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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.11 by root, Thu Nov 17 04:17:43 2011 UTC vs.
Revision 1.70 by root, Sat May 5 02:33:55 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;
26 62
27typedef SV *FUTURE; 63typedef SV *FUTURE;
28 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;
89
29/*****************************************************************************/ 90/*****************************************************************************/
30 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
31/* up to two temporary buffers */ 137/* up to two temporary buffers */
32static void * 138static void * ecb_noinline
33tmpbuf (size_t size) 139tmpbuf (size_t size)
34{ 140{
141 enum { buffers = 4 };
35 static int idx; 142 static int idx;
36 static void *buf [2]; 143 static void *buf [buffers];
37 static size_t len [2]; 144 static size_t len [buffers];
38 145
39 idx ^= 1; 146 idx = (idx + 1) % buffers;
40 147
41 if (len [idx] < size) 148 if (len [idx] < size)
42 { 149 {
43 free (buf [idx]); 150 free (buf [idx]);
44 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
45 buf [idx] = malloc (len [idx]); 152 buf [idx] = malloc (len [idx]);
46 } 153 }
47 154
48 return buf [idx]; 155 return buf [idx];
156}
157
158static const char * ecb_noinline
159cv_get_name (CV *cv)
160{
161 static char fullname [256];
162
163 GV *gv = CvGV (cv); // gv better be nonzero
164
165 HV *stash = GvSTASH (gv);
166 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
167 const char *gvname = GvNAME (gv);
168
169 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
170 return fullname;
49} 171}
50 172
51/*****************************************************************************/ 173/*****************************************************************************/
52 174
53typedef struct 175typedef struct
112 if (res) \ 234 if (res) \
113 FAIL (name); 235 FAIL (name);
114 236
115/*****************************************************************************/ 237/*****************************************************************************/
116 238
117#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
118#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
119
120static void * 239static SV *
240new_clobj (HV *stash, IV id)
241{
242 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
243}
244
245#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
246#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
247
248/* cl objects are either \$iv, or [$iv, ...] */
249/* they can be upgraded at runtime to the array form */
250static void * ecb_noinline
121SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 251SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
122{ 252{
253 // sv_derived_from is quite slow :(
123 if (SvROK (sv) && sv_derived_from (sv, pkg)) 254 if (SvROK (sv) && sv_derived_from (sv, pkg))
124 return (void *)SvIV (SvRV (sv)); 255 return (void *)SvIV (SvRV (sv));
125 256
126 croak ("%s: %s is not of type %s", func, svname, pkg); 257 croak ("%s: %s is not of type %s", func, svname, pkg);
258}
259
260// the "no-inherit" version of the above
261static void * ecb_noinline
262SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
263{
264 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
265 return (void *)SvIV (SvRV (sv));
266
267 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
127} 268}
128 269
129/*****************************************************************************/ 270/*****************************************************************************/
271
272static cl_context_properties * ecb_noinline
273SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
274{
275 if (!sv || !SvOK (sv))
276 if (extra)
277 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
278 else
279 return 0;
280
281 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
282 {
283 AV *av = (AV *)SvRV (sv);
284 int i, len = av_len (av) + 1;
285 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
286 cl_context_properties *l = p;
287
288 if (len & 1)
289 croak ("%s: %s is not a property list (must contain an even number of elements)", func, svname);
290
291 while (extracount--)
292 *l++ = *extra++;
293
294 for (i = 0; i < len; i += 2)
295 {
296 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
297 SV *p_sv = *av_fetch (av, i + 1, 0);
298 cl_context_properties v = SvIV (p_sv); // code below can override
299
300 switch (t)
301 {
302 case CL_CONTEXT_PLATFORM:
303 if (SvROK (p_sv))
304 v = (cl_context_properties)SvCLOBJ (func, svname, p_sv, "OpenCL::Platform");
305 break;
306
307 case CL_GLX_DISPLAY_KHR:
308 if (!SvOK (p_sv))
309 {
310 void *func = glsym ("_glXGetCurrentDisplay");
311 if (func)
312 v = (cl_context_properties)((void *(*)(void))func)();
313 }
314 break;
315
316 case CL_GL_CONTEXT_KHR:
317 if (!SvOK (p_sv))
318 {
319 void *func = glsym ("_glXGetCurrentContext");
320 if (func)
321 v = (cl_context_properties)((void *(*)(void))func)();
322 }
323 break;
324
325 default:
326 /* unknown property, treat as int */
327 break;
328 }
329
330 *l++ = t;
331 *l++ = v;
332 }
333
334 *l = 0;
335
336 return p;
337 }
338
339 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
340}
341
342static void * ecb_noinline
343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
344{
345 const char *funcname = cv_get_name (cv);
346
347 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
348 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
349 funcname, argname, or_undef ? "undef or " : "", klass);
350
351 AV *av = (AV *)SvRV (arg);
352 void **list = 0;
353 cl_uint count = av_len (av) + 1;
354
355 if (count)
356 {
357 list = tmpbuf (sizeof (*list) * count);
358 int i;
359 for (i = 0; i < count; ++i)
360 list [i] = SvCLOBJ (funcname, argname, *av_fetch (av, i, 1), klass);
361 }
362
363 if (!count && !or_undef)
364 croak ("%s: '%s' must contain at least one %s object",
365 funcname, argname, klass);
366
367 *rcount = count;
368 return (void *)list;
369}
370
371/*****************************************************************************/
372/* callback stuff */
373
374/* default context callback, log to stderr */
375static void CL_CALLBACK
376context_default_notify (const char *msg, const void *info, size_t cb, void *data)
377{
378 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
379}
380
381typedef struct
382{
383 int free_cb;
384 void (*push)(void *data1, void *data2, void *data3);
385} eq_vtbl;
386
387typedef struct eq_item
388{
389 struct eq_item *next;
390 eq_vtbl *vtbl;
391 SV *cb;
392 void *data1, *data2, *data3;
393} eq_item;
394
395static void (*eq_signal_func)(void *signal_arg, int value);
396static void *eq_signal_arg;
397static xmutex_t eq_lock = X_MUTEX_INIT;
398static eq_item *eq_head, *eq_tail;
399
400static void ecb_noinline
401eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
402{
403 eq_item *item = malloc (sizeof (eq_item));
404
405 item->next = 0;
406 item->vtbl = vtbl;
407 item->cb = cb;
408 item->data1 = data1;
409 item->data2 = data2;
410 item->data3 = data3;
411
412 X_LOCK (eq_lock);
413
414 *(eq_head ? &eq_tail->next : &eq_head) = item;
415 eq_tail = item;
416
417 X_UNLOCK (eq_lock);
418
419 eq_signal_func (eq_signal_arg, 0);
420}
421
422static eq_item *
423eq_dec (void)
424{
425 eq_item *res;
426
427 X_LOCK (eq_lock);
428
429 res = eq_head;
430 if (res)
431 eq_head = res->next;
432
433 X_UNLOCK (eq_lock);
434
435 return res;
436}
437
438static void
439eq_poll (void)
440{
441 eq_item *item;
442
443 while ((item = eq_dec ()))
444 {
445 ENTER;
446 SAVETMPS;
447
448 dSP;
449 PUSHMARK (SP);
450 EXTEND (SP, 2);
451
452 if (item->vtbl->free_cb)
453 sv_2mortal (item->cb);
454
455 PUTBACK;
456 item->vtbl->push (item->data1, item->data2, item->data3);
457
458 SV *cb = item->cb;
459 free (item);
460
461 call_sv (cb, G_DISCARD | G_VOID);
462
463 FREETMPS;
464 LEAVE;
465 }
466}
467
468static void
469eq_poll_interrupt (pTHX_ void *c_arg, int value)
470{
471 eq_poll ();
472}
473
474/*****************************************************************************/
475/* context notify */
476
477static void ecb_noinline
478eq_context_push (void *data1, void *data2, void *data3)
479{
480 dSP;
481 PUSHs (sv_2mortal (newSVpv (data1, 0)));
482 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
483 PUTBACK;
484
485 free (data1);
486 free (data2);
487}
488
489static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
490
491static void CL_CALLBACK
492eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
493{
494 void *pvt_copy = malloc (cb);
495 memcpy (pvt_copy, pvt, cb);
496 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
497}
498
499#define CONTEXT_NOTIFY_CALLBACK \
500 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
501 void *user_data = 0; \
502 \
503 if (SvOK (notify)) \
504 { \
505 pfn_notify = eq_context_notify; \
506 user_data = s_get_cv (notify); \
507 }
508
509static SV * ecb_noinline
510new_clobj_context (cl_context ctx, void *user_data)
511{
512 SV *sv = new_clobj (stash_context, (IV)ctx);
513
514 if (user_data)
515 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
516
517 return sv;
518}
519
520#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
521
522/*****************************************************************************/
523/* build/compile/link notify */
524
525static void
526eq_program_push (void *data1, void *data2, void *data3)
527{
528 dSP;
529 PUSH_CLOBJ (stash_program, data1);
530 PUTBACK;
531}
532
533static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
534
535static void CL_CALLBACK
536eq_program_notify (cl_program program, void *user_data)
537{
538 clRetainProgram (program);
539
540 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
541}
542
543typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
544
545static program_callback ecb_noinline
546make_program_callback (SV *notify, void **ruser_data)
547{
548 if (SvOK (notify))
549 {
550 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
551 return eq_program_notify;
552 }
553 else
554 {
555 *ruser_data = 0;
556 return 0;
557 }
558}
559
560struct build_args
561{
562 cl_program program;
563 char *options;
564 void *user_data;
565 cl_uint num_devices;
566};
567
568X_THREAD_PROC (build_program_thread)
569{
570 struct build_args *arg = thr_arg;
571
572 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
573
574 if (arg->user_data)
575 eq_program_notify (arg->program, arg->user_data);
576 else
577 clReleaseProgram (arg->program);
578
579 free (arg->options);
580 free (arg);
581
582 return 0;
583}
584
585static void
586build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
587{
588 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
589
590 arg->program = program;
591 arg->options = strdup (options);
592 arg->user_data = user_data;
593 arg->num_devices = num_devices;
594 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
595
596 xthread_t id;
597 thread_create (&id, build_program_thread, arg);
598}
599
600/*****************************************************************************/
601/* event objects */
602
603static void
604eq_event_push (void *data1, void *data2, void *data3)
605{
606 dSP;
607 PUSH_CLOBJ (stash_event, data1);
608 PUSHs (sv_2mortal (newSViv ((IV)data2)));
609 PUTBACK;
610}
611
612static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
613
614static void CL_CALLBACK
615eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
616{
617 clRetainEvent (event);
618 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
619}
620
621/*****************************************************************************/
622/* utilities for XS code */
130 623
131static size_t 624static size_t
132img_row_pitch (cl_mem img) 625img_row_pitch (cl_mem img)
133{ 626{
134 size_t res; 627 size_t res;
135 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 628 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
136 return res; 629 return res;
137} 630}
138 631
139static cl_event * 632static cl_event * ecb_noinline
140event_list (SV **items, int count) 633event_list (SV **items, cl_uint *rcount, cl_event extra)
141{ 634{
635 cl_uint count = *rcount;
636
637 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
638 *rcount = count = 0;
639
640 if (!count && !extra)
641 return 0;
642
142 cl_event *list = tmpbuf (sizeof (cl_event) * count); 643 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
644 int i = 0;
143 645
144 while (count--) 646 while (count--)
647 if (SvOK (items [count]))
145 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 648 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
146 649
650 if (extra)
651 list [i++] = extra;
652
653 *rcount = i;
654
147 return list; 655 return i ? list : 0;
148} 656}
149 657
150#define EVENT_LIST(items,count) \ 658#define EVENT_LIST(skip) \
151 cl_uint event_list_count = (count); \ 659 cl_uint event_list_count = items - (skip); \
152 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 660 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
153 661
154#define INFO(class) \ 662#define INFO(class) \
155{ \ 663{ \
156 size_t size; \ 664 size_t size; \
157 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 665 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
158 SV *sv = sv_2mortal (newSV (size)); \ 666 SV *sv = sv_2mortal (newSV (size)); \
159 SvUPGRADE (sv, SVt_PV); \ 667 SvUPGRADE (sv, SVt_PV); \
160 SvPOK_only (sv); \ 668 SvPOK_only (sv); \
161 SvCUR_set (sv, size); \ 669 SvCUR_set (sv, size); \
162 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 670 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
163 XPUSHs (sv); \ 671 XPUSHs (sv); \
164} 672}
165 673
674/*****************************************************************************/
675/* mapped_xxx */
676
677static OpenCL__Mapped
678SvMAPPED (SV *sv)
679{
680 // no typechecking atm., keep your fingers crossed
681 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
682}
683
684struct mapped
685{
686 cl_command_queue queue;
687 cl_mem memobj;
688 void *ptr;
689 size_t cb;
690 cl_event event;
691 size_t row_pitch;
692 size_t slice_pitch;
693};
694
695static SV *
696mapped_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)
697{
698 SV *data = newSV (0);
699 SvUPGRADE (data, SVt_PVMG);
700
701 OpenCL__Mapped mapped;
702 New (0, mapped, 1, struct mapped);
703
704 clRetainCommandQueue (queue);
705
706 mapped->queue = queue;
707 mapped->memobj = memobj;
708 mapped->ptr = ptr;
709 mapped->cb = cb;
710 mapped->event = ev;
711 mapped->row_pitch = row_pitch;
712 mapped->slice_pitch = slice_pitch;
713
714 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
715
716 if (SvLEN (data))
717 Safefree (data);
718
719 SvPVX (data) = (char *)ptr;
720 SvCUR_set (data, cb);
721 SvLEN_set (data, 0);
722 SvPOK_only (data);
723
724 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
725
726 if (!(flags & CL_MAP_WRITE))
727 SvREADONLY_on (data);
728
729 return obj;
730}
731
732static void
733mapped_detach (SV *sv, OpenCL__Mapped mapped)
734{
735 SV *data = SvRV (sv);
736
737 // the next check checks both before AND after detach, where SvPVX should STILL be 0
738 if (SvPVX (data) != (char *)mapped->ptr)
739 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
740 else
741 {
742 SvREADONLY_off (data);
743 SvCUR_set (data, 0);
744 SvPVX (data) = 0;
745 SvOK_off (data);
746 }
747
748 mapped->ptr = 0;
749}
750
751static void
752mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
753{
754 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
755 cl_event ev;
756
757 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
758
759 clReleaseEvent (mapped->event);
760 mapped->event = ev;
761
762 mapped_detach (self, mapped);
763}
764
765/*****************************************************************************/
766
166MODULE = OpenCL PACKAGE = OpenCL 767MODULE = OpenCL PACKAGE = OpenCL
167 768
168PROTOTYPES: ENABLE 769PROTOTYPES: ENABLE
169 770
771void
772poll ()
773 CODE:
774 eq_poll ();
775
776void
777_eq_initialise (IV func, IV arg)
778 CODE:
779 eq_signal_func = (void (*)(void *, int))func;
780 eq_signal_arg = (void*)arg;
781
170BOOT: 782BOOT:
171{ 783{
172 HV *stash = gv_stashpv ("OpenCL", 1); 784 HV *stash = gv_stashpv ("OpenCL", 1);
173 static const ivstr *civ, const_iv[] = { 785 static const ivstr *civ, const_iv[] = {
174 { sizeof (cl_char ), "SIZEOF_CHAR" }, 786 { sizeof (cl_char ), "SIZEOF_CHAR" },
175 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 787 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
176 { sizeof (cl_short ), "SIZEOF_SHORT" }, 788 { sizeof (cl_short ), "SIZEOF_SHORT" },
177 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 789 { sizeof (cl_ushort), "SIZEOF_USHORT" },
178 { sizeof (cl_int ), "SIZEOF_INT" }, 790 { sizeof (cl_int ), "SIZEOF_INT" },
179 { sizeof (cl_uint ), "SIZEOF_UINT" }, 791 { sizeof (cl_uint ), "SIZEOF_UINT" },
180 { sizeof (cl_long ), "SIZEOF_LONG" }, 792 { sizeof (cl_long ), "SIZEOF_LONG" },
181 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 793 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
182 { sizeof (cl_half ), "SIZEOF_HALF" }, 794 { sizeof (cl_half ), "SIZEOF_HALF" },
183 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 795 { sizeof (cl_float ), "SIZEOF_FLOAT" },
184 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 796 { sizeof (cl_double), "SIZEOF_DOUBLE" },
185#include "constiv.h" 797#include "constiv.h"
186 }; 798 };
799
187 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 800 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
188 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 801 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
802
803 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
804 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
805 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
806 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
807 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
808 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
809 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
810 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
811 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
812 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
813 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
814 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
815 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
816 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
817 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
818 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
819 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
820 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
821 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
822 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
823 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
824 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
825 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
826
827 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
189} 828}
190 829
191cl_int 830cl_int
192errno () 831errno ()
193 CODE: 832 CODE:
194 errno = res; 833 RETVAL = res;
834 OUTPUT:
835 RETVAL
195 836
196const char * 837const char *
197err2str (cl_int err) 838err2str (cl_int err = res)
198 839
199const char * 840const char *
200enum2str (cl_uint value) 841enum2str (cl_uint value)
201 842
202void 843void
210 list = tmpbuf (sizeof (*list) * count); 851 list = tmpbuf (sizeof (*list) * count);
211 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 852 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
212 853
213 EXTEND (SP, count); 854 EXTEND (SP, count);
214 for (i = 0; i < count; ++i) 855 for (i = 0; i < count; ++i)
215 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 856 PUSH_CLOBJ (stash_platform, list [i]);
216 857
217void 858void
218context_from_type (FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 859context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
219 PPCODE: 860 PPCODE:
861 CONTEXT_NOTIFY_CALLBACK;
220 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (0, type, 0, 0, &res)); 862 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
221 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 863 XPUSH_CLOBJ_CONTEXT;
222 864
223void 865void
224context (FUTURE properties, FUTURE devices, FUTURE notify = 0) 866context (FUTURE properties, FUTURE devices, FUTURE notify)
225 PPCODE: 867 PPCODE:
226 /* der Gipfel der Kunst */ 868 /* der Gipfel der Kunst */
227 869
228void 870void
229wait_for_events (...) 871wait_for_events (...)
230 CODE: 872 CODE:
231 EVENT_LIST (0, items); 873 EVENT_LIST (0);
232 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 874 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
233 875
234PROTOTYPES: DISABLE 876PROTOTYPES: DISABLE
235 877
236MODULE = OpenCL PACKAGE = OpenCL::Platform 878MODULE = OpenCL PACKAGE = OpenCL::Platform
237 879
238void 880void
239info (OpenCL::Platform this, cl_platform_info name) 881info (OpenCL::Platform self, cl_platform_info name)
240 PPCODE: 882 PPCODE:
241 INFO (Platform) 883 INFO (Platform)
242 884
243void 885void
886unload_compiler (OpenCL::Platform self)
887 CODE:
888#if CL_VERSION_1_2
889 clUnloadPlatformCompiler (self);
890#endif
891
892#BEGIN:platform
893
894void
895profile (OpenCL::Platform self)
896 ALIAS:
897 profile = CL_PLATFORM_PROFILE
898 version = CL_PLATFORM_VERSION
899 name = CL_PLATFORM_NAME
900 vendor = CL_PLATFORM_VENDOR
901 extensions = CL_PLATFORM_EXTENSIONS
902 PPCODE:
903 size_t size;
904 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
905 char *value = tmpbuf (size);
906 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
907 EXTEND (SP, 1);
908 const int i = 0;
909 PUSHs (sv_2mortal (newSVpv (value, 0)));
910
911#END:platform
912
913void
244devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 914devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
245 PPCODE: 915 PPCODE:
246 cl_device_id *list; 916 cl_device_id *list;
247 cl_uint count; 917 cl_uint count;
248 int i; 918 int i;
249 919
250 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 920 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
251 list = tmpbuf (sizeof (*list) * count); 921 list = tmpbuf (sizeof (*list) * count);
252 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 922 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
253 923
254 EXTEND (SP, count); 924 EXTEND (SP, count);
255 for (i = 0; i < count; ++i) 925 for (i = 0; i < count; ++i)
256 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 926 PUSH_CLOBJ (stash_device, list [i]);
257 927
258void 928void
259context (OpenCL::Platform this, FUTURE properties, SV *devices, FUTURE notify = 0) 929context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
260 PPCODE: 930 PPCODE:
931 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
932 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
933
934 cl_uint device_count = 0;
935 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
936
937 CONTEXT_NOTIFY_CALLBACK;
938 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
939 XPUSH_CLOBJ_CONTEXT;
940
941void
942context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
943 PPCODE:
944 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
945 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
946
947 CONTEXT_NOTIFY_CALLBACK;
948 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
949 XPUSH_CLOBJ_CONTEXT;
950
951MODULE = OpenCL PACKAGE = OpenCL::Device
952
953void
954info (OpenCL::Device self, cl_device_info name)
955 PPCODE:
956 INFO (Device)
957
958#if CL_VERSION_1_2
959
960void
961sub_devices (OpenCL::Device self, SV *properties)
962 PPCODE:
261 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 963 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
262 croak ("OpenCL::Platform argument 'device' must be an arrayref with device objects, in call"); 964 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
263 965
264 AV *av = (AV *)SvRV (devices); 966 properties = SvRV (properties);
265 cl_uint num_devices = av_len (av) + 1; 967
266 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 968 cl_uint count = av_len ((AV *)properties) + 1;
969 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
970
267 int i; 971 int i;
972 for (i = 0; i < count; ++i)
973 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
268 974
269 for (i = num_devices; i--; ) 975 props [count] = 0;
270 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
271 976
272 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (0, num_devices, device_list, 0, 0, &res)); 977 cl_uint num_devices;
273 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 978 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
979 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
980 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
274 981
982 EXTEND (SP, num_devices);
983 for (i = 0; i < count; ++i)
984 PUSH_CLOBJ (stash_subdevice, list [i]);
985
986#endif
987
988#BEGIN:device
989
275void 990void
276context_from_type (OpenCL::Platform this, FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 991type (OpenCL::Device self)
277 PPCODE: 992 PPCODE:
278 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 993 cl_device_type value [1];
279 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 994 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
280 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 995 EXTEND (SP, 1);
996 const int i = 0;
997 PUSHs (sv_2mortal (newSVuv (value [i])));
281 998
999void
1000vendor_id (OpenCL::Device self)
1001 ALIAS:
1002 vendor_id = CL_DEVICE_VENDOR_ID
1003 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
1004 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
1005 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
1006 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
1007 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
1008 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
1009 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
1010 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
1011 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
1012 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
1013 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
1014 image_support = CL_DEVICE_IMAGE_SUPPORT
1015 max_samplers = CL_DEVICE_MAX_SAMPLERS
1016 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
1017 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1018 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
1019 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1020 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
1021 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
1022 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
1023 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
1024 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
1025 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
1026 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
1027 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
1028 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
1029 PPCODE:
1030 cl_uint value [1];
1031 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1032 EXTEND (SP, 1);
1033 const int i = 0;
1034 PUSHs (sv_2mortal (newSVuv (value [i])));
1035
1036void
1037max_work_group_size (OpenCL::Device self)
1038 ALIAS:
1039 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
1040 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
1041 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
1042 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
1043 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
1044 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
1045 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
1046 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
1047 PPCODE:
1048 size_t value [1];
1049 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1050 EXTEND (SP, 1);
1051 const int i = 0;
1052 PUSHs (sv_2mortal (newSVuv (value [i])));
1053
1054void
1055max_work_item_sizes (OpenCL::Device self)
1056 PPCODE:
1057 size_t size;
1058 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
1059 size_t *value = tmpbuf (size);
1060 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
1061 int i, n = size / sizeof (*value);
1062 EXTEND (SP, n);
1063 for (i = 0; i < n; ++i)
1064 PUSHs (sv_2mortal (newSVuv (value [i])));
1065
1066void
1067address_bits (OpenCL::Device self)
1068 PPCODE:
1069 cl_bitfield value [1];
1070 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
1071 EXTEND (SP, 1);
1072 const int i = 0;
1073 PUSHs (sv_2mortal (newSVuv (value [i])));
1074
1075void
1076max_mem_alloc_size (OpenCL::Device self)
1077 ALIAS:
1078 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1079 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1080 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1081 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1082 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1083 PPCODE:
1084 cl_ulong value [1];
1085 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1086 EXTEND (SP, 1);
1087 const int i = 0;
1088 PUSHs (sv_2mortal (newSVuv (value [i])));
1089
1090void
1091single_fp_config (OpenCL::Device self)
1092 ALIAS:
1093 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1094 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1095 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
1096 PPCODE:
1097 cl_device_fp_config value [1];
1098 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1099 EXTEND (SP, 1);
1100 const int i = 0;
1101 PUSHs (sv_2mortal (newSVuv (value [i])));
1102
1103void
1104global_mem_cache_type (OpenCL::Device self)
1105 PPCODE:
1106 cl_device_mem_cache_type value [1];
1107 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
1108 EXTEND (SP, 1);
1109 const int i = 0;
1110 PUSHs (sv_2mortal (newSVuv (value [i])));
1111
1112void
1113local_mem_type (OpenCL::Device self)
1114 PPCODE:
1115 cl_device_local_mem_type value [1];
1116 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
1117 EXTEND (SP, 1);
1118 const int i = 0;
1119 PUSHs (sv_2mortal (newSVuv (value [i])));
1120
1121void
1122error_correction_support (OpenCL::Device self)
1123 ALIAS:
1124 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1125 endian_little = CL_DEVICE_ENDIAN_LITTLE
1126 available = CL_DEVICE_AVAILABLE
1127 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1128 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1129 PPCODE:
1130 cl_bool value [1];
1131 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1132 EXTEND (SP, 1);
1133 const int i = 0;
1134 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1135
1136void
1137execution_capabilities (OpenCL::Device self)
1138 PPCODE:
1139 cl_device_exec_capabilities value [1];
1140 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1141 EXTEND (SP, 1);
1142 const int i = 0;
1143 PUSHs (sv_2mortal (newSVuv (value [i])));
1144
1145void
1146properties (OpenCL::Device self)
1147 PPCODE:
1148 cl_command_queue_properties value [1];
1149 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1150 EXTEND (SP, 1);
1151 const int i = 0;
1152 PUSHs (sv_2mortal (newSVuv (value [i])));
1153
1154void
1155platform (OpenCL::Device self)
1156 PPCODE:
1157 cl_platform_id value [1];
1158 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1159 EXTEND (SP, 1);
1160 const int i = 0;
1161 {
1162 PUSH_CLOBJ (stash_platform, value [i]);
1163 }
1164
1165void
1166name (OpenCL::Device self)
1167 ALIAS:
1168 name = CL_DEVICE_NAME
1169 vendor = CL_DEVICE_VENDOR
1170 driver_version = CL_DRIVER_VERSION
1171 profile = CL_DEVICE_PROFILE
1172 version = CL_DEVICE_VERSION
1173 extensions = CL_DEVICE_EXTENSIONS
1174 PPCODE:
1175 size_t size;
1176 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1177 char *value = tmpbuf (size);
1178 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1179 EXTEND (SP, 1);
1180 const int i = 0;
1181 PUSHs (sv_2mortal (newSVpv (value, 0)));
1182
1183void
1184parent_device_ext (OpenCL::Device self)
1185 PPCODE:
1186 cl_device_id value [1];
1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1188 EXTEND (SP, 1);
1189 const int i = 0;
1190 {
1191 PUSH_CLOBJ (stash_device, value [i]);
1192 }
1193
1194void
1195partition_types_ext (OpenCL::Device self)
1196 ALIAS:
1197 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1198 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
1199 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
1200 PPCODE:
1201 size_t size;
1202 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1203 cl_device_partition_property_ext *value = tmpbuf (size);
1204 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1205 int i, n = size / sizeof (*value);
1206 EXTEND (SP, n);
1207 for (i = 0; i < n; ++i)
1208 PUSHs (sv_2mortal (newSVuv (value [i])));
1209
1210#END:device
1211
282MODULE = OpenCL PACKAGE = OpenCL::Device 1212MODULE = OpenCL PACKAGE = OpenCL::SubDevice
283 1213
1214#if CL_VERSION_1_2
1215
284void 1216void
285info (OpenCL::Device this, cl_device_info name) 1217DESTROY (OpenCL::SubDevice self)
286 PPCODE: 1218 CODE:
287 INFO (Device) 1219 clReleaseDevice (self);
1220
1221#endif
288 1222
289MODULE = OpenCL PACKAGE = OpenCL::Context 1223MODULE = OpenCL PACKAGE = OpenCL::Context
290 1224
291void 1225void
292DESTROY (OpenCL::Context context) 1226DESTROY (OpenCL::Context self)
293 CODE: 1227 CODE:
294 clReleaseContext (context); 1228 clReleaseContext (self);
295 1229
296void 1230void
297info (OpenCL::Context this, cl_context_info name) 1231info (OpenCL::Context self, cl_context_info name)
298 PPCODE: 1232 PPCODE:
299 INFO (Context) 1233 INFO (Context)
300 1234
301void 1235void
302queue (OpenCL::Context this, OpenCL::Device device, cl_command_queue_properties properties = 0) 1236queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
303 PPCODE: 1237 PPCODE:
304 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (this, device, properties, &res)); 1238 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
305 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1239 XPUSH_CLOBJ (stash_queue, queue);
306 1240
307void 1241void
308user_event (OpenCL::Context this) 1242user_event (OpenCL::Context self)
309 PPCODE: 1243 PPCODE:
310 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (this, &res)); 1244 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
311 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1245 XPUSH_CLOBJ (stash_userevent, ev);
312 1246
313void 1247void
314buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1248buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
315 PPCODE: 1249 PPCODE:
316 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1250 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
317 croak ("clCreateBuffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1251 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
318 1252
319 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (this, flags, len, 0, &res)); 1253 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
320 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1254 XPUSH_CLOBJ (stash_bufferobj, mem);
321 1255
322void 1256void
323buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1257buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
324 PPCODE: 1258 PPCODE:
325 STRLEN len; 1259 STRLEN len;
326 char *ptr = SvPVbyte (data, len); 1260 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
327
328 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1261 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
329 croak ("clCreateBuffer: have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1262 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
330
331 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (this, flags, len, ptr, &res)); 1263 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
332 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1264 XPUSH_CLOBJ (stash_bufferobj, mem);
333 1265
1266#if CL_VERSION_1_2
1267
334void 1268void
335image2d (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 row_pitch, SV *data) 1269image (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)
336 PPCODE: 1270 PPCODE:
337 STRLEN len; 1271 STRLEN len;
338 char *ptr = SvPVbyte (data, len); 1272 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
339 const cl_image_format format = { channel_order, channel_type }; 1273 const cl_image_format format = { channel_order, channel_type };
1274 const cl_image_desc desc = {
1275 type,
1276 width, height, depth,
1277 array_size, row_pitch, slice_pitch,
1278 num_mip_level, num_samples,
1279 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1280 };
340 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (this, flags, &format, width, height, row_pitch, ptr, &res)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
341 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1282 HV *stash = stash_image;
1283 switch (type)
1284 {
1285 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1286 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1287 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1288 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1289 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1290 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1291 }
1292 XPUSH_CLOBJ (stash, mem);
342 1293
1294#endif
1295
343void 1296void
344image3d (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 row_pitch, size_t slice_pitch, SV *data) 1297image2d (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)
345 PPCODE: 1298 PPCODE:
346 STRLEN len; 1299 STRLEN len;
347 char *ptr = SvPVbyte (data, len); 1300 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
348 const cl_image_format format = { channel_order, channel_type }; 1301 const cl_image_format format = { channel_order, channel_type };
1302#if PREFER_1_1
1303 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1304#else
1305 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1306 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1307#endif
1308 XPUSH_CLOBJ (stash_image2d, mem);
1309
1310void
1311image3d (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)
1312 PPCODE:
1313 STRLEN len;
1314 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1315 const cl_image_format format = { channel_order, channel_type };
1316#if PREFER_1_1
349 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (this, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1317 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
350 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1318#else
1319 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1320 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1321#endif
1322 XPUSH_CLOBJ (stash_image3d, mem);
351 1323
1324#if cl_apple_gl_sharing || cl_khr_gl_sharing
1325
352void 1326void
1327gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1328 PPCODE:
1329 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1330 XPUSH_CLOBJ (stash_bufferobj, mem);
1331
1332void
1333gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1334 PPCODE:
1335 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1336 XPUSH_CLOBJ (stash_image2d, mem);
1337
1338#if CL_VERSION_1_2
1339
1340void
1341gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1342 ALIAS:
1343 PPCODE:
1344 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1345 cl_gl_object_type type;
1346 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1347 HV *stash = stash_memory;
1348 switch (type)
1349 {
1350 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1351 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1352 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1353 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1354 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1355 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1356 }
1357 XPUSH_CLOBJ (stash, mem);
1358
1359#endif
1360
1361void
1362gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1363 PPCODE:
1364#if PREFER_1_1
1365 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1366#else
1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1368#endif
1369 XPUSH_CLOBJ (stash_image2d, mem);
1370
1371void
1372gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1373 PPCODE:
1374#if PREFER_1_1
1375 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1376#else
1377 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1378#endif
1379 XPUSH_CLOBJ (stash_image3d, mem);
1380
1381#endif
1382
1383void
353supported_image_formats (OpenCL::Context this, cl_mem_flags flags, cl_mem_object_type image_type) 1384supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
354 PPCODE: 1385 PPCODE:
355{ 1386{
356 cl_uint count; 1387 cl_uint count;
357 cl_image_format *list; 1388 cl_image_format *list;
358 int i; 1389 int i;
359 1390
360 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1391 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
361 Newx (list, count, cl_image_format); 1392 Newx (list, count, cl_image_format);
362 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1393 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
363 1394
364 EXTEND (SP, count); 1395 EXTEND (SP, count);
365 for (i = 0; i < count; ++i) 1396 for (i = 0; i < count; ++i)
366 { 1397 {
367 AV *av = newAV (); 1398 AV *av = newAV ();
370 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1401 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
371 } 1402 }
372} 1403}
373 1404
374void 1405void
375sampler (OpenCL::Context this, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1406sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
376 PPCODE: 1407 PPCODE:
377 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (this, normalized_coords, addressing_mode, filter_mode, &res)); 1408 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
378 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1409 XPUSH_CLOBJ (stash_sampler, sampler);
379 1410
380void 1411void
381program_with_source (OpenCL::Context this, SV *program) 1412program_with_source (OpenCL::Context self, SV *program)
382 PPCODE: 1413 PPCODE:
383 STRLEN len; 1414 STRLEN len;
384 size_t len2; 1415 size_t len2;
385 const char *ptr = SvPVbyte (program, len); 1416 const char *ptr = SvPVbyte (program, len);
386 1417
387 len2 = len; 1418 len2 = len;
388 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (this, 1, &ptr, &len2, &res)); 1419 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
389 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1420 XPUSH_CLOBJ (stash_program, prog);
1421
1422void
1423program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1424 PPCODE:
1425 cl_uint device_count;
1426 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1427
1428 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1429 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1430
1431 binaries = SvRV (binaries);
1432
1433 if (device_count != av_len ((AV *)binaries) + 1)
1434 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1435
1436 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1437 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1438 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1439
1440 int i;
1441 for (i = 0; i < device_count; ++i)
1442 {
1443 STRLEN len;
1444 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1445 length_list [i] = len;
1446 }
1447
1448 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1449 length_list, binary_list,
1450 GIMME_V == G_ARRAY ? status_list : 0, &res));
1451
1452 EXTEND (SP, 2);
1453 PUSH_CLOBJ (stash_program, prog);
1454
1455 if (GIMME_V == G_ARRAY)
1456 {
1457 AV *av = newAV ();
1458 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1459
1460 for (i = device_count; i--; )
1461 av_store (av, i, newSViv (status_list [i]));
1462 }
1463
1464#if CL_VERSION_1_2
1465
1466void
1467program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1468 PPCODE:
1469 cl_uint device_count = 0;
1470 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1471
1472 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1473
1474 XPUSH_CLOBJ (stash_program, prog);
1475
1476void
1477link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1478 CODE:
1479 cl_uint device_count = 0;
1480 cl_device_id *device_list = 0;
1481
1482 if (SvOK (devices))
1483 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1484
1485 cl_uint program_count;
1486 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1487
1488 void *user_data;
1489 program_callback pfn_notify = make_program_callback (notify, &user_data);
1490
1491 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1492 program_count, program_list, pfn_notify, user_data, &res));
1493
1494 XPUSH_CLOBJ (stash_program, prog);
1495
1496#endif
1497
1498#BEGIN:context
1499
1500void
1501reference_count (OpenCL::Context self)
1502 ALIAS:
1503 reference_count = CL_CONTEXT_REFERENCE_COUNT
1504 num_devices = CL_CONTEXT_NUM_DEVICES
1505 PPCODE:
1506 cl_uint value [1];
1507 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1508 EXTEND (SP, 1);
1509 const int i = 0;
1510 PUSHs (sv_2mortal (newSVuv (value [i])));
1511
1512void
1513devices (OpenCL::Context self)
1514 PPCODE:
1515 size_t size;
1516 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1517 cl_device_id *value = tmpbuf (size);
1518 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1519 int i, n = size / sizeof (*value);
1520 EXTEND (SP, n);
1521 for (i = 0; i < n; ++i)
1522 {
1523 PUSH_CLOBJ (stash_device, value [i]);
1524 }
1525
1526void
1527properties (OpenCL::Context self)
1528 PPCODE:
1529 size_t size;
1530 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1531 cl_context_properties *value = tmpbuf (size);
1532 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1533 int i, n = size / sizeof (*value);
1534 EXTEND (SP, n);
1535 for (i = 0; i < n; ++i)
1536 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1537
1538#END:context
390 1539
391MODULE = OpenCL PACKAGE = OpenCL::Queue 1540MODULE = OpenCL PACKAGE = OpenCL::Queue
392 1541
393void 1542void
394DESTROY (OpenCL::Queue this) 1543DESTROY (OpenCL::Queue self)
395 CODE: 1544 CODE:
396 clReleaseCommandQueue (this); 1545 clReleaseCommandQueue (self);
397 1546
398void 1547void
399info (OpenCL::Queue this, cl_command_queue_info name)
400 PPCODE:
401 INFO (CommandQueue)
402
403void
404enqueue_read_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1548read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1549 ALIAS:
1550 enqueue_read_buffer = 0
405 PPCODE: 1551 PPCODE:
406 cl_event ev = 0;
407 EVENT_LIST (6, items - 6); 1552 EVENT_LIST (6);
408 1553
409 SvUPGRADE (data, SVt_PV); 1554 SvUPGRADE (data, SVt_PV);
410 SvGROW (data, len); 1555 SvGROW (data, len);
411 SvPOK_only (data); 1556 SvPOK_only (data);
412 SvCUR_set (data, len); 1557 SvCUR_set (data, len);
1558
1559 cl_event ev = 0;
413 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1560 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
414 1561
415 if (ev) 1562 if (ev)
416 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1563 XPUSH_CLOBJ (stash_event, ev);
417 1564
418void 1565void
419enqueue_write_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1566write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1567 ALIAS:
1568 enqueue_write_buffer = 0
420 PPCODE: 1569 PPCODE:
421 cl_event ev = 0; 1570 EVENT_LIST (5);
1571
422 STRLEN len; 1572 STRLEN len;
423 char *ptr = SvPVbyte (data, len); 1573 char *ptr = SvPVbyte (data, len);
424 EVENT_LIST (5, items - 5);
425 1574
1575 cl_event ev = 0;
426 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
427 1577
428 if (ev) 1578 if (ev)
429 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
430 1580
1581#if CL_VERSION_1_2
1582
431void 1583void
432enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1584fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1585 ALIAS:
1586 enqueue_fill_buffer = 0
433 PPCODE: 1587 PPCODE:
1588 EVENT_LIST (5);
1589
1590 STRLEN len;
1591 char *ptr = SvPVbyte (data, len);
1592
434 cl_event ev = 0; 1593 cl_event ev = 0;
435 EVENT_LIST (6, items - 6);
436
437 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1594 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
438 1595
439 if (ev) 1596 if (ev)
440 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1597 XPUSH_CLOBJ (stash_event, ev);
441 1598
442void 1599void
443enqueue_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, ...) 1600fill_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, ...)
1601 ALIAS:
1602 enqueue_fill_image = 0
444 PPCODE: 1603 PPCODE:
1604 EVENT_LIST (12);
1605
1606 const size_t origin [3] = { x, y, z };
1607 const size_t region [3] = { width, height, depth };
1608
1609 const cl_float c_f [4] = { r, g, b, a };
1610 const cl_uint c_u [4] = { r, g, b, a };
1611 const cl_int c_s [4] = { r, g, b, a };
1612 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1613 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1614 cl_image_format format;
1615 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1616 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1617 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1618 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1619
445 cl_event ev = 0; 1620 cl_event ev = 0;
1621 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1622 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1623
1624 if (ev)
1625 XPUSH_CLOBJ (stash_event, ev);
1626
1627#endif
1628
1629void
1630copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1631 ALIAS:
1632 enqueue_copy_buffer = 0
1633 PPCODE:
1634 EVENT_LIST (6);
1635
1636 cl_event ev = 0;
1637 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1638
1639 if (ev)
1640 XPUSH_CLOBJ (stash_event, ev);
1641
1642void
1643read_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, ...)
1644 ALIAS:
1645 enqueue_read_buffer_rect = 0
1646 PPCODE:
1647 EVENT_LIST (17);
1648
1649 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
446 const size_t src_origin[3] = { src_x, src_y, src_z }; 1650 const size_t host_origin[3] = { host_x, host_y, host_z };
447 const size_t region[3] = { width, height, depth }; 1651 const size_t region[3] = { width, height, depth };
448 EVENT_LIST (12, items - 12);
449 1652
450 if (!row_pitch) 1653 if (!buf_row_pitch)
451 row_pitch = img_row_pitch (src); 1654 buf_row_pitch = region [0];
452 1655
453 if (depth > 1 && !slice_pitch) 1656 if (!buf_slice_pitch)
454 slice_pitch = row_pitch * height; 1657 buf_slice_pitch = region [1] * buf_row_pitch;
455 1658
456 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1659 if (!host_row_pitch)
1660 host_row_pitch = region [0];
1661
1662 if (!host_slice_pitch)
1663 host_slice_pitch = region [1] * host_row_pitch;
1664
1665 size_t len = host_row_pitch * host_slice_pitch * region [2];
457 1666
458 SvUPGRADE (data, SVt_PV); 1667 SvUPGRADE (data, SVt_PV);
459 SvGROW (data, len); 1668 SvGROW (data, len);
460 SvPOK_only (data); 1669 SvPOK_only (data);
461 SvCUR_set (data, len); 1670 SvCUR_set (data, len);
462 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)); 1671
1672 cl_event ev = 0;
1673 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));
463 1674
464 if (ev) 1675 if (ev)
465 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1676 XPUSH_CLOBJ (stash_event, ev);
466 1677
467void 1678void
468enqueue_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, size_t slice_pitch, SV *data, ...) 1679write_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, ...)
1680 ALIAS:
1681 enqueue_write_buffer_rect = 0
469 PPCODE: 1682 PPCODE:
1683 EVENT_LIST (17);
1684
1685 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1686 const size_t host_origin[3] = { host_x, host_y, host_z };
1687 const size_t region[3] = { width, height, depth };
1688 STRLEN len;
1689 char *ptr = SvPVbyte (data, len);
1690
1691 if (!buf_row_pitch)
1692 buf_row_pitch = region [0];
1693
1694 if (!buf_slice_pitch)
1695 buf_slice_pitch = region [1] * buf_row_pitch;
1696
1697 if (!host_row_pitch)
1698 host_row_pitch = region [0];
1699
1700 if (!host_slice_pitch)
1701 host_slice_pitch = region [1] * host_row_pitch;
1702
1703 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1704
1705 if (len < min_len)
1706 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1707
470 cl_event ev = 0; 1708 cl_event ev = 0;
1709 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));
1710
1711 if (ev)
1712 XPUSH_CLOBJ (stash_event, ev);
1713
1714void
1715copy_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, ...)
1716 ALIAS:
1717 enqueue_copy_buffer_rect = 0
1718 PPCODE:
1719 EVENT_LIST (16);
1720
1721 const size_t src_origin[3] = { src_x, src_y, src_z };
1722 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1723 const size_t region[3] = { width, height, depth };
1724
1725 cl_event ev = 0;
1726 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));
1727
1728 if (ev)
1729 XPUSH_CLOBJ (stash_event, ev);
1730
1731void
1732read_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, ...)
1733 ALIAS:
1734 enqueue_read_image = 0
1735 PPCODE:
1736 EVENT_LIST (12);
1737
1738 const size_t src_origin[3] = { src_x, src_y, src_z };
1739 const size_t region[3] = { width, height, depth };
1740
1741 if (!row_pitch)
1742 row_pitch = img_row_pitch (src);
1743
1744 if (depth > 1 && !slice_pitch)
1745 slice_pitch = row_pitch * height;
1746
1747 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1748
1749 SvUPGRADE (data, SVt_PV);
1750 SvGROW (data, len);
1751 SvPOK_only (data);
1752 SvCUR_set (data, len);
1753
1754 cl_event ev = 0;
1755 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));
1756
1757 if (ev)
1758 XPUSH_CLOBJ (stash_event, ev);
1759
1760void
1761write_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, ...)
1762 ALIAS:
1763 enqueue_write_image = 0
1764 PPCODE:
1765 EVENT_LIST (12);
1766
471 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1767 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
472 const size_t region[3] = { width, height, depth }; 1768 const size_t region[3] = { width, height, depth };
473 STRLEN len; 1769 STRLEN len;
474 char *ptr = SvPVbyte (data, len); 1770 char *ptr = SvPVbyte (data, len);
475 EVENT_LIST (12, items - 12);
476 1771
477 if (!row_pitch) 1772 if (!row_pitch)
478 row_pitch = img_row_pitch (dst); 1773 row_pitch = img_row_pitch (dst);
479 1774
480 if (depth > 1 && !slice_pitch) 1775 if (depth > 1 && !slice_pitch)
483 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1778 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
484 1779
485 if (len < min_len) 1780 if (len < min_len)
486 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1781 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
487 1782
1783 cl_event ev = 0;
488 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)); 1784 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));
489 1785
490 if (ev) 1786 if (ev)
491 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1787 XPUSH_CLOBJ (stash_event, ev);
492 1788
493void 1789void
494enqueue_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, ...) 1790copy_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, ...)
1791 ALIAS:
1792 enqueue_copy_image = 0
495 PPCODE: 1793 PPCODE:
496 cl_event ev = 0; 1794 EVENT_LIST (12);
1795
497 const size_t src_origin[3] = { src_x, src_y, src_z }; 1796 const size_t src_origin[3] = { src_x, src_y, src_z };
498 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1797 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
499 const size_t region[3] = { width, height, depth }; 1798 const size_t region[3] = { width, height, depth };
500 EVENT_LIST (16, items - 16);
501 1799
1800 cl_event ev = 0;
502 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)); 1801 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
503 1802
504 if (ev) 1803 if (ev)
505 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1804 XPUSH_CLOBJ (stash_event, ev);
506 1805
507void 1806void
508enqueue_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, ...) 1807copy_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, ...)
1808 ALIAS:
1809 enqueue_copy_image_to_buffer = 0
509 PPCODE: 1810 PPCODE:
1811 EVENT_LIST (10);
1812
1813 const size_t src_origin[3] = { src_x, src_y, src_z };
1814 const size_t region [3] = { width, height, depth };
1815
510 cl_event ev = 0; 1816 cl_event ev = 0;
1817 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1818
1819 if (ev)
1820 XPUSH_CLOBJ (stash_event, ev);
1821
1822void
1823copy_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, ...)
1824 ALIAS:
1825 enqueue_copy_buffer_to_image = 0
1826 PPCODE:
1827 EVENT_LIST (10);
1828
511 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1829 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1830 const size_t region [3] = { width, height, depth };
1831
1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839map_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, ...)
1840 ALIAS:
1841 enqueue_map_buffer = 0
1842 PPCODE:
1843 EVENT_LIST (6);
1844
1845 size_t cb = SvIV (cb_);
1846
1847 if (!SvOK (cb_))
1848 {
1849 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1850 cb -= offset;
1851 }
1852
1853 cl_event ev;
1854 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1855 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1856
1857void
1858map_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, ...)
1859 ALIAS:
1860 enqueue_map_image = 0
1861 PPCODE:
1862 size_t width = SvIV (width_);
1863 if (!SvOK (width_))
1864 {
1865 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1866 width -= x;
1867 }
1868
1869 size_t height = SvIV (width_);
1870 if (!SvOK (height_))
1871 {
1872 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1873 height -= y;
1874 }
1875
1876 size_t depth = SvIV (width_);
1877 if (!SvOK (depth_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1880 depth -= z;
1881
1882 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1883 if (!depth)
1884 depth = 1;
1885 }
1886
1887 const size_t origin[3] = { x, y, z };
512 const size_t region[3] = { width, height, depth }; 1888 const size_t region[3] = { width, height, depth };
1889 size_t row_pitch, slice_pitch;
513 EVENT_LIST (10, items - 10); 1890 EVENT_LIST (10);
514 1891
515 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1892 cl_event ev;
1893 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));
1894
1895 size_t cb = slice_pitch ? slice_pitch * region [2]
1896 : row_pitch ? row_pitch * region [1]
1897 : region [0];
1898
1899 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1900
1901void
1902unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1903 PPCODE:
1904 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1905 if (GIMME_V != G_VOID)
1906 {
1907 clRetainEvent (mapped->event);
1908 XPUSH_CLOBJ (stash_event, mapped->event);
1909 }
1910
1911void
1912task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1913 ALIAS:
1914 enqueue_task = 0
1915 PPCODE:
1916 EVENT_LIST (2);
1917
1918 cl_event ev = 0;
1919 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
516 1920
517 if (ev) 1921 if (ev)
518 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1922 XPUSH_CLOBJ (stash_event, ev);
519 1923
520void 1924void
521enqueue_copy_image (OpenCL::Queue this, 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, ...)
522 PPCODE:
523 cl_event ev = 0;
524 const size_t src_origin[3] = { src_x, src_y, src_z };
525 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
526 const size_t region[3] = { width, height, depth };
527 EVENT_LIST (12, items - 12);
528
529 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
530
531 if (ev)
532 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
533
534void
535enqueue_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, ...)
536 PPCODE:
537 cl_event ev = 0;
538 const size_t src_origin[3] = { src_x, src_y, src_z };
539 const size_t region[3] = { width, height, depth };
540 EVENT_LIST (10, items - 10);
541
542 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
543
544 if (ev)
545 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
546
547void
548enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
549 PPCODE:
550 cl_event ev = 0;
551 EVENT_LIST (2, items - 2);
552
553 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
554
555 if (ev)
556 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
557
558void
559enqueue_nd_range_kernel (OpenCL::Queue this, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1925nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1926 ALIAS:
1927 enqueue_nd_range_kernel = 0
560 PPCODE: 1928 PPCODE:
561 cl_event ev = 0; 1929 EVENT_LIST (5);
1930
562 size_t *gwo = 0, *gws, *lws = 0; 1931 size_t *gwo = 0, *gws, *lws = 0;
563 int gws_len; 1932 int gws_len;
564 size_t *lists; 1933 size_t *lists;
565 int i; 1934 int i;
566 EVENT_LIST (5, items - 5);
567 1935
568 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1936 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
569 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1937 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
570 1938
571 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1939 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
572 1940
573 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1941 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
574 1942
575 gws = lists + gws_len * 0; 1943 gws = lists + gws_len * 0;
576 for (i = 0; i < gws_len; ++i) 1944 for (i = 0; i < gws_len; ++i)
1945 {
577 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1946 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1947 // at least nvidia crashes for 0-sized work group sizes, work around
1948 if (!gws [i])
1949 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1950 }
578 1951
579 if (SvOK (global_work_offset)) 1952 if (SvOK (global_work_offset))
580 { 1953 {
581 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1954 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
582 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
589 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1962 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
590 } 1963 }
591 1964
592 if (SvOK (local_work_size)) 1965 if (SvOK (local_work_size))
593 { 1966 {
594 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1967 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
595 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1968 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
596 1969
597 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1970 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
598 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1971 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
599 1972
600 lws = lists + gws_len * 2; 1973 lws = lists + gws_len * 2;
601 for (i = 0; i < gws_len; ++i) 1974 for (i = 0; i < gws_len; ++i)
1975 {
602 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1976 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1977 // at least nvidia crashes for 0-sized work group sizes, work around
1978 if (!lws [i])
1979 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1980 }
603 } 1981 }
604 1982
1983 cl_event ev = 0;
605 NEED_SUCCESS (EnqueueNDRangeKernel, (this, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1984 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
606 1985
607 if (ev) 1986 if (ev)
608 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1987 XPUSH_CLOBJ (stash_event, ev);
609 1988
1989#if CL_VERSION_1_2
1990
610void 1991void
611enqueue_marker (OpenCL::Queue this) 1992migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1993 ALIAS:
1994 enqueue_migrate_mem_objects = 0
612 PPCODE: 1995 PPCODE:
1996 EVENT_LIST (3);
1997
1998 cl_uint obj_count;
1999 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2000
2001 cl_event ev = 0;
2002 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2003
2004 if (ev)
2005 XPUSH_CLOBJ (stash_event, ev);
2006
2007#endif
2008
2009#if cl_apple_gl_sharing || cl_khr_gl_sharing
2010
2011void
2012acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2013 ALIAS:
2014 release_gl_objects = 1
2015 enqueue_acquire_gl_objects = 0
2016 enqueue_release_gl_objects = 1
2017 PPCODE:
2018 EVENT_LIST (2);
2019
2020 cl_uint obj_count;
2021 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2022
613 cl_event ev; 2023 cl_event ev = 0;
614 NEED_SUCCESS (EnqueueMarker, (this, &ev));
615 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
616 2024
2025 if (ix)
2026 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2027 else
2028 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2029
2030 if (ev)
2031 XPUSH_CLOBJ (stash_event, ev);
2032
2033#endif
2034
617void 2035void
618enqueue_wait_for_events (OpenCL::Queue this, ...) 2036wait_for_events (OpenCL::Queue self, ...)
2037 ALIAS:
2038 enqueue_wait_for_events = 0
619 CODE: 2039 CODE:
620 EVENT_LIST (1, items - 1); 2040 EVENT_LIST (1);
2041#if PREFER_1_1
621 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 2042 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2043#else
2044 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2045#endif
622 2046
623void 2047void
624enqueue_barrier (OpenCL::Queue this) 2048marker (OpenCL::Queue self, ...)
2049 ALIAS:
2050 enqueue_marker = 0
625 CODE: 2051 PPCODE:
2052 EVENT_LIST (1);
2053 cl_event ev = 0;
2054#if PREFER_1_1
2055 if (!event_list_count)
2056 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2057 else
2058#if CL_VERSION_1_2
2059 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2060#else
2061 {
2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2063 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2064 }
2065#endif
2066#else
2067 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2068#endif
2069 if (ev)
2070 XPUSH_CLOBJ (stash_event, ev);
2071
2072void
2073barrier (OpenCL::Queue self, ...)
2074 ALIAS:
2075 enqueue_barrier = 0
2076 PPCODE:
2077 EVENT_LIST (1);
2078 cl_event ev = 0;
2079#if PREFER_1_1
2080 if (!event_list_count && GIMME_V == G_VOID)
626 NEED_SUCCESS (EnqueueBarrier, (this)); 2081 NEED_SUCCESS (EnqueueBarrier, (self));
2082 else
2083#if CL_VERSION_1_2
2084 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2085#else
2086 {
2087 if (event_list_count)
2088 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
627 2089
2090 if (GIMME_V != G_VOID)
2091 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2092 }
2093#endif
2094#else
2095 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2096#endif
2097 if (ev)
2098 XPUSH_CLOBJ (stash_event, ev);
2099
628void 2100void
629flush (OpenCL::Queue this) 2101flush (OpenCL::Queue self)
630 CODE: 2102 CODE:
631 NEED_SUCCESS (Flush, (this)); 2103 NEED_SUCCESS (Flush, (self));
632 2104
633void 2105void
634finish (OpenCL::Queue this) 2106finish (OpenCL::Queue self)
635 CODE: 2107 CODE:
636 NEED_SUCCESS (Finish, (this)); 2108 NEED_SUCCESS (Finish, (self));
2109
2110void
2111info (OpenCL::Queue self, cl_command_queue_info name)
2112 PPCODE:
2113 INFO (CommandQueue)
2114
2115#BEGIN:command_queue
2116
2117void
2118context (OpenCL::Queue self)
2119 PPCODE:
2120 cl_context value [1];
2121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
2122 EXTEND (SP, 1);
2123 const int i = 0;
2124 {
2125 NEED_SUCCESS (RetainContext, (value [i]));
2126 PUSH_CLOBJ (stash_context, value [i]);
2127 }
2128
2129void
2130device (OpenCL::Queue self)
2131 PPCODE:
2132 cl_device_id value [1];
2133 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
2134 EXTEND (SP, 1);
2135 const int i = 0;
2136 {
2137 PUSH_CLOBJ (stash_device, value [i]);
2138 }
2139
2140void
2141reference_count (OpenCL::Queue self)
2142 PPCODE:
2143 cl_uint value [1];
2144 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
2145 EXTEND (SP, 1);
2146 const int i = 0;
2147 PUSHs (sv_2mortal (newSVuv (value [i])));
2148
2149void
2150properties (OpenCL::Queue self)
2151 PPCODE:
2152 cl_command_queue_properties value [1];
2153 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
2154 EXTEND (SP, 1);
2155 const int i = 0;
2156 PUSHs (sv_2mortal (newSVuv (value [i])));
2157
2158#END:command_queue
637 2159
638MODULE = OpenCL PACKAGE = OpenCL::Memory 2160MODULE = OpenCL PACKAGE = OpenCL::Memory
639 2161
640void 2162void
641DESTROY (OpenCL::Memory this) 2163DESTROY (OpenCL::Memory self)
642 CODE: 2164 CODE:
643 clReleaseMemObject (this); 2165 clReleaseMemObject (self);
644 2166
645void 2167void
646info (OpenCL::Memory this, cl_mem_info name) 2168info (OpenCL::Memory self, cl_mem_info name)
647 PPCODE: 2169 PPCODE:
648 INFO (MemObject) 2170 INFO (MemObject)
649 2171
2172#BEGIN:mem
2173
2174void
2175type (OpenCL::Memory self)
2176 PPCODE:
2177 cl_mem_object_type value [1];
2178 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
2179 EXTEND (SP, 1);
2180 const int i = 0;
2181 PUSHs (sv_2mortal (newSVuv (value [i])));
2182
2183void
2184flags (OpenCL::Memory self)
2185 PPCODE:
2186 cl_mem_flags value [1];
2187 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
2188 EXTEND (SP, 1);
2189 const int i = 0;
2190 PUSHs (sv_2mortal (newSVuv (value [i])));
2191
2192void
2193size (OpenCL::Memory self)
2194 ALIAS:
2195 size = CL_MEM_SIZE
2196 offset = CL_MEM_OFFSET
2197 PPCODE:
2198 size_t value [1];
2199 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2200 EXTEND (SP, 1);
2201 const int i = 0;
2202 PUSHs (sv_2mortal (newSVuv (value [i])));
2203
2204void
2205host_ptr (OpenCL::Memory self)
2206 PPCODE:
2207 void * value [1];
2208 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
2209 EXTEND (SP, 1);
2210 const int i = 0;
2211 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
2212
2213void
2214map_count (OpenCL::Memory self)
2215 ALIAS:
2216 map_count = CL_MEM_MAP_COUNT
2217 reference_count = CL_MEM_REFERENCE_COUNT
2218 PPCODE:
2219 cl_uint value [1];
2220 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2221 EXTEND (SP, 1);
2222 const int i = 0;
2223 PUSHs (sv_2mortal (newSVuv (value [i])));
2224
2225void
2226context (OpenCL::Memory self)
2227 PPCODE:
2228 cl_context value [1];
2229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
2230 EXTEND (SP, 1);
2231 const int i = 0;
2232 {
2233 NEED_SUCCESS (RetainContext, (value [i]));
2234 PUSH_CLOBJ (stash_context, value [i]);
2235 }
2236
2237void
2238associated_memobject (OpenCL::Memory self)
2239 PPCODE:
2240 cl_mem value [1];
2241 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
2242 EXTEND (SP, 1);
2243 const int i = 0;
2244 {
2245 NEED_SUCCESS (RetainMemObject, (value [i]));
2246 PUSH_CLOBJ (stash_memory, value [i]);
2247 }
2248
2249#END:mem
2250
2251#if cl_apple_gl_sharing || cl_khr_gl_sharing
2252
2253void
2254gl_object_info (OpenCL::Memory self)
2255 PPCODE:
2256 cl_gl_object_type type;
2257 cl_GLuint name;
2258 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2259 EXTEND (SP, 2);
2260 PUSHs (sv_2mortal (newSVuv (type)));
2261 PUSHs (sv_2mortal (newSVuv (name)));
2262
2263#endif
2264
2265MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2266
2267void
2268sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2269 PPCODE:
2270 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2271 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2272
2273 cl_buffer_region crdata = { origin, size };
2274
2275 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2276 XPUSH_CLOBJ (stash_buffer, mem);
2277
2278MODULE = OpenCL PACKAGE = OpenCL::Image
2279
2280void
2281image_info (OpenCL::Image self, cl_image_info name)
2282 PPCODE:
2283 INFO (Image)
2284
2285void
2286format (OpenCL::Image self)
2287 PPCODE:
2288 cl_image_format format;
2289 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2290 EXTEND (SP, 2);
2291 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2292 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2293
2294#BEGIN:image
2295
2296void
2297element_size (OpenCL::Image self)
2298 ALIAS:
2299 element_size = CL_IMAGE_ELEMENT_SIZE
2300 row_pitch = CL_IMAGE_ROW_PITCH
2301 slice_pitch = CL_IMAGE_SLICE_PITCH
2302 width = CL_IMAGE_WIDTH
2303 height = CL_IMAGE_HEIGHT
2304 depth = CL_IMAGE_DEPTH
2305 PPCODE:
2306 size_t value [1];
2307 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
2308 EXTEND (SP, 1);
2309 const int i = 0;
2310 PUSHs (sv_2mortal (newSVuv (value [i])));
2311
2312#END:image
2313
2314#if cl_apple_gl_sharing || cl_khr_gl_sharing
2315
2316#BEGIN:gl_texture
2317
2318void
2319target (OpenCL::Image self)
2320 PPCODE:
2321 cl_GLenum value [1];
2322 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2323 EXTEND (SP, 1);
2324 const int i = 0;
2325 PUSHs (sv_2mortal (newSVuv (value [i])));
2326
2327void
2328gl_mipmap_level (OpenCL::Image self)
2329 PPCODE:
2330 cl_GLint value [1];
2331 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2332 EXTEND (SP, 1);
2333 const int i = 0;
2334 PUSHs (sv_2mortal (newSViv (value [i])));
2335
2336#END:gl_texture
2337
2338#endif
2339
650MODULE = OpenCL PACKAGE = OpenCL::Sampler 2340MODULE = OpenCL PACKAGE = OpenCL::Sampler
651 2341
652void 2342void
653DESTROY (OpenCL::Sampler this) 2343DESTROY (OpenCL::Sampler self)
654 CODE: 2344 CODE:
655 clReleaseSampler (this); 2345 clReleaseSampler (self);
656 2346
657void 2347void
658info (OpenCL::Sampler this, cl_sampler_info name) 2348info (OpenCL::Sampler self, cl_sampler_info name)
659 PPCODE: 2349 PPCODE:
660 INFO (Sampler) 2350 INFO (Sampler)
661 2351
2352#BEGIN:sampler
2353
2354void
2355reference_count (OpenCL::Sampler self)
2356 PPCODE:
2357 cl_uint value [1];
2358 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0));
2359 EXTEND (SP, 1);
2360 const int i = 0;
2361 PUSHs (sv_2mortal (newSVuv (value [i])));
2362
2363void
2364context (OpenCL::Sampler self)
2365 PPCODE:
2366 cl_context value [1];
2367 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2368 EXTEND (SP, 1);
2369 const int i = 0;
2370 {
2371 NEED_SUCCESS (RetainContext, (value [i]));
2372 PUSH_CLOBJ (stash_context, value [i]);
2373 }
2374
2375void
2376normalized_coords (OpenCL::Sampler self)
2377 PPCODE:
2378 cl_addressing_mode value [1];
2379 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2380 EXTEND (SP, 1);
2381 const int i = 0;
2382 PUSHs (sv_2mortal (newSVuv (value [i])));
2383
2384void
2385addressing_mode (OpenCL::Sampler self)
2386 PPCODE:
2387 cl_filter_mode value [1];
2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2389 EXTEND (SP, 1);
2390 const int i = 0;
2391 PUSHs (sv_2mortal (newSVuv (value [i])));
2392
2393void
2394filter_mode (OpenCL::Sampler self)
2395 PPCODE:
2396 cl_bool value [1];
2397 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2398 EXTEND (SP, 1);
2399 const int i = 0;
2400 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2401
2402#END:sampler
2403
662MODULE = OpenCL PACKAGE = OpenCL::Program 2404MODULE = OpenCL PACKAGE = OpenCL::Program
663 2405
664void 2406void
665DESTROY (OpenCL::Program this) 2407DESTROY (OpenCL::Program self)
666 CODE: 2408 CODE:
667 clReleaseProgram (this); 2409 clReleaseProgram (self);
668 2410
669void 2411void
670info (OpenCL::Program this, cl_program_info name) 2412build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2413 ALIAS:
2414 build_async = 1
671 PPCODE: 2415 CODE:
672 INFO (Program) 2416 cl_uint device_count = 0;
2417 cl_device_id *device_list = 0;
673 2418
2419 if (SvOK (devices))
2420 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2421
2422 void *user_data;
2423 program_callback pfn_notify = make_program_callback (notify, &user_data);
2424
2425 if (ix)
2426 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2427 else
2428 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2429
2430#if CL_VERSION_1_2
2431
674void 2432void
675build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2433compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
676 CODE: 2434 CODE:
2435 cl_uint device_count = 0;
2436 cl_device_id *device_list = 0;
2437
2438 if (SvOK (devices))
2439 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2440
2441 cl_uint header_count = 0;
2442 cl_program *header_list = 0;
2443 const char **header_name = 0;
2444
2445 if (SvOK (headers))
2446 {
2447 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2448 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2449
2450 HV *hv = (HV *)SvRV (devices);
2451
2452 header_count = hv_iterinit (hv);
2453 header_list = tmpbuf (sizeof (*header_list) * header_count);
2454 header_name = tmpbuf (sizeof (*header_name) * header_count);
2455
2456 HE *he;
2457 int i = 0;
2458 while (he = hv_iternext (hv))
2459 {
2460 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2461 header_list [i] = SvCLOBJ ("clCompileProgram", "headers", HeVAL (he), "OpenCL::Program");
2462 ++i;
2463 }
2464 }
2465
2466 void *user_data;
2467 program_callback pfn_notify = make_program_callback (notify, &user_data);
2468
677 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2469 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2470 header_count, header_list, header_name, pfn_notify, user_data));
678 2471
2472#endif
2473
679void 2474void
680build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2475build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
681 PPCODE: 2476 PPCODE:
682 size_t size; 2477 size_t size;
683 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2478 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
684 SV *sv = sv_2mortal (newSV (size)); 2479 SV *sv = sv_2mortal (newSV (size));
685 SvUPGRADE (sv, SVt_PV); 2480 SvUPGRADE (sv, SVt_PV);
686 SvPOK_only (sv); 2481 SvPOK_only (sv);
687 SvCUR_set (sv, size); 2482 SvCUR_set (sv, size);
688 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2483 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
689 XPUSHs (sv); 2484 XPUSHs (sv);
690 2485
2486#BEGIN:program_build
2487
2488void
2489build_status (OpenCL::Program self, OpenCL::Device device)
2490 PPCODE:
2491 cl_build_status value [1];
2492 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2493 EXTEND (SP, 1);
2494 const int i = 0;
2495 PUSHs (sv_2mortal (newSViv (value [i])));
2496
2497void
2498build_options (OpenCL::Program self, OpenCL::Device device)
2499 ALIAS:
2500 build_options = CL_PROGRAM_BUILD_OPTIONS
2501 build_log = CL_PROGRAM_BUILD_LOG
2502 PPCODE:
2503 size_t size;
2504 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2505 char *value = tmpbuf (size);
2506 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2507 EXTEND (SP, 1);
2508 const int i = 0;
2509 PUSHs (sv_2mortal (newSVpv (value, 0)));
2510
2511#END:program_build
2512
691void 2513void
692kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
693 PPCODE: 2515 PPCODE:
694 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
695 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
2518
2519void
2520kernels_in_program (OpenCL::Program program)
2521 PPCODE:
2522 cl_uint num_kernels;
2523 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2524 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2526
2527 int i;
2528 EXTEND (SP, num_kernels);
2529 for (i = 0; i < num_kernels; ++i)
2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
2531
2532void
2533info (OpenCL::Program self, cl_program_info name)
2534 PPCODE:
2535 INFO (Program)
2536
2537void
2538binaries (OpenCL::Program self)
2539 PPCODE:
2540 cl_uint n, i;
2541 size_t size;
2542
2543 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2544 if (!n) XSRETURN_EMPTY;
2545
2546 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2547 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2548 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2549 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2550
2551 EXTEND (SP, n);
2552 for (i = 0; i < n; ++i)
2553 {
2554 SV *sv = sv_2mortal (newSV (sizes [i]));
2555 SvUPGRADE (sv, SVt_PV);
2556 SvPOK_only (sv);
2557 SvCUR_set (sv, sizes [i]);
2558 ptrs [i] = (void *)SvPVX (sv);
2559 PUSHs (sv);
2560 }
2561
2562 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2563 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2564
2565#BEGIN:program
2566
2567void
2568reference_count (OpenCL::Program self)
2569 ALIAS:
2570 reference_count = CL_PROGRAM_REFERENCE_COUNT
2571 num_devices = CL_PROGRAM_NUM_DEVICES
2572 PPCODE:
2573 cl_uint value [1];
2574 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2575 EXTEND (SP, 1);
2576 const int i = 0;
2577 PUSHs (sv_2mortal (newSVuv (value [i])));
2578
2579void
2580context (OpenCL::Program self)
2581 PPCODE:
2582 cl_context value [1];
2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2584 EXTEND (SP, 1);
2585 const int i = 0;
2586 {
2587 NEED_SUCCESS (RetainContext, (value [i]));
2588 PUSH_CLOBJ (stash_context, value [i]);
2589 }
2590
2591void
2592devices (OpenCL::Program self)
2593 PPCODE:
2594 size_t size;
2595 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2596 cl_device_id *value = tmpbuf (size);
2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2598 int i, n = size / sizeof (*value);
2599 EXTEND (SP, n);
2600 for (i = 0; i < n; ++i)
2601 {
2602 PUSH_CLOBJ (stash_device, value [i]);
2603 }
2604
2605void
2606source (OpenCL::Program self)
2607 PPCODE:
2608 size_t size;
2609 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2610 char *value = tmpbuf (size);
2611 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2612 EXTEND (SP, 1);
2613 const int i = 0;
2614 PUSHs (sv_2mortal (newSVpv (value, 0)));
2615
2616void
2617binary_sizes (OpenCL::Program self)
2618 PPCODE:
2619 size_t size;
2620 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2621 size_t *value = tmpbuf (size);
2622 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2623 int i, n = size / sizeof (*value);
2624 EXTEND (SP, n);
2625 for (i = 0; i < n; ++i)
2626 PUSHs (sv_2mortal (newSVuv (value [i])));
2627
2628#END:program
696 2629
697MODULE = OpenCL PACKAGE = OpenCL::Kernel 2630MODULE = OpenCL PACKAGE = OpenCL::Kernel
698 2631
699void 2632void
700DESTROY (OpenCL::Kernel this) 2633DESTROY (OpenCL::Kernel self)
701 CODE: 2634 CODE:
702 clReleaseKernel (this); 2635 clReleaseKernel (self);
703 2636
704void 2637void
2638setf (OpenCL::Kernel self, const char *format, ...)
2639 CODE:
2640 int i;
2641 for (i = 2; ; ++i)
2642 {
2643 while (*format == ' ')
2644 ++format;
2645
2646 char type = *format++;
2647
2648 if (!type)
2649 break;
2650
2651 if (i >= items)
2652 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2653
2654 SV *sv = ST (i);
2655
2656 union
2657 {
2658 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2659 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2660 cl_half ch; cl_float cf; cl_double cd;
2661 cl_mem cm;
2662 cl_sampler ca;
2663 size_t cz;
2664 cl_event ce;
2665 } arg;
2666 size_t size;
2667 int nullarg = 0;
2668
2669 switch (type)
2670 {
2671 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2672 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2673 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2674 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2675 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2676 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2677 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2678 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2679
2680 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2681 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2682 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2683
2684 case 'z': nullarg = 1; size = SvIV (sv); break;
2685
2686 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2687 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2688 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2689
2690 default:
2691 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2692 }
2693
2694 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2695 if (res)
2696 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2697 }
2698
2699 if (i != items)
2700 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2701
2702void
2703set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2704 CODE:
2705 clSetKernelArg (self, idx, sizeof (value), &value);
2706
2707void
2708set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2709 CODE:
2710 clSetKernelArg (self, idx, sizeof (value), &value);
2711
2712void
2713set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2714 CODE:
2715 clSetKernelArg (self, idx, sizeof (value), &value);
2716
2717void
2718set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2719 CODE:
2720 clSetKernelArg (self, idx, sizeof (value), &value);
2721
2722void
2723set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2724 CODE:
2725 clSetKernelArg (self, idx, sizeof (value), &value);
2726
2727void
2728set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2729 CODE:
2730 clSetKernelArg (self, idx, sizeof (value), &value);
2731
2732void
2733set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2734 CODE:
2735 clSetKernelArg (self, idx, sizeof (value), &value);
2736
2737void
2738set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2739 CODE:
2740 clSetKernelArg (self, idx, sizeof (value), &value);
2741
2742void
2743set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2744 CODE:
2745 clSetKernelArg (self, idx, sizeof (value), &value);
2746
2747void
2748set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2749 CODE:
2750 clSetKernelArg (self, idx, sizeof (value), &value);
2751
2752void
2753set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2754 CODE:
2755 clSetKernelArg (self, idx, sizeof (value), &value);
2756
2757void
2758set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2759 CODE:
2760 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2761
2762void
2763set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2764 CODE:
2765 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2766
2767void
2768set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2769 ALIAS:
2770 set_image2d = 0
2771 set_image3d = 0
2772 CODE:
2773 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2774
2775void
2776set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2777 CODE:
2778 clSetKernelArg (self, idx, sizeof (value), &value);
2779
2780void
2781set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2782 CODE:
2783 clSetKernelArg (self, idx, size, 0);
2784
2785void
2786set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2787 CODE:
2788 clSetKernelArg (self, idx, sizeof (value), &value);
2789
2790void
705info (OpenCL::Kernel this, cl_kernel_info name) 2791info (OpenCL::Kernel self, cl_kernel_info name)
706 PPCODE: 2792 PPCODE:
707 INFO (Kernel) 2793 INFO (Kernel)
708 2794
2795#BEGIN:kernel
2796
709void 2797void
710set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2798function_name (OpenCL::Kernel self)
2799 PPCODE:
2800 size_t size;
2801 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2802 char *value = tmpbuf (size);
2803 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2804 EXTEND (SP, 1);
2805 const int i = 0;
2806 PUSHs (sv_2mortal (newSVpv (value, 0)));
2807
2808void
2809num_args (OpenCL::Kernel self)
2810 ALIAS:
2811 num_args = CL_KERNEL_NUM_ARGS
2812 reference_count = CL_KERNEL_REFERENCE_COUNT
2813 PPCODE:
2814 cl_uint value [1];
2815 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2816 EXTEND (SP, 1);
2817 const int i = 0;
2818 PUSHs (sv_2mortal (newSVuv (value [i])));
2819
2820void
2821context (OpenCL::Kernel self)
2822 PPCODE:
2823 cl_context value [1];
2824 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2825 EXTEND (SP, 1);
2826 const int i = 0;
2827 {
2828 NEED_SUCCESS (RetainContext, (value [i]));
2829 PUSH_CLOBJ (stash_context, value [i]);
2830 }
2831
2832void
2833program (OpenCL::Kernel self)
2834 PPCODE:
2835 cl_program value [1];
2836 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2837 EXTEND (SP, 1);
2838 const int i = 0;
2839 {
2840 NEED_SUCCESS (RetainProgram, (value [i]));
2841 PUSH_CLOBJ (stash_program, value [i]);
2842 }
2843
2844#END:kernel
2845
2846void
2847work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
711 CODE: 2848 PPCODE:
712 clSetKernelArg (this, idx, sizeof (value), &value); 2849 size_t size;
2850 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2851 SV *sv = sv_2mortal (newSV (size));
2852 SvUPGRADE (sv, SVt_PV);
2853 SvPOK_only (sv);
2854 SvCUR_set (sv, size);
2855 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2856 XPUSHs (sv);
713 2857
2858#BEGIN:kernel_work_group
2859
714void 2860void
715set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value) 2861work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2862 ALIAS:
2863 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2864 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2865 PPCODE:
2866 size_t value [1];
2867 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2868 EXTEND (SP, 1);
2869 const int i = 0;
2870 PUSHs (sv_2mortal (newSVuv (value [i])));
2871
2872void
2873compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2874 PPCODE:
2875 size_t size;
2876 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2877 size_t *value = tmpbuf (size);
2878 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2879 int i, n = size / sizeof (*value);
2880 EXTEND (SP, n);
2881 for (i = 0; i < n; ++i)
2882 PUSHs (sv_2mortal (newSVuv (value [i])));
2883
2884void
2885local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
2886 ALIAS:
2887 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2888 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2889 PPCODE:
2890 cl_ulong value [1];
2891 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2892 EXTEND (SP, 1);
2893 const int i = 0;
2894 PUSHs (sv_2mortal (newSVuv (value [i])));
2895
2896#END:kernel_work_group
2897
2898#if CL_VERSION_1_2
2899
2900void
2901arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
716 CODE: 2902 PPCODE:
717 clSetKernelArg (this, idx, sizeof (value), &value); 2903 size_t size;
2904 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2905 SV *sv = sv_2mortal (newSV (size));
2906 SvUPGRADE (sv, SVt_PV);
2907 SvPOK_only (sv);
2908 SvCUR_set (sv, size);
2909 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2910 XPUSHs (sv);
718 2911
719void 2912#BEGIN:kernel_arg
720set_short (OpenCL::Kernel this, cl_uint idx, cl_short value)
721 CODE:
722 clSetKernelArg (this, idx, sizeof (value), &value);
723 2913
724void 2914void
725set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2915arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
726 CODE: 2916 PPCODE:
727 clSetKernelArg (this, idx, sizeof (value), &value); 2917 cl_kernel_arg_address_qualifier value [1];
2918 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2919 EXTEND (SP, 1);
2920 const int i = 0;
2921 PUSHs (sv_2mortal (newSVuv (value [i])));
728 2922
729void 2923void
730set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2924arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
731 CODE: 2925 PPCODE:
732 clSetKernelArg (this, idx, sizeof (value), &value); 2926 cl_kernel_arg_access_qualifier value [1];
2927 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2928 EXTEND (SP, 1);
2929 const int i = 0;
2930 PUSHs (sv_2mortal (newSVuv (value [i])));
733 2931
734void 2932void
735set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value) 2933arg_type_name (OpenCL::Kernel self, cl_uint idx)
736 CODE: 2934 ALIAS:
737 clSetKernelArg (this, idx, sizeof (value), &value); 2935 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2936 arg_name = CL_KERNEL_ARG_NAME
2937 PPCODE:
2938 size_t size;
2939 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2940 char *value = tmpbuf (size);
2941 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2942 EXTEND (SP, 1);
2943 const int i = 0;
2944 PUSHs (sv_2mortal (newSVpv (value, 0)));
738 2945
739void 2946void
740set_long (OpenCL::Kernel this, cl_uint idx, cl_long value) 2947arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
741 CODE: 2948 PPCODE:
742 clSetKernelArg (this, idx, sizeof (value), &value); 2949 cl_kernel_arg_type_qualifier value [1];
2950 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2951 EXTEND (SP, 1);
2952 const int i = 0;
2953 PUSHs (sv_2mortal (newSVuv (value [i])));
743 2954
744void 2955#END:kernel_arg
745set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
746 CODE:
747 clSetKernelArg (this, idx, sizeof (value), &value);
748 2956
749void 2957#endif
750set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
751 CODE:
752 clSetKernelArg (this, idx, sizeof (value), &value);
753
754void
755set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
756 CODE:
757 clSetKernelArg (this, idx, sizeof (value), &value);
758
759void
760set_double (OpenCL::Kernel this, cl_uint idx, cl_double value)
761 CODE:
762 clSetKernelArg (this, idx, sizeof (value), &value);
763
764void
765set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
766 CODE:
767 clSetKernelArg (this, idx, sizeof (value), &value);
768
769void
770set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
771 CODE:
772 clSetKernelArg (this, idx, sizeof (value), &value);
773
774void
775set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
776 CODE:
777 clSetKernelArg (this, idx, sizeof (value), &value);
778
779void
780set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
781 CODE:
782 clSetKernelArg (this, idx, sizeof (value), &value);
783
784void
785set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
786 CODE:
787 clSetKernelArg (this, idx, sizeof (value), &value);
788
789void
790set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
791 CODE:
792 clSetKernelArg (this, idx, sizeof (value), &value);
793 2958
794MODULE = OpenCL PACKAGE = OpenCL::Event 2959MODULE = OpenCL PACKAGE = OpenCL::Event
795 2960
796void 2961void
797DESTROY (OpenCL::Event this) 2962DESTROY (OpenCL::Event self)
798 CODE: 2963 CODE:
799 clReleaseEvent (this); 2964 clReleaseEvent (self);
800 2965
801void 2966void
2967wait (OpenCL::Event self)
2968 CODE:
2969 clWaitForEvents (1, &self);
2970
2971void
2972cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2973 CODE:
2974 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2975
2976void
802info (OpenCL::Event this, cl_event_info name) 2977info (OpenCL::Event self, cl_event_info name)
803 PPCODE: 2978 PPCODE:
804 INFO (Event) 2979 INFO (Event)
805 2980
2981#BEGIN:event
2982
806void 2983void
807wait (OpenCL::Event this) 2984command_queue (OpenCL::Event self)
2985 PPCODE:
2986 cl_command_queue value [1];
2987 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2988 EXTEND (SP, 1);
2989 const int i = 0;
2990 {
2991 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2992 PUSH_CLOBJ (stash_queue, value [i]);
2993 }
2994
2995void
2996command_type (OpenCL::Event self)
2997 PPCODE:
2998 cl_command_type value [1];
2999 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
3000 EXTEND (SP, 1);
3001 const int i = 0;
3002 PUSHs (sv_2mortal (newSVuv (value [i])));
3003
3004void
3005reference_count (OpenCL::Event self)
3006 ALIAS:
3007 reference_count = CL_EVENT_REFERENCE_COUNT
3008 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
3009 PPCODE:
3010 cl_uint value [1];
3011 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
3012 EXTEND (SP, 1);
3013 const int i = 0;
3014 PUSHs (sv_2mortal (newSVuv (value [i])));
3015
3016void
3017context (OpenCL::Event self)
3018 PPCODE:
3019 cl_context value [1];
3020 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
3021 EXTEND (SP, 1);
3022 const int i = 0;
3023 {
3024 NEED_SUCCESS (RetainContext, (value [i]));
3025 PUSH_CLOBJ (stash_context, value [i]);
3026 }
3027
3028#END:event
3029
3030void
3031profiling_info (OpenCL::Event self, cl_profiling_info name)
808 CODE: 3032 PPCODE:
809 clWaitForEvents (1, &this); 3033 INFO (EventProfiling)
3034
3035#BEGIN:profiling
3036
3037void
3038profiling_command_queued (OpenCL::Event self)
3039 ALIAS:
3040 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
3041 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
3042 profiling_command_start = CL_PROFILING_COMMAND_START
3043 profiling_command_end = CL_PROFILING_COMMAND_END
3044 PPCODE:
3045 cl_ulong value [1];
3046 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
3047 EXTEND (SP, 1);
3048 const int i = 0;
3049 PUSHs (sv_2mortal (newSVuv (value [i])));
3050
3051#END:profiling
810 3052
811MODULE = OpenCL PACKAGE = OpenCL::UserEvent 3053MODULE = OpenCL PACKAGE = OpenCL::UserEvent
812 3054
813void 3055void
814set_status (OpenCL::UserEvent this, cl_int execution_status) 3056set_status (OpenCL::UserEvent self, cl_int execution_status)
815 CODE: 3057 CODE:
816 clSetUserEventStatus (this, execution_status); 3058 clSetUserEventStatus (self, execution_status);
817 3059
3060MODULE = OpenCL PACKAGE = OpenCL::Mapped
3061
3062void
3063DESTROY (SV *self)
3064 CODE:
3065 OpenCL__Mapped mapped = SvMAPPED (self);
3066
3067 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3068 mapped_detach (self, mapped);
3069
3070 clReleaseCommandQueue (mapped->queue);
3071 clReleaseEvent (mapped->event);
3072 Safefree (mapped);
3073
3074void
3075unmap (OpenCL::Mapped self, ...)
3076 CODE:
3077 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
3078
3079bool
3080mapped (OpenCL::Mapped self)
3081 CODE:
3082 RETVAL = !!self->ptr;
3083 OUTPUT:
3084 RETVAL
3085
3086void
3087wait (OpenCL::Mapped self)
3088 PPCODE:
3089 if (self->event)
3090 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3091
3092void
3093event (OpenCL::Mapped self)
3094 PPCODE:
3095 if (!self->event)
3096 XSRETURN_UNDEF;
3097
3098 clRetainEvent (self->event);
3099 XPUSH_CLOBJ (stash_event, self->event);
3100
3101size_t
3102size (OpenCL::Mapped self)
3103 CODE:
3104 RETVAL = self->cb;
3105 OUTPUT:
3106 RETVAL
3107
3108IV
3109ptr (OpenCL::Mapped self)
3110 CODE:
3111 RETVAL = PTR2IV (self->ptr);
3112 OUTPUT:
3113 RETVAL
3114
3115void
3116set (OpenCL::Mapped self, size_t offset, SV *data)
3117 CODE:
3118 STRLEN len;
3119 const char *ptr = SvPVbyte (data, len);
3120
3121 if (offset + len > self->cb)
3122 croak ("OpenCL::Mapped::set out of bound condition detected");
3123
3124 memcpy (offset + (char *)self->ptr, ptr, len);
3125
3126MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3127
3128MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3129
3130IV
3131row_pitch (OpenCL::Mapped self)
3132 ALIAS:
3133 slice_pitch = 1
3134 CODE:
3135 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3136 OUTPUT:
3137 RETVAL
3138
3139
3140

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines