (res)
445 FAIL (CreateProgramWithSource, res);
446
447 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1365 XPUSH_CLOBJ (stash_program, prog);
448} 1366
1367void
1368program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1369 PPCODE:
1370 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1371 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects");
1372
1373 devices = SvRV (devices);
1374
1375 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1376 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1377
1378 binaries = SvRV (binaries);
1379
1380 if (av_len ((AV *)devices) != av_len ((AV *)binaries))
1381 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1382
1383 int count = av_len ((AV *)devices) + 1;
1384 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1385 size_t *length_list = tmpbuf (sizeof (*length_list) * count);
1386 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count);
1387 cl_int *status_list = tmpbuf (sizeof (*status_list) * count);
1388
1389 int i;
1390 for (i = 0; i < count; ++i)
1391 {
1392 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1393 STRLEN len;
1394 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1395 length_list [i] = len;
1396 }
1397
1398 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res));
1399
1400 EXTEND (SP, 2);
1401 PUSH_CLOBJ (stash_program, prog);
1402
1403 if (GIMME_V == G_ARRAY)
1404 {
1405 AV *av = newAV ();
1406 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1407
1408 for (i = count; i--; )
1409 av_store (av, i, newSViv (status_list [i]));
1410 }
1411
1412#if CL_VERSION_1_2
1413
1414void
1415program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1416 PPCODE:
1417 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1418 croak ("OpenCL::Context::program_with_built_in_kernels: devices must be specified as reference to an array of device objects");
1419
1420 devices = SvRV (devices);
1421
1422 int count = av_len ((AV *)devices) + 1;
1423 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1424
1425 int i;
1426 for (i = 0; i < count; ++i)
1427 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_built_in_kernels", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1428
1429 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, count, device_list, SvPVbyte_nolen (kernel_names), &res));
1430
1431 XPUSH_CLOBJ (stash_program, prog);
1432
1433#endif
1434
1435#BEGIN:context
1436
1437void
1438reference_count (OpenCL::Context self)
1439 ALIAS:
1440 reference_count = CL_CONTEXT_REFERENCE_COUNT
1441 num_devices = CL_CONTEXT_NUM_DEVICES
1442 PPCODE:
1443 cl_uint value [1];
1444 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1445 EXTEND (SP, 1);
1446 const int i = 0;
1447 PUSHs (sv_2mortal (newSVuv (value [i])));
1448
1449void
1450devices (OpenCL::Context self)
1451 PPCODE:
1452 size_t size;
1453 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1454 cl_device_id *value = tmpbuf (size);
1455 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1456 int i, n = size / sizeof (*value);
1457 EXTEND (SP, n);
1458 for (i = 0; i < n; ++i)
1459 {
1460 PUSH_CLOBJ (stash_device, value [i]);
1461 }
1462
1463void
1464properties (OpenCL::Context self)
1465 PPCODE:
1466 size_t size;
1467 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1468 cl_context_properties *value = tmpbuf (size);
1469 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1470 int i, n = size / sizeof (*value);
1471 EXTEND (SP, n);
1472 for (i = 0; i < n; ++i)
1473 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1474
1475#END:context
449 1476
450MODULE = OpenCL PACKAGE = OpenCL::Queue 1477MODULE = OpenCL PACKAGE = OpenCL::Queue
451 1478
452void 1479void
453DESTROY (OpenCL::Queue this) 1480DESTROY (OpenCL::Queue self)
454 CODE: 1481 CODE:
455 clReleaseCommandQueue (this); 1482 clReleaseCommandQueue (self);
456 1483
457void 1484void
458info (OpenCL::Queue this, cl_command_queue_info name)
459 PPCODE:
460 INFO (CommandQueue)
461
462void
463enqueue_read_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1485read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1486 ALIAS:
1487 enqueue_read_buffer = 0
464 PPCODE: 1488 PPCODE:
465{
466 cl_event ev = 0; 1489 cl_event ev = 0;
467 EVENT_LIST (6, items - 6); 1490 EVENT_LIST (6);
468 1491
469 SvUPGRADE (data, SVt_PV); 1492 SvUPGRADE (data, SVt_PV);
470 SvGROW (data, len); 1493 SvGROW (data, len);
471 SvPOK_only (data); 1494 SvPOK_only (data);
472 SvCUR_set (data, len); 1495 SvCUR_set (data, len);
473 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1496 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
474 1497
475 if (ev) 1498 if (ev)
476 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1499 XPUSH_CLOBJ (stash_event, ev);
477}
478 1500
479void 1501void
480enqueue_write_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1502write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1503 ALIAS:
1504 enqueue_write_buffer = 0
481 PPCODE: 1505 PPCODE:
482{
483 cl_event ev = 0; 1506 cl_event ev = 0;
484 STRLEN len; 1507 STRLEN len;
485 char *ptr = SvPVbyte (data, len); 1508 char *ptr = SvPVbyte (data, len);
486 EVENT_LIST (5, items - 5); 1509 EVENT_LIST (5);
487 1510
488 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1511 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
489 1512
490 if (ev) 1513 if (ev)
491 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1514 XPUSH_CLOBJ (stash_event, ev);
492}
493 1515
1516#if CL_VERSION_1_2
1517
494void 1518void
495enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1519fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1520 ALIAS:
1521 enqueue_fill_buffer = 0
496 PPCODE: 1522 PPCODE:
497{
498 cl_event ev = 0; 1523 cl_event ev = 0;
1524 STRLEN len;
1525 char *ptr = SvPVbyte (data, len);
499 EVENT_LIST (6, items - 6); 1526 EVENT_LIST (5);
500 1527
501 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1528 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
502 1529
503 if (ev) 1530 if (ev)
504 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1531 XPUSH_CLOBJ (stash_event, ev);
505}
506 1532
507 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html */
508 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html */
509
510void 1533void
511enqueue_read_image (OpenCL::Queue this, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1534fill_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, ...)
1535 ALIAS:
1536 enqueue_fill_image = 0
512 PPCODE: 1537 PPCODE:
513{
514 cl_event ev = 0; 1538 cl_event ev = 0;
1539 const size_t origin [3] = { x, y, z };
1540 const size_t region [3] = { width, height, depth };
1541 EVENT_LIST (12);
1542
1543 const cl_float c_f [4] = { r, g, b, a };
1544 const cl_uint c_u [4] = { r, g, b, a };
1545 const cl_int c_s [4] = { r, g, b, a };
1546 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1547 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1548 cl_image_format format;
1549 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1550 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1551 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1552 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1553
1554 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1555 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1556
1557 if (ev)
1558 XPUSH_CLOBJ (stash_event, ev);
1559
1560#endif
1561
1562void
1563copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1564 ALIAS:
1565 enqueue_copy_buffer = 0
1566 PPCODE:
1567 cl_event ev = 0;
1568 EVENT_LIST (6);
1569
1570 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1571
1572 if (ev)
1573 XPUSH_CLOBJ (stash_event, ev);
1574
1575void
1576read_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, ...)
1577 ALIAS:
1578 enqueue_read_buffer_rect = 0
1579 PPCODE:
1580 cl_event ev = 0;
1581 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
515 const size_t src_origin[3] = { src_x, src_y, src_z }; 1582 const size_t host_origin[3] = { host_x, host_y, host_z };
516 const size_t region[3] = { width, height, depth }; 1583 const size_t region[3] = { width, height, depth };
1584 EVENT_LIST (17);
1585
1586 if (!buf_row_pitch)
1587 buf_row_pitch = region [0];
1588
1589 if (!buf_slice_pitch)
1590 buf_slice_pitch = region [1] * buf_row_pitch;
1591
1592 if (!host_row_pitch)
1593 host_row_pitch = region [0];
1594
1595 if (!host_slice_pitch)
1596 host_slice_pitch = region [1] * host_row_pitch;
1597
517 size_t len = row_pitch * slice_pitch * depth; 1598 size_t len = host_row_pitch * host_slice_pitch * region [2];
518 EVENT_LIST (11, items - 11);
519 1599
520 SvUPGRADE (data, SVt_PV); 1600 SvUPGRADE (data, SVt_PV);
521 SvGROW (data, len); 1601 SvGROW (data, len);
522 SvPOK_only (data); 1602 SvPOK_only (data);
523 SvCUR_set (data, len); 1603 SvCUR_set (data, len);
524 NEED_SUCCESS (EnqueueReadImage, (this, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1604 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
525 1605
526 if (ev) 1606 if (ev)
527 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1607 XPUSH_CLOBJ (stash_event, ev);
528}
529 1608
530void 1609void
1610write_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, ...)
1611 ALIAS:
1612 enqueue_write_buffer_rect = 0
1613 PPCODE:
1614 cl_event ev = 0;
1615 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1616 const size_t host_origin[3] = { host_x, host_y, host_z };
1617 const size_t region[3] = { width, height, depth };
1618 STRLEN len;
1619 char *ptr = SvPVbyte (data, len);
1620 EVENT_LIST (17);
1621
1622 if (!buf_row_pitch)
1623 buf_row_pitch = region [0];
1624
1625 if (!buf_slice_pitch)
1626 buf_slice_pitch = region [1] * buf_row_pitch;
1627
1628 if (!host_row_pitch)
1629 host_row_pitch = region [0];
1630
1631 if (!host_slice_pitch)
1632 host_slice_pitch = region [1] * host_row_pitch;
1633
1634 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1635
1636 if (len < min_len)
1637 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1638
1639 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));
1640
1641 if (ev)
1642 XPUSH_CLOBJ (stash_event, ev);
1643
1644void
1645copy_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, ...)
1646 ALIAS:
1647 enqueue_copy_buffer_rect = 0
1648 PPCODE:
1649 cl_event ev = 0;
1650 const size_t src_origin[3] = { src_x, src_y, src_z };
1651 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1652 const size_t region[3] = { width, height, depth };
1653 EVENT_LIST (16);
1654
1655 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));
1656
1657 if (ev)
1658 XPUSH_CLOBJ (stash_event, ev);
1659
1660void
1661read_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, ...)
1662 ALIAS:
1663 enqueue_read_image = 0
1664 PPCODE:
1665 cl_event ev = 0;
1666 const size_t src_origin[3] = { src_x, src_y, src_z };
1667 const size_t region[3] = { width, height, depth };
1668 EVENT_LIST (12);
1669
1670 if (!row_pitch)
1671 row_pitch = img_row_pitch (src);
1672
1673 if (depth > 1 && !slice_pitch)
1674 slice_pitch = row_pitch * height;
1675
1676 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1677
1678 SvUPGRADE (data, SVt_PV);
1679 SvGROW (data, len);
1680 SvPOK_only (data);
1681 SvCUR_set (data, len);
1682 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));
1683
1684 if (ev)
1685 XPUSH_CLOBJ (stash_event, ev);
1686
1687void
531enqueue_write_image (OpenCL::Queue this, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, SV *data, ...) 1688write_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, ...)
1689 ALIAS:
1690 enqueue_write_image = 0
532 PPCODE: 1691 PPCODE:
533{
534 cl_event ev = 0; 1692 cl_event ev = 0;
535 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1693 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
536 const size_t region[3] = { width, height, depth }; 1694 const size_t region[3] = { width, height, depth };
537 STRLEN len; 1695 STRLEN len;
538 char *ptr = SvPVbyte (data, len); 1696 char *ptr = SvPVbyte (data, len);
1697 EVENT_LIST (12);
1698
1699 if (!row_pitch)
1700 row_pitch = img_row_pitch (dst);
1701
1702 if (depth > 1 && !slice_pitch)
539 size_t slice_pitch = len / (row_pitch * height); 1703 slice_pitch = row_pitch * height;
540 EVENT_LIST (11, items - 11);
541 1704
1705 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1706
1707 if (len < min_len)
1708 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1709
542 NEED_SUCCESS (EnqueueWriteImage, (this, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1710 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
543 1711
544 if (ev) 1712 if (ev)
545 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1713 XPUSH_CLOBJ (stash_event, ev);
546}
547 1714
548void 1715void
549enqueue_copy_buffer_rect (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1716copy_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, ...)
1717 ALIAS:
1718 enqueue_copy_image = 0
550 PPCODE: 1719 PPCODE:
551{
552 cl_event ev = 0; 1720 cl_event ev = 0;
553 const size_t src_origin[3] = { src_x, src_y, src_z }; 1721 const size_t src_origin[3] = { src_x, src_y, src_z };
554 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1722 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
555 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
556 EVENT_LIST (16, items - 16); 1724 EVENT_LIST (12);
557 1725
558 NEED_SUCCESS (EnqueueCopyBufferRect, (this, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1726 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
559 1727
560 if (ev) 1728 if (ev)
561 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
562}
563 1730
564void 1731void
565enqueue_copy_buffer_to_image (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1732copy_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, ...)
1733 ALIAS:
1734 enqueue_copy_image_to_buffer = 0
566 PPCODE: 1735 PPCODE:
567{
568 cl_event ev = 0; 1736 cl_event ev = 0;
1737 const size_t src_origin[3] = { src_x, src_y, src_z };
1738 const size_t region [3] = { width, height, depth };
1739 EVENT_LIST (10);
1740
1741 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1742
1743 if (ev)
1744 XPUSH_CLOBJ (stash_event, ev);
1745
1746void
1747copy_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, ...)
1748 ALIAS:
1749 enqueue_copy_buffer_to_image = 0
1750 PPCODE:
1751 cl_event ev = 0;
569 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1752 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1753 const size_t region [3] = { width, height, depth };
1754 EVENT_LIST (10);
1755
1756 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1757
1758 if (ev)
1759 XPUSH_CLOBJ (stash_event, ev);
1760
1761void
1762map_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, ...)
1763 ALIAS:
1764 enqueue_map_buffer = 0
1765 PPCODE:
1766 cl_event ev;
1767 EVENT_LIST (6);
1768 size_t cb = SvIV (cb_);
1769
1770 if (!SvOK (cb_))
1771 {
1772 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1773 cb -= offset;
1774 }
1775
1776 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1777 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1778
1779void
1780map_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, ...)
1781 ALIAS:
1782 enqueue_map_image = 0
1783 PPCODE:
1784 size_t width = SvIV (width_);
1785 if (!SvOK (width_))
1786 {
1787 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1788 width -= x;
1789 }
1790
1791 size_t height = SvIV (width_);
1792 if (!SvOK (height_))
1793 {
1794 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1795 height -= y;
1796 }
1797
1798 size_t depth = SvIV (width_);
1799 if (!SvOK (depth_))
1800 {
1801 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1802 depth -= z;
1803
1804 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1805 if (!depth)
1806 depth = 1;
1807 }
1808
1809 const size_t origin[3] = { x, y, z };
570 const size_t region[3] = { width, height, depth }; 1810 const size_t region[3] = { width, height, depth };
1811 size_t row_pitch, slice_pitch;
571 EVENT_LIST (10, items - 10); 1812 EVENT_LIST (10);
572 1813
573 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1814 cl_event ev;
1815 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));
1816
1817 size_t cb = slice_pitch ? slice_pitch * region [2]
1818 : row_pitch ? row_pitch * region [1]
1819 : region [0];
1820
1821 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1822
1823void
1824unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1825 PPCODE:
1826 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1827 if (GIMME_V != G_VOID)
1828 {
1829 clRetainEvent (mapped->event);
1830 XPUSH_CLOBJ (stash_event, mapped->event);
1831 }
1832
1833void
1834task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1835 ALIAS:
1836 enqueue_task = 0
1837 PPCODE:
1838 cl_event ev = 0;
1839 EVENT_LIST (2);
1840
1841 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
574 1842
575 if (ev) 1843 if (ev)
576 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1844 XPUSH_CLOBJ (stash_event, ev);
577}
578 1845
579void 1846void
580enqueue_copy_image (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
581 PPCODE:
582{
583 cl_event ev = 0;
584 const size_t src_origin[3] = { src_x, src_y, src_z };
585 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
586 const size_t region[3] = { width, height, depth };
587 EVENT_LIST (12, items - 12);
588
589 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
590
591 if (ev)
592 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
593}
594
595void
596enqueue_copy_image_to_buffer (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
597 PPCODE:
598{
599 cl_event ev = 0;
600 const size_t src_origin[3] = { src_x, src_y, src_z };
601 const size_t region[3] = { width, height, depth };
602 EVENT_LIST (10, items - 10);
603
604 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
605
606 if (ev)
607 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
608}
609
610void
611enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
612 PPCODE:
613{
614 cl_event ev = 0;
615 EVENT_LIST (2, items - 2);
616
617 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
618
619 if (ev)
620 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
621}
622
623void
624enqueue_nd_range_kernel (OpenCL::Queue this, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1847nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1848 ALIAS:
1849 enqueue_nd_range_kernel = 0
625 PPCODE: 1850 PPCODE:
626{
627 cl_event ev = 0; 1851 cl_event ev = 0;
628 size_t *gwo = 0, *gws, *lws = 0; 1852 size_t *gwo = 0, *gws, *lws = 0;
629 int gws_len; 1853 int gws_len;
630 size_t *lists; 1854 size_t *lists;
631 int i; 1855 int i;
632 EVENT_LIST (5, items - 5); 1856 EVENT_LIST (5);
633 1857
634 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1858 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
635 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1859 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
636 1860
637 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1861 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
638 1862
639 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1863 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
640 1864
641 gws = lists + gws_len * 0; 1865 gws = lists + gws_len * 0;
642 for (i = 0; i < gws_len; ++i) 1866 for (i = 0; i < gws_len; ++i)
1867 {
643 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1868 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1869 // at least nvidia crashes for 0-sized work group sizes, work around
1870 if (!gws [i])
1871 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1872 }
644 1873
645 if (SvOK (global_work_offset)) 1874 if (SvOK (global_work_offset))
646 { 1875 {
647 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1876 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
648 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1877 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
655 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1884 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
656 } 1885 }
657 1886
658 if (SvOK (local_work_size)) 1887 if (SvOK (local_work_size))
659 { 1888 {
660 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1889 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
661 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1890 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
662 1891
663 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1892 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
664 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1893 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
665 1894
666 lws = lists + gws_len * 2; 1895 lws = lists + gws_len * 2;
667 for (i = 0; i < gws_len; ++i) 1896 for (i = 0; i < gws_len; ++i)
1897 {
668 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1898 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1899 // at least nvidia crashes for 0-sized work group sizes, work around
1900 if (!lws [i])
1901 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1902 }
669 } 1903 }
670 1904
671 NEED_SUCCESS (EnqueueNDRangeKernel, (this, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1905 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
672 1906
673 if (ev) 1907 if (ev)
674 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1908 XPUSH_CLOBJ (stash_event, ev);
675}
676 1909
1910#if CL_VERSION_1_2
1911
677void 1912void
678enqueue_marker (OpenCL::Queue this) 1913migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1914 ALIAS:
1915 enqueue_migrate_mem_objects = 0
679 PPCODE: 1916 PPCODE:
680{ 1917 EVENT_LIST (3);
1918
1919 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1920 croak ("OpenCL::Queue::migrate_mem_objects: objects must be an array reference with OpenCL::Memory objects");
1921
1922 objects = SvRV (objects);
1923
1924 cl_uint object_count = av_len ((AV *)objects) + 1;
1925 cl_mem *object_list = tmpbuf (sizeof (*object_list) * object_count);
1926
1927 int i;
1928 for (i = object_count; i--; )
1929 object_list [i] = SvCLOBJ ("OpenCL::Queue::migrate_mem_objects", "objects", *av_fetch ((AV *)objects, i, 0), "OpenCL::Memory");
1930
1931 cl_event ev = 0;
1932 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, object_count, object_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1933
1934 if (ev)
1935 XPUSH_CLOBJ (stash_event, ev);
1936
1937#endif
1938
1939#if cl_apple_gl_sharing || cl_khr_gl_sharing
1940
1941void
1942acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1943 ALIAS:
1944 release_gl_objects = 1
1945 enqueue_acquire_gl_objects = 0
1946 enqueue_release_gl_objects = 1
1947 PPCODE:
1948 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1949 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1950
681 cl_event ev; 1951 cl_event ev = 0;
682 NEED_SUCCESS (EnqueueMarker, (this, &ev)); 1952 EVENT_LIST (2);
683 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1953 AV *av = (AV *)SvRV (objects);
684} 1954 cl_uint num_objects = av_len (av) + 1;
1955 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
685 1956
1957 int i;
1958 for (i = num_objects; i--; )
1959 object_list [i] = SvCLOBJ ("OpenCL::Queue::acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1960
1961 if (ix)
1962 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1963 else
1964 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1965
1966 if (ev)
1967 XPUSH_CLOBJ (stash_event, ev);
1968
1969#endif
1970
686void 1971void
687enqueue_wait_for_events (OpenCL::Queue this, ...) 1972wait_for_events (OpenCL::Queue self, ...)
1973 ALIAS:
1974 enqueue_wait_for_events = 0
688 CODE: 1975 CODE:
689{
690 EVENT_LIST (1, items - 1); 1976 EVENT_LIST (1);
1977#if PREFER_1_1
691 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 1978 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
692} 1979#else
1980 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1981#endif
693 1982
694void 1983void
695enqueue_barrier (OpenCL::Queue this) 1984marker (OpenCL::Queue self, ...)
1985 ALIAS:
1986 enqueue_marker = 0
696 CODE: 1987 PPCODE:
1988 cl_event ev = 0;
1989 EVENT_LIST (1);
1990#if PREFER_1_1
1991 if (!event_list_count)
1992 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1993 else
1994#if CL_VERSION_1_2
1995 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1996#else
1997 {
1998 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1999 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2000 }
2001#endif
2002#else
2003 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2004#endif
2005 if (ev)
2006 XPUSH_CLOBJ (stash_event, ev);
2007
2008void
2009barrier (OpenCL::Queue self, ...)
2010 ALIAS:
2011 enqueue_barrier = 0
2012 PPCODE:
2013 cl_event ev = 0;
2014 EVENT_LIST (1);
2015#if PREFER_1_1
2016 if (!event_list_count && GIMME_V == G_VOID)
697 NEED_SUCCESS (EnqueueBarrier, (this)); 2017 NEED_SUCCESS (EnqueueBarrier, (self));
2018 else
2019#if CL_VERSION_1_2
2020 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2021#else
2022 {
2023 if (event_list_count)
2024 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
698 2025
2026 if (GIMME_V != G_VOID)
2027 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2028 }
2029#endif
2030#else
2031 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2032#endif
2033 if (ev)
2034 XPUSH_CLOBJ (stash_event, ev);
2035
699void 2036void
700flush (OpenCL::Queue this) 2037flush (OpenCL::Queue self)
701 CODE: 2038 CODE:
702 NEED_SUCCESS (Flush, (this)); 2039 NEED_SUCCESS (Flush, (self));
703 2040
704void 2041void
705finish (OpenCL::Queue this) 2042finish (OpenCL::Queue self)
706 CODE: 2043 CODE:
707 NEED_SUCCESS (Finish, (this)); 2044 NEED_SUCCESS (Finish, (self));
2045
2046void
2047info (OpenCL::Queue self, cl_command_queue_info name)
2048 PPCODE:
2049 INFO (CommandQueue)
2050
2051#BEGIN:command_queue
2052
2053void
2054context (OpenCL::Queue self)
2055 PPCODE:
2056 cl_context value [1];
2057 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
2058 EXTEND (SP, 1);
2059 const int i = 0;
2060 {
2061 NEED_SUCCESS (RetainContext, (value [i]));
2062 PUSH_CLOBJ (stash_context, value [i]);
2063 }
2064
2065void
2066device (OpenCL::Queue self)
2067 PPCODE:
2068 cl_device_id value [1];
2069 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
2070 EXTEND (SP, 1);
2071 const int i = 0;
2072 {
2073 PUSH_CLOBJ (stash_device, value [i]);
2074 }
2075
2076void
2077reference_count (OpenCL::Queue self)
2078 PPCODE:
2079 cl_uint value [1];
2080 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
2081 EXTEND (SP, 1);
2082 const int i = 0;
2083 PUSHs (sv_2mortal (newSVuv (value [i])));
2084
2085void
2086properties (OpenCL::Queue self)
2087 PPCODE:
2088 cl_command_queue_properties value [1];
2089 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
2090 EXTEND (SP, 1);
2091 const int i = 0;
2092 PUSHs (sv_2mortal (newSVuv (value [i])));
2093
2094#END:command_queue
708 2095
709MODULE = OpenCL PACKAGE = OpenCL::Memory 2096MODULE = OpenCL PACKAGE = OpenCL::Memory
710 2097
711void 2098void
712DESTROY (OpenCL::Memory this) 2099DESTROY (OpenCL::Memory self)
713 CODE: 2100 CODE:
714 clReleaseMemObject (this); 2101 clReleaseMemObject (self);
715 2102
716void 2103void
717info (OpenCL::Memory this, cl_mem_info name) 2104info (OpenCL::Memory self, cl_mem_info name)
718 PPCODE: 2105 PPCODE:
719 INFO (MemObject) 2106 INFO (MemObject)
720 2107
2108#BEGIN:mem
2109
2110void
2111type (OpenCL::Memory self)
2112 PPCODE:
2113 cl_mem_object_type value [1];
2114 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
2115 EXTEND (SP, 1);
2116 const int i = 0;
2117 PUSHs (sv_2mortal (newSVuv (value [i])));
2118
2119void
2120flags (OpenCL::Memory self)
2121 PPCODE:
2122 cl_mem_flags value [1];
2123 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
2124 EXTEND (SP, 1);
2125 const int i = 0;
2126 PUSHs (sv_2mortal (newSVuv (value [i])));
2127
2128void
2129size (OpenCL::Memory self)
2130 ALIAS:
2131 size = CL_MEM_SIZE
2132 offset = CL_MEM_OFFSET
2133 PPCODE:
2134 size_t value [1];
2135 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2136 EXTEND (SP, 1);
2137 const int i = 0;
2138 PUSHs (sv_2mortal (newSVuv (value [i])));
2139
2140void
2141host_ptr (OpenCL::Memory self)
2142 PPCODE:
2143 void * value [1];
2144 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
2145 EXTEND (SP, 1);
2146 const int i = 0;
2147 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
2148
2149void
2150map_count (OpenCL::Memory self)
2151 ALIAS:
2152 map_count = CL_MEM_MAP_COUNT
2153 reference_count = CL_MEM_REFERENCE_COUNT
2154 PPCODE:
2155 cl_uint value [1];
2156 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2157 EXTEND (SP, 1);
2158 const int i = 0;
2159 PUSHs (sv_2mortal (newSVuv (value [i])));
2160
2161void
2162context (OpenCL::Memory self)
2163 PPCODE:
2164 cl_context value [1];
2165 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
2166 EXTEND (SP, 1);
2167 const int i = 0;
2168 {
2169 NEED_SUCCESS (RetainContext, (value [i]));
2170 PUSH_CLOBJ (stash_context, value [i]);
2171 }
2172
2173void
2174associated_memobject (OpenCL::Memory self)
2175 PPCODE:
2176 cl_mem value [1];
2177 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
2178 EXTEND (SP, 1);
2179 const int i = 0;
2180 {
2181 NEED_SUCCESS (RetainMemObject, (value [i]));
2182 PUSH_CLOBJ (stash_memory, value [i]);
2183 }
2184
2185#END:mem
2186
2187#if cl_apple_gl_sharing || cl_khr_gl_sharing
2188
2189void
2190gl_object_info (OpenCL::Memory self)
2191 PPCODE:
2192 cl_gl_object_type type;
2193 cl_GLuint name;
2194 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2195 EXTEND (SP, 2);
2196 PUSHs (sv_2mortal (newSVuv (type)));
2197 PUSHs (sv_2mortal (newSVuv (name)));
2198
2199#endif
2200
2201MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2202
2203void
2204sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2205 PPCODE:
2206 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2207 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2208
2209 cl_buffer_region crdata = { origin, size };
2210
2211 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2212 XPUSH_CLOBJ (stash_buffer, mem);
2213
2214MODULE = OpenCL PACKAGE = OpenCL::Image
2215
2216void
2217image_info (OpenCL::Image self, cl_image_info name)
2218 PPCODE:
2219 INFO (Image)
2220
2221void
2222format (OpenCL::Image self)
2223 PPCODE:
2224 cl_image_format format;
2225 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2226 EXTEND (SP, 2);
2227 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2228 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2229
2230#BEGIN:image
2231
2232void
2233element_size (OpenCL::Image self)
2234 ALIAS:
2235 element_size = CL_IMAGE_ELEMENT_SIZE
2236 row_pitch = CL_IMAGE_ROW_PITCH
2237 slice_pitch = CL_IMAGE_SLICE_PITCH
2238 width = CL_IMAGE_WIDTH
2239 height = CL_IMAGE_HEIGHT
2240 depth = CL_IMAGE_DEPTH
2241 PPCODE:
2242 size_t value [1];
2243 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
2244 EXTEND (SP, 1);
2245 const int i = 0;
2246 PUSHs (sv_2mortal (newSVuv (value [i])));
2247
2248#END:image
2249
2250#if cl_apple_gl_sharing || cl_khr_gl_sharing
2251
2252#BEGIN:gl_texture
2253
2254void
2255target (OpenCL::Image self)
2256 PPCODE:
2257 cl_GLenum value [1];
2258 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2259 EXTEND (SP, 1);
2260 const int i = 0;
2261 PUSHs (sv_2mortal (newSVuv (value [i])));
2262
2263void
2264gl_mipmap_level (OpenCL::Image self)
2265 PPCODE:
2266 cl_GLint value [1];
2267 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2268 EXTEND (SP, 1);
2269 const int i = 0;
2270 PUSHs (sv_2mortal (newSViv (value [i])));
2271
2272#END:gl_texture
2273
2274#endif
2275
721MODULE = OpenCL PACKAGE = OpenCL::Sampler 2276MODULE = OpenCL PACKAGE = OpenCL::Sampler
722 2277
723void 2278void
724DESTROY (OpenCL::Sampler this) 2279DESTROY (OpenCL::Sampler self)
725 CODE: 2280 CODE:
726 clReleaseSampler (this); 2281 clReleaseSampler (self);
727 2282
728void 2283void
729info (OpenCL::Sampler this, cl_sampler_info name) 2284info (OpenCL::Sampler self, cl_sampler_info name)
730 PPCODE: 2285 PPCODE:
731 INFO (Sampler) 2286 INFO (Sampler)
732 2287
2288#BEGIN:sampler
2289
2290void
2291reference_count (OpenCL::Sampler self)
2292 PPCODE:
2293 cl_uint value [1];
2294 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0));
2295 EXTEND (SP, 1);
2296 const int i = 0;
2297 PUSHs (sv_2mortal (newSVuv (value [i])));
2298
2299void
2300context (OpenCL::Sampler self)
2301 PPCODE:
2302 cl_context value [1];
2303 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2304 EXTEND (SP, 1);
2305 const int i = 0;
2306 {
2307 NEED_SUCCESS (RetainContext, (value [i]));
2308 PUSH_CLOBJ (stash_context, value [i]);
2309 }
2310
2311void
2312normalized_coords (OpenCL::Sampler self)
2313 PPCODE:
2314 cl_addressing_mode value [1];
2315 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2316 EXTEND (SP, 1);
2317 const int i = 0;
2318 PUSHs (sv_2mortal (newSVuv (value [i])));
2319
2320void
2321addressing_mode (OpenCL::Sampler self)
2322 PPCODE:
2323 cl_filter_mode value [1];
2324 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2325 EXTEND (SP, 1);
2326 const int i = 0;
2327 PUSHs (sv_2mortal (newSVuv (value [i])));
2328
2329void
2330filter_mode (OpenCL::Sampler self)
2331 PPCODE:
2332 cl_bool value [1];
2333 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2334 EXTEND (SP, 1);
2335 const int i = 0;
2336 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2337
2338#END:sampler
2339
733MODULE = OpenCL PACKAGE = OpenCL::Program 2340MODULE = OpenCL PACKAGE = OpenCL::Program
734 2341
735void 2342void
736DESTROY (OpenCL::Program this) 2343DESTROY (OpenCL::Program self)
737 CODE: 2344 CODE:
738 clReleaseProgram (this); 2345 clReleaseProgram (self);
739 2346
740void 2347void
741info (OpenCL::Program this, cl_program_info name) 2348build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2349 ALIAS:
2350 build_async = 1
742 PPCODE: 2351 CODE:
743 INFO (Program) 2352 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2353 void *user_data = 0;
2354 cl_uint num_devices = 0;
2355 cl_device_id *device_list = 0;
744 2356
745void 2357 if (SvOK (devices))
746build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2358 {
747 CODE: 2359 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2360 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2361
2362 AV *av = (AV *)SvRV (devices);
2363 num_devices = av_len (av) + 1;
2364
2365 if (num_devices)
2366 {
2367 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2368 int count;
2369 for (count = 0; count < num_devices; ++count)
2370 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2371 }
2372 }
2373
2374 if (SvOK (notify))
2375 {
2376 NEED_SUCCESS (RetainProgram, (self));
2377 pfn_notify = eq_program_notify;
2378 user_data = SvREFCNT_inc (s_get_cv (notify));
2379 }
2380
2381 if (ix)
2382 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2383 else
748 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2384 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
749 2385
750void 2386void
751build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2387build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
752 PPCODE: 2388 PPCODE:
753{
754 size_t size; 2389 size_t size;
755 SV *sv;
756
757 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2390 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
758 sv = sv_2mortal (newSV (size)); 2391 SV *sv = sv_2mortal (newSV (size));
759 SvUPGRADE (sv, SVt_PV); 2392 SvUPGRADE (sv, SVt_PV);
760 SvPOK_only (sv); 2393 SvPOK_only (sv);
761 SvCUR_set (sv, size); 2394 SvCUR_set (sv, size);
762 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2395 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
763 XPUSHs (sv); 2396 XPUSHs (sv);
764} 2397
2398#BEGIN:program_build
2399
2400void
2401build_status (OpenCL::Program self, OpenCL::Device device)
2402 PPCODE:
2403 cl_build_status value [1];
2404 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2405 EXTEND (SP, 1);
2406 const int i = 0;
2407 PUSHs (sv_2mortal (newSViv (value [i])));
2408
2409void
2410build_options (OpenCL::Program self, OpenCL::Device device)
2411 ALIAS:
2412 build_options = CL_PROGRAM_BUILD_OPTIONS
2413 build_log = CL_PROGRAM_BUILD_LOG
2414 PPCODE:
2415 size_t size;
2416 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2417 char *value = tmpbuf (size);
2418 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2419 EXTEND (SP, 1);
2420 const int i = 0;
2421 PUSHs (sv_2mortal (newSVpv (value, 0)));
2422
2423#END:program_build
765 2424
766void 2425void
767kernel (OpenCL::Program program, SV *function) 2426kernel (OpenCL::Program program, SV *function)
768 PPCODE: 2427 PPCODE:
769{
770 cl_int res;
771 cl_kernel kernel = clCreateKernel (program, SvPVbyte_nolen (function), &res); 2428 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2429 XPUSH_CLOBJ (stash_kernel, kernel);
772 2430
773 if (res) 2431void
774 FAIL (CreateKernel, res); 2432kernels_in_program (OpenCL::Program program)
2433 PPCODE:
2434 cl_uint num_kernels;
2435 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2436 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2437 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
775 2438
776 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2439 int i;
777} 2440 EXTEND (SP, num_kernels);
2441 for (i = 0; i < num_kernels; ++i)
2442 PUSH_CLOBJ (stash_kernel, kernels [i]);
2443
2444void
2445info (OpenCL::Program self, cl_program_info name)
2446 PPCODE:
2447 INFO (Program)
2448
2449void
2450binaries (OpenCL::Program self)
2451 PPCODE:
2452 cl_uint n, i;
2453 size_t size;
2454
2455 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2456 if (!n) XSRETURN_EMPTY;
2457
2458 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2459 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2460 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2461 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2462
2463 EXTEND (SP, n);
2464 for (i = 0; i < n; ++i)
2465 {
2466 SV *sv = sv_2mortal (newSV (sizes [i]));
2467 SvUPGRADE (sv, SVt_PV);
2468 SvPOK_only (sv);
2469 SvCUR_set (sv, sizes [i]);
2470 ptrs [i] = (void *)SvPVX (sv);
2471 PUSHs (sv);
2472 }
2473
2474 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2475 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2476
2477#BEGIN:program
2478
2479void
2480reference_count (OpenCL::Program self)
2481 ALIAS:
2482 reference_count = CL_PROGRAM_REFERENCE_COUNT
2483 num_devices = CL_PROGRAM_NUM_DEVICES
2484 PPCODE:
2485 cl_uint value [1];
2486 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2487 EXTEND (SP, 1);
2488 const int i = 0;
2489 PUSHs (sv_2mortal (newSVuv (value [i])));
2490
2491void
2492context (OpenCL::Program self)
2493 PPCODE:
2494 cl_context value [1];
2495 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2496 EXTEND (SP, 1);
2497 const int i = 0;
2498 {
2499 NEED_SUCCESS (RetainContext, (value [i]));
2500 PUSH_CLOBJ (stash_context, value [i]);
2501 }
2502
2503void
2504devices (OpenCL::Program self)
2505 PPCODE:
2506 size_t size;
2507 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2508 cl_device_id *value = tmpbuf (size);
2509 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2510 int i, n = size / sizeof (*value);
2511 EXTEND (SP, n);
2512 for (i = 0; i < n; ++i)
2513 {
2514 PUSH_CLOBJ (stash_device, value [i]);
2515 }
2516
2517void
2518source (OpenCL::Program self)
2519 PPCODE:
2520 size_t size;
2521 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2522 char *value = tmpbuf (size);
2523 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2524 EXTEND (SP, 1);
2525 const int i = 0;
2526 PUSHs (sv_2mortal (newSVpv (value, 0)));
2527
2528void
2529binary_sizes (OpenCL::Program self)
2530 PPCODE:
2531 size_t size;
2532 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2533 size_t *value = tmpbuf (size);
2534 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2535 int i, n = size / sizeof (*value);
2536 EXTEND (SP, n);
2537 for (i = 0; i < n; ++i)
2538 PUSHs (sv_2mortal (newSVuv (value [i])));
2539
2540#END:program
778 2541
779MODULE = OpenCL PACKAGE = OpenCL::Kernel 2542MODULE = OpenCL PACKAGE = OpenCL::Kernel
780 2543
781void 2544void
782DESTROY (OpenCL::Kernel this) 2545DESTROY (OpenCL::Kernel self)
783 CODE: 2546 CODE:
784 clReleaseKernel (this); 2547 clReleaseKernel (self);
785 2548
786void 2549void
2550setf (OpenCL::Kernel self, const char *format, ...)
2551 CODE:
2552 int i;
2553 for (i = 2; ; ++i)
2554 {
2555 while (*format == ' ')
2556 ++format;
2557
2558 char type = *format++;
2559
2560 if (!type)
2561 break;
2562
2563 if (i >= items)
2564 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2565
2566 SV *sv = ST (i);
2567
2568 union
2569 {
2570 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2571 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2572 cl_half ch; cl_float cf; cl_double cd;
2573 cl_mem cm;
2574 cl_sampler ca;
2575 size_t cz;
2576 cl_event ce;
2577 } arg;
2578 size_t size;
2579 int nullarg = 0;
2580
2581 switch (type)
2582 {
2583 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2584 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2585 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2586 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2587 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2588 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2589 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2590 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2591
2592 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2593 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2594 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2595
2596 case 'z': nullarg = 1; size = SvIV (sv); break;
2597
2598 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2599 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2600 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2601
2602 default:
2603 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2604 }
2605
2606 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2607 if (res)
2608 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2609 }
2610
2611 if (i != items)
2612 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2613
2614void
2615set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2616 CODE:
2617 clSetKernelArg (self, idx, sizeof (value), &value);
2618
2619void
2620set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2621 CODE:
2622 clSetKernelArg (self, idx, sizeof (value), &value);
2623
2624void
2625set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2626 CODE:
2627 clSetKernelArg (self, idx, sizeof (value), &value);
2628
2629void
2630set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2631 CODE:
2632 clSetKernelArg (self, idx, sizeof (value), &value);
2633
2634void
2635set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2636 CODE:
2637 clSetKernelArg (self, idx, sizeof (value), &value);
2638
2639void
2640set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2641 CODE:
2642 clSetKernelArg (self, idx, sizeof (value), &value);
2643
2644void
2645set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2646 CODE:
2647 clSetKernelArg (self, idx, sizeof (value), &value);
2648
2649void
2650set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2651 CODE:
2652 clSetKernelArg (self, idx, sizeof (value), &value);
2653
2654void
2655set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2656 CODE:
2657 clSetKernelArg (self, idx, sizeof (value), &value);
2658
2659void
2660set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2661 CODE:
2662 clSetKernelArg (self, idx, sizeof (value), &value);
2663
2664void
2665set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2666 CODE:
2667 clSetKernelArg (self, idx, sizeof (value), &value);
2668
2669void
2670set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2671 CODE:
2672 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2673
2674void
2675set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2676 CODE:
2677 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2678
2679void
2680set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2681 ALIAS:
2682 set_image2d = 0
2683 set_image3d = 0
2684 CODE:
2685 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2686
2687void
2688set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2689 CODE:
2690 clSetKernelArg (self, idx, sizeof (value), &value);
2691
2692void
2693set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2694 CODE:
2695 clSetKernelArg (self, idx, size, 0);
2696
2697void
2698set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2699 CODE:
2700 clSetKernelArg (self, idx, sizeof (value), &value);
2701
2702void
787info (OpenCL::Kernel this, cl_kernel_info name) 2703info (OpenCL::Kernel self, cl_kernel_info name)
788 PPCODE: 2704 PPCODE:
789 INFO (Kernel) 2705 INFO (Kernel)
790 2706
2707#BEGIN:kernel
2708
791void 2709void
792set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2710function_name (OpenCL::Kernel self)
2711 PPCODE:
2712 size_t size;
2713 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2714 char *value = tmpbuf (size);
2715 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2716 EXTEND (SP, 1);
2717 const int i = 0;
2718 PUSHs (sv_2mortal (newSVpv (value, 0)));
2719
2720void
2721num_args (OpenCL::Kernel self)
2722 ALIAS:
2723 num_args = CL_KERNEL_NUM_ARGS
2724 reference_count = CL_KERNEL_REFERENCE_COUNT
2725 PPCODE:
2726 cl_uint value [1];
2727 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2728 EXTEND (SP, 1);
2729 const int i = 0;
2730 PUSHs (sv_2mortal (newSVuv (value [i])));
2731
2732void
2733context (OpenCL::Kernel self)
2734 PPCODE:
2735 cl_context value [1];
2736 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2737 EXTEND (SP, 1);
2738 const int i = 0;
2739 {
2740 NEED_SUCCESS (RetainContext, (value [i]));
2741 PUSH_CLOBJ (stash_context, value [i]);
2742 }
2743
2744void
2745program (OpenCL::Kernel self)
2746 PPCODE:
2747 cl_program value [1];
2748 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2749 EXTEND (SP, 1);
2750 const int i = 0;
2751 {
2752 NEED_SUCCESS (RetainProgram, (value [i]));
2753 PUSH_CLOBJ (stash_program, value [i]);
2754 }
2755
2756#END:kernel
2757
2758void
2759work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
793 CODE: 2760 PPCODE:
794 clSetKernelArg (this, idx, sizeof (value), &value); 2761 size_t size;
2762 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2763 SV *sv = sv_2mortal (newSV (size));
2764 SvUPGRADE (sv, SVt_PV);
2765 SvPOK_only (sv);
2766 SvCUR_set (sv, size);
2767 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2768 XPUSHs (sv);
795 2769
2770#BEGIN:kernel_work_group
2771
796void 2772void
797set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value) 2773work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2774 ALIAS:
2775 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2776 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2777 PPCODE:
2778 size_t value [1];
2779 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2780 EXTEND (SP, 1);
2781 const int i = 0;
2782 PUSHs (sv_2mortal (newSVuv (value [i])));
2783
2784void
2785compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2786 PPCODE:
2787 size_t size;
2788 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2789 size_t *value = tmpbuf (size);
2790 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2791 int i, n = size / sizeof (*value);
2792 EXTEND (SP, n);
2793 for (i = 0; i < n; ++i)
2794 PUSHs (sv_2mortal (newSVuv (value [i])));
2795
2796void
2797local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
2798 ALIAS:
2799 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2800 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2801 PPCODE:
2802 cl_ulong value [1];
2803 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2804 EXTEND (SP, 1);
2805 const int i = 0;
2806 PUSHs (sv_2mortal (newSVuv (value [i])));
2807
2808#END:kernel_work_group
2809
2810#if CL_VERSION_1_2
2811
2812void
2813arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
798 CODE: 2814 PPCODE:
799 clSetKernelArg (this, idx, sizeof (value), &value); 2815 size_t size;
2816 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2817 SV *sv = sv_2mortal (newSV (size));
2818 SvUPGRADE (sv, SVt_PV);
2819 SvPOK_only (sv);
2820 SvCUR_set (sv, size);
2821 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2822 XPUSHs (sv);
800 2823
801void 2824#BEGIN:kernel_arg
802set_short (OpenCL::Kernel this, cl_uint idx, cl_short value)
803 CODE:
804 clSetKernelArg (this, idx, sizeof (value), &value);
805 2825
806void 2826#END:kernel_arg
807set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value)
808 CODE:
809 clSetKernelArg (this, idx, sizeof (value), &value);
810 2827
811void 2828#endif
812set_int (OpenCL::Kernel this, cl_uint idx, cl_int value)
813 CODE:
814 clSetKernelArg (this, idx, sizeof (value), &value);
815
816void
817set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value)
818 CODE:
819 clSetKernelArg (this, idx, sizeof (value), &value);
820
821void
822set_long (OpenCL::Kernel this, cl_uint idx, cl_long value)
823 CODE:
824 clSetKernelArg (this, idx, sizeof (value), &value);
825
826void
827set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
828 CODE:
829 clSetKernelArg (this, idx, sizeof (value), &value);
830
831void
832set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
833 CODE:
834 clSetKernelArg (this, idx, sizeof (value), &value);
835
836void
837set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
838 CODE:
839 clSetKernelArg (this, idx, sizeof (value), &value);
840
841void
842set_double (OpenCL::Kernel this, cl_uint idx, cl_double value)
843 CODE:
844 clSetKernelArg (this, idx, sizeof (value), &value);
845
846void
847set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
848 CODE:
849 clSetKernelArg (this, idx, sizeof (value), &value);
850
851void
852set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
853 CODE:
854 clSetKernelArg (this, idx, sizeof (value), &value);
855
856void
857set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
858 CODE:
859 clSetKernelArg (this, idx, sizeof (value), &value);
860
861void
862set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
863 CODE:
864 clSetKernelArg (this, idx, sizeof (value), &value);
865
866void
867set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
868 CODE:
869 clSetKernelArg (this, idx, sizeof (value), &value);
870
871void
872set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
873 CODE:
874 clSetKernelArg (this, idx, sizeof (value), &value);
875 2829
876MODULE = OpenCL PACKAGE = OpenCL::Event 2830MODULE = OpenCL PACKAGE = OpenCL::Event
877 2831
878void 2832void
879DESTROY (OpenCL::Event this) 2833DESTROY (OpenCL::Event self)
880 CODE: 2834 CODE:
881 clReleaseEvent (this); 2835 clReleaseEvent (self);
882 2836
883void 2837void
2838wait (OpenCL::Event self)
2839 CODE:
2840 clWaitForEvents (1, &self);
2841
2842void
2843cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2844 CODE:
2845 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2846
2847void
884info (OpenCL::Event this, cl_event_info name) 2848info (OpenCL::Event self, cl_event_info name)
885 PPCODE: 2849 PPCODE:
886 INFO (Event) 2850 INFO (Event)
887 2851
2852#BEGIN:event
2853
888void 2854void
889wait (OpenCL::Event this) 2855command_queue (OpenCL::Event self)
2856 PPCODE:
2857 cl_command_queue value [1];
2858 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2859 EXTEND (SP, 1);
2860 const int i = 0;
2861 {
2862 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2863 PUSH_CLOBJ (stash_queue, value [i]);
2864 }
2865
2866void
2867command_type (OpenCL::Event self)
2868 PPCODE:
2869 cl_command_type value [1];
2870 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2871 EXTEND (SP, 1);
2872 const int i = 0;
2873 PUSHs (sv_2mortal (newSVuv (value [i])));
2874
2875void
2876reference_count (OpenCL::Event self)
2877 ALIAS:
2878 reference_count = CL_EVENT_REFERENCE_COUNT
2879 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2880 PPCODE:
2881 cl_uint value [1];
2882 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2883 EXTEND (SP, 1);
2884 const int i = 0;
2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2886
2887void
2888context (OpenCL::Event self)
2889 PPCODE:
2890 cl_context value [1];
2891 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2892 EXTEND (SP, 1);
2893 const int i = 0;
2894 {
2895 NEED_SUCCESS (RetainContext, (value [i]));
2896 PUSH_CLOBJ (stash_context, value [i]);
2897 }
2898
2899#END:event
2900
2901void
2902profiling_info (OpenCL::Event self, cl_profiling_info name)
890 CODE: 2903 PPCODE:
891 clWaitForEvents (1, &this); 2904 INFO (EventProfiling)
2905
2906#BEGIN:profiling
2907
2908void
2909profiling_command_queued (OpenCL::Event self)
2910 ALIAS:
2911 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
2912 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
2913 profiling_command_start = CL_PROFILING_COMMAND_START
2914 profiling_command_end = CL_PROFILING_COMMAND_END
2915 PPCODE:
2916 cl_ulong value [1];
2917 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
2918 EXTEND (SP, 1);
2919 const int i = 0;
2920 PUSHs (sv_2mortal (newSVuv (value [i])));
2921
2922#END:profiling
892 2923
893MODULE = OpenCL PACKAGE = OpenCL::UserEvent 2924MODULE = OpenCL PACKAGE = OpenCL::UserEvent
894 2925
895void 2926void
896set_status (OpenCL::UserEvent this, cl_int execution_status) 2927set_status (OpenCL::UserEvent self, cl_int execution_status)
897 CODE: 2928 CODE:
898 clSetUserEventStatus (this, execution_status); 2929 clSetUserEventStatus (self, execution_status);
899 2930
2931MODULE = OpenCL PACKAGE = OpenCL::Mapped
2932
2933void
2934DESTROY (SV *self)
2935 CODE:
2936 OpenCL__Mapped mapped = SvMAPPED (self);
2937
2938 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2939 mapped_detach (self, mapped);
2940
2941 clReleaseCommandQueue (mapped->queue);
2942 clReleaseEvent (mapped->event);
2943 Safefree (mapped);
2944
2945void
2946unmap (OpenCL::Mapped self, ...)
2947 CODE:
2948 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2949
2950bool
2951mapped (OpenCL::Mapped self)
2952 CODE:
2953 RETVAL = !!self->ptr;
2954 OUTPUT:
2955 RETVAL
2956
2957void
2958wait (OpenCL::Mapped self)
2959 PPCODE:
2960 if (self->event)
2961 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2962
2963void
2964event (OpenCL::Mapped self)
2965 PPCODE:
2966 if (!self->event)
2967 XSRETURN_UNDEF;
2968
2969 clRetainEvent (self->event);
2970 XPUSH_CLOBJ (stash_event, self->event);
2971
2972size_t
2973size (OpenCL::Mapped self)
2974 CODE:
2975 RETVAL = self->cb;
2976 OUTPUT:
2977 RETVAL
2978
2979IV
2980ptr (OpenCL::Mapped self)
2981 CODE:
2982 RETVAL = PTR2IV (self->ptr);
2983 OUTPUT:
2984 RETVAL
2985
2986void
2987set (OpenCL::Mapped self, size_t offset, SV *data)
2988 CODE:
2989 STRLEN len;
2990 const char *ptr = SvPVbyte (data, len);
2991
2992 if (offset + len > self->cb)
2993 croak ("OpenCL::Mapped::set out of bound condition detected");
2994
2995 memcpy (offset + (char *)self->ptr, ptr, len);
2996
2997MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2998
2999MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3000
3001IV
3002row_pitch (OpenCL::Mapped self)
3003 ALIAS:
3004 slice_pitch = 1
3005 CODE:
3006 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3007 OUTPUT:
3008 RETVAL
3009
3010
3011

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines