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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.12 by root, Thu Nov 17 06:22:29 2011 UTC vs.
Revision 1.62 by root, Tue May 1 22:04:56 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#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h"
7#include "schmorp.h"
8
9#ifdef I_DLFCN
10 #include <dlfcn.h>
11#endif
12
13// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
14// but then fail to define the api functions unless you ALSO define
15// this. This breaks 100% of the opencl 1.1 apps, for what reason?
16// after all, the functions are deprecated, not removed.
17// in addition, you cannot test for this in any future-proof way.
18// each time a new opencl version comes out, you need to make a new
19// release.
20#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
21
22#ifndef PREFER_1_1
23 #define PREFER_1_1 1
24#endif
25
26#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif
29
30#ifdef __APPLE__
31 #include <OpenCL/opencl.h>
32#else
5#include <CL/opencl.h> 33 #include <CL/opencl.h>
34#endif
6 35
36#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1
38 #define PREFER_1_1 1
39#endif
40
7typedef cl_platform_id OpenCL__Platform; 41typedef cl_platform_id OpenCL__Platform;
8typedef cl_device_id OpenCL__Device; 42typedef cl_device_id OpenCL__Device;
9typedef cl_context OpenCL__Context; 43typedef cl_context OpenCL__Context;
10typedef cl_command_queue OpenCL__Queue; 44typedef cl_command_queue OpenCL__Queue;
11typedef cl_mem OpenCL__Memory; 45typedef cl_mem OpenCL__Memory;
12typedef cl_mem OpenCL__Buffer; 46typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj;
13typedef cl_mem OpenCL__Image; 48typedef cl_mem OpenCL__Image;
14typedef cl_mem OpenCL__Image2D; 49typedef cl_mem OpenCL__Image2D;
15typedef cl_mem OpenCL__Image3D; 50typedef cl_mem OpenCL__Image3D;
16typedef cl_mem OpenCL__Memory_ornull; 51typedef cl_mem OpenCL__Memory_ornull;
17typedef cl_mem OpenCL__Buffer_ornull; 52typedef cl_mem OpenCL__Buffer_ornull;
18typedef cl_mem OpenCL__Image_ornull; 53typedef cl_mem OpenCL__Image_ornull;
19typedef cl_mem OpenCL__Image2D_ornull; 54typedef cl_mem OpenCL__Image2D_ornull;
20typedef cl_mem OpenCL__Image3D_ornull; 55typedef cl_mem OpenCL__Image3D_ornull;
21typedef cl_sampler OpenCL__Sampler; 56typedef cl_sampler OpenCL__Sampler;
22typedef cl_program OpenCL__Program; 57typedef cl_program OpenCL__Program;
23typedef cl_kernel OpenCL__Kernel; 58typedef cl_kernel OpenCL__Kernel;
24typedef cl_event OpenCL__Event; 59typedef cl_event OpenCL__Event;
25typedef cl_event OpenCL__UserEvent; 60typedef cl_event OpenCL__UserEvent;
61
62typedef struct mapped * OpenCL__Mapped;
26 63
27typedef SV *FUTURE; 64typedef SV *FUTURE;
65
66static HV
67 *stash_platform,
68 *stash_device,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
89
90/*****************************************************************************/
91
92// name must include a leading underscore
93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
94// for perl. doh.
95static void *
96glsym (const char *name)
97{
98 void *fun = 0;
99
100 #if defined I_DLFCN && defined RTLD_DEFAULT
101 fun = dlsym (RTLD_DEFAULT, name + 1);
102 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
103
104 if (!fun)
105 {
106 static void *libgl;
107 static const char *glso[] = {
108 "libGL.so.1",
109 "libGL.so.3",
110 "libGL.so.4.0",
111 "libGL.so",
112 "/usr/lib/libGL.so",
113 "/usr/X11R6/lib/libGL.1.dylib"
114 };
115 int i;
116
117 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
118 {
119 libgl = dlopen (glso [i], RTLD_LAZY);
120 if (libgl)
121 break;
122 }
123
124 if (libgl)
125 {
126 fun = dlsym (libgl, name + 1);
127 if (!fun) fun = dlsym (libgl, name);
128 }
129 }
130 #endif
131
132 return fun;
133}
28 134
29/*****************************************************************************/ 135/*****************************************************************************/
30 136
31/* up to two temporary buffers */ 137/* up to two temporary buffers */
32static void * 138static void *
33tmpbuf (size_t size) 139tmpbuf (size_t size)
34{ 140{
141 enum { buffers = 3 };
35 static int idx; 142 static int idx;
36 static void *buf [2]; 143 static void *buf [buffers];
37 static size_t len [2]; 144 static size_t len [buffers];
38 145
39 idx ^= 1; 146 idx = (idx + 1) % buffers;
40 147
41 if (len [idx] < size) 148 if (len [idx] < size)
42 { 149 {
43 free (buf [idx]); 150 free (buf [idx]);
44 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
112 if (res) \ 219 if (res) \
113 FAIL (name); 220 FAIL (name);
114 221
115/*****************************************************************************/ 222/*****************************************************************************/
116 223
117#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 224static cl_context_properties *
118#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr)) 225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
226{
227 if (!sv || !SvOK (sv))
228 if (extra)
229 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
230 else
231 return 0;
119 232
233 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
234 {
235 AV *av = (AV *)SvRV (sv);
236 int i, len = av_len (av) + 1;
237 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
238 cl_context_properties *l = p;
239
240 if (len & 1)
241 croak ("%s: %s is not a property list (must be even number of elements)", func, svname);
242
243 while (extracount--)
244 *l++ = *extra++;
245
246 for (i = 0; i < len; i += 2)
247 {
248 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
249 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override
251
252 switch (t)
253 {
254 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv))
256 {
257 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func)
259 v = (cl_context_properties)((void *(*)(void))func)();
260 }
261 break;
262
263 case CL_GL_CONTEXT_KHR:
264 if (!SvOK (p_sv))
265 {
266 void *func = glsym ("_glXGetCurrentContext");
267 if (func)
268 v = (cl_context_properties)((void *(*)(void))func)();
269 }
270 break;
271
272 default:
273 /* unknown property, treat as int */
274 break;
275 }
276
277 *l++ = t;
278 *l++ = v;
279 }
280
281 *l = 0;
282
283 return p;
284 }
285
286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
287}
288
289/*****************************************************************************/
290
291static SV *
292new_clobj (HV *stash, IV id)
293{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
295}
296
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
299
300/* cl objects are either \$iv, or [$iv, ...] */
301/* they can be upgraded at runtime to the array form */
120static void * 302static void *
121SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
122{ 304{
123 if (SvROK (sv) && sv_derived_from (sv, pkg)) 305 if (SvROK (sv) && sv_derived_from (sv, pkg))
124 return (void *)SvIV (SvRV (sv)); 306 return (void *)SvIV (SvRV (sv));
125 307
126 croak ("%s: %s is not of type %s", func, svname, pkg); 308 croak ("%s: %s is not of type %s", func, svname, pkg);
127} 309}
128 310
129/*****************************************************************************/ 311/*****************************************************************************/
312/* callback stuff */
313
314/* default context callback, log to stderr */
315static void CL_CALLBACK
316context_default_notify (const char *msg, const void *info, size_t cb, void *data)
317{
318 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
319}
320
321typedef struct
322{
323 int free_cb;
324 void (*push)(void *data1, void *data2, void *data3);
325} eq_vtbl;
326
327typedef struct eq_item
328{
329 struct eq_item *next;
330 eq_vtbl *vtbl;
331 SV *cb;
332 void *data1, *data2, *data3;
333} eq_item;
334
335static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail;
339
340static void
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{
343 eq_item *item = malloc (sizeof (eq_item));
344
345 item->next = 0;
346 item->vtbl = vtbl;
347 item->cb = cb;
348 item->data1 = data1;
349 item->data2 = data2;
350 item->data3 = data3;
351
352 X_LOCK (eq_lock);
353
354 *(eq_head ? &eq_tail->next : &eq_head) = item;
355 eq_tail = item;
356
357 X_UNLOCK (eq_lock);
358
359 eq_signal_func (eq_signal_arg, 0);
360}
361
362static eq_item *
363eq_dec (void)
364{
365 eq_item *res;
366
367 X_LOCK (eq_lock);
368
369 res = eq_head;
370 if (res)
371 eq_head = res->next;
372
373 X_UNLOCK (eq_lock);
374
375 return res;
376}
377
378static void
379eq_poll (void)
380{
381 eq_item *item;
382
383 while ((item = eq_dec ()))
384 {
385 ENTER;
386 SAVETMPS;
387
388 dSP;
389 PUSHMARK (SP);
390 EXTEND (SP, 2);
391
392 if (item->vtbl->free_cb)
393 sv_2mortal (item->cb);
394
395 PUTBACK;
396 item->vtbl->push (item->data1, item->data2, item->data3);
397
398 SV *cb = item->cb;
399 free (item);
400
401 call_sv (cb, G_DISCARD | G_VOID);
402
403 FREETMPS;
404 LEAVE;
405 }
406}
407
408static void
409eq_poll_interrupt (pTHX_ void *c_arg, int value)
410{
411 eq_poll ();
412}
413
414/*****************************************************************************/
415/* context notify */
416
417static void
418eq_context_push (void *data1, void *data2, void *data3)
419{
420 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
423 PUTBACK;
424
425 free (data1);
426 free (data2);
427}
428
429static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
430
431static void CL_CALLBACK
432eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
433{
434 void *pvt_copy = malloc (cb);
435 memcpy (pvt_copy, pvt, cb);
436 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
437}
438
439#define CONTEXT_NOTIFY_CALLBACK \
440 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
441 void *user_data = 0; \
442 \
443 if (SvOK (notify)) \
444 { \
445 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \
447 }
448
449static SV *
450new_clobj_context (cl_context ctx, void *user_data)
451{
452 SV *sv = new_clobj (stash_context, (IV)ctx);
453
454 if (user_data)
455 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
456
457 return sv;
458}
459
460#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
461
462/*****************************************************************************/
463/* build/compile/link notify */
464
465static void
466eq_program_push (void *data1, void *data2, void *data3)
467{
468 dSP;
469 PUSH_CLOBJ (stash_program, data1);
470 PUTBACK;
471}
472
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474
475static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data)
477{
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
479}
480
481struct build_args
482{
483 cl_program program;
484 char *options;
485 void *user_data;
486 cl_uint num_devices;
487};
488
489X_THREAD_PROC (build_program_thread)
490{
491 struct build_args *arg = thr_arg;
492
493 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
494
495 if (arg->user_data)
496 eq_program_notify (arg->program, arg->user_data);
497 else
498 clReleaseProgram (arg->program);
499
500 free (arg->options);
501 free (arg);
502}
503
504static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{
507 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
508
509 arg->program = program;
510 arg->options = strdup (options);
511 arg->user_data = user_data;
512 arg->num_devices = num_devices;
513 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
514
515 xthread_t id;
516 thread_create (&id, build_program_thread, arg);
517}
518
519/*****************************************************************************/
520/* event objects */
521
522static void
523eq_event_push (void *data1, void *data2, void *data3)
524{
525 dSP;
526 PUSH_CLOBJ (stash_event, data1);
527 PUSHs (sv_2mortal (newSViv ((IV)data2)));
528 PUTBACK;
529}
530
531static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
532
533static void CL_CALLBACK
534eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
535{
536 clRetainEvent (event);
537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
538}
539
540/*****************************************************************************/
541/* utilities for XS code */
130 542
131static size_t 543static size_t
132img_row_pitch (cl_mem img) 544img_row_pitch (cl_mem img)
133{ 545{
134 size_t res; 546 size_t res;
135 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
136 return res; 548 return res;
137} 549}
138 550
139static cl_event * 551static cl_event *
140event_list (SV **items, int count) 552event_list (SV **items, cl_uint *rcount, cl_event extra)
141{ 553{
554 cl_uint count = *rcount;
555
556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed
557 count = 0;
558
559 if (!count && !extra)
560 return 0;
561
142 cl_event *list = tmpbuf (sizeof (cl_event) * count); 562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
563 int i = 0;
143 564
144 while (count--) 565 while (count--)
566 if (SvOK (items [count]))
145 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
146 568
569 if (extra)
570 list [i++] = extra;
571
572 *rcount = i;
573
147 return list; 574 return i ? list : 0;
148} 575}
149 576
150#define EVENT_LIST(items,count) \ 577#define EVENT_LIST(skip) \
151 cl_uint event_list_count = (count); \ 578 cl_uint event_list_count = items - (skip); \
152 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
153 580
154#define INFO(class) \ 581#define INFO(class) \
155{ \ 582{ \
156 size_t size; \ 583 size_t size; \
157 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
158 SV *sv = sv_2mortal (newSV (size)); \ 585 SV *sv = sv_2mortal (newSV (size)); \
159 SvUPGRADE (sv, SVt_PV); \ 586 SvUPGRADE (sv, SVt_PV); \
160 SvPOK_only (sv); \ 587 SvPOK_only (sv); \
161 SvCUR_set (sv, size); \ 588 SvCUR_set (sv, size); \
162 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 589 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
163 XPUSHs (sv); \ 590 XPUSHs (sv); \
164} 591}
165 592
593/*****************************************************************************/
594/* mapped_xxx */
595
596static OpenCL__Mapped
597SvMAPPED (SV *sv)
598{
599 // no typechecking atm., keep your fingers crossed
600 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
601}
602
603struct mapped
604{
605 cl_command_queue queue;
606 cl_mem memobj;
607 void *ptr;
608 size_t cb;
609 cl_event event;
610 size_t row_pitch;
611 size_t slice_pitch;
612};
613
614static SV *
615mapped_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)
616{
617 SV *data = newSV (0);
618 SvUPGRADE (data, SVt_PVMG);
619
620 OpenCL__Mapped mapped;
621 New (0, mapped, 1, struct mapped);
622
623 clRetainCommandQueue (queue);
624
625 mapped->queue = queue;
626 mapped->memobj = memobj;
627 mapped->ptr = ptr;
628 mapped->cb = cb;
629 mapped->event = ev;
630 mapped->row_pitch = row_pitch;
631 mapped->slice_pitch = slice_pitch;
632
633 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
634
635 if (SvLEN (data))
636 Safefree (data);
637
638 SvPVX (data) = (char *)ptr;
639 SvCUR_set (data, cb);
640 SvLEN_set (data, 0);
641 SvPOK_only (data);
642
643 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
644
645 if (!(flags & CL_MAP_WRITE))
646 SvREADONLY_on (data);
647
648 return obj;
649}
650
651static void
652mapped_detach (SV *sv, OpenCL__Mapped mapped)
653{
654 SV *data = SvRV (sv);
655
656 // the next check checks both before AND after detach, where SvPVX should STILL be 0
657 if (SvPVX (data) != (char *)mapped->ptr)
658 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
659 else
660 {
661 SvREADONLY_off (data);
662 SvCUR_set (data, 0);
663 SvPVX (data) = 0;
664 SvOK_off (data);
665 }
666
667 mapped->ptr = 0;
668}
669
670static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
674 cl_event ev;
675
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677
678 clReleaseEvent (mapped->event);
679 mapped->event = ev;
680
681 mapped_detach (self, mapped);
682}
683
684/*****************************************************************************/
685
166MODULE = OpenCL PACKAGE = OpenCL 686MODULE = OpenCL PACKAGE = OpenCL
167 687
168PROTOTYPES: ENABLE 688PROTOTYPES: ENABLE
169 689
690void
691poll ()
692 CODE:
693 eq_poll ();
694
695void
696_eq_initialise (IV func, IV arg)
697 CODE:
698 eq_signal_func = (void (*)(void *, int))func;
699 eq_signal_arg = (void*)arg;
700
170BOOT: 701BOOT:
171{ 702{
172 HV *stash = gv_stashpv ("OpenCL", 1); 703 HV *stash = gv_stashpv ("OpenCL", 1);
173 static const ivstr *civ, const_iv[] = { 704 static const ivstr *civ, const_iv[] = {
174 { sizeof (cl_char ), "SIZEOF_CHAR" }, 705 { sizeof (cl_char ), "SIZEOF_CHAR" },
175 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 706 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
176 { sizeof (cl_short ), "SIZEOF_SHORT" }, 707 { sizeof (cl_short ), "SIZEOF_SHORT" },
177 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 708 { sizeof (cl_ushort), "SIZEOF_USHORT" },
178 { sizeof (cl_int ), "SIZEOF_INT" }, 709 { sizeof (cl_int ), "SIZEOF_INT" },
179 { sizeof (cl_uint ), "SIZEOF_UINT" }, 710 { sizeof (cl_uint ), "SIZEOF_UINT" },
180 { sizeof (cl_long ), "SIZEOF_LONG" }, 711 { sizeof (cl_long ), "SIZEOF_LONG" },
181 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 712 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
182 { sizeof (cl_half ), "SIZEOF_HALF" }, 713 { sizeof (cl_half ), "SIZEOF_HALF" },
183 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 714 { sizeof (cl_float ), "SIZEOF_FLOAT" },
184 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 715 { sizeof (cl_double), "SIZEOF_DOUBLE" },
185#include "constiv.h" 716#include "constiv.h"
186 }; 717 };
718
187 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
188 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
729 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
730 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
731 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
732 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
733 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
734 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
735 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
736 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
737 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
738 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
739 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
740 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
741 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
742 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
743 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
744
745 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
189} 746}
190 747
191cl_int 748cl_int
192errno () 749errno ()
193 CODE: 750 CODE:
194 errno = res; 751 RETVAL = res;
752 OUTPUT:
753 RETVAL
195 754
196const char * 755const char *
197err2str (cl_int err) 756err2str (cl_int err = res)
198 757
199const char * 758const char *
200enum2str (cl_uint value) 759enum2str (cl_uint value)
201 760
202void 761void
210 list = tmpbuf (sizeof (*list) * count); 769 list = tmpbuf (sizeof (*list) * count);
211 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
212 771
213 EXTEND (SP, count); 772 EXTEND (SP, count);
214 for (i = 0; i < count; ++i) 773 for (i = 0; i < count; ++i)
215 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 774 PUSH_CLOBJ (stash_platform, (IV)list [i]);
216 775
217void 776void
218context_from_type (FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
219 PPCODE: 778 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK;
220 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (0, type, 0, 0, &res)); 780 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res));
221 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 781 XPUSH_CLOBJ_CONTEXT;
222 782
223void 783void
224context (FUTURE properties, FUTURE devices, FUTURE notify = 0) 784context (FUTURE properties, FUTURE devices, FUTURE notify)
225 PPCODE: 785 PPCODE:
226 /* der Gipfel der Kunst */ 786 /* der Gipfel der Kunst */
227 787
228void 788void
229wait_for_events (...) 789wait_for_events (...)
230 CODE: 790 CODE:
231 EVENT_LIST (0, items); 791 EVENT_LIST (0);
232 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 792 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
233 793
234PROTOTYPES: DISABLE 794PROTOTYPES: DISABLE
235 795
236MODULE = OpenCL PACKAGE = OpenCL::Platform 796MODULE = OpenCL PACKAGE = OpenCL::Platform
237 797
238void 798void
239info (OpenCL::Platform this, cl_platform_info name) 799info (OpenCL::Platform self, cl_platform_info name)
240 PPCODE: 800 PPCODE:
241 INFO (Platform) 801 INFO (Platform)
242 802
243void 803void
804unload_compiler (OpenCL::Platform self)
805 CODE:
806#if CL_VERSION_1_2
807 clUnloadPlatformCompiler (self);
808#endif
809
810#BEGIN:platform
811
812void
813profile (OpenCL::Platform self)
814 ALIAS:
815 profile = CL_PLATFORM_PROFILE
816 version = CL_PLATFORM_VERSION
817 name = CL_PLATFORM_NAME
818 vendor = CL_PLATFORM_VENDOR
819 extensions = CL_PLATFORM_EXTENSIONS
820 PPCODE:
821 size_t size;
822 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
823 char *value = tmpbuf (size);
824 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
825 EXTEND (SP, 1);
826 const int i = 0;
827 PUSHs (sv_2mortal (newSVpv (value, 0)));
828
829#END:platform
830
831void
244devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 832devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
245 PPCODE: 833 PPCODE:
246 cl_device_id *list; 834 cl_device_id *list;
247 cl_uint count; 835 cl_uint count;
248 int i; 836 int i;
249 837
250 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 838 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
251 list = tmpbuf (sizeof (*list) * count); 839 list = tmpbuf (sizeof (*list) * count);
252 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 840 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
253 841
254 EXTEND (SP, count); 842 EXTEND (SP, count);
255 for (i = 0; i < count; ++i) 843 for (i = 0; i < count; ++i)
256 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 844 PUSH_CLOBJ (stash_device, list [i]);
257 845
258void 846void
259context (OpenCL::Platform this, FUTURE properties, SV *devices, FUTURE notify = 0) 847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
260 PPCODE: 848 PPCODE:
261 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
262 croak ("OpenCL::Platform argument 'device' must be an arrayref with device objects, in call"); 850 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
263 851
264 AV *av = (AV *)SvRV (devices); 852 AV *av = (AV *)SvRV (devices);
265 cl_uint num_devices = av_len (av) + 1; 853 cl_uint num_devices = av_len (av) + 1;
266 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
855
267 int i; 856 int i;
268
269 for (i = num_devices; i--; ) 857 for (i = num_devices; i--; )
270 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
271 859
860 CONTEXT_NOTIFY_CALLBACK;
272 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (0, num_devices, device_list, 0, 0, &res)); 861 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
273 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 862 XPUSH_CLOBJ_CONTEXT;
274 863
275void 864void
276context_from_type (OpenCL::Platform this, FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 865context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
277 PPCODE: 866 PPCODE:
278 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
868 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
869
870 CONTEXT_NOTIFY_CALLBACK;
279 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 871 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
280 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 872 XPUSH_CLOBJ_CONTEXT;
281 873
282MODULE = OpenCL PACKAGE = OpenCL::Device 874MODULE = OpenCL PACKAGE = OpenCL::Device
283 875
284void 876void
285info (OpenCL::Device this, cl_device_info name) 877info (OpenCL::Device self, cl_device_info name)
286 PPCODE: 878 PPCODE:
287 INFO (Device) 879 INFO (Device)
288 880
881#BEGIN:device
882
883void
884type (OpenCL::Device self)
885 PPCODE:
886 cl_device_type value [1];
887 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
888 EXTEND (SP, 1);
889 const int i = 0;
890 PUSHs (sv_2mortal (newSVuv (value [i])));
891
892void
893vendor_id (OpenCL::Device self)
894 ALIAS:
895 vendor_id = CL_DEVICE_VENDOR_ID
896 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
897 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
898 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
899 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
900 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
901 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
902 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
903 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
904 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
905 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
906 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
907 image_support = CL_DEVICE_IMAGE_SUPPORT
908 max_samplers = CL_DEVICE_MAX_SAMPLERS
909 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
910 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
911 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
912 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
913 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
914 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
915 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
916 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
917 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
918 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
919 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
920 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
921 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
922 PPCODE:
923 cl_uint value [1];
924 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
925 EXTEND (SP, 1);
926 const int i = 0;
927 PUSHs (sv_2mortal (newSVuv (value [i])));
928
929void
930max_work_group_size (OpenCL::Device self)
931 ALIAS:
932 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
933 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
934 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
935 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
936 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
937 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
938 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
939 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
940 PPCODE:
941 size_t value [1];
942 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
943 EXTEND (SP, 1);
944 const int i = 0;
945 PUSHs (sv_2mortal (newSVuv (value [i])));
946
947void
948max_work_item_sizes (OpenCL::Device self)
949 PPCODE:
950 size_t size;
951 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
952 size_t *value = tmpbuf (size);
953 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
954 int i, n = size / sizeof (*value);
955 EXTEND (SP, n);
956 for (i = 0; i < n; ++i)
957 PUSHs (sv_2mortal (newSVuv (value [i])));
958
959void
960address_bits (OpenCL::Device self)
961 PPCODE:
962 cl_bitfield value [1];
963 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
964 EXTEND (SP, 1);
965 const int i = 0;
966 PUSHs (sv_2mortal (newSVuv (value [i])));
967
968void
969max_mem_alloc_size (OpenCL::Device self)
970 ALIAS:
971 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
972 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
973 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
974 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
975 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
976 PPCODE:
977 cl_ulong value [1];
978 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
979 EXTEND (SP, 1);
980 const int i = 0;
981 PUSHs (sv_2mortal (newSVuv (value [i])));
982
983void
984single_fp_config (OpenCL::Device self)
985 ALIAS:
986 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
987 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
988 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
989 PPCODE:
990 cl_device_fp_config value [1];
991 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
992 EXTEND (SP, 1);
993 const int i = 0;
994 PUSHs (sv_2mortal (newSVuv (value [i])));
995
996void
997global_mem_cache_type (OpenCL::Device self)
998 PPCODE:
999 cl_device_mem_cache_type value [1];
1000 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
1001 EXTEND (SP, 1);
1002 const int i = 0;
1003 PUSHs (sv_2mortal (newSVuv (value [i])));
1004
1005void
1006local_mem_type (OpenCL::Device self)
1007 PPCODE:
1008 cl_device_local_mem_type value [1];
1009 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
1010 EXTEND (SP, 1);
1011 const int i = 0;
1012 PUSHs (sv_2mortal (newSVuv (value [i])));
1013
1014void
1015error_correction_support (OpenCL::Device self)
1016 ALIAS:
1017 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1018 endian_little = CL_DEVICE_ENDIAN_LITTLE
1019 available = CL_DEVICE_AVAILABLE
1020 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1021 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1022 PPCODE:
1023 cl_bool value [1];
1024 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1025 EXTEND (SP, 1);
1026 const int i = 0;
1027 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1028
1029void
1030execution_capabilities (OpenCL::Device self)
1031 PPCODE:
1032 cl_device_exec_capabilities value [1];
1033 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1034 EXTEND (SP, 1);
1035 const int i = 0;
1036 PUSHs (sv_2mortal (newSVuv (value [i])));
1037
1038void
1039properties (OpenCL::Device self)
1040 PPCODE:
1041 cl_command_queue_properties value [1];
1042 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1043 EXTEND (SP, 1);
1044 const int i = 0;
1045 PUSHs (sv_2mortal (newSVuv (value [i])));
1046
1047void
1048platform (OpenCL::Device self)
1049 PPCODE:
1050 cl_platform_id value [1];
1051 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1052 EXTEND (SP, 1);
1053 const int i = 0;
1054 {
1055 PUSH_CLOBJ (stash_platform, value [i]);
1056 }
1057
1058void
1059name (OpenCL::Device self)
1060 ALIAS:
1061 name = CL_DEVICE_NAME
1062 vendor = CL_DEVICE_VENDOR
1063 driver_version = CL_DRIVER_VERSION
1064 profile = CL_DEVICE_PROFILE
1065 version = CL_DEVICE_VERSION
1066 extensions = CL_DEVICE_EXTENSIONS
1067 PPCODE:
1068 size_t size;
1069 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1070 char *value = tmpbuf (size);
1071 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1072 EXTEND (SP, 1);
1073 const int i = 0;
1074 PUSHs (sv_2mortal (newSVpv (value, 0)));
1075
1076void
1077parent_device_ext (OpenCL::Device self)
1078 PPCODE:
1079 cl_device_id value [1];
1080 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1081 EXTEND (SP, 1);
1082 const int i = 0;
1083 {
1084 PUSH_CLOBJ (stash_device, value [i]);
1085 }
1086
1087void
1088partition_types_ext (OpenCL::Device self)
1089 ALIAS:
1090 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1091 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
1092 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
1093 PPCODE:
1094 size_t size;
1095 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1096 cl_device_partition_property_ext *value = tmpbuf (size);
1097 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1098 int i, n = size / sizeof (*value);
1099 EXTEND (SP, n);
1100 for (i = 0; i < n; ++i)
1101 PUSHs (sv_2mortal (newSVuv (value [i])));
1102
1103#END:device
1104
289MODULE = OpenCL PACKAGE = OpenCL::Context 1105MODULE = OpenCL PACKAGE = OpenCL::Context
290 1106
291void 1107void
292DESTROY (OpenCL::Context context) 1108DESTROY (OpenCL::Context context)
293 CODE: 1109 CODE:
294 clReleaseContext (context); 1110 clReleaseContext (context);
295 1111
296void 1112void
297info (OpenCL::Context this, cl_context_info name) 1113info (OpenCL::Context self, cl_context_info name)
298 PPCODE: 1114 PPCODE:
299 INFO (Context) 1115 INFO (Context)
300 1116
301void 1117void
302queue (OpenCL::Context this, OpenCL::Device device, cl_command_queue_properties properties = 0) 1118queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
303 PPCODE: 1119 PPCODE:
304 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (this, device, properties, &res)); 1120 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
305 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1121 XPUSH_CLOBJ (stash_queue, queue);
306 1122
307void 1123void
308user_event (OpenCL::Context this) 1124user_event (OpenCL::Context self)
309 PPCODE: 1125 PPCODE:
310 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (this, &res)); 1126 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
311 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1127 XPUSH_CLOBJ (stash_userevent, ev);
312 1128
313void 1129void
314buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1130buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
315 PPCODE: 1131 PPCODE:
316 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1132 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?"); 1133 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
318 1134
319 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (this, flags, len, 0, &res)); 1135 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
320 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1136 XPUSH_CLOBJ (stash_bufferobj, mem);
321 1137
322void 1138void
323buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1139buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
324 PPCODE: 1140 PPCODE:
325 STRLEN len; 1141 STRLEN len;
326 char *ptr = SvPVbyte (data, len); 1142 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
327
328 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1143 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?"); 1144 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)); 1145 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
332 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1146 XPUSH_CLOBJ (stash_bufferobj, mem);
333 1147
1148#if CL_VERSION_1_2
1149
334void 1150void
335image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1151image (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: 1152 PPCODE:
337 STRLEN len; 1153 STRLEN len;
338 char *ptr = SvPVbyte (data, len); 1154 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
339 const cl_image_format format = { channel_order, channel_type }; 1155 const cl_image_format format = { channel_order, channel_type };
1156 const cl_image_desc desc = {
1157 type,
1158 width, height, depth,
1159 array_size, row_pitch, slice_pitch,
1160 num_mip_level, num_samples,
1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1162 };
340 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (this, flags, &format, width, height, row_pitch, ptr, &res)); 1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
341 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1164 HV *stash = stash_image;
1165 switch (type)
1166 {
1167 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1168 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1169 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1170 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1171 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1172 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1173 }
1174 XPUSH_CLOBJ (stash, mem);
342 1175
1176#endif
1177
343void 1178void
344image3d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1179image2d (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: 1180 PPCODE:
346 STRLEN len; 1181 STRLEN len;
347 char *ptr = SvPVbyte (data, len); 1182 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
348 const cl_image_format format = { channel_order, channel_type }; 1183 const cl_image_format format = { channel_order, channel_type };
1184#if PREFER_1_1
1185 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1186#else
1187 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1188 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1189#endif
1190 XPUSH_CLOBJ (stash_image2d, mem);
1191
1192void
1193image3d (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)
1194 PPCODE:
1195 STRLEN len;
1196 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1197 const cl_image_format format = { channel_order, channel_type };
1198#if PREFER_1_1
349 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (this, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1199 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
350 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1200#else
1201 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1202 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1203#endif
1204 XPUSH_CLOBJ (stash_image3d, mem);
351 1205
1206#if cl_apple_gl_sharing || cl_khr_gl_sharing
1207
352void 1208void
1209gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1210 PPCODE:
1211 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1212 XPUSH_CLOBJ (stash_bufferobj, mem);
1213
1214void
1215gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1216 PPCODE:
1217 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1218 XPUSH_CLOBJ (stash_image2d, mem);
1219
1220#if CL_VERSION_1_2
1221
1222void
1223gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1224 ALIAS:
1225 PPCODE:
1226 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1227 cl_gl_object_type type;
1228 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1229 HV *stash = stash_memory;
1230 switch (type)
1231 {
1232 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1233 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1234 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1235 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1236 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1237 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1238 }
1239 XPUSH_CLOBJ (stash, mem);
1240
1241#endif
1242
1243void
1244gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1245 PPCODE:
1246#if PREFER_1_1
1247 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1248#else
1249 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1250#endif
1251 XPUSH_CLOBJ (stash_image2d, mem);
1252
1253void
1254gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1255 PPCODE:
1256#if PREFER_1_1
1257 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1258#else
1259 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1260#endif
1261 XPUSH_CLOBJ (stash_image3d, mem);
1262
1263#endif
1264
1265void
353supported_image_formats (OpenCL::Context this, cl_mem_flags flags, cl_mem_object_type image_type) 1266supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
354 PPCODE: 1267 PPCODE:
355{ 1268{
356 cl_uint count; 1269 cl_uint count;
357 cl_image_format *list; 1270 cl_image_format *list;
358 int i; 1271 int i;
359 1272
360 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1273 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
361 Newx (list, count, cl_image_format); 1274 Newx (list, count, cl_image_format);
362 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1275 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
363 1276
364 EXTEND (SP, count); 1277 EXTEND (SP, count);
365 for (i = 0; i < count; ++i) 1278 for (i = 0; i < count; ++i)
366 { 1279 {
367 AV *av = newAV (); 1280 AV *av = newAV ();
370 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1283 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
371 } 1284 }
372} 1285}
373 1286
374void 1287void
375sampler (OpenCL::Context this, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1288sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
376 PPCODE: 1289 PPCODE:
377 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (this, normalized_coords, addressing_mode, filter_mode, &res)); 1290 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
378 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1291 XPUSH_CLOBJ (stash_sampler, sampler);
379 1292
380void 1293void
381program_with_source (OpenCL::Context this, SV *program) 1294program_with_source (OpenCL::Context self, SV *program)
382 PPCODE: 1295 PPCODE:
383 STRLEN len; 1296 STRLEN len;
384 size_t len2; 1297 size_t len2;
385 const char *ptr = SvPVbyte (program, len); 1298 const char *ptr = SvPVbyte (program, len);
386 1299
387 len2 = len; 1300 len2 = len;
388 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (this, 1, &ptr, &len2, &res)); 1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
389 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1302 XPUSH_CLOBJ (stash_program, prog);
1303
1304#BEGIN:context
1305
1306void
1307reference_count (OpenCL::Context self)
1308 ALIAS:
1309 reference_count = CL_CONTEXT_REFERENCE_COUNT
1310 num_devices = CL_CONTEXT_NUM_DEVICES
1311 PPCODE:
1312 cl_uint value [1];
1313 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1314 EXTEND (SP, 1);
1315 const int i = 0;
1316 PUSHs (sv_2mortal (newSVuv (value [i])));
1317
1318void
1319devices (OpenCL::Context self)
1320 PPCODE:
1321 size_t size;
1322 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1323 cl_device_id *value = tmpbuf (size);
1324 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1325 int i, n = size / sizeof (*value);
1326 EXTEND (SP, n);
1327 for (i = 0; i < n; ++i)
1328 {
1329 PUSH_CLOBJ (stash_device, value [i]);
1330 }
1331
1332void
1333properties (OpenCL::Context self)
1334 PPCODE:
1335 size_t size;
1336 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1337 cl_context_properties *value = tmpbuf (size);
1338 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1339 int i, n = size / sizeof (*value);
1340 EXTEND (SP, n);
1341 for (i = 0; i < n; ++i)
1342 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1343
1344#END:context
390 1345
391MODULE = OpenCL PACKAGE = OpenCL::Queue 1346MODULE = OpenCL PACKAGE = OpenCL::Queue
392 1347
393void 1348void
394DESTROY (OpenCL::Queue this) 1349DESTROY (OpenCL::Queue self)
395 CODE: 1350 CODE:
396 clReleaseCommandQueue (this); 1351 clReleaseCommandQueue (self);
397 1352
398void 1353void
399info (OpenCL::Queue this, cl_command_queue_info name)
400 PPCODE:
401 INFO (CommandQueue)
402
403void
404enqueue_read_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS:
1356 enqueue_read_buffer = 0
405 PPCODE: 1357 PPCODE:
406 cl_event ev = 0; 1358 cl_event ev = 0;
407 EVENT_LIST (6, items - 6); 1359 EVENT_LIST (6);
408 1360
409 SvUPGRADE (data, SVt_PV); 1361 SvUPGRADE (data, SVt_PV);
410 SvGROW (data, len); 1362 SvGROW (data, len);
411 SvPOK_only (data); 1363 SvPOK_only (data);
412 SvCUR_set (data, len); 1364 SvCUR_set (data, len);
413 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1365 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
414 1366
415 if (ev) 1367 if (ev)
416 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1368 XPUSH_CLOBJ (stash_event, ev);
417 1369
418void 1370void
419enqueue_write_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS:
1373 enqueue_write_buffer = 0
420 PPCODE: 1374 PPCODE:
421 cl_event ev = 0; 1375 cl_event ev = 0;
422 STRLEN len; 1376 STRLEN len;
423 char *ptr = SvPVbyte (data, len); 1377 char *ptr = SvPVbyte (data, len);
424 EVENT_LIST (5, items - 5); 1378 EVENT_LIST (5);
425 1379
426 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1380 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
427 1381
428 if (ev) 1382 if (ev)
429 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1383 XPUSH_CLOBJ (stash_event, ev);
430 1384
1385#if CL_VERSION_1_2
1386
431void 1387void
432enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS:
1390 enqueue_fill_buffer = 0
433 PPCODE: 1391 PPCODE:
434 cl_event ev = 0; 1392 cl_event ev = 0;
1393 STRLEN len;
1394 char *ptr = SvPVbyte (data, len);
435 EVENT_LIST (6, items - 6); 1395 EVENT_LIST (5);
436 1396
437 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1397 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
438 1398
439 if (ev) 1399 if (ev)
440 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1400 XPUSH_CLOBJ (stash_event, ev);
441 1401
442void 1402void
443enqueue_read_image (OpenCL::Queue this, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1403fill_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, ...)
1404 ALIAS:
1405 enqueue_fill_image = 0
444 PPCODE: 1406 PPCODE:
445 cl_event ev = 0; 1407 cl_event ev = 0;
1408 STRLEN len;
1409 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412
1413 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1417 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1418 cl_image_format format;
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1421 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1422 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1425 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426
1427 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev);
1429
1430#endif
1431
1432void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS:
1435 enqueue_copy_buffer = 0
1436 PPCODE:
1437 cl_event ev = 0;
1438 EVENT_LIST (6);
1439
1440 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441
1442 if (ev)
1443 XPUSH_CLOBJ (stash_event, ev);
1444
1445void
1446read_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, ...)
1447 ALIAS:
1448 enqueue_read_buffer_rect = 0
1449 PPCODE:
1450 cl_event ev = 0;
1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
446 const size_t src_origin[3] = { src_x, src_y, src_z }; 1452 const size_t host_origin[3] = { host_x, host_y, host_z };
447 const size_t region[3] = { width, height, depth }; 1453 const size_t region[3] = { width, height, depth };
448 EVENT_LIST (12, items - 12); 1454 EVENT_LIST (17);
449 1455
450 if (!row_pitch) 1456 if (!buf_row_pitch)
451 row_pitch = img_row_pitch (src); 1457 buf_row_pitch = region [0];
452 1458
453 if (depth > 1 && !slice_pitch) 1459 if (!buf_slice_pitch)
454 slice_pitch = row_pitch * height; 1460 buf_slice_pitch = region [1] * buf_row_pitch;
455 1461
456 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1462 if (!host_row_pitch)
1463 host_row_pitch = region [0];
1464
1465 if (!host_slice_pitch)
1466 host_slice_pitch = region [1] * host_row_pitch;
1467
1468 size_t len = host_row_pitch * host_slice_pitch * region [2];
457 1469
458 SvUPGRADE (data, SVt_PV); 1470 SvUPGRADE (data, SVt_PV);
459 SvGROW (data, len); 1471 SvGROW (data, len);
460 SvPOK_only (data); 1472 SvPOK_only (data);
461 SvCUR_set (data, len); 1473 SvCUR_set (data, len);
462 NEED_SUCCESS (EnqueueReadImage, (this, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1474 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
463 1475
464 if (ev) 1476 if (ev)
465 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1477 XPUSH_CLOBJ (stash_event, ev);
466 1478
467void 1479void
1480write_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, ...)
1481 ALIAS:
1482 enqueue_write_buffer_rect = 0
1483 PPCODE:
1484 cl_event ev = 0;
1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1486 const size_t host_origin[3] = { host_x, host_y, host_z };
1487 const size_t region[3] = { width, height, depth };
1488 STRLEN len;
1489 char *ptr = SvPVbyte (data, len);
1490 EVENT_LIST (17);
1491
1492 if (!buf_row_pitch)
1493 buf_row_pitch = region [0];
1494
1495 if (!buf_slice_pitch)
1496 buf_slice_pitch = region [1] * buf_row_pitch;
1497
1498 if (!host_row_pitch)
1499 host_row_pitch = region [0];
1500
1501 if (!host_slice_pitch)
1502 host_slice_pitch = region [1] * host_row_pitch;
1503
1504 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1505
1506 if (len < min_len)
1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1508
1509 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));
1510
1511 if (ev)
1512 XPUSH_CLOBJ (stash_event, ev);
1513
1514void
1515copy_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, ...)
1516 ALIAS:
1517 enqueue_copy_buffer_rect = 0
1518 PPCODE:
1519 cl_event ev = 0;
1520 const size_t src_origin[3] = { src_x, src_y, src_z };
1521 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1522 const size_t region[3] = { width, height, depth };
1523 EVENT_LIST (16);
1524
1525 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));
1526
1527 if (ev)
1528 XPUSH_CLOBJ (stash_event, ev);
1529
1530void
1531read_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, ...)
1532 ALIAS:
1533 enqueue_read_image = 0
1534 PPCODE:
1535 cl_event ev = 0;
1536 const size_t src_origin[3] = { src_x, src_y, src_z };
1537 const size_t region[3] = { width, height, depth };
1538 EVENT_LIST (12);
1539
1540 if (!row_pitch)
1541 row_pitch = img_row_pitch (src);
1542
1543 if (depth > 1 && !slice_pitch)
1544 slice_pitch = row_pitch * height;
1545
1546 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1547
1548 SvUPGRADE (data, SVt_PV);
1549 SvGROW (data, len);
1550 SvPOK_only (data);
1551 SvCUR_set (data, len);
1552 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));
1553
1554 if (ev)
1555 XPUSH_CLOBJ (stash_event, ev);
1556
1557void
468enqueue_write_image (OpenCL::Queue this, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1558write_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, ...)
1559 ALIAS:
1560 enqueue_write_image = 0
469 PPCODE: 1561 PPCODE:
470 cl_event ev = 0; 1562 cl_event ev = 0;
471 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
472 const size_t region[3] = { width, height, depth }; 1564 const size_t region[3] = { width, height, depth };
473 STRLEN len; 1565 STRLEN len;
474 char *ptr = SvPVbyte (data, len); 1566 char *ptr = SvPVbyte (data, len);
475 EVENT_LIST (12, items - 12); 1567 EVENT_LIST (12);
476 1568
477 if (!row_pitch) 1569 if (!row_pitch)
478 row_pitch = img_row_pitch (dst); 1570 row_pitch = img_row_pitch (dst);
479 1571
480 if (depth > 1 && !slice_pitch) 1572 if (depth > 1 && !slice_pitch)
483 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1575 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
484 1576
485 if (len < min_len) 1577 if (len < min_len)
486 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
487 1579
488 NEED_SUCCESS (EnqueueWriteImage, (this, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1580 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
489 1581
490 if (ev) 1582 if (ev)
491 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1583 XPUSH_CLOBJ (stash_event, ev);
492 1584
493void 1585void
494enqueue_copy_buffer_rect (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1586copy_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, ...)
1587 ALIAS:
1588 enqueue_copy_image = 0
495 PPCODE: 1589 PPCODE:
496 cl_event ev = 0; 1590 cl_event ev = 0;
497 const size_t src_origin[3] = { src_x, src_y, src_z }; 1591 const size_t src_origin[3] = { src_x, src_y, src_z };
498 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1592 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
499 const size_t region[3] = { width, height, depth }; 1593 const size_t region[3] = { width, height, depth };
500 EVENT_LIST (16, items - 16); 1594 EVENT_LIST (12);
501 1595
502 NEED_SUCCESS (EnqueueCopyBufferRect, (this, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1596 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
503 1597
504 if (ev) 1598 if (ev)
505 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1599 XPUSH_CLOBJ (stash_event, ev);
506 1600
507void 1601void
508enqueue_copy_buffer_to_image (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1602copy_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, ...)
1603 ALIAS:
1604 enqueue_copy_image_to_buffer = 0
509 PPCODE: 1605 PPCODE:
510 cl_event ev = 0; 1606 cl_event ev = 0;
1607 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610
1611 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612
1613 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev);
1615
1616void
1617copy_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, ...)
1618 ALIAS:
1619 enqueue_copy_buffer_to_image = 0
1620 PPCODE:
1621 cl_event ev = 0;
511 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625
1626 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627
1628 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev);
1630
1631void
1632map_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, size_t cb = 0, ...)
1633 ALIAS:
1634 enqueue_map_buffer = 0
1635 PPCODE:
1636 cl_event ev;
1637 EVENT_LIST (6);
1638
1639 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1640 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1641
1642void
1643map_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, size_t width = 0, size_t height = 0, size_t depth = 0, ...)
1644 ALIAS:
1645 enqueue_map_image = 0
1646 PPCODE:
1647 cl_event ev;
1648 const size_t origin[3] = { x, y, z };
512 const size_t region[3] = { width, height, depth }; 1649 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch;
513 EVENT_LIST (10, items - 10); 1651 EVENT_LIST (10);
514 1652
515 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1653 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));
1654
1655 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1]
1657 : region [0];
1658
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1660
1661void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID)
1666 {
1667 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event);
1669 }
1670
1671void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS:
1674 enqueue_task = 0
1675 PPCODE:
1676 cl_event ev = 0;
1677 EVENT_LIST (2);
1678
1679 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
516 1680
517 if (ev) 1681 if (ev)
518 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1682 XPUSH_CLOBJ (stash_event, ev);
519 1683
520void 1684void
521enqueue_copy_image (OpenCL::Queue this, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
522 PPCODE:
523 cl_event ev = 0;
524 const size_t src_origin[3] = { src_x, src_y, src_z };
525 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
526 const size_t region[3] = { width, height, depth };
527 EVENT_LIST (12, items - 12);
528
529 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
530
531 if (ev)
532 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
533
534void
535enqueue_copy_image_to_buffer (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
536 PPCODE:
537 cl_event ev = 0;
538 const size_t src_origin[3] = { src_x, src_y, src_z };
539 const size_t region[3] = { width, height, depth };
540 EVENT_LIST (10, items - 10);
541
542 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
543
544 if (ev)
545 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
546
547void
548enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
549 PPCODE:
550 cl_event ev = 0;
551 EVENT_LIST (2, items - 2);
552
553 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
554
555 if (ev)
556 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
557
558void
559enqueue_nd_range_kernel (OpenCL::Queue this, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1686 ALIAS:
1687 enqueue_nd_range_kernel = 0
560 PPCODE: 1688 PPCODE:
561 cl_event ev = 0; 1689 cl_event ev = 0;
562 size_t *gwo = 0, *gws, *lws = 0; 1690 size_t *gwo = 0, *gws, *lws = 0;
563 int gws_len; 1691 int gws_len;
564 size_t *lists; 1692 size_t *lists;
565 int i; 1693 int i;
566 EVENT_LIST (5, items - 5); 1694 EVENT_LIST (5);
567 1695
568 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
569 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
570 1698
571 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
572 1700
573 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1701 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
574 1702
575 gws = lists + gws_len * 0; 1703 gws = lists + gws_len * 0;
576 for (i = 0; i < gws_len; ++i) 1704 for (i = 0; i < gws_len; ++i)
1705 {
577 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1706 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1707 // at least nvidia crashes for 0-sized work group sizes, work around
1708 if (!gws [i])
1709 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1710 }
578 1711
579 if (SvOK (global_work_offset)) 1712 if (SvOK (global_work_offset))
580 { 1713 {
581 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1714 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
582 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1715 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
589 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1722 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
590 } 1723 }
591 1724
592 if (SvOK (local_work_size)) 1725 if (SvOK (local_work_size))
593 { 1726 {
594 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1727 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
595 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1728 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
596 1729
597 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1730 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
598 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1731 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
599 1732
600 lws = lists + gws_len * 2; 1733 lws = lists + gws_len * 2;
601 for (i = 0; i < gws_len; ++i) 1734 for (i = 0; i < gws_len; ++i)
1735 {
602 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1736 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1737 // at least nvidia crashes for 0-sized work group sizes, work around
1738 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 }
603 } 1741 }
604 1742
605 NEED_SUCCESS (EnqueueNDRangeKernel, (this, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1743 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
606 1744
607 if (ev) 1745 if (ev)
608 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1746 XPUSH_CLOBJ (stash_event, ev);
609 1747
1748#if cl_apple_gl_sharing || cl_khr_gl_sharing
1749
610void 1750void
611enqueue_marker (OpenCL::Queue this) 1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1752 ALIAS:
1753 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0
1755 enqueue_release_gl_objects = 1
612 PPCODE: 1756 PPCODE:
1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1759
613 cl_event ev; 1760 cl_event ev = 0;
614 NEED_SUCCESS (EnqueueMarker, (this, &ev)); 1761 EVENT_LIST (2);
615 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1762 AV *av = (AV *)SvRV (objects);
1763 cl_uint num_objects = av_len (av) + 1;
1764 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1765 int i;
616 1766
1767 for (i = num_objects; i--; )
1768 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1769
1770 if (ix)
1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1772 else
1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1774
1775 if (ev)
1776 XPUSH_CLOBJ (stash_event, ev);
1777
1778#endif
1779
617void 1780void
618enqueue_wait_for_events (OpenCL::Queue this, ...) 1781wait_for_events (OpenCL::Queue self, ...)
1782 ALIAS:
1783 enqueue_wait_for_events = 0
619 CODE: 1784 CODE:
620 EVENT_LIST (1, items - 1); 1785 EVENT_LIST (1);
1786#if PREFER_1_1
621 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 1787 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1788#else
1789 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1790#endif
622 1791
623void 1792void
624enqueue_barrier (OpenCL::Queue this) 1793marker (OpenCL::Queue self, ...)
1794 ALIAS:
1795 enqueue_marker = 0
625 CODE: 1796 PPCODE:
1797 cl_event ev = 0;
1798 EVENT_LIST (1);
1799#if PREFER_1_1
1800 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else
1803#if CL_VERSION_1_2
1804 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1805#else
1806 {
1807 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1808 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1809 }
1810#endif
1811#else
1812 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1813#endif
1814 if (ev)
1815 XPUSH_CLOBJ (stash_event, ev);
1816
1817void
1818barrier (OpenCL::Queue self, ...)
1819 ALIAS:
1820 enqueue_barrier = 0
1821 PPCODE:
1822 cl_event ev = 0;
1823 EVENT_LIST (1);
1824#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID)
626 NEED_SUCCESS (EnqueueBarrier, (this)); 1826 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else
1828#if CL_VERSION_1_2
1829 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1830#else
1831 {
1832 if (event_list_count)
1833 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
627 1834
1835 if (GIMME_V != G_VOID)
1836 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1837 }
1838#endif
1839#else
1840 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1841#endif
1842 if (ev)
1843 XPUSH_CLOBJ (stash_event, ev);
1844
628void 1845void
629flush (OpenCL::Queue this) 1846flush (OpenCL::Queue self)
630 CODE: 1847 CODE:
631 NEED_SUCCESS (Flush, (this)); 1848 NEED_SUCCESS (Flush, (self));
632 1849
633void 1850void
634finish (OpenCL::Queue this) 1851finish (OpenCL::Queue self)
635 CODE: 1852 CODE:
636 NEED_SUCCESS (Finish, (this)); 1853 NEED_SUCCESS (Finish, (self));
1854
1855void
1856info (OpenCL::Queue self, cl_command_queue_info name)
1857 PPCODE:
1858 INFO (CommandQueue)
1859
1860#BEGIN:command_queue
1861
1862void
1863context (OpenCL::Queue self)
1864 PPCODE:
1865 cl_context value [1];
1866 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1867 EXTEND (SP, 1);
1868 const int i = 0;
1869 {
1870 NEED_SUCCESS (RetainContext, (value [i]));
1871 PUSH_CLOBJ (stash_context, value [i]);
1872 }
1873
1874void
1875device (OpenCL::Queue self)
1876 PPCODE:
1877 cl_device_id value [1];
1878 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1879 EXTEND (SP, 1);
1880 const int i = 0;
1881 {
1882 PUSH_CLOBJ (stash_device, value [i]);
1883 }
1884
1885void
1886reference_count (OpenCL::Queue self)
1887 PPCODE:
1888 cl_uint value [1];
1889 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
1890 EXTEND (SP, 1);
1891 const int i = 0;
1892 PUSHs (sv_2mortal (newSVuv (value [i])));
1893
1894void
1895properties (OpenCL::Queue self)
1896 PPCODE:
1897 cl_command_queue_properties value [1];
1898 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1899 EXTEND (SP, 1);
1900 const int i = 0;
1901 PUSHs (sv_2mortal (newSVuv (value [i])));
1902
1903#END:command_queue
637 1904
638MODULE = OpenCL PACKAGE = OpenCL::Memory 1905MODULE = OpenCL PACKAGE = OpenCL::Memory
639 1906
640void 1907void
641DESTROY (OpenCL::Memory this) 1908DESTROY (OpenCL::Memory self)
642 CODE: 1909 CODE:
643 clReleaseMemObject (this); 1910 clReleaseMemObject (self);
644 1911
645void 1912void
646info (OpenCL::Memory this, cl_mem_info name) 1913info (OpenCL::Memory self, cl_mem_info name)
647 PPCODE: 1914 PPCODE:
648 INFO (MemObject) 1915 INFO (MemObject)
649 1916
1917#BEGIN:mem
1918
1919void
1920type (OpenCL::Memory self)
1921 PPCODE:
1922 cl_mem_object_type value [1];
1923 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1924 EXTEND (SP, 1);
1925 const int i = 0;
1926 PUSHs (sv_2mortal (newSVuv (value [i])));
1927
1928void
1929flags (OpenCL::Memory self)
1930 PPCODE:
1931 cl_mem_flags value [1];
1932 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1933 EXTEND (SP, 1);
1934 const int i = 0;
1935 PUSHs (sv_2mortal (newSVuv (value [i])));
1936
1937void
1938size (OpenCL::Memory self)
1939 ALIAS:
1940 size = CL_MEM_SIZE
1941 offset = CL_MEM_OFFSET
1942 PPCODE:
1943 size_t value [1];
1944 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1945 EXTEND (SP, 1);
1946 const int i = 0;
1947 PUSHs (sv_2mortal (newSVuv (value [i])));
1948
1949void
1950host_ptr (OpenCL::Memory self)
1951 PPCODE:
1952 void * value [1];
1953 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
1954 EXTEND (SP, 1);
1955 const int i = 0;
1956 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1957
1958void
1959map_count (OpenCL::Memory self)
1960 ALIAS:
1961 map_count = CL_MEM_MAP_COUNT
1962 reference_count = CL_MEM_REFERENCE_COUNT
1963 PPCODE:
1964 cl_uint value [1];
1965 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1966 EXTEND (SP, 1);
1967 const int i = 0;
1968 PUSHs (sv_2mortal (newSVuv (value [i])));
1969
1970void
1971context (OpenCL::Memory self)
1972 PPCODE:
1973 cl_context value [1];
1974 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1975 EXTEND (SP, 1);
1976 const int i = 0;
1977 {
1978 NEED_SUCCESS (RetainContext, (value [i]));
1979 PUSH_CLOBJ (stash_context, value [i]);
1980 }
1981
1982void
1983associated_memobject (OpenCL::Memory self)
1984 PPCODE:
1985 cl_mem value [1];
1986 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1987 EXTEND (SP, 1);
1988 const int i = 0;
1989 {
1990 NEED_SUCCESS (RetainMemObject, (value [i]));
1991 PUSH_CLOBJ (stash_memory, value [i]);
1992 }
1993
1994#END:mem
1995
1996#if cl_apple_gl_sharing || cl_khr_gl_sharing
1997
1998void
1999gl_object_info (OpenCL::Memory self)
2000 PPCODE:
2001 cl_gl_object_type type;
2002 cl_GLuint name;
2003 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2004 EXTEND (SP, 2);
2005 PUSHs (sv_2mortal (newSVuv (type)));
2006 PUSHs (sv_2mortal (newSVuv (name)));
2007
2008#endif
2009
2010MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2011
2012void
2013sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2014 PPCODE:
2015 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2016 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2017
2018 cl_buffer_region crdata = { origin, size };
2019
2020 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2021 XPUSH_CLOBJ (stash_buffer, mem);
2022
2023MODULE = OpenCL PACKAGE = OpenCL::Image
2024
2025void
2026image_info (OpenCL::Image self, cl_image_info name)
2027 PPCODE:
2028 INFO (Image)
2029
2030void
2031format (OpenCL::Image self)
2032 PPCODE:
2033 cl_image_format format;
2034 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2035 EXTEND (SP, 2);
2036 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2037 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2038
2039#BEGIN:image
2040
2041void
2042element_size (OpenCL::Image self)
2043 ALIAS:
2044 element_size = CL_IMAGE_ELEMENT_SIZE
2045 row_pitch = CL_IMAGE_ROW_PITCH
2046 slice_pitch = CL_IMAGE_SLICE_PITCH
2047 width = CL_IMAGE_WIDTH
2048 height = CL_IMAGE_HEIGHT
2049 depth = CL_IMAGE_DEPTH
2050 PPCODE:
2051 size_t value [1];
2052 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
2053 EXTEND (SP, 1);
2054 const int i = 0;
2055 PUSHs (sv_2mortal (newSVuv (value [i])));
2056
2057#END:image
2058
2059#if cl_apple_gl_sharing || cl_khr_gl_sharing
2060
2061#BEGIN:gl_texture
2062
2063void
2064target (OpenCL::Image self)
2065 PPCODE:
2066 cl_GLenum value [1];
2067 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2068 EXTEND (SP, 1);
2069 const int i = 0;
2070 PUSHs (sv_2mortal (newSVuv (value [i])));
2071
2072void
2073gl_mipmap_level (OpenCL::Image self)
2074 PPCODE:
2075 cl_GLint value [1];
2076 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2077 EXTEND (SP, 1);
2078 const int i = 0;
2079 PUSHs (sv_2mortal (newSViv (value [i])));
2080
2081#END:gl_texture
2082
2083#endif
2084
650MODULE = OpenCL PACKAGE = OpenCL::Sampler 2085MODULE = OpenCL PACKAGE = OpenCL::Sampler
651 2086
652void 2087void
653DESTROY (OpenCL::Sampler this) 2088DESTROY (OpenCL::Sampler self)
654 CODE: 2089 CODE:
655 clReleaseSampler (this); 2090 clReleaseSampler (self);
656 2091
657void 2092void
658info (OpenCL::Sampler this, cl_sampler_info name) 2093info (OpenCL::Sampler self, cl_sampler_info name)
659 PPCODE: 2094 PPCODE:
660 INFO (Sampler) 2095 INFO (Sampler)
661 2096
2097#BEGIN:sampler
2098
2099void
2100reference_count (OpenCL::Sampler self)
2101 PPCODE:
2102 cl_uint value [1];
2103 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0));
2104 EXTEND (SP, 1);
2105 const int i = 0;
2106 PUSHs (sv_2mortal (newSVuv (value [i])));
2107
2108void
2109context (OpenCL::Sampler self)
2110 PPCODE:
2111 cl_context value [1];
2112 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2113 EXTEND (SP, 1);
2114 const int i = 0;
2115 {
2116 NEED_SUCCESS (RetainContext, (value [i]));
2117 PUSH_CLOBJ (stash_context, value [i]);
2118 }
2119
2120void
2121normalized_coords (OpenCL::Sampler self)
2122 PPCODE:
2123 cl_addressing_mode value [1];
2124 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2125 EXTEND (SP, 1);
2126 const int i = 0;
2127 PUSHs (sv_2mortal (newSVuv (value [i])));
2128
2129void
2130addressing_mode (OpenCL::Sampler self)
2131 PPCODE:
2132 cl_filter_mode value [1];
2133 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2134 EXTEND (SP, 1);
2135 const int i = 0;
2136 PUSHs (sv_2mortal (newSVuv (value [i])));
2137
2138void
2139filter_mode (OpenCL::Sampler self)
2140 PPCODE:
2141 cl_bool value [1];
2142 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2143 EXTEND (SP, 1);
2144 const int i = 0;
2145 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2146
2147#END:sampler
2148
662MODULE = OpenCL PACKAGE = OpenCL::Program 2149MODULE = OpenCL PACKAGE = OpenCL::Program
663 2150
664void 2151void
665DESTROY (OpenCL::Program this) 2152DESTROY (OpenCL::Program self)
666 CODE: 2153 CODE:
667 clReleaseProgram (this); 2154 clReleaseProgram (self);
668 2155
669void 2156void
670info (OpenCL::Program this, cl_program_info name) 2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS:
2159 build_async = 1
671 PPCODE: 2160 CODE:
672 INFO (Program) 2161 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2162 void *user_data = 0;
2163 cl_uint num_devices = 0;
2164 cl_device_id *device_list = 0;
673 2165
674void 2166 if (SvOK (devices))
675build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2167 {
676 CODE: 2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2170
2171 AV *av = (AV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173
2174 if (num_devices)
2175 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2177 int count;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 }
2181 }
2182
2183 if (SvOK (notify))
2184 {
2185 NEED_SUCCESS (RetainProgram, (self));
2186 pfn_notify = eq_program_notify;
2187 user_data = SvREFCNT_inc (s_get_cv (notify));
2188 }
2189
2190 if (ix)
2191 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2192 else
677 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2193 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
678 2194
679void 2195void
680build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
681 PPCODE: 2197 PPCODE:
682 size_t size; 2198 size_t size;
683 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2199 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
684 SV *sv = sv_2mortal (newSV (size)); 2200 SV *sv = sv_2mortal (newSV (size));
685 SvUPGRADE (sv, SVt_PV); 2201 SvUPGRADE (sv, SVt_PV);
686 SvPOK_only (sv); 2202 SvPOK_only (sv);
687 SvCUR_set (sv, size); 2203 SvCUR_set (sv, size);
688 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2204 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
689 XPUSHs (sv); 2205 XPUSHs (sv);
690 2206
2207#BEGIN:program_build
2208
2209void
2210build_status (OpenCL::Program self, OpenCL::Device device)
2211 PPCODE:
2212 cl_build_status value [1];
2213 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2214 EXTEND (SP, 1);
2215 const int i = 0;
2216 PUSHs (sv_2mortal (newSViv (value [i])));
2217
2218void
2219build_options (OpenCL::Program self, OpenCL::Device device)
2220 ALIAS:
2221 build_options = CL_PROGRAM_BUILD_OPTIONS
2222 build_log = CL_PROGRAM_BUILD_LOG
2223 PPCODE:
2224 size_t size;
2225 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2226 char *value = tmpbuf (size);
2227 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2228 EXTEND (SP, 1);
2229 const int i = 0;
2230 PUSHs (sv_2mortal (newSVpv (value, 0)));
2231
2232#END:program_build
2233
691void 2234void
692kernel (OpenCL::Program program, SV *function) 2235kernel (OpenCL::Program program, SV *function)
693 PPCODE: 2236 PPCODE:
694 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2237 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
695 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2238 XPUSH_CLOBJ (stash_kernel, kernel);
2239
2240void
2241kernels_in_program (OpenCL::Program program)
2242 PPCODE:
2243 cl_uint num_kernels;
2244 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2245 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2246 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2247
2248 int i;
2249 EXTEND (SP, num_kernels);
2250 for (i = 0; i < num_kernels; ++i)
2251 PUSH_CLOBJ (stash_kernel, kernels [i]);
2252
2253void
2254info (OpenCL::Program self, cl_program_info name)
2255 PPCODE:
2256 INFO (Program)
2257
2258void
2259binaries (OpenCL::Program self)
2260 PPCODE:
2261 cl_uint n, i;
2262 size_t size;
2263
2264 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2265 if (!n) XSRETURN_EMPTY;
2266
2267 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2268 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2269 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2270 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2271
2272 EXTEND (SP, n);
2273 for (i = 0; i < n; ++i)
2274 {
2275 SV *sv = sv_2mortal (newSV (sizes [i]));
2276 SvUPGRADE (sv, SVt_PV);
2277 SvPOK_only (sv);
2278 SvCUR_set (sv, sizes [i]);
2279 ptrs [i] = (void *)SvPVX (sv);
2280 PUSHs (sv);
2281 }
2282
2283 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2284 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2285
2286#BEGIN:program
2287
2288void
2289reference_count (OpenCL::Program self)
2290 ALIAS:
2291 reference_count = CL_PROGRAM_REFERENCE_COUNT
2292 num_devices = CL_PROGRAM_NUM_DEVICES
2293 PPCODE:
2294 cl_uint value [1];
2295 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2296 EXTEND (SP, 1);
2297 const int i = 0;
2298 PUSHs (sv_2mortal (newSVuv (value [i])));
2299
2300void
2301context (OpenCL::Program self)
2302 PPCODE:
2303 cl_context value [1];
2304 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2305 EXTEND (SP, 1);
2306 const int i = 0;
2307 {
2308 NEED_SUCCESS (RetainContext, (value [i]));
2309 PUSH_CLOBJ (stash_context, value [i]);
2310 }
2311
2312void
2313devices (OpenCL::Program self)
2314 PPCODE:
2315 size_t size;
2316 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2317 cl_device_id *value = tmpbuf (size);
2318 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2319 int i, n = size / sizeof (*value);
2320 EXTEND (SP, n);
2321 for (i = 0; i < n; ++i)
2322 {
2323 PUSH_CLOBJ (stash_device, value [i]);
2324 }
2325
2326void
2327source (OpenCL::Program self)
2328 PPCODE:
2329 size_t size;
2330 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2331 char *value = tmpbuf (size);
2332 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2333 EXTEND (SP, 1);
2334 const int i = 0;
2335 PUSHs (sv_2mortal (newSVpv (value, 0)));
2336
2337void
2338binary_sizes (OpenCL::Program self)
2339 PPCODE:
2340 size_t size;
2341 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2342 size_t *value = tmpbuf (size);
2343 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2344 int i, n = size / sizeof (*value);
2345 EXTEND (SP, n);
2346 for (i = 0; i < n; ++i)
2347 PUSHs (sv_2mortal (newSVuv (value [i])));
2348
2349#END:program
696 2350
697MODULE = OpenCL PACKAGE = OpenCL::Kernel 2351MODULE = OpenCL PACKAGE = OpenCL::Kernel
698 2352
699void 2353void
700DESTROY (OpenCL::Kernel this) 2354DESTROY (OpenCL::Kernel self)
701 CODE: 2355 CODE:
702 clReleaseKernel (this); 2356 clReleaseKernel (self);
703 2357
704void 2358void
2359setf (OpenCL::Kernel self, const char *format, ...)
2360 CODE:
2361 int i;
2362 for (i = 2; ; ++i)
2363 {
2364 while (*format == ' ')
2365 ++format;
2366
2367 char type = *format++;
2368
2369 if (!type)
2370 break;
2371
2372 if (i >= items)
2373 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2374
2375 SV *sv = ST (i);
2376
2377 union
2378 {
2379 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2380 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2381 cl_half ch; cl_float cf; cl_double cd;
2382 cl_mem cm;
2383 cl_sampler ca;
2384 size_t cz;
2385 cl_event ce;
2386 } arg;
2387 size_t size;
2388 int nullarg = 0;
2389
2390 switch (type)
2391 {
2392 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2393 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2394 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2395 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2396 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2397 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2398 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2399 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2400
2401 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404
2405 case 'z': nullarg = 1; size = SvIV (sv); break;
2406
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2408 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2409 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410
2411 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 }
2414
2415 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2416 if (res)
2417 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2418 }
2419
2420 if (i != items)
2421 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2422
2423void
2424set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2425 CODE:
2426 clSetKernelArg (self, idx, sizeof (value), &value);
2427
2428void
2429set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2430 CODE:
2431 clSetKernelArg (self, idx, sizeof (value), &value);
2432
2433void
2434set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2435 CODE:
2436 clSetKernelArg (self, idx, sizeof (value), &value);
2437
2438void
2439set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2440 CODE:
2441 clSetKernelArg (self, idx, sizeof (value), &value);
2442
2443void
2444set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2445 CODE:
2446 clSetKernelArg (self, idx, sizeof (value), &value);
2447
2448void
2449set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2450 CODE:
2451 clSetKernelArg (self, idx, sizeof (value), &value);
2452
2453void
2454set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2455 CODE:
2456 clSetKernelArg (self, idx, sizeof (value), &value);
2457
2458void
2459set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2460 CODE:
2461 clSetKernelArg (self, idx, sizeof (value), &value);
2462
2463void
2464set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2465 CODE:
2466 clSetKernelArg (self, idx, sizeof (value), &value);
2467
2468void
2469set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2470 CODE:
2471 clSetKernelArg (self, idx, sizeof (value), &value);
2472
2473void
2474set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2475 CODE:
2476 clSetKernelArg (self, idx, sizeof (value), &value);
2477
2478void
2479set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2480 CODE:
2481 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2482
2483void
2484set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2485 CODE:
2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2487
2488void
2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
2493 CODE:
2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2495
2496void
2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2498 CODE:
2499 clSetKernelArg (self, idx, sizeof (value), &value);
2500
2501void
2502set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2503 CODE:
2504 clSetKernelArg (self, idx, size, 0);
2505
2506void
2507set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2508 CODE:
2509 clSetKernelArg (self, idx, sizeof (value), &value);
2510
2511void
705info (OpenCL::Kernel this, cl_kernel_info name) 2512info (OpenCL::Kernel self, cl_kernel_info name)
706 PPCODE: 2513 PPCODE:
707 INFO (Kernel) 2514 INFO (Kernel)
708 2515
2516#BEGIN:kernel
2517
709void 2518void
710set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2519function_name (OpenCL::Kernel self)
2520 PPCODE:
2521 size_t size;
2522 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2523 char *value = tmpbuf (size);
2524 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2525 EXTEND (SP, 1);
2526 const int i = 0;
2527 PUSHs (sv_2mortal (newSVpv (value, 0)));
2528
2529void
2530num_args (OpenCL::Kernel self)
2531 ALIAS:
2532 num_args = CL_KERNEL_NUM_ARGS
2533 reference_count = CL_KERNEL_REFERENCE_COUNT
2534 PPCODE:
2535 cl_uint value [1];
2536 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2537 EXTEND (SP, 1);
2538 const int i = 0;
2539 PUSHs (sv_2mortal (newSVuv (value [i])));
2540
2541void
2542context (OpenCL::Kernel self)
2543 PPCODE:
2544 cl_context value [1];
2545 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2546 EXTEND (SP, 1);
2547 const int i = 0;
2548 {
2549 NEED_SUCCESS (RetainContext, (value [i]));
2550 PUSH_CLOBJ (stash_context, value [i]);
2551 }
2552
2553void
2554program (OpenCL::Kernel self)
2555 PPCODE:
2556 cl_program value [1];
2557 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2558 EXTEND (SP, 1);
2559 const int i = 0;
2560 {
2561 NEED_SUCCESS (RetainProgram, (value [i]));
2562 PUSH_CLOBJ (stash_program, value [i]);
2563 }
2564
2565#END:kernel
2566
2567void
2568work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
711 CODE: 2569 PPCODE:
712 clSetKernelArg (this, idx, sizeof (value), &value); 2570 size_t size;
2571 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2572 SV *sv = sv_2mortal (newSV (size));
2573 SvUPGRADE (sv, SVt_PV);
2574 SvPOK_only (sv);
2575 SvCUR_set (sv, size);
2576 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2577 XPUSHs (sv);
713 2578
714void 2579#BEGIN:kernel_work_group
715set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value)
716 CODE:
717 clSetKernelArg (this, idx, sizeof (value), &value);
718 2580
719void 2581void
720set_short (OpenCL::Kernel this, cl_uint idx, cl_short value) 2582work_group_size (OpenCL::Kernel self, OpenCL::Device device)
721 CODE: 2583 ALIAS:
722 clSetKernelArg (this, idx, sizeof (value), &value); 2584 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2585 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2586 PPCODE:
2587 size_t value [1];
2588 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2589 EXTEND (SP, 1);
2590 const int i = 0;
2591 PUSHs (sv_2mortal (newSVuv (value [i])));
723 2592
724void 2593void
725set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2594compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
726 CODE: 2595 PPCODE:
727 clSetKernelArg (this, idx, sizeof (value), &value); 2596 size_t size;
2597 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2598 size_t *value = tmpbuf (size);
2599 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2600 int i, n = size / sizeof (*value);
2601 EXTEND (SP, n);
2602 for (i = 0; i < n; ++i)
2603 PUSHs (sv_2mortal (newSVuv (value [i])));
728 2604
729void 2605void
730set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2606local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
731 CODE: 2607 ALIAS:
732 clSetKernelArg (this, idx, sizeof (value), &value); 2608 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2609 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2610 PPCODE:
2611 cl_ulong value [1];
2612 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2613 EXTEND (SP, 1);
2614 const int i = 0;
2615 PUSHs (sv_2mortal (newSVuv (value [i])));
733 2616
734void 2617#END:kernel_work_group
735set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value)
736 CODE:
737 clSetKernelArg (this, idx, sizeof (value), &value);
738
739void
740set_long (OpenCL::Kernel this, cl_uint idx, cl_long value)
741 CODE:
742 clSetKernelArg (this, idx, sizeof (value), &value);
743
744void
745set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
746 CODE:
747 clSetKernelArg (this, idx, sizeof (value), &value);
748
749void
750set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
751 CODE:
752 clSetKernelArg (this, idx, sizeof (value), &value);
753
754void
755set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
756 CODE:
757 clSetKernelArg (this, idx, sizeof (value), &value);
758
759void
760set_double (OpenCL::Kernel this, cl_uint idx, cl_double value)
761 CODE:
762 clSetKernelArg (this, idx, sizeof (value), &value);
763
764void
765set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
766 CODE:
767 clSetKernelArg (this, idx, sizeof (value), &value);
768
769void
770set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
771 CODE:
772 clSetKernelArg (this, idx, sizeof (value), &value);
773
774void
775set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
776 CODE:
777 clSetKernelArg (this, idx, sizeof (value), &value);
778
779void
780set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
781 CODE:
782 clSetKernelArg (this, idx, sizeof (value), &value);
783
784void
785set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
786 CODE:
787 clSetKernelArg (this, idx, sizeof (value), &value);
788
789void
790set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
791 CODE:
792 clSetKernelArg (this, idx, sizeof (value), &value);
793 2618
794MODULE = OpenCL PACKAGE = OpenCL::Event 2619MODULE = OpenCL PACKAGE = OpenCL::Event
795 2620
796void 2621void
797DESTROY (OpenCL::Event this) 2622DESTROY (OpenCL::Event self)
798 CODE: 2623 CODE:
799 clReleaseEvent (this); 2624 clReleaseEvent (self);
800 2625
801void 2626void
2627wait (OpenCL::Event self)
2628 CODE:
2629 clWaitForEvents (1, &self);
2630
2631void
2632cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2633 CODE:
2634 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2635
2636void
802info (OpenCL::Event this, cl_event_info name) 2637info (OpenCL::Event self, cl_event_info name)
803 PPCODE: 2638 PPCODE:
804 INFO (Event) 2639 INFO (Event)
805 2640
2641#BEGIN:event
2642
806void 2643void
807wait (OpenCL::Event this) 2644command_queue (OpenCL::Event self)
2645 PPCODE:
2646 cl_command_queue value [1];
2647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2648 EXTEND (SP, 1);
2649 const int i = 0;
2650 {
2651 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2652 PUSH_CLOBJ (stash_queue, value [i]);
2653 }
2654
2655void
2656command_type (OpenCL::Event self)
2657 PPCODE:
2658 cl_command_type value [1];
2659 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2660 EXTEND (SP, 1);
2661 const int i = 0;
2662 PUSHs (sv_2mortal (newSVuv (value [i])));
2663
2664void
2665reference_count (OpenCL::Event self)
2666 ALIAS:
2667 reference_count = CL_EVENT_REFERENCE_COUNT
2668 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2669 PPCODE:
2670 cl_uint value [1];
2671 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2672 EXTEND (SP, 1);
2673 const int i = 0;
2674 PUSHs (sv_2mortal (newSVuv (value [i])));
2675
2676void
2677context (OpenCL::Event self)
2678 PPCODE:
2679 cl_context value [1];
2680 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2681 EXTEND (SP, 1);
2682 const int i = 0;
2683 {
2684 NEED_SUCCESS (RetainContext, (value [i]));
2685 PUSH_CLOBJ (stash_context, value [i]);
2686 }
2687
2688#END:event
2689
2690void
2691profiling_info (OpenCL::Event self, cl_profiling_info name)
808 CODE: 2692 PPCODE:
809 clWaitForEvents (1, &this); 2693 INFO (EventProfiling)
2694
2695#BEGIN:profiling
2696
2697void
2698profiling_command_queued (OpenCL::Event self)
2699 ALIAS:
2700 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
2701 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
2702 profiling_command_start = CL_PROFILING_COMMAND_START
2703 profiling_command_end = CL_PROFILING_COMMAND_END
2704 PPCODE:
2705 cl_ulong value [1];
2706 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
2707 EXTEND (SP, 1);
2708 const int i = 0;
2709 PUSHs (sv_2mortal (newSVuv (value [i])));
2710
2711#END:profiling
810 2712
811MODULE = OpenCL PACKAGE = OpenCL::UserEvent 2713MODULE = OpenCL PACKAGE = OpenCL::UserEvent
812 2714
813void 2715void
814set_status (OpenCL::UserEvent this, cl_int execution_status) 2716set_status (OpenCL::UserEvent self, cl_int execution_status)
815 CODE: 2717 CODE:
816 clSetUserEventStatus (this, execution_status); 2718 clSetUserEventStatus (self, execution_status);
817 2719
2720MODULE = OpenCL PACKAGE = OpenCL::Mapped
2721
2722void
2723DESTROY (SV *self)
2724 CODE:
2725 OpenCL__Mapped mapped = SvMAPPED (self);
2726
2727 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2728 mapped_detach (self, mapped);
2729
2730 clReleaseCommandQueue (mapped->queue);
2731 clReleaseEvent (mapped->event);
2732 Safefree (mapped);
2733
2734void
2735unmap (OpenCL::Mapped self, ...)
2736 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2738
2739bool
2740mapped (OpenCL::Mapped self)
2741 CODE:
2742 RETVAL = !!self->ptr;
2743 OUTPUT:
2744 RETVAL
2745
2746void
2747wait (OpenCL::Mapped self)
2748 PPCODE:
2749 if (self->event)
2750 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2751
2752void
2753event (OpenCL::Mapped self)
2754 PPCODE:
2755 if (!self->event)
2756 XSRETURN_UNDEF;
2757
2758 clRetainEvent (self->event);
2759 XPUSH_CLOBJ (stash_event, self->event);
2760
2761size_t
2762size (OpenCL::Mapped self)
2763 CODE:
2764 RETVAL = self->cb;
2765 OUTPUT:
2766 RETVAL
2767
2768IV
2769ptr (OpenCL::Mapped self)
2770 CODE:
2771 RETVAL = PTR2IV (self->ptr);
2772 OUTPUT:
2773 RETVAL
2774
2775MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2776
2777MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2778
2779IV
2780row_pitch (OpenCL::Mapped self)
2781 ALIAS:
2782 slice_pitch = 1
2783 CODE:
2784 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2785 OUTPUT:
2786 RETVAL
2787
2788
2789

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines