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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.8 by root, Thu Nov 17 02:54:14 2011 UTC vs.
Revision 1.71 by root, Sat May 5 02:45:38 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;
26 60
27typedef SV *FUTURE; 61typedef struct mapped * OpenCL__Mapped;
62
63static HV
64 *stash_platform,
65 *stash_device,
66 *stash_subdevice,
67 *stash_context,
68 *stash_queue,
69 *stash_program,
70 *stash_kernel,
71 *stash_sampler,
72 *stash_event,
73 *stash_userevent,
74 *stash_memory,
75 *stash_buffer,
76 *stash_bufferobj,
77 *stash_image,
78 *stash_image1d,
79 *stash_image1darray,
80 *stash_image1dbuffer,
81 *stash_image2d,
82 *stash_image2darray,
83 *stash_image3d,
84 *stash_mapped,
85 *stash_mappedbuffer,
86 *stash_mappedimage;
28 87
29/*****************************************************************************/ 88/*****************************************************************************/
30 89
90// name must include a leading underscore
91// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
92// for perl. doh.
93static void *
94glsym (const char *name)
95{
96 void *fun = 0;
97
98 #if defined I_DLFCN && defined RTLD_DEFAULT
99 fun = dlsym (RTLD_DEFAULT, name + 1);
100 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
101
102 if (!fun)
103 {
104 static void *libgl;
105 static const char *glso[] = {
106 "libGL.so.1",
107 "libGL.so.3",
108 "libGL.so.4.0",
109 "libGL.so",
110 "/usr/lib/libGL.so",
111 "/usr/X11R6/lib/libGL.1.dylib"
112 };
113 int i;
114
115 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
116 {
117 libgl = dlopen (glso [i], RTLD_LAZY);
118 if (libgl)
119 break;
120 }
121
122 if (libgl)
123 {
124 fun = dlsym (libgl, name + 1);
125 if (!fun) fun = dlsym (libgl, name);
126 }
127 }
128 #endif
129
130 return fun;
131}
132
133/*****************************************************************************/
134
31/* up to two temporary buffers */ 135/* up to two temporary buffers */
32static void * 136static void * ecb_noinline
33tmpbuf (size_t size) 137tmpbuf (size_t size)
34{ 138{
139 enum { buffers = 4 };
35 static int idx; 140 static int idx;
36 static void *buf [2]; 141 static void *buf [buffers];
37 static size_t len [2]; 142 static size_t len [buffers];
38 143
39 idx ^= 1; 144 idx = (idx + 1) % buffers;
40 145
41 if (len [idx] < size) 146 if (len [idx] < size)
42 { 147 {
43 free (buf [idx]); 148 free (buf [idx]);
44 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 149 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
45 buf [idx] = malloc (len [idx]); 150 buf [idx] = malloc (len [idx]);
46 } 151 }
47 152
48 return buf [idx]; 153 return buf [idx];
154}
155
156static const char * ecb_noinline
157cv_get_name (CV *cv)
158{
159 static char fullname [256];
160
161 GV *gv = CvGV (cv); // gv better be nonzero
162
163 HV *stash = GvSTASH (gv);
164 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
165 const char *gvname = GvNAME (gv);
166
167 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
168 return fullname;
49} 169}
50 170
51/*****************************************************************************/ 171/*****************************************************************************/
52 172
53typedef struct 173typedef struct
112 if (res) \ 232 if (res) \
113 FAIL (name); 233 FAIL (name);
114 234
115/*****************************************************************************/ 235/*****************************************************************************/
116 236
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 * 237static SV *
238new_clobj (HV *stash, IV id)
239{
240 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
241}
242
243#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
244#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
245
246/* cl objects are either \$iv, or [$iv, ...] */
247/* they can be upgraded at runtime to the array form */
248static void * ecb_noinline
121SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 249SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
122{ 250{
251 // sv_derived_from is quite slow :(
123 if (SvROK (sv) && sv_derived_from (sv, pkg)) 252 if (SvROK (sv) && sv_derived_from (sv, pkg))
124 return (void *)SvIV (SvRV (sv)); 253 return (void *)SvIV (SvRV (sv));
125 254
126 croak ("%s: %s is not of type %s", func, svname, pkg); 255 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
256}
257
258// the "no-inherit" version of the above
259static void * ecb_noinline
260SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
261{
262 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
263 return (void *)SvIV (SvRV (sv));
264
265 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
127} 266}
128 267
129/*****************************************************************************/ 268/*****************************************************************************/
130 269
131static cl_event * 270static cl_context_properties * ecb_noinline
132event_list (SV **items, int count) 271SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
133{ 272{
273 if (!sv || !SvOK (sv))
274 if (extra)
275 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
276 else
277 return 0;
278
279 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
280 {
281 AV *av = (AV *)SvRV (sv);
282 int i, len = av_len (av) + 1;
283 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
284 cl_context_properties *l = p;
285
286 if (len & 1)
287 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
288
289 while (extracount--)
290 *l++ = *extra++;
291
292 for (i = 0; i < len; i += 2)
293 {
294 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
295 SV *p_sv = *av_fetch (av, i + 1, 0);
296 cl_context_properties v = SvIV (p_sv); // code below can override
297
298 switch (t)
299 {
300 case CL_CONTEXT_PLATFORM:
301 if (SvROK (p_sv))
302 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
303 break;
304
305 case CL_GLX_DISPLAY_KHR:
306 if (!SvOK (p_sv))
307 {
308 void *func = glsym ("_glXGetCurrentDisplay");
309 if (func)
310 v = (cl_context_properties)((void *(*)(void))func)();
311 }
312 break;
313
314 case CL_GL_CONTEXT_KHR:
315 if (!SvOK (p_sv))
316 {
317 void *func = glsym ("_glXGetCurrentContext");
318 if (func)
319 v = (cl_context_properties)((void *(*)(void))func)();
320 }
321 break;
322
323 default:
324 /* unknown property, treat as int */
325 break;
326 }
327
328 *l++ = t;
329 *l++ = v;
330 }
331
332 *l = 0;
333
334 return p;
335 }
336
337 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
338}
339
340// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
341// is a pointer (and only on well-behaved systems).
342static void * ecb_noinline
343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
344{
345 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
346 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
347 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
348
349 AV *av = (AV *)SvRV (arg);
350 void **list = 0;
351 cl_uint count = av_len (av) + 1;
352
353 if (count)
354 {
355 list = tmpbuf (sizeof (*list) * count);
356 int i;
357 for (i = 0; i < count; ++i)
358 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
359 }
360
361 if (!count && !or_undef)
362 croak ("%s: '%s' must contain at least one %s object",
363 cv_get_name (cv), argname, klass);
364
365 *rcount = count;
366 return (void *)list;
367}
368
369/*****************************************************************************/
370/* callback stuff */
371
372/* default context callback, log to stderr */
373static void CL_CALLBACK
374context_default_notify (const char *msg, const void *info, size_t cb, void *data)
375{
376 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
377}
378
379typedef struct
380{
381 int free_cb;
382 void (*push)(void *data1, void *data2, void *data3);
383} eq_vtbl;
384
385typedef struct eq_item
386{
387 struct eq_item *next;
388 eq_vtbl *vtbl;
389 SV *cb;
390 void *data1, *data2, *data3;
391} eq_item;
392
393static void (*eq_signal_func)(void *signal_arg, int value);
394static void *eq_signal_arg;
395static xmutex_t eq_lock = X_MUTEX_INIT;
396static eq_item *eq_head, *eq_tail;
397
398static void ecb_noinline
399eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
400{
401 eq_item *item = malloc (sizeof (eq_item));
402
403 item->next = 0;
404 item->vtbl = vtbl;
405 item->cb = cb;
406 item->data1 = data1;
407 item->data2 = data2;
408 item->data3 = data3;
409
410 X_LOCK (eq_lock);
411
412 *(eq_head ? &eq_tail->next : &eq_head) = item;
413 eq_tail = item;
414
415 X_UNLOCK (eq_lock);
416
417 eq_signal_func (eq_signal_arg, 0);
418}
419
420static eq_item *
421eq_dec (void)
422{
423 eq_item *res;
424
425 X_LOCK (eq_lock);
426
427 res = eq_head;
428 if (res)
429 eq_head = res->next;
430
431 X_UNLOCK (eq_lock);
432
433 return res;
434}
435
436static void
437eq_poll (void)
438{
439 eq_item *item;
440
441 while ((item = eq_dec ()))
442 {
443 ENTER;
444 SAVETMPS;
445
446 dSP;
447 PUSHMARK (SP);
448 EXTEND (SP, 2);
449
450 if (item->vtbl->free_cb)
451 sv_2mortal (item->cb);
452
453 PUTBACK;
454 item->vtbl->push (item->data1, item->data2, item->data3);
455
456 SV *cb = item->cb;
457 free (item);
458
459 call_sv (cb, G_DISCARD | G_VOID);
460
461 FREETMPS;
462 LEAVE;
463 }
464}
465
466static void
467eq_poll_interrupt (pTHX_ void *c_arg, int value)
468{
469 eq_poll ();
470}
471
472/*****************************************************************************/
473/* context notify */
474
475static void ecb_noinline
476eq_context_push (void *data1, void *data2, void *data3)
477{
478 dSP;
479 PUSHs (sv_2mortal (newSVpv (data1, 0)));
480 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
481 PUTBACK;
482
483 free (data1);
484 free (data2);
485}
486
487static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
488
489static void CL_CALLBACK
490eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
491{
492 void *pvt_copy = malloc (cb);
493 memcpy (pvt_copy, pvt, cb);
494 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
495}
496
497#define CONTEXT_NOTIFY_CALLBACK \
498 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
499 void *user_data = 0; \
500 \
501 if (SvOK (notify)) \
502 { \
503 pfn_notify = eq_context_notify; \
504 user_data = s_get_cv (notify); \
505 }
506
507static SV * ecb_noinline
508new_clobj_context (cl_context ctx, void *user_data)
509{
510 SV *sv = new_clobj (stash_context, (IV)ctx);
511
512 if (user_data)
513 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
514
515 return sv;
516}
517
518#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
519
520/*****************************************************************************/
521/* build/compile/link notify */
522
523static void
524eq_program_push (void *data1, void *data2, void *data3)
525{
526 dSP;
527 PUSH_CLOBJ (stash_program, data1);
528 PUTBACK;
529}
530
531static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
532
533static void CL_CALLBACK
534eq_program_notify (cl_program program, void *user_data)
535{
536 clRetainProgram (program);
537
538 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
539}
540
541typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
542
543static program_callback ecb_noinline
544make_program_callback (SV *notify, void **ruser_data)
545{
546 if (SvOK (notify))
547 {
548 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
549 return eq_program_notify;
550 }
551 else
552 {
553 *ruser_data = 0;
554 return 0;
555 }
556}
557
558struct build_args
559{
560 cl_program program;
561 char *options;
562 void *user_data;
563 cl_uint num_devices;
564};
565
566X_THREAD_PROC (build_program_thread)
567{
568 struct build_args *arg = thr_arg;
569
570 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
571
572 if (arg->user_data)
573 eq_program_notify (arg->program, arg->user_data);
574 else
575 clReleaseProgram (arg->program);
576
577 free (arg->options);
578 free (arg);
579
580 return 0;
581}
582
583static void
584build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
585{
586 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
587
588 arg->program = program;
589 arg->options = strdup (options);
590 arg->user_data = user_data;
591 arg->num_devices = num_devices;
592 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
593
594 xthread_t id;
595 thread_create (&id, build_program_thread, arg);
596}
597
598/*****************************************************************************/
599/* event objects */
600
601static void
602eq_event_push (void *data1, void *data2, void *data3)
603{
604 dSP;
605 PUSH_CLOBJ (stash_event, data1);
606 PUSHs (sv_2mortal (newSViv ((IV)data2)));
607 PUTBACK;
608}
609
610static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
611
612static void CL_CALLBACK
613eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
614{
615 clRetainEvent (event);
616 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
617}
618
619/*****************************************************************************/
620/* utilities for XS code */
621
622static size_t
623img_row_pitch (cl_mem img)
624{
625 size_t res;
626 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
627 return res;
628}
629
630static cl_event * ecb_noinline
631event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
632{
633 cl_uint count = *rcount;
634
635 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
636 *rcount = count = 0;
637
638 if (!count && !extra)
639 return 0;
640
134 cl_event *list = tmpbuf (sizeof (cl_event) * count); 641 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
642 int i = 0;
135 643
136 while (count--) 644 while (count--)
645 if (SvOK (items [count]))
137 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 646 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
138 647
648 if (extra)
649 list [i++] = extra;
650
651 *rcount = i;
652
139 return list; 653 return i ? list : 0;
140} 654}
141 655
142#define EVENT_LIST(items,count) \ 656#define EVENT_LIST(skip) \
143 cl_uint event_list_count = (count); \ 657 cl_uint event_list_count = items - (skip); \
144 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 658 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
145 659
146#define INFO(class) \ 660#define INFO(class) \
147{ \ 661{ \
148 size_t size; \ 662 size_t size; \
149 SV *sv; \
150 \
151 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 663 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
152 sv = sv_2mortal (newSV (size)); \ 664 SV *sv = sv_2mortal (newSV (size)); \
153 SvUPGRADE (sv, SVt_PV); \ 665 SvUPGRADE (sv, SVt_PV); \
154 SvPOK_only (sv); \ 666 SvPOK_only (sv); \
155 SvCUR_set (sv, size); \ 667 SvCUR_set (sv, size); \
156 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 668 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
157 XPUSHs (sv); \ 669 XPUSHs (sv); \
158} 670}
159 671
672/*****************************************************************************/
673/* mapped_xxx */
674
675static OpenCL__Mapped
676SvMAPPED (SV *sv)
677{
678 // no typechecking atm., keep your fingers crossed
679 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
680}
681
682struct mapped
683{
684 cl_command_queue queue;
685 cl_mem memobj;
686 void *ptr;
687 size_t cb;
688 cl_event event;
689 size_t row_pitch;
690 size_t slice_pitch;
691};
692
693static SV *
694mapped_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)
695{
696 SV *data = newSV (0);
697 SvUPGRADE (data, SVt_PVMG);
698
699 OpenCL__Mapped mapped;
700 New (0, mapped, 1, struct mapped);
701
702 clRetainCommandQueue (queue);
703
704 mapped->queue = queue;
705 mapped->memobj = memobj;
706 mapped->ptr = ptr;
707 mapped->cb = cb;
708 mapped->event = ev;
709 mapped->row_pitch = row_pitch;
710 mapped->slice_pitch = slice_pitch;
711
712 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
713
714 if (SvLEN (data))
715 Safefree (data);
716
717 SvPVX (data) = (char *)ptr;
718 SvCUR_set (data, cb);
719 SvLEN_set (data, 0);
720 SvPOK_only (data);
721
722 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
723
724 if (!(flags & CL_MAP_WRITE))
725 SvREADONLY_on (data);
726
727 return obj;
728}
729
730static void
731mapped_detach (SV *sv, OpenCL__Mapped mapped)
732{
733 SV *data = SvRV (sv);
734
735 // the next check checks both before AND after detach, where SvPVX should STILL be 0
736 if (SvPVX (data) != (char *)mapped->ptr)
737 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
738 else
739 {
740 SvREADONLY_off (data);
741 SvCUR_set (data, 0);
742 SvPVX (data) = 0;
743 SvOK_off (data);
744 }
745
746 mapped->ptr = 0;
747}
748
749static void
750mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
751{
752 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
753 cl_event ev;
754
755 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
756
757 clReleaseEvent (mapped->event);
758 mapped->event = ev;
759
760 mapped_detach (self, mapped);
761}
762
763/*****************************************************************************/
764
160MODULE = OpenCL PACKAGE = OpenCL 765MODULE = OpenCL PACKAGE = OpenCL
161 766
162PROTOTYPES: ENABLE 767PROTOTYPES: ENABLE
163 768
769void
770poll ()
771 CODE:
772 eq_poll ();
773
774void
775_eq_initialise (IV func, IV arg)
776 CODE:
777 eq_signal_func = (void (*)(void *, int))func;
778 eq_signal_arg = (void*)arg;
779
164BOOT: 780BOOT:
165{ 781{
166 HV *stash = gv_stashpv ("OpenCL", 1); 782 HV *stash = gv_stashpv ("OpenCL", 1);
167 static const ivstr *civ, const_iv[] = { 783 static const ivstr *civ, const_iv[] = {
168 { sizeof (cl_char ), "SIZEOF_CHAR" }, 784 { sizeof (cl_char ), "SIZEOF_CHAR" },
169 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 785 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
170 { sizeof (cl_short ), "SIZEOF_SHORT" }, 786 { sizeof (cl_short ), "SIZEOF_SHORT" },
171 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 787 { sizeof (cl_ushort), "SIZEOF_USHORT" },
172 { sizeof (cl_int ), "SIZEOF_INT" }, 788 { sizeof (cl_int ), "SIZEOF_INT" },
173 { sizeof (cl_uint ), "SIZEOF_UINT" }, 789 { sizeof (cl_uint ), "SIZEOF_UINT" },
174 { sizeof (cl_long ), "SIZEOF_LONG" }, 790 { sizeof (cl_long ), "SIZEOF_LONG" },
175 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 791 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
176 { sizeof (cl_half ), "SIZEOF_HALF" }, 792 { sizeof (cl_half ), "SIZEOF_HALF" },
177 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 793 { sizeof (cl_float ), "SIZEOF_FLOAT" },
178 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 794 { sizeof (cl_double), "SIZEOF_DOUBLE" },
179#include "constiv.h" 795#include "constiv.h"
180 }; 796 };
797
181 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 798 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
182 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 799 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
800
801 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
802 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
803 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
804 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
805 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
806 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
807 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
808 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
809 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
810 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
811 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
812 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
813 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
814 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
815 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
816 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
817 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
818 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
819 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
820 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
821 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
822 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
823 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
824
825 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
183} 826}
184 827
185cl_int 828cl_int
186errno () 829errno ()
187 CODE: 830 CODE:
188 errno = res; 831 RETVAL = res;
832 OUTPUT:
833 RETVAL
189 834
190const char * 835const char *
191err2str (cl_int err) 836err2str (cl_int err = res)
192 837
193const char * 838const char *
194enum2str (cl_uint value) 839enum2str (cl_uint value)
195 840
196void 841void
197platforms () 842platforms ()
198 PPCODE: 843 PPCODE:
199{
200 cl_platform_id *list; 844 cl_platform_id *list;
201 cl_uint count; 845 cl_uint count;
202 int i; 846 int i;
203 847
204 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count)); 848 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count));
205 list = tmpbuf (sizeof (*list) * count); 849 list = tmpbuf (sizeof (*list) * count);
206 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 850 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
207 851
208 EXTEND (SP, count); 852 EXTEND (SP, count);
209 for (i = 0; i < count; ++i) 853 for (i = 0; i < count; ++i)
210 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 854 PUSH_CLOBJ (stash_platform, list [i]);
211}
212 855
213void 856void
214context_from_type (FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 857context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
215 PPCODE: 858 PPCODE:
859 CONTEXT_NOTIFY_CALLBACK;
216 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (0, type, 0, 0, &res)); 860 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
217 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 861 XPUSH_CLOBJ_CONTEXT;
218 862
219void 863void
220context (FUTURE properties, FUTURE devices, FUTURE notify = 0) 864context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
221 PPCODE: 865 PPCODE:
222 /* der Gipfel der Kunst */ 866 cl_uint device_count;
867 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
868
869 CONTEXT_NOTIFY_CALLBACK;
870 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
871 XPUSH_CLOBJ_CONTEXT;
223 872
224void 873void
225wait_for_events (...) 874wait_for_events (...)
226 CODE: 875 CODE:
227{
228 EVENT_LIST (0, items); 876 EVENT_LIST (0);
229 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 877 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
230}
231 878
232PROTOTYPES: DISABLE 879PROTOTYPES: DISABLE
233 880
234MODULE = OpenCL PACKAGE = OpenCL::Platform 881MODULE = OpenCL PACKAGE = OpenCL::Platform
235 882
236void 883void
237info (OpenCL::Platform this, cl_platform_info name) 884info (OpenCL::Platform self, cl_platform_info name)
238 PPCODE: 885 PPCODE:
239 INFO (Platform) 886 INFO (Platform)
240 887
241void 888void
889unload_compiler (OpenCL::Platform self)
890 CODE:
891#if CL_VERSION_1_2
892 clUnloadPlatformCompiler (self);
893#endif
894
895#BEGIN:platform
896
897void
898profile (OpenCL::Platform self)
899 ALIAS:
900 profile = CL_PLATFORM_PROFILE
901 version = CL_PLATFORM_VERSION
902 name = CL_PLATFORM_NAME
903 vendor = CL_PLATFORM_VENDOR
904 extensions = CL_PLATFORM_EXTENSIONS
905 PPCODE:
906 size_t size;
907 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
908 char *value = tmpbuf (size);
909 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
910 EXTEND (SP, 1);
911 const int i = 0;
912 PUSHs (sv_2mortal (newSVpv (value, 0)));
913
914#END:platform
915
916void
242devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 917devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
243 PPCODE: 918 PPCODE:
244{
245 cl_device_id *list; 919 cl_device_id *list;
246 cl_uint count; 920 cl_uint count;
247 int i; 921 int i;
248 922
249 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 923 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
250 list = tmpbuf (sizeof (*list) * count); 924 list = tmpbuf (sizeof (*list) * count);
251 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 925 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
252 926
253 EXTEND (SP, count); 927 EXTEND (SP, count);
254 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
255 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 929 PUSH_CLOBJ (stash_device, list [i]);
256}
257 930
258void 931void
259context (OpenCL::Platform this, FUTURE properties, SV *devices, FUTURE notify = 0) 932context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
260 PPCODE: 933 PPCODE:
934 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
935 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
936
937 cl_uint device_count;
938 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
939
940 CONTEXT_NOTIFY_CALLBACK;
941 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
942 XPUSH_CLOBJ_CONTEXT;
943
944void
945context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
946 PPCODE:
947 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
948 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
949
950 CONTEXT_NOTIFY_CALLBACK;
951 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
952 XPUSH_CLOBJ_CONTEXT;
953
954MODULE = OpenCL PACKAGE = OpenCL::Device
955
956void
957info (OpenCL::Device self, cl_device_info name)
958 PPCODE:
959 INFO (Device)
960
961#if CL_VERSION_1_2
962
963void
964sub_devices (OpenCL::Device self, SV *properties)
965 PPCODE:
261 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 966 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
262 croak ("OpenCL::Platform argument 'device' must be an arrayref with device objects, in call"); 967 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
263 968
264 AV *av = (SV *)SvRV (devices); 969 properties = SvRV (properties);
265 cl_uint num_devices = av_len (av) + 1; 970
266 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 971 cl_uint count = av_len ((AV *)properties) + 1;
972 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
973
267 int i; 974 int i;
975 for (i = 0; i < count; ++i)
976 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
268 977
269 for (i = num_devices; i--; ) 978 props [count] = 0;
270 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
271 979
272 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (0, num_devices, device_list, 0, 0, &res)); 980 cl_uint num_devices;
273 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 981 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
982 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
983 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
274 984
985 EXTEND (SP, num_devices);
986 for (i = 0; i < count; ++i)
987 PUSH_CLOBJ (stash_subdevice, list [i]);
988
989#endif
990
991#BEGIN:device
992
275void 993void
276context_from_type (OpenCL::Platform this, FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 994type (OpenCL::Device self)
277 PPCODE: 995 PPCODE:
278 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 996 cl_device_type value [1];
279 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 997 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
280 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 998 EXTEND (SP, 1);
999 const int i = 0;
1000 PUSHs (sv_2mortal (newSVuv (value [i])));
281 1001
1002void
1003vendor_id (OpenCL::Device self)
1004 ALIAS:
1005 vendor_id = CL_DEVICE_VENDOR_ID
1006 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
1007 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
1008 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
1009 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
1010 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
1011 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
1012 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
1013 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
1014 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
1015 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
1016 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
1017 image_support = CL_DEVICE_IMAGE_SUPPORT
1018 max_samplers = CL_DEVICE_MAX_SAMPLERS
1019 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
1020 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1021 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
1022 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1023 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
1024 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
1025 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
1026 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
1027 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
1028 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
1029 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
1030 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
1031 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
1032 PPCODE:
1033 cl_uint value [1];
1034 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1035 EXTEND (SP, 1);
1036 const int i = 0;
1037 PUSHs (sv_2mortal (newSVuv (value [i])));
1038
1039void
1040max_work_group_size (OpenCL::Device self)
1041 ALIAS:
1042 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
1043 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
1044 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
1045 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
1046 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
1047 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
1048 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
1049 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
1050 PPCODE:
1051 size_t value [1];
1052 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1053 EXTEND (SP, 1);
1054 const int i = 0;
1055 PUSHs (sv_2mortal (newSVuv (value [i])));
1056
1057void
1058max_work_item_sizes (OpenCL::Device self)
1059 PPCODE:
1060 size_t size;
1061 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
1062 size_t *value = tmpbuf (size);
1063 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
1064 int i, n = size / sizeof (*value);
1065 EXTEND (SP, n);
1066 for (i = 0; i < n; ++i)
1067 PUSHs (sv_2mortal (newSVuv (value [i])));
1068
1069void
1070address_bits (OpenCL::Device self)
1071 PPCODE:
1072 cl_bitfield value [1];
1073 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
1074 EXTEND (SP, 1);
1075 const int i = 0;
1076 PUSHs (sv_2mortal (newSVuv (value [i])));
1077
1078void
1079max_mem_alloc_size (OpenCL::Device self)
1080 ALIAS:
1081 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1082 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1083 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1084 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1085 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1086 PPCODE:
1087 cl_ulong value [1];
1088 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1089 EXTEND (SP, 1);
1090 const int i = 0;
1091 PUSHs (sv_2mortal (newSVuv (value [i])));
1092
1093void
1094single_fp_config (OpenCL::Device self)
1095 ALIAS:
1096 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1097 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1098 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
1099 PPCODE:
1100 cl_device_fp_config value [1];
1101 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1102 EXTEND (SP, 1);
1103 const int i = 0;
1104 PUSHs (sv_2mortal (newSVuv (value [i])));
1105
1106void
1107global_mem_cache_type (OpenCL::Device self)
1108 PPCODE:
1109 cl_device_mem_cache_type value [1];
1110 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
1111 EXTEND (SP, 1);
1112 const int i = 0;
1113 PUSHs (sv_2mortal (newSVuv (value [i])));
1114
1115void
1116local_mem_type (OpenCL::Device self)
1117 PPCODE:
1118 cl_device_local_mem_type value [1];
1119 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
1120 EXTEND (SP, 1);
1121 const int i = 0;
1122 PUSHs (sv_2mortal (newSVuv (value [i])));
1123
1124void
1125error_correction_support (OpenCL::Device self)
1126 ALIAS:
1127 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1128 endian_little = CL_DEVICE_ENDIAN_LITTLE
1129 available = CL_DEVICE_AVAILABLE
1130 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1131 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1132 PPCODE:
1133 cl_bool value [1];
1134 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1135 EXTEND (SP, 1);
1136 const int i = 0;
1137 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1138
1139void
1140execution_capabilities (OpenCL::Device self)
1141 PPCODE:
1142 cl_device_exec_capabilities value [1];
1143 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1144 EXTEND (SP, 1);
1145 const int i = 0;
1146 PUSHs (sv_2mortal (newSVuv (value [i])));
1147
1148void
1149properties (OpenCL::Device self)
1150 PPCODE:
1151 cl_command_queue_properties value [1];
1152 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1153 EXTEND (SP, 1);
1154 const int i = 0;
1155 PUSHs (sv_2mortal (newSVuv (value [i])));
1156
1157void
1158platform (OpenCL::Device self)
1159 PPCODE:
1160 cl_platform_id value [1];
1161 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1162 EXTEND (SP, 1);
1163 const int i = 0;
1164 {
1165 PUSH_CLOBJ (stash_platform, value [i]);
1166 }
1167
1168void
1169name (OpenCL::Device self)
1170 ALIAS:
1171 name = CL_DEVICE_NAME
1172 vendor = CL_DEVICE_VENDOR
1173 driver_version = CL_DRIVER_VERSION
1174 profile = CL_DEVICE_PROFILE
1175 version = CL_DEVICE_VERSION
1176 extensions = CL_DEVICE_EXTENSIONS
1177 PPCODE:
1178 size_t size;
1179 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1180 char *value = tmpbuf (size);
1181 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1182 EXTEND (SP, 1);
1183 const int i = 0;
1184 PUSHs (sv_2mortal (newSVpv (value, 0)));
1185
1186void
1187parent_device_ext (OpenCL::Device self)
1188 PPCODE:
1189 cl_device_id value [1];
1190 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1191 EXTEND (SP, 1);
1192 const int i = 0;
1193 {
1194 PUSH_CLOBJ (stash_device, value [i]);
1195 }
1196
1197void
1198partition_types_ext (OpenCL::Device self)
1199 ALIAS:
1200 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1201 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
1202 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
1203 PPCODE:
1204 size_t size;
1205 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1206 cl_device_partition_property_ext *value = tmpbuf (size);
1207 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1208 int i, n = size / sizeof (*value);
1209 EXTEND (SP, n);
1210 for (i = 0; i < n; ++i)
1211 PUSHs (sv_2mortal (newSVuv (value [i])));
1212
1213#END:device
1214
282MODULE = OpenCL PACKAGE = OpenCL::Device 1215MODULE = OpenCL PACKAGE = OpenCL::SubDevice
283 1216
1217#if CL_VERSION_1_2
1218
284void 1219void
285info (OpenCL::Device this, cl_device_info name) 1220DESTROY (OpenCL::SubDevice self)
286 PPCODE: 1221 CODE:
287 INFO (Device) 1222 clReleaseDevice (self);
1223
1224#endif
288 1225
289MODULE = OpenCL PACKAGE = OpenCL::Context 1226MODULE = OpenCL PACKAGE = OpenCL::Context
290 1227
291void 1228void
292DESTROY (OpenCL::Context context) 1229DESTROY (OpenCL::Context self)
293 CODE: 1230 CODE:
294 clReleaseContext (context); 1231 clReleaseContext (self);
295 1232
296void 1233void
297info (OpenCL::Context this, cl_context_info name) 1234info (OpenCL::Context self, cl_context_info name)
298 PPCODE: 1235 PPCODE:
299 INFO (Context) 1236 INFO (Context)
300 1237
301void 1238void
302queue (OpenCL::Context this, OpenCL::Device device, cl_command_queue_properties properties = 0) 1239queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
303 PPCODE: 1240 PPCODE:
304 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (this, device, properties, &res)); 1241 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
305 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1242 XPUSH_CLOBJ (stash_queue, queue);
306 1243
307void 1244void
308user_event (OpenCL::Context this) 1245user_event (OpenCL::Context self)
309 PPCODE: 1246 PPCODE:
310 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (this, &res)); 1247 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
311 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1248 XPUSH_CLOBJ (stash_userevent, ev);
312 1249
313void 1250void
314buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1251buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
315 PPCODE: 1252 PPCODE:
316 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1253 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?"); 1254 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
318 1255
319 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (this, flags, len, 0, &res)); 1256 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
320 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1257 XPUSH_CLOBJ (stash_bufferobj, mem);
321 1258
322void 1259void
323buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1260buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
324 PPCODE: 1261 PPCODE:
325 STRLEN len; 1262 STRLEN len;
326 char *ptr = SvPVbyte (data, len); 1263 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
327
328 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1264 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?"); 1265 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)); 1266 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
332 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1267 XPUSH_CLOBJ (stash_bufferobj, mem);
333 1268
1269#if CL_VERSION_1_2
1270
334void 1271void
335image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, SV *data) 1272image (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: 1273 PPCODE:
337 STRLEN len; 1274 STRLEN len;
338 char *ptr = SvPVbyte (data, len); 1275 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
339 const cl_image_format format = { channel_order, channel_type }; 1276 const cl_image_format format = { channel_order, channel_type };
1277 const cl_image_desc desc = {
1278 type,
1279 width, height, depth,
1280 array_size, row_pitch, slice_pitch,
1281 num_mip_level, num_samples,
1282 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1283 };
340 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (this, flags, &format, width, height, len / height, ptr, &res)); 1284 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
341 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1285 HV *stash = stash_image;
1286 switch (type)
1287 {
1288 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1289 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1290 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1291 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1292 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1293 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1294 }
1295 XPUSH_CLOBJ (stash, mem);
342 1296
1297#endif
1298
343void 1299void
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 slice_pitch, SV *data) 1300image2d (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: 1301 PPCODE:
346 STRLEN len; 1302 STRLEN len;
347 char *ptr = SvPVbyte (data, len); 1303 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
348 const cl_image_format format = { channel_order, channel_type }; 1304 const cl_image_format format = { channel_order, channel_type };
1305#if PREFER_1_1
349 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (this, flags, &format, width, height, 1306 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
350 depth, len / (height * slice_pitch), slice_pitch, ptr, &res)); 1307#else
351 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1308 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1309 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1310#endif
1311 XPUSH_CLOBJ (stash_image2d, mem);
352 1312
353void 1313void
1314image3d (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)
1315 PPCODE:
1316 STRLEN len;
1317 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1318 const cl_image_format format = { channel_order, channel_type };
1319#if PREFER_1_1
1320 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1321#else
1322 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1323 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1324#endif
1325 XPUSH_CLOBJ (stash_image3d, mem);
1326
1327#if cl_apple_gl_sharing || cl_khr_gl_sharing
1328
1329void
1330gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1331 PPCODE:
1332 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1333 XPUSH_CLOBJ (stash_bufferobj, mem);
1334
1335void
1336gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1337 PPCODE:
1338 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1339 XPUSH_CLOBJ (stash_image2d, mem);
1340
1341#if CL_VERSION_1_2
1342
1343void
1344gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1345 ALIAS:
1346 PPCODE:
1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1348 cl_gl_object_type type;
1349 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1350 HV *stash = stash_memory;
1351 switch (type)
1352 {
1353 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1354 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1355 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1356 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1357 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1358 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1359 }
1360 XPUSH_CLOBJ (stash, mem);
1361
1362#endif
1363
1364void
1365gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1366 PPCODE:
1367#if PREFER_1_1
1368 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1369#else
1370 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1371#endif
1372 XPUSH_CLOBJ (stash_image2d, mem);
1373
1374void
1375gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1376 PPCODE:
1377#if PREFER_1_1
1378 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1379#else
1380 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1381#endif
1382 XPUSH_CLOBJ (stash_image3d, mem);
1383
1384#endif
1385
1386void
354supported_image_formats (OpenCL::Context this, cl_mem_flags flags, cl_mem_object_type image_type) 1387supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
355 PPCODE: 1388 PPCODE:
356{ 1389{
357 cl_uint count; 1390 cl_uint count;
358 cl_image_format *list; 1391 cl_image_format *list;
359 int i; 1392 int i;
360 1393
361 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1394 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
362 Newx (list, count, cl_image_format); 1395 Newx (list, count, cl_image_format);
363 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1396 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
364 1397
365 EXTEND (SP, count); 1398 EXTEND (SP, count);
366 for (i = 0; i < count; ++i) 1399 for (i = 0; i < count; ++i)
367 { 1400 {
368 AV *av = newAV (); 1401 AV *av = newAV ();
371 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1404 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
372 } 1405 }
373} 1406}
374 1407
375void 1408void
376sampler (OpenCL::Context this, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1409sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
377 PPCODE: 1410 PPCODE:
378 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (this, normalized_coords, addressing_mode, filter_mode, &res)); 1411 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
379 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1412 XPUSH_CLOBJ (stash_sampler, sampler);
380 1413
381void 1414void
382program_with_source (OpenCL::Context this, SV *program) 1415program_with_source (OpenCL::Context self, SV *program)
383 PPCODE: 1416 PPCODE:
384 STRLEN len; 1417 STRLEN len;
385 size_t len2; 1418 size_t len2;
386 const char *ptr = SvPVbyte (program, len); 1419 const char *ptr = SvPVbyte (program, len);
387 1420
388 len2 = len; 1421 len2 = len;
389 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (this, 1, &ptr, &len2, &res)); 1422 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
390 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1423 XPUSH_CLOBJ (stash_program, prog);
1424
1425void
1426program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1427 PPCODE:
1428 cl_uint device_count;
1429 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1430
1431 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1432 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1433
1434 binaries = SvRV (binaries);
1435
1436 if (device_count != av_len ((AV *)binaries) + 1)
1437 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1438
1439 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1440 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1441 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1442
1443 int i;
1444 for (i = 0; i < device_count; ++i)
1445 {
1446 STRLEN len;
1447 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1448 length_list [i] = len;
1449 }
1450
1451 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1452 length_list, binary_list,
1453 GIMME_V == G_ARRAY ? status_list : 0, &res));
1454
1455 EXTEND (SP, 2);
1456 PUSH_CLOBJ (stash_program, prog);
1457
1458 if (GIMME_V == G_ARRAY)
1459 {
1460 AV *av = newAV ();
1461 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1462
1463 for (i = device_count; i--; )
1464 av_store (av, i, newSViv (status_list [i]));
1465 }
1466
1467#if CL_VERSION_1_2
1468
1469void
1470program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1471 PPCODE:
1472 cl_uint device_count;
1473 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1474
1475 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1476
1477 XPUSH_CLOBJ (stash_program, prog);
1478
1479void
1480link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1481 CODE:
1482 cl_uint device_count = 0;
1483 cl_device_id *device_list = 0;
1484
1485 if (SvOK (devices))
1486 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 cl_uint program_count;
1489 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1490
1491 void *user_data;
1492 program_callback pfn_notify = make_program_callback (notify, &user_data);
1493
1494 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1495 program_count, program_list, pfn_notify, user_data, &res));
1496
1497 XPUSH_CLOBJ (stash_program, prog);
1498
1499#endif
1500
1501#BEGIN:context
1502
1503void
1504reference_count (OpenCL::Context self)
1505 ALIAS:
1506 reference_count = CL_CONTEXT_REFERENCE_COUNT
1507 num_devices = CL_CONTEXT_NUM_DEVICES
1508 PPCODE:
1509 cl_uint value [1];
1510 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1511 EXTEND (SP, 1);
1512 const int i = 0;
1513 PUSHs (sv_2mortal (newSVuv (value [i])));
1514
1515void
1516devices (OpenCL::Context self)
1517 PPCODE:
1518 size_t size;
1519 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1520 cl_device_id *value = tmpbuf (size);
1521 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1522 int i, n = size / sizeof (*value);
1523 EXTEND (SP, n);
1524 for (i = 0; i < n; ++i)
1525 {
1526 PUSH_CLOBJ (stash_device, value [i]);
1527 }
1528
1529void
1530properties (OpenCL::Context self)
1531 PPCODE:
1532 size_t size;
1533 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1534 cl_context_properties *value = tmpbuf (size);
1535 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1536 int i, n = size / sizeof (*value);
1537 EXTEND (SP, n);
1538 for (i = 0; i < n; ++i)
1539 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1540
1541#END:context
391 1542
392MODULE = OpenCL PACKAGE = OpenCL::Queue 1543MODULE = OpenCL PACKAGE = OpenCL::Queue
393 1544
394void 1545void
395DESTROY (OpenCL::Queue this) 1546DESTROY (OpenCL::Queue self)
396 CODE: 1547 CODE:
397 clReleaseCommandQueue (this); 1548 clReleaseCommandQueue (self);
398 1549
399void 1550void
400info (OpenCL::Queue this, cl_command_queue_info name)
401 PPCODE:
402 INFO (CommandQueue)
403
404void
405enqueue_read_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1551read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1552 ALIAS:
1553 enqueue_read_buffer = 0
406 PPCODE: 1554 PPCODE:
407{
408 cl_event ev = 0;
409 EVENT_LIST (6, items - 6); 1555 EVENT_LIST (6);
410 1556
411 SvUPGRADE (data, SVt_PV); 1557 SvUPGRADE (data, SVt_PV);
412 SvGROW (data, len); 1558 SvGROW (data, len);
413 SvPOK_only (data); 1559 SvPOK_only (data);
414 SvCUR_set (data, len); 1560 SvCUR_set (data, len);
1561
1562 cl_event ev = 0;
415 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1563 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
416 1564
417 if (ev) 1565 if (ev)
418 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1566 XPUSH_CLOBJ (stash_event, ev);
419}
420 1567
421void 1568void
422enqueue_write_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1569write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1570 ALIAS:
1571 enqueue_write_buffer = 0
423 PPCODE: 1572 PPCODE:
424{ 1573 EVENT_LIST (5);
425 cl_event ev = 0; 1574
426 STRLEN len; 1575 STRLEN len;
427 char *ptr = SvPVbyte (data, len); 1576 char *ptr = SvPVbyte (data, len);
428 EVENT_LIST (5, items - 5);
429 1577
1578 cl_event ev = 0;
430 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1579 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
431 1580
432 if (ev) 1581 if (ev)
433 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1582 XPUSH_CLOBJ (stash_event, ev);
434}
435 1583
1584#if CL_VERSION_1_2
1585
436void 1586void
437enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1587fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1588 ALIAS:
1589 enqueue_fill_buffer = 0
438 PPCODE: 1590 PPCODE:
439{ 1591 EVENT_LIST (5);
1592
1593 STRLEN len;
1594 char *ptr = SvPVbyte (data, len);
1595
440 cl_event ev = 0; 1596 cl_event ev = 0;
441 EVENT_LIST (6, items - 6);
442
443 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1597 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
444 1598
445 if (ev) 1599 if (ev)
446 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1600 XPUSH_CLOBJ (stash_event, ev);
447}
448 1601
449 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html */
450 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html */
451
452void 1602void
453enqueue_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, ...) 1603fill_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, ...)
1604 ALIAS:
1605 enqueue_fill_image = 0
454 PPCODE: 1606 PPCODE:
455{ 1607 EVENT_LIST (12);
1608
1609 const size_t origin [3] = { x, y, z };
1610 const size_t region [3] = { width, height, depth };
1611
1612 const cl_float c_f [4] = { r, g, b, a };
1613 const cl_uint c_u [4] = { r, g, b, a };
1614 const cl_int c_s [4] = { r, g, b, a };
1615 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1616 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1617 cl_image_format format;
1618 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1619 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1620 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1621 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1622
456 cl_event ev = 0; 1623 cl_event ev = 0;
1624 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1625 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1626
1627 if (ev)
1628 XPUSH_CLOBJ (stash_event, ev);
1629
1630#endif
1631
1632void
1633copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1634 ALIAS:
1635 enqueue_copy_buffer = 0
1636 PPCODE:
1637 EVENT_LIST (6);
1638
1639 cl_event ev = 0;
1640 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1641
1642 if (ev)
1643 XPUSH_CLOBJ (stash_event, ev);
1644
1645void
1646read_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, ...)
1647 ALIAS:
1648 enqueue_read_buffer_rect = 0
1649 PPCODE:
1650 EVENT_LIST (17);
1651
1652 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
457 const size_t src_origin[3] = { src_x, src_y, src_z }; 1653 const size_t host_origin[3] = { host_x, host_y, host_z };
458 const size_t region[3] = { width, height, depth }; 1654 const size_t region[3] = { width, height, depth };
1655
1656 if (!buf_row_pitch)
1657 buf_row_pitch = region [0];
1658
1659 if (!buf_slice_pitch)
1660 buf_slice_pitch = region [1] * buf_row_pitch;
1661
1662 if (!host_row_pitch)
1663 host_row_pitch = region [0];
1664
1665 if (!host_slice_pitch)
1666 host_slice_pitch = region [1] * host_row_pitch;
1667
459 size_t len = row_pitch * slice_pitch * depth; 1668 size_t len = host_row_pitch * host_slice_pitch * region [2];
460 EVENT_LIST (11, items - 11);
461 1669
462 SvUPGRADE (data, SVt_PV); 1670 SvUPGRADE (data, SVt_PV);
463 SvGROW (data, len); 1671 SvGROW (data, len);
464 SvPOK_only (data); 1672 SvPOK_only (data);
465 SvCUR_set (data, len); 1673 SvCUR_set (data, len);
466 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)); 1674
1675 cl_event ev = 0;
1676 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));
467 1677
468 if (ev) 1678 if (ev)
469 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1679 XPUSH_CLOBJ (stash_event, ev);
470}
471 1680
472void 1681void
473enqueue_write_image (OpenCL::Queue this, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, SV *data, ...) 1682write_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, ...)
1683 ALIAS:
1684 enqueue_write_buffer_rect = 0
474 PPCODE: 1685 PPCODE:
475{ 1686 EVENT_LIST (17);
1687
1688 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1689 const size_t host_origin[3] = { host_x, host_y, host_z };
1690 const size_t region[3] = { width, height, depth };
1691 STRLEN len;
1692 char *ptr = SvPVbyte (data, len);
1693
1694 if (!buf_row_pitch)
1695 buf_row_pitch = region [0];
1696
1697 if (!buf_slice_pitch)
1698 buf_slice_pitch = region [1] * buf_row_pitch;
1699
1700 if (!host_row_pitch)
1701 host_row_pitch = region [0];
1702
1703 if (!host_slice_pitch)
1704 host_slice_pitch = region [1] * host_row_pitch;
1705
1706 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1707
1708 if (len < min_len)
1709 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1710
476 cl_event ev = 0; 1711 cl_event ev = 0;
1712 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));
1713
1714 if (ev)
1715 XPUSH_CLOBJ (stash_event, ev);
1716
1717void
1718copy_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, ...)
1719 ALIAS:
1720 enqueue_copy_buffer_rect = 0
1721 PPCODE:
1722 EVENT_LIST (16);
1723
1724 const size_t src_origin[3] = { src_x, src_y, src_z };
1725 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1726 const size_t region[3] = { width, height, depth };
1727
1728 cl_event ev = 0;
1729 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));
1730
1731 if (ev)
1732 XPUSH_CLOBJ (stash_event, ev);
1733
1734void
1735read_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, ...)
1736 ALIAS:
1737 enqueue_read_image = 0
1738 PPCODE:
1739 EVENT_LIST (12);
1740
1741 const size_t src_origin[3] = { src_x, src_y, src_z };
1742 const size_t region[3] = { width, height, depth };
1743
1744 if (!row_pitch)
1745 row_pitch = img_row_pitch (src);
1746
1747 if (depth > 1 && !slice_pitch)
1748 slice_pitch = row_pitch * height;
1749
1750 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1751
1752 SvUPGRADE (data, SVt_PV);
1753 SvGROW (data, len);
1754 SvPOK_only (data);
1755 SvCUR_set (data, len);
1756
1757 cl_event ev = 0;
1758 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));
1759
1760 if (ev)
1761 XPUSH_CLOBJ (stash_event, ev);
1762
1763void
1764write_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, ...)
1765 ALIAS:
1766 enqueue_write_image = 0
1767 PPCODE:
1768 EVENT_LIST (12);
1769
477 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1770 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
478 const size_t region[3] = { width, height, depth }; 1771 const size_t region[3] = { width, height, depth };
479 STRLEN len; 1772 STRLEN len;
480 char *ptr = SvPVbyte (data, len); 1773 char *ptr = SvPVbyte (data, len);
1774
1775 if (!row_pitch)
1776 row_pitch = img_row_pitch (dst);
1777
1778 if (depth > 1 && !slice_pitch)
481 size_t slice_pitch = len / (row_pitch * height); 1779 slice_pitch = row_pitch * height;
482 EVENT_LIST (11, items - 11);
483 1780
1781 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1782
1783 if (len < min_len)
1784 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1785
1786 cl_event ev = 0;
484 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)); 1787 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));
485 1788
486 if (ev) 1789 if (ev)
487 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1790 XPUSH_CLOBJ (stash_event, ev);
488}
489 1791
490void 1792void
491enqueue_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, ...) 1793copy_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, ...)
1794 ALIAS:
1795 enqueue_copy_image = 0
492 PPCODE: 1796 PPCODE:
493{ 1797 EVENT_LIST (12);
494 cl_event ev = 0; 1798
495 const size_t src_origin[3] = { src_x, src_y, src_z }; 1799 const size_t src_origin[3] = { src_x, src_y, src_z };
496 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1800 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
497 const size_t region[3] = { width, height, depth }; 1801 const size_t region[3] = { width, height, depth };
498 EVENT_LIST (16, items - 16);
499 1802
1803 cl_event ev = 0;
500 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)); 1804 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
501 1805
502 if (ev) 1806 if (ev)
503 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1807 XPUSH_CLOBJ (stash_event, ev);
504}
505 1808
506void 1809void
507enqueue_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, ...) 1810copy_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, ...)
1811 ALIAS:
1812 enqueue_copy_image_to_buffer = 0
508 PPCODE: 1813 PPCODE:
509{ 1814 EVENT_LIST (10);
1815
1816 const size_t src_origin[3] = { src_x, src_y, src_z };
1817 const size_t region [3] = { width, height, depth };
1818
510 cl_event ev = 0; 1819 cl_event ev = 0;
1820 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1821
1822 if (ev)
1823 XPUSH_CLOBJ (stash_event, ev);
1824
1825void
1826copy_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, ...)
1827 ALIAS:
1828 enqueue_copy_buffer_to_image = 0
1829 PPCODE:
1830 EVENT_LIST (10);
1831
511 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1832 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1833 const size_t region [3] = { width, height, depth };
1834
1835 cl_event ev = 0;
1836 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1837
1838 if (ev)
1839 XPUSH_CLOBJ (stash_event, ev);
1840
1841void
1842map_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, ...)
1843 ALIAS:
1844 enqueue_map_buffer = 0
1845 PPCODE:
1846 EVENT_LIST (6);
1847
1848 size_t cb = SvIV (cb_);
1849
1850 if (!SvOK (cb_))
1851 {
1852 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1853 cb -= offset;
1854 }
1855
1856 cl_event ev;
1857 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1858 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1859
1860void
1861map_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, ...)
1862 ALIAS:
1863 enqueue_map_image = 0
1864 PPCODE:
1865 size_t width = SvIV (width_);
1866 if (!SvOK (width_))
1867 {
1868 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1869 width -= x;
1870 }
1871
1872 size_t height = SvIV (width_);
1873 if (!SvOK (height_))
1874 {
1875 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1876 height -= y;
1877 }
1878
1879 size_t depth = SvIV (width_);
1880 if (!SvOK (depth_))
1881 {
1882 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1883 depth -= z;
1884
1885 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1886 if (!depth)
1887 depth = 1;
1888 }
1889
1890 const size_t origin[3] = { x, y, z };
512 const size_t region[3] = { width, height, depth }; 1891 const size_t region[3] = { width, height, depth };
1892 size_t row_pitch, slice_pitch;
513 EVENT_LIST (10, items - 10); 1893 EVENT_LIST (10);
514 1894
515 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1895 cl_event ev;
1896 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));
1897
1898 size_t cb = slice_pitch ? slice_pitch * region [2]
1899 : row_pitch ? row_pitch * region [1]
1900 : region [0];
1901
1902 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1903
1904void
1905unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1906 PPCODE:
1907 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1908 if (GIMME_V != G_VOID)
1909 {
1910 clRetainEvent (mapped->event);
1911 XPUSH_CLOBJ (stash_event, mapped->event);
1912 }
1913
1914void
1915task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1916 ALIAS:
1917 enqueue_task = 0
1918 PPCODE:
1919 EVENT_LIST (2);
1920
1921 cl_event ev = 0;
1922 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
516 1923
517 if (ev) 1924 if (ev)
518 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1925 XPUSH_CLOBJ (stash_event, ev);
519}
520 1926
521void 1927void
522enqueue_copy_image (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
523 PPCODE:
524{
525 cl_event ev = 0;
526 const size_t src_origin[3] = { src_x, src_y, src_z };
527 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
528 const size_t region[3] = { width, height, depth };
529 EVENT_LIST (12, items - 12);
530
531 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
532
533 if (ev)
534 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
535}
536
537void
538enqueue_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, ...)
539 PPCODE:
540{
541 cl_event ev = 0;
542 const size_t src_origin[3] = { src_x, src_y, src_z };
543 const size_t region[3] = { width, height, depth };
544 EVENT_LIST (10, items - 10);
545
546 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
547
548 if (ev)
549 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
550}
551
552void
553enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
554 PPCODE:
555{
556 cl_event ev = 0;
557 EVENT_LIST (2, items - 2);
558
559 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
560
561 if (ev)
562 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
563}
564
565void
566enqueue_nd_range_kernel (OpenCL::Queue this, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1928nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1929 ALIAS:
1930 enqueue_nd_range_kernel = 0
567 PPCODE: 1931 PPCODE:
568{ 1932 EVENT_LIST (5);
569 cl_event ev = 0; 1933
570 size_t *gwo = 0, *gws, *lws = 0; 1934 size_t *gwo = 0, *gws, *lws = 0;
571 int gws_len; 1935 int gws_len;
572 size_t *lists; 1936 size_t *lists;
573 int i; 1937 int i;
574 EVENT_LIST (5, items - 5);
575 1938
576 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1939 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
577 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1940 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
578 1941
579 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1942 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
580 1943
581 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1944 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
582 1945
583 gws = lists + gws_len * 0; 1946 gws = lists + gws_len * 0;
584 for (i = 0; i < gws_len; ++i) 1947 for (i = 0; i < gws_len; ++i)
1948 {
585 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1949 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1950 // at least nvidia crashes for 0-sized work group sizes, work around
1951 if (!gws [i])
1952 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1953 }
586 1954
587 if (SvOK (global_work_offset)) 1955 if (SvOK (global_work_offset))
588 { 1956 {
589 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1957 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
590 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1958 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
597 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1965 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
598 } 1966 }
599 1967
600 if (SvOK (local_work_size)) 1968 if (SvOK (local_work_size))
601 { 1969 {
602 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1970 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
603 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1971 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
604 1972
605 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1973 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
606 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1974 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
607 1975
608 lws = lists + gws_len * 2; 1976 lws = lists + gws_len * 2;
609 for (i = 0; i < gws_len; ++i) 1977 for (i = 0; i < gws_len; ++i)
1978 {
610 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1979 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1980 // at least nvidia crashes for 0-sized work group sizes, work around
1981 if (!lws [i])
1982 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1983 }
611 } 1984 }
612 1985
1986 cl_event ev = 0;
613 NEED_SUCCESS (EnqueueNDRangeKernel, (this, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1987 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
614 1988
615 if (ev) 1989 if (ev)
616 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1990 XPUSH_CLOBJ (stash_event, ev);
617}
618 1991
1992#if CL_VERSION_1_2
1993
619void 1994void
620enqueue_marker (OpenCL::Queue this) 1995migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1996 ALIAS:
1997 enqueue_migrate_mem_objects = 0
621 PPCODE: 1998 PPCODE:
622{ 1999 EVENT_LIST (3);
2000
2001 cl_uint obj_count;
2002 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2003
2004 cl_event ev = 0;
2005 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2006
2007 if (ev)
2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#endif
2011
2012#if cl_apple_gl_sharing || cl_khr_gl_sharing
2013
2014void
2015acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2016 ALIAS:
2017 release_gl_objects = 1
2018 enqueue_acquire_gl_objects = 0
2019 enqueue_release_gl_objects = 1
2020 PPCODE:
2021 EVENT_LIST (2);
2022
2023 cl_uint obj_count;
2024 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2025
623 cl_event ev; 2026 cl_event ev = 0;
624 NEED_SUCCESS (EnqueueMarker, (this, &ev));
625 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
626}
627 2027
2028 if (ix)
2029 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2030 else
2031 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2032
2033 if (ev)
2034 XPUSH_CLOBJ (stash_event, ev);
2035
2036#endif
2037
628void 2038void
629enqueue_wait_for_events (OpenCL::Queue this, ...) 2039wait_for_events (OpenCL::Queue self, ...)
2040 ALIAS:
2041 enqueue_wait_for_events = 0
630 CODE: 2042 CODE:
631{
632 EVENT_LIST (1, items - 1); 2043 EVENT_LIST (1);
2044#if PREFER_1_1
633 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 2045 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
634} 2046#else
2047 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2048#endif
635 2049
636void 2050void
637enqueue_barrier (OpenCL::Queue this) 2051marker (OpenCL::Queue self, ...)
2052 ALIAS:
2053 enqueue_marker = 0
638 CODE: 2054 PPCODE:
2055 EVENT_LIST (1);
2056 cl_event ev = 0;
2057#if PREFER_1_1
2058 if (!event_list_count)
2059 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2060 else
2061#if CL_VERSION_1_2
2062 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2063#else
2064 {
2065 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2066 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2067 }
2068#endif
2069#else
2070 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2071#endif
2072 if (ev)
2073 XPUSH_CLOBJ (stash_event, ev);
2074
2075void
2076barrier (OpenCL::Queue self, ...)
2077 ALIAS:
2078 enqueue_barrier = 0
2079 PPCODE:
2080 EVENT_LIST (1);
2081 cl_event ev = 0;
2082#if PREFER_1_1
2083 if (!event_list_count && GIMME_V == G_VOID)
639 NEED_SUCCESS (EnqueueBarrier, (this)); 2084 NEED_SUCCESS (EnqueueBarrier, (self));
2085 else
2086#if CL_VERSION_1_2
2087 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2088#else
2089 {
2090 if (event_list_count)
2091 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
640 2092
2093 if (GIMME_V != G_VOID)
2094 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2095 }
2096#endif
2097#else
2098 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2099#endif
2100 if (ev)
2101 XPUSH_CLOBJ (stash_event, ev);
2102
641void 2103void
642flush (OpenCL::Queue this) 2104flush (OpenCL::Queue self)
643 CODE: 2105 CODE:
644 NEED_SUCCESS (Flush, (this)); 2106 NEED_SUCCESS (Flush, (self));
645 2107
646void 2108void
647finish (OpenCL::Queue this) 2109finish (OpenCL::Queue self)
648 CODE: 2110 CODE:
649 NEED_SUCCESS (Finish, (this)); 2111 NEED_SUCCESS (Finish, (self));
2112
2113void
2114info (OpenCL::Queue self, cl_command_queue_info name)
2115 PPCODE:
2116 INFO (CommandQueue)
2117
2118#BEGIN:command_queue
2119
2120void
2121context (OpenCL::Queue self)
2122 PPCODE:
2123 cl_context value [1];
2124 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
2125 EXTEND (SP, 1);
2126 const int i = 0;
2127 {
2128 NEED_SUCCESS (RetainContext, (value [i]));
2129 PUSH_CLOBJ (stash_context, value [i]);
2130 }
2131
2132void
2133device (OpenCL::Queue self)
2134 PPCODE:
2135 cl_device_id value [1];
2136 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
2137 EXTEND (SP, 1);
2138 const int i = 0;
2139 {
2140 PUSH_CLOBJ (stash_device, value [i]);
2141 }
2142
2143void
2144reference_count (OpenCL::Queue self)
2145 PPCODE:
2146 cl_uint value [1];
2147 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
2148 EXTEND (SP, 1);
2149 const int i = 0;
2150 PUSHs (sv_2mortal (newSVuv (value [i])));
2151
2152void
2153properties (OpenCL::Queue self)
2154 PPCODE:
2155 cl_command_queue_properties value [1];
2156 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
2157 EXTEND (SP, 1);
2158 const int i = 0;
2159 PUSHs (sv_2mortal (newSVuv (value [i])));
2160
2161#END:command_queue
650 2162
651MODULE = OpenCL PACKAGE = OpenCL::Memory 2163MODULE = OpenCL PACKAGE = OpenCL::Memory
652 2164
653void 2165void
654DESTROY (OpenCL::Memory this) 2166DESTROY (OpenCL::Memory self)
655 CODE: 2167 CODE:
656 clReleaseMemObject (this); 2168 clReleaseMemObject (self);
657 2169
658void 2170void
659info (OpenCL::Memory this, cl_mem_info name) 2171info (OpenCL::Memory self, cl_mem_info name)
660 PPCODE: 2172 PPCODE:
661 INFO (MemObject) 2173 INFO (MemObject)
662 2174
2175#BEGIN:mem
2176
2177void
2178type (OpenCL::Memory self)
2179 PPCODE:
2180 cl_mem_object_type value [1];
2181 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
2182 EXTEND (SP, 1);
2183 const int i = 0;
2184 PUSHs (sv_2mortal (newSVuv (value [i])));
2185
2186void
2187flags (OpenCL::Memory self)
2188 PPCODE:
2189 cl_mem_flags value [1];
2190 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
2191 EXTEND (SP, 1);
2192 const int i = 0;
2193 PUSHs (sv_2mortal (newSVuv (value [i])));
2194
2195void
2196size (OpenCL::Memory self)
2197 ALIAS:
2198 size = CL_MEM_SIZE
2199 offset = CL_MEM_OFFSET
2200 PPCODE:
2201 size_t value [1];
2202 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2203 EXTEND (SP, 1);
2204 const int i = 0;
2205 PUSHs (sv_2mortal (newSVuv (value [i])));
2206
2207void
2208host_ptr (OpenCL::Memory self)
2209 PPCODE:
2210 void * value [1];
2211 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
2212 EXTEND (SP, 1);
2213 const int i = 0;
2214 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
2215
2216void
2217map_count (OpenCL::Memory self)
2218 ALIAS:
2219 map_count = CL_MEM_MAP_COUNT
2220 reference_count = CL_MEM_REFERENCE_COUNT
2221 PPCODE:
2222 cl_uint value [1];
2223 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2224 EXTEND (SP, 1);
2225 const int i = 0;
2226 PUSHs (sv_2mortal (newSVuv (value [i])));
2227
2228void
2229context (OpenCL::Memory self)
2230 PPCODE:
2231 cl_context value [1];
2232 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
2233 EXTEND (SP, 1);
2234 const int i = 0;
2235 {
2236 NEED_SUCCESS (RetainContext, (value [i]));
2237 PUSH_CLOBJ (stash_context, value [i]);
2238 }
2239
2240void
2241associated_memobject (OpenCL::Memory self)
2242 PPCODE:
2243 cl_mem value [1];
2244 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
2245 EXTEND (SP, 1);
2246 const int i = 0;
2247 {
2248 NEED_SUCCESS (RetainMemObject, (value [i]));
2249 PUSH_CLOBJ (stash_memory, value [i]);
2250 }
2251
2252#END:mem
2253
2254#if cl_apple_gl_sharing || cl_khr_gl_sharing
2255
2256void
2257gl_object_info (OpenCL::Memory self)
2258 PPCODE:
2259 cl_gl_object_type type;
2260 cl_GLuint name;
2261 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2262 EXTEND (SP, 2);
2263 PUSHs (sv_2mortal (newSVuv (type)));
2264 PUSHs (sv_2mortal (newSVuv (name)));
2265
2266#endif
2267
2268MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2269
2270void
2271sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2272 PPCODE:
2273 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2274 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2275
2276 cl_buffer_region crdata = { origin, size };
2277
2278 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2279 XPUSH_CLOBJ (stash_buffer, mem);
2280
2281MODULE = OpenCL PACKAGE = OpenCL::Image
2282
2283void
2284image_info (OpenCL::Image self, cl_image_info name)
2285 PPCODE:
2286 INFO (Image)
2287
2288void
2289format (OpenCL::Image self)
2290 PPCODE:
2291 cl_image_format format;
2292 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2293 EXTEND (SP, 2);
2294 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2295 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2296
2297#BEGIN:image
2298
2299void
2300element_size (OpenCL::Image self)
2301 ALIAS:
2302 element_size = CL_IMAGE_ELEMENT_SIZE
2303 row_pitch = CL_IMAGE_ROW_PITCH
2304 slice_pitch = CL_IMAGE_SLICE_PITCH
2305 width = CL_IMAGE_WIDTH
2306 height = CL_IMAGE_HEIGHT
2307 depth = CL_IMAGE_DEPTH
2308 PPCODE:
2309 size_t value [1];
2310 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
2311 EXTEND (SP, 1);
2312 const int i = 0;
2313 PUSHs (sv_2mortal (newSVuv (value [i])));
2314
2315#END:image
2316
2317#if cl_apple_gl_sharing || cl_khr_gl_sharing
2318
2319#BEGIN:gl_texture
2320
2321void
2322target (OpenCL::Image self)
2323 PPCODE:
2324 cl_GLenum value [1];
2325 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2326 EXTEND (SP, 1);
2327 const int i = 0;
2328 PUSHs (sv_2mortal (newSVuv (value [i])));
2329
2330void
2331gl_mipmap_level (OpenCL::Image self)
2332 PPCODE:
2333 cl_GLint value [1];
2334 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2335 EXTEND (SP, 1);
2336 const int i = 0;
2337 PUSHs (sv_2mortal (newSViv (value [i])));
2338
2339#END:gl_texture
2340
2341#endif
2342
663MODULE = OpenCL PACKAGE = OpenCL::Sampler 2343MODULE = OpenCL PACKAGE = OpenCL::Sampler
664 2344
665void 2345void
666DESTROY (OpenCL::Sampler this) 2346DESTROY (OpenCL::Sampler self)
667 CODE: 2347 CODE:
668 clReleaseSampler (this); 2348 clReleaseSampler (self);
669 2349
670void 2350void
671info (OpenCL::Sampler this, cl_sampler_info name) 2351info (OpenCL::Sampler self, cl_sampler_info name)
672 PPCODE: 2352 PPCODE:
673 INFO (Sampler) 2353 INFO (Sampler)
674 2354
2355#BEGIN:sampler
2356
2357void
2358reference_count (OpenCL::Sampler self)
2359 PPCODE:
2360 cl_uint value [1];
2361 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0));
2362 EXTEND (SP, 1);
2363 const int i = 0;
2364 PUSHs (sv_2mortal (newSVuv (value [i])));
2365
2366void
2367context (OpenCL::Sampler self)
2368 PPCODE:
2369 cl_context value [1];
2370 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2371 EXTEND (SP, 1);
2372 const int i = 0;
2373 {
2374 NEED_SUCCESS (RetainContext, (value [i]));
2375 PUSH_CLOBJ (stash_context, value [i]);
2376 }
2377
2378void
2379normalized_coords (OpenCL::Sampler self)
2380 PPCODE:
2381 cl_addressing_mode value [1];
2382 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2383 EXTEND (SP, 1);
2384 const int i = 0;
2385 PUSHs (sv_2mortal (newSVuv (value [i])));
2386
2387void
2388addressing_mode (OpenCL::Sampler self)
2389 PPCODE:
2390 cl_filter_mode value [1];
2391 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2392 EXTEND (SP, 1);
2393 const int i = 0;
2394 PUSHs (sv_2mortal (newSVuv (value [i])));
2395
2396void
2397filter_mode (OpenCL::Sampler self)
2398 PPCODE:
2399 cl_bool value [1];
2400 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2401 EXTEND (SP, 1);
2402 const int i = 0;
2403 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2404
2405#END:sampler
2406
675MODULE = OpenCL PACKAGE = OpenCL::Program 2407MODULE = OpenCL PACKAGE = OpenCL::Program
676 2408
677void 2409void
678DESTROY (OpenCL::Program this) 2410DESTROY (OpenCL::Program self)
679 CODE: 2411 CODE:
680 clReleaseProgram (this); 2412 clReleaseProgram (self);
681 2413
682void 2414void
683info (OpenCL::Program this, cl_program_info name) 2415build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2416 ALIAS:
2417 build_async = 1
684 PPCODE: 2418 CODE:
685 INFO (Program) 2419 cl_uint device_count = 0;
2420 cl_device_id *device_list = 0;
686 2421
2422 if (SvOK (devices))
2423 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2424
2425 void *user_data;
2426 program_callback pfn_notify = make_program_callback (notify, &user_data);
2427
2428 if (ix)
2429 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2430 else
2431 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2432
2433#if CL_VERSION_1_2
2434
687void 2435void
688build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2436compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
689 CODE: 2437 CODE:
2438 cl_uint device_count = 0;
2439 cl_device_id *device_list = 0;
2440
2441 if (SvOK (devices))
2442 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2443
2444 cl_uint header_count = 0;
2445 cl_program *header_list = 0;
2446 const char **header_name = 0;
2447
2448 if (SvOK (headers))
2449 {
2450 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2451 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2452
2453 HV *hv = (HV *)SvRV (devices);
2454
2455 header_count = hv_iterinit (hv);
2456 header_list = tmpbuf (sizeof (*header_list) * header_count);
2457 header_name = tmpbuf (sizeof (*header_name) * header_count);
2458
2459 HE *he;
2460 int i = 0;
2461 while (he = hv_iternext (hv))
2462 {
2463 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2464 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2465 ++i;
2466 }
2467 }
2468
2469 void *user_data;
2470 program_callback pfn_notify = make_program_callback (notify, &user_data);
2471
690 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2472 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2473 header_count, header_list, header_name, pfn_notify, user_data));
691 2474
2475#endif
2476
692void 2477void
693build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2478build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
694 PPCODE: 2479 PPCODE:
695{
696 size_t size; 2480 size_t size;
697 SV *sv;
698
699 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2481 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
700 sv = sv_2mortal (newSV (size)); 2482 SV *sv = sv_2mortal (newSV (size));
701 SvUPGRADE (sv, SVt_PV); 2483 SvUPGRADE (sv, SVt_PV);
702 SvPOK_only (sv); 2484 SvPOK_only (sv);
703 SvCUR_set (sv, size); 2485 SvCUR_set (sv, size);
704 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2486 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
705 XPUSHs (sv); 2487 XPUSHs (sv);
706} 2488
2489#BEGIN:program_build
2490
2491void
2492build_status (OpenCL::Program self, OpenCL::Device device)
2493 PPCODE:
2494 cl_build_status value [1];
2495 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2496 EXTEND (SP, 1);
2497 const int i = 0;
2498 PUSHs (sv_2mortal (newSViv (value [i])));
2499
2500void
2501build_options (OpenCL::Program self, OpenCL::Device device)
2502 ALIAS:
2503 build_options = CL_PROGRAM_BUILD_OPTIONS
2504 build_log = CL_PROGRAM_BUILD_LOG
2505 PPCODE:
2506 size_t size;
2507 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2508 char *value = tmpbuf (size);
2509 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2510 EXTEND (SP, 1);
2511 const int i = 0;
2512 PUSHs (sv_2mortal (newSVpv (value, 0)));
2513
2514#END:program_build
707 2515
708void 2516void
709kernel (OpenCL::Program program, SV *function) 2517kernel (OpenCL::Program program, SV *function)
710 PPCODE: 2518 PPCODE:
711 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2519 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
712 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2520 XPUSH_CLOBJ (stash_kernel, kernel);
2521
2522void
2523kernels_in_program (OpenCL::Program program)
2524 PPCODE:
2525 cl_uint num_kernels;
2526 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2527 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2528 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2529
2530 int i;
2531 EXTEND (SP, num_kernels);
2532 for (i = 0; i < num_kernels; ++i)
2533 PUSH_CLOBJ (stash_kernel, kernels [i]);
2534
2535void
2536info (OpenCL::Program self, cl_program_info name)
2537 PPCODE:
2538 INFO (Program)
2539
2540void
2541binaries (OpenCL::Program self)
2542 PPCODE:
2543 cl_uint n, i;
2544 size_t size;
2545
2546 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2547 if (!n) XSRETURN_EMPTY;
2548
2549 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2550 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2551 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2552 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2553
2554 EXTEND (SP, n);
2555 for (i = 0; i < n; ++i)
2556 {
2557 SV *sv = sv_2mortal (newSV (sizes [i]));
2558 SvUPGRADE (sv, SVt_PV);
2559 SvPOK_only (sv);
2560 SvCUR_set (sv, sizes [i]);
2561 ptrs [i] = (void *)SvPVX (sv);
2562 PUSHs (sv);
2563 }
2564
2565 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2566 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2567
2568#BEGIN:program
2569
2570void
2571reference_count (OpenCL::Program self)
2572 ALIAS:
2573 reference_count = CL_PROGRAM_REFERENCE_COUNT
2574 num_devices = CL_PROGRAM_NUM_DEVICES
2575 PPCODE:
2576 cl_uint value [1];
2577 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2578 EXTEND (SP, 1);
2579 const int i = 0;
2580 PUSHs (sv_2mortal (newSVuv (value [i])));
2581
2582void
2583context (OpenCL::Program self)
2584 PPCODE:
2585 cl_context value [1];
2586 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2587 EXTEND (SP, 1);
2588 const int i = 0;
2589 {
2590 NEED_SUCCESS (RetainContext, (value [i]));
2591 PUSH_CLOBJ (stash_context, value [i]);
2592 }
2593
2594void
2595devices (OpenCL::Program self)
2596 PPCODE:
2597 size_t size;
2598 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2599 cl_device_id *value = tmpbuf (size);
2600 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2601 int i, n = size / sizeof (*value);
2602 EXTEND (SP, n);
2603 for (i = 0; i < n; ++i)
2604 {
2605 PUSH_CLOBJ (stash_device, value [i]);
2606 }
2607
2608void
2609source (OpenCL::Program self)
2610 PPCODE:
2611 size_t size;
2612 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2613 char *value = tmpbuf (size);
2614 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2615 EXTEND (SP, 1);
2616 const int i = 0;
2617 PUSHs (sv_2mortal (newSVpv (value, 0)));
2618
2619void
2620binary_sizes (OpenCL::Program self)
2621 PPCODE:
2622 size_t size;
2623 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2624 size_t *value = tmpbuf (size);
2625 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2626 int i, n = size / sizeof (*value);
2627 EXTEND (SP, n);
2628 for (i = 0; i < n; ++i)
2629 PUSHs (sv_2mortal (newSVuv (value [i])));
2630
2631#END:program
713 2632
714MODULE = OpenCL PACKAGE = OpenCL::Kernel 2633MODULE = OpenCL PACKAGE = OpenCL::Kernel
715 2634
716void 2635void
717DESTROY (OpenCL::Kernel this) 2636DESTROY (OpenCL::Kernel self)
718 CODE: 2637 CODE:
719 clReleaseKernel (this); 2638 clReleaseKernel (self);
720 2639
721void 2640void
2641setf (OpenCL::Kernel self, const char *format, ...)
2642 CODE:
2643 int i;
2644 for (i = 2; ; ++i)
2645 {
2646 while (*format == ' ')
2647 ++format;
2648
2649 char type = *format++;
2650
2651 if (!type)
2652 break;
2653
2654 if (i >= items)
2655 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2656
2657 SV *sv = ST (i);
2658
2659 union
2660 {
2661 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2662 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2663 cl_half ch; cl_float cf; cl_double cd;
2664 cl_mem cm;
2665 cl_sampler ca;
2666 size_t cz;
2667 cl_event ce;
2668 } arg;
2669 size_t size;
2670 int nullarg = 0;
2671
2672 switch (type)
2673 {
2674 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2675 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2676 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2677 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2678 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2679 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2680 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2681 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2682
2683 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2684 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2685 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2686
2687 case 'z': nullarg = 1; size = SvIV (sv); break;
2688
2689 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2690 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2691 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2692
2693 default:
2694 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2695 }
2696
2697 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2698 if (res)
2699 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2700 }
2701
2702 if (i != items)
2703 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2704
2705void
2706set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2707 CODE:
2708 clSetKernelArg (self, idx, sizeof (value), &value);
2709
2710void
2711set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2712 CODE:
2713 clSetKernelArg (self, idx, sizeof (value), &value);
2714
2715void
2716set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2717 CODE:
2718 clSetKernelArg (self, idx, sizeof (value), &value);
2719
2720void
2721set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2722 CODE:
2723 clSetKernelArg (self, idx, sizeof (value), &value);
2724
2725void
2726set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2727 CODE:
2728 clSetKernelArg (self, idx, sizeof (value), &value);
2729
2730void
2731set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2732 CODE:
2733 clSetKernelArg (self, idx, sizeof (value), &value);
2734
2735void
2736set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2737 CODE:
2738 clSetKernelArg (self, idx, sizeof (value), &value);
2739
2740void
2741set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2742 CODE:
2743 clSetKernelArg (self, idx, sizeof (value), &value);
2744
2745void
2746set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2747 CODE:
2748 clSetKernelArg (self, idx, sizeof (value), &value);
2749
2750void
2751set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2752 CODE:
2753 clSetKernelArg (self, idx, sizeof (value), &value);
2754
2755void
2756set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2757 CODE:
2758 clSetKernelArg (self, idx, sizeof (value), &value);
2759
2760void
2761set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2762 CODE:
2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2764
2765void
2766set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2767 CODE:
2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2769
2770void
2771set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2772 ALIAS:
2773 set_image2d = 0
2774 set_image3d = 0
2775 CODE:
2776 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2777
2778void
2779set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2780 CODE:
2781 clSetKernelArg (self, idx, sizeof (value), &value);
2782
2783void
2784set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2785 CODE:
2786 clSetKernelArg (self, idx, size, 0);
2787
2788void
2789set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2790 CODE:
2791 clSetKernelArg (self, idx, sizeof (value), &value);
2792
2793void
722info (OpenCL::Kernel this, cl_kernel_info name) 2794info (OpenCL::Kernel self, cl_kernel_info name)
723 PPCODE: 2795 PPCODE:
724 INFO (Kernel) 2796 INFO (Kernel)
725 2797
2798#BEGIN:kernel
2799
726void 2800void
727set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2801function_name (OpenCL::Kernel self)
2802 PPCODE:
2803 size_t size;
2804 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2805 char *value = tmpbuf (size);
2806 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2807 EXTEND (SP, 1);
2808 const int i = 0;
2809 PUSHs (sv_2mortal (newSVpv (value, 0)));
2810
2811void
2812num_args (OpenCL::Kernel self)
2813 ALIAS:
2814 num_args = CL_KERNEL_NUM_ARGS
2815 reference_count = CL_KERNEL_REFERENCE_COUNT
2816 PPCODE:
2817 cl_uint value [1];
2818 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2819 EXTEND (SP, 1);
2820 const int i = 0;
2821 PUSHs (sv_2mortal (newSVuv (value [i])));
2822
2823void
2824context (OpenCL::Kernel self)
2825 PPCODE:
2826 cl_context value [1];
2827 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2828 EXTEND (SP, 1);
2829 const int i = 0;
2830 {
2831 NEED_SUCCESS (RetainContext, (value [i]));
2832 PUSH_CLOBJ (stash_context, value [i]);
2833 }
2834
2835void
2836program (OpenCL::Kernel self)
2837 PPCODE:
2838 cl_program value [1];
2839 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2840 EXTEND (SP, 1);
2841 const int i = 0;
2842 {
2843 NEED_SUCCESS (RetainProgram, (value [i]));
2844 PUSH_CLOBJ (stash_program, value [i]);
2845 }
2846
2847#END:kernel
2848
2849void
2850work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
728 CODE: 2851 PPCODE:
729 clSetKernelArg (this, idx, sizeof (value), &value); 2852 size_t size;
2853 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2854 SV *sv = sv_2mortal (newSV (size));
2855 SvUPGRADE (sv, SVt_PV);
2856 SvPOK_only (sv);
2857 SvCUR_set (sv, size);
2858 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2859 XPUSHs (sv);
730 2860
2861#BEGIN:kernel_work_group
2862
731void 2863void
732set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value) 2864work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2865 ALIAS:
2866 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2867 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2868 PPCODE:
2869 size_t value [1];
2870 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2871 EXTEND (SP, 1);
2872 const int i = 0;
2873 PUSHs (sv_2mortal (newSVuv (value [i])));
2874
2875void
2876compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2877 PPCODE:
2878 size_t size;
2879 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2880 size_t *value = tmpbuf (size);
2881 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2882 int i, n = size / sizeof (*value);
2883 EXTEND (SP, n);
2884 for (i = 0; i < n; ++i)
2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2886
2887void
2888local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
2889 ALIAS:
2890 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2891 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2892 PPCODE:
2893 cl_ulong value [1];
2894 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2895 EXTEND (SP, 1);
2896 const int i = 0;
2897 PUSHs (sv_2mortal (newSVuv (value [i])));
2898
2899#END:kernel_work_group
2900
2901#if CL_VERSION_1_2
2902
2903void
2904arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
733 CODE: 2905 PPCODE:
734 clSetKernelArg (this, idx, sizeof (value), &value); 2906 size_t size;
2907 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2908 SV *sv = sv_2mortal (newSV (size));
2909 SvUPGRADE (sv, SVt_PV);
2910 SvPOK_only (sv);
2911 SvCUR_set (sv, size);
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2913 XPUSHs (sv);
735 2914
736void 2915#BEGIN:kernel_arg
737set_short (OpenCL::Kernel this, cl_uint idx, cl_short value)
738 CODE:
739 clSetKernelArg (this, idx, sizeof (value), &value);
740 2916
741void 2917void
742set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2918arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
743 CODE: 2919 PPCODE:
744 clSetKernelArg (this, idx, sizeof (value), &value); 2920 cl_kernel_arg_address_qualifier value [1];
2921 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2922 EXTEND (SP, 1);
2923 const int i = 0;
2924 PUSHs (sv_2mortal (newSVuv (value [i])));
745 2925
746void 2926void
747set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2927arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
748 CODE: 2928 PPCODE:
749 clSetKernelArg (this, idx, sizeof (value), &value); 2929 cl_kernel_arg_access_qualifier value [1];
2930 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2931 EXTEND (SP, 1);
2932 const int i = 0;
2933 PUSHs (sv_2mortal (newSVuv (value [i])));
750 2934
751void 2935void
752set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value) 2936arg_type_name (OpenCL::Kernel self, cl_uint idx)
753 CODE: 2937 ALIAS:
754 clSetKernelArg (this, idx, sizeof (value), &value); 2938 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2939 arg_name = CL_KERNEL_ARG_NAME
2940 PPCODE:
2941 size_t size;
2942 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2943 char *value = tmpbuf (size);
2944 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2945 EXTEND (SP, 1);
2946 const int i = 0;
2947 PUSHs (sv_2mortal (newSVpv (value, 0)));
755 2948
756void 2949void
757set_long (OpenCL::Kernel this, cl_uint idx, cl_long value) 2950arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
758 CODE: 2951 PPCODE:
759 clSetKernelArg (this, idx, sizeof (value), &value); 2952 cl_kernel_arg_type_qualifier value [1];
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2954 EXTEND (SP, 1);
2955 const int i = 0;
2956 PUSHs (sv_2mortal (newSVuv (value [i])));
760 2957
761void 2958#END:kernel_arg
762set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
763 CODE:
764 clSetKernelArg (this, idx, sizeof (value), &value);
765 2959
766void 2960#endif
767set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
768 CODE:
769 clSetKernelArg (this, idx, sizeof (value), &value);
770
771void
772set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
773 CODE:
774 clSetKernelArg (this, idx, sizeof (value), &value);
775
776void
777set_double (OpenCL::Kernel this, cl_uint idx, cl_double value)
778 CODE:
779 clSetKernelArg (this, idx, sizeof (value), &value);
780
781void
782set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
783 CODE:
784 clSetKernelArg (this, idx, sizeof (value), &value);
785
786void
787set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
788 CODE:
789 clSetKernelArg (this, idx, sizeof (value), &value);
790
791void
792set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
793 CODE:
794 clSetKernelArg (this, idx, sizeof (value), &value);
795
796void
797set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
798 CODE:
799 clSetKernelArg (this, idx, sizeof (value), &value);
800
801void
802set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
803 CODE:
804 clSetKernelArg (this, idx, sizeof (value), &value);
805
806void
807set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
808 CODE:
809 clSetKernelArg (this, idx, sizeof (value), &value);
810 2961
811MODULE = OpenCL PACKAGE = OpenCL::Event 2962MODULE = OpenCL PACKAGE = OpenCL::Event
812 2963
813void 2964void
814DESTROY (OpenCL::Event this) 2965DESTROY (OpenCL::Event self)
815 CODE: 2966 CODE:
816 clReleaseEvent (this); 2967 clReleaseEvent (self);
817 2968
818void 2969void
2970wait (OpenCL::Event self)
2971 CODE:
2972 clWaitForEvents (1, &self);
2973
2974void
2975cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2976 CODE:
2977 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2978
2979void
819info (OpenCL::Event this, cl_event_info name) 2980info (OpenCL::Event self, cl_event_info name)
820 PPCODE: 2981 PPCODE:
821 INFO (Event) 2982 INFO (Event)
822 2983
2984#BEGIN:event
2985
823void 2986void
824wait (OpenCL::Event this) 2987command_queue (OpenCL::Event self)
2988 PPCODE:
2989 cl_command_queue value [1];
2990 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2991 EXTEND (SP, 1);
2992 const int i = 0;
2993 {
2994 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2995 PUSH_CLOBJ (stash_queue, value [i]);
2996 }
2997
2998void
2999command_type (OpenCL::Event self)
3000 PPCODE:
3001 cl_command_type value [1];
3002 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
3003 EXTEND (SP, 1);
3004 const int i = 0;
3005 PUSHs (sv_2mortal (newSVuv (value [i])));
3006
3007void
3008reference_count (OpenCL::Event self)
3009 ALIAS:
3010 reference_count = CL_EVENT_REFERENCE_COUNT
3011 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
3012 PPCODE:
3013 cl_uint value [1];
3014 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
3015 EXTEND (SP, 1);
3016 const int i = 0;
3017 PUSHs (sv_2mortal (newSVuv (value [i])));
3018
3019void
3020context (OpenCL::Event self)
3021 PPCODE:
3022 cl_context value [1];
3023 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
3024 EXTEND (SP, 1);
3025 const int i = 0;
3026 {
3027 NEED_SUCCESS (RetainContext, (value [i]));
3028 PUSH_CLOBJ (stash_context, value [i]);
3029 }
3030
3031#END:event
3032
3033void
3034profiling_info (OpenCL::Event self, cl_profiling_info name)
825 CODE: 3035 PPCODE:
826 clWaitForEvents (1, &this); 3036 INFO (EventProfiling)
3037
3038#BEGIN:profiling
3039
3040void
3041profiling_command_queued (OpenCL::Event self)
3042 ALIAS:
3043 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
3044 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
3045 profiling_command_start = CL_PROFILING_COMMAND_START
3046 profiling_command_end = CL_PROFILING_COMMAND_END
3047 PPCODE:
3048 cl_ulong value [1];
3049 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
3050 EXTEND (SP, 1);
3051 const int i = 0;
3052 PUSHs (sv_2mortal (newSVuv (value [i])));
3053
3054#END:profiling
827 3055
828MODULE = OpenCL PACKAGE = OpenCL::UserEvent 3056MODULE = OpenCL PACKAGE = OpenCL::UserEvent
829 3057
830void 3058void
831set_status (OpenCL::UserEvent this, cl_int execution_status) 3059set_status (OpenCL::UserEvent self, cl_int execution_status)
832 CODE: 3060 CODE:
833 clSetUserEventStatus (this, execution_status); 3061 clSetUserEventStatus (self, execution_status);
834 3062
3063MODULE = OpenCL PACKAGE = OpenCL::Mapped
3064
3065void
3066DESTROY (SV *self)
3067 CODE:
3068 OpenCL__Mapped mapped = SvMAPPED (self);
3069
3070 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3071 mapped_detach (self, mapped);
3072
3073 clReleaseCommandQueue (mapped->queue);
3074 clReleaseEvent (mapped->event);
3075 Safefree (mapped);
3076
3077void
3078unmap (OpenCL::Mapped self, ...)
3079 CODE:
3080 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3081
3082bool
3083mapped (OpenCL::Mapped self)
3084 CODE:
3085 RETVAL = !!self->ptr;
3086 OUTPUT:
3087 RETVAL
3088
3089void
3090wait (OpenCL::Mapped self)
3091 PPCODE:
3092 if (self->event)
3093 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3094
3095void
3096event (OpenCL::Mapped self)
3097 PPCODE:
3098 if (!self->event)
3099 XSRETURN_UNDEF;
3100
3101 clRetainEvent (self->event);
3102 XPUSH_CLOBJ (stash_event, self->event);
3103
3104size_t
3105size (OpenCL::Mapped self)
3106 CODE:
3107 RETVAL = self->cb;
3108 OUTPUT:
3109 RETVAL
3110
3111IV
3112ptr (OpenCL::Mapped self)
3113 CODE:
3114 RETVAL = PTR2IV (self->ptr);
3115 OUTPUT:
3116 RETVAL
3117
3118void
3119set (OpenCL::Mapped self, size_t offset, SV *data)
3120 CODE:
3121 STRLEN len;
3122 const char *ptr = SvPVbyte (data, len);
3123
3124 if (offset + len > self->cb)
3125 croak ("OpenCL::Mapped::set out of bound condition detected");
3126
3127 memcpy (offset + (char *)self->ptr, ptr, len);
3128
3129MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3130
3131MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3132
3133IV
3134row_pitch (OpenCL::Mapped self)
3135 ALIAS:
3136 slice_pitch = 1
3137 CODE:
3138 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3139 OUTPUT:
3140 RETVAL
3141
3142
3143

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines