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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.30 by root, Thu Apr 19 14:34:56 2012 UTC vs.
Revision 1.64 by root, Thu May 3 23:30:08 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
10
5#ifdef I_DLFCN 11#ifdef I_DLFCN
6 #include <dlfcn.h> 12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
7#endif 30#endif
8 31
9#ifdef __APPLE__ 32#ifdef __APPLE__
10 #include <OpenCL/opencl.h> 33 #include <OpenCL/opencl.h>
11#else 34#else
12 #include <CL/opencl.h> 35 #include <CL/opencl.h>
36#endif
37
38#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1
40 #define PREFER_1_1 1
13#endif 41#endif
14 42
15typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
16typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
17typedef cl_context OpenCL__Context; 45typedef cl_context OpenCL__Context;
18typedef cl_command_queue OpenCL__Queue; 46typedef cl_command_queue OpenCL__Queue;
19typedef cl_mem OpenCL__Memory; 47typedef cl_mem OpenCL__Memory;
20typedef cl_mem OpenCL__Buffer; 48typedef cl_mem OpenCL__Buffer;
21typedef cl_mem OpenCL__BufferObj; 49typedef cl_mem OpenCL__BufferObj;
22typedef cl_mem OpenCL__Image; 50typedef cl_mem OpenCL__Image;
23typedef cl_mem OpenCL__Image2D;
24typedef cl_mem OpenCL__Image3D;
25typedef cl_mem OpenCL__Memory_ornull; 51typedef cl_mem OpenCL__Memory_ornull;
26typedef cl_mem OpenCL__Buffer_ornull; 52typedef cl_mem OpenCL__Buffer_ornull;
27typedef cl_mem OpenCL__Image_ornull; 53typedef cl_mem OpenCL__Image_ornull;
28typedef cl_mem OpenCL__Image2D_ornull;
29typedef cl_mem OpenCL__Image3D_ornull;
30typedef cl_sampler OpenCL__Sampler; 54typedef cl_sampler OpenCL__Sampler;
31typedef cl_program OpenCL__Program; 55typedef cl_program OpenCL__Program;
32typedef cl_kernel OpenCL__Kernel; 56typedef cl_kernel OpenCL__Kernel;
33typedef cl_event OpenCL__Event; 57typedef cl_event OpenCL__Event;
34typedef cl_event OpenCL__UserEvent; 58typedef cl_event OpenCL__UserEvent;
35 59
60typedef struct mapped * OpenCL__Mapped;
61
36typedef SV *FUTURE; 62typedef SV *FUTURE;
37 63
64static HV
65 *stash_platform,
66 *stash_device,
67 *stash_context,
68 *stash_queue,
69 *stash_program,
70 *stash_kernel,
71 *stash_sampler,
72 *stash_event,
73 *stash_userevent,
74 *stash_memory,
75 *stash_buffer,
76 *stash_bufferobj,
77 *stash_image,
78 *stash_image1d,
79 *stash_image1darray,
80 *stash_image1dbuffer,
81 *stash_image2d,
82 *stash_image2darray,
83 *stash_image3d,
84 *stash_mapped,
85 *stash_mappedbuffer,
86 *stash_mappedimage;
87
38/*****************************************************************************/ 88/*****************************************************************************/
39 89
40// name must include a leading underscore 90// name must include a leading underscore
91// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
92// for perl. doh.
41static void * 93static void *
42getsym (const char *name) 94glsym (const char *name)
43{ 95{
96 void *fun = 0;
97
44 #if defined I_DLFCN && defined RTLD_DEFAULT 98 #if defined I_DLFCN && defined RTLD_DEFAULT
45 #if !DLSYM_NEEDS_UNDERSCORE 99 fun = dlsym (RTLD_DEFAULT, name + 1);
46 ++name; // skip _
47 #endif
48 return dlsym (RTLD_DEFAULT, name); 100 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
49 #else 101
50 return 0; 102 if (!fun)
103 {
104 static void *libgl;
105 static const char *glso[] = {
106 "libGL.so.1",
107 "libGL.so.3",
108 "libGL.so.4.0",
109 "libGL.so",
110 "/usr/lib/libGL.so",
111 "/usr/X11R6/lib/libGL.1.dylib"
112 };
113 int i;
114
115 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
116 {
117 libgl = dlopen (glso [i], RTLD_LAZY);
118 if (libgl)
119 break;
120 }
121
122 if (libgl)
123 {
124 fun = dlsym (libgl, name + 1);
125 if (!fun) fun = dlsym (libgl, name);
126 }
127 }
51 #endif 128 #endif
129
130 return fun;
52} 131}
53 132
54/*****************************************************************************/ 133/*****************************************************************************/
55 134
56/* up to two temporary buffers */ 135/* up to two temporary buffers */
57static void * 136static void * ecb_noinline
58tmpbuf (size_t size) 137tmpbuf (size_t size)
59{ 138{
60 enum { buffers = 3 }; 139 enum { buffers = 4 };
61 static int idx; 140 static int idx;
62 static void *buf [buffers]; 141 static void *buf [buffers];
63 static size_t len [buffers]; 142 static size_t len [buffers];
64 143
65 idx = ++idx % buffers; 144 idx = (idx + 1) % buffers;
66 145
67 if (len [idx] < size) 146 if (len [idx] < size)
68 { 147 {
69 free (buf [idx]); 148 free (buf [idx]);
70 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 149 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
138 if (res) \ 217 if (res) \
139 FAIL (name); 218 FAIL (name);
140 219
141/*****************************************************************************/ 220/*****************************************************************************/
142 221
143#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
144#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
145
146static void *
147SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg)
148{
149 if (SvROK (sv) && sv_derived_from (sv, pkg))
150 return (void *)SvIV (SvRV (sv));
151
152 croak ("%s: %s is not of type %s", func, svname, pkg);
153}
154
155/*****************************************************************************/
156
157static cl_context_properties * 222static cl_context_properties * ecb_noinline
158SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 223SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
159{ 224{
160 if (!sv || !SvOK (sv)) 225 if (!sv || !SvOK (sv))
161 if (extra) 226 if (extra)
162 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 227 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
178 243
179 for (i = 0; i < len; i += 2) 244 for (i = 0; i < len; i += 2)
180 { 245 {
181 cl_context_properties t = SvIV (*av_fetch (av, i , 0)); 246 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
182 SV *p_sv = *av_fetch (av, i + 1, 0); 247 SV *p_sv = *av_fetch (av, i + 1, 0);
183 cl_context_properties v; 248 cl_context_properties v = SvIV (p_sv); // code below can override
184 249
185 switch (t) 250 switch (t)
186 { 251 {
252 case CL_GLX_DISPLAY_KHR:
253 if (!SvOK (p_sv))
254 {
255 void *func = glsym ("_glXGetCurrentDisplay");
256 if (func)
257 v = (cl_context_properties)((void *(*)(void))func)();
258 }
259 break;
260
261 case CL_GL_CONTEXT_KHR:
262 if (!SvOK (p_sv))
263 {
264 void *func = glsym ("_glXGetCurrentContext");
265 if (func)
266 v = (cl_context_properties)((void *(*)(void))func)();
267 }
268 break;
269
187 default: 270 default:
188 /* unknown property, treat as int */ 271 /* unknown property, treat as int */
189 v = SvIV (p_sv);
190 break; 272 break;
191 } 273 }
192 274
193 *l++ = t; 275 *l++ = t;
194 *l++ = v; 276 *l++ = v;
202 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 284 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
203} 285}
204 286
205/*****************************************************************************/ 287/*****************************************************************************/
206 288
289static SV *
290new_clobj (HV *stash, IV id)
291{
292 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
293}
294
295#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
296#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
297
298/* cl objects are either \$iv, or [$iv, ...] */
299/* they can be upgraded at runtime to the array form */
300static void * ecb_noinline
301SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
302{
303 // sv_derived_from is quite slow :(
304 if (SvROK (sv) && sv_derived_from (sv, pkg))
305 return (void *)SvIV (SvRV (sv));
306
307 croak ("%s: %s is not of type %s", func, svname, pkg);
308}
309
310// the "no-inherit" version of the above
311static void * ecb_noinline
312SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
313{
314 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
315 return (void *)SvIV (SvRV (sv));
316
317 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
318}
319
320/*****************************************************************************/
321/* callback stuff */
322
323/* default context callback, log to stderr */
324static void CL_CALLBACK
325context_default_notify (const char *msg, const void *info, size_t cb, void *data)
326{
327 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
328}
329
330typedef struct
331{
332 int free_cb;
333 void (*push)(void *data1, void *data2, void *data3);
334} eq_vtbl;
335
336typedef struct eq_item
337{
338 struct eq_item *next;
339 eq_vtbl *vtbl;
340 SV *cb;
341 void *data1, *data2, *data3;
342} eq_item;
343
344static void (*eq_signal_func)(void *signal_arg, int value);
345static void *eq_signal_arg;
346static xmutex_t eq_lock = X_MUTEX_INIT;
347static eq_item *eq_head, *eq_tail;
348
349static void ecb_noinline
350eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
351{
352 eq_item *item = malloc (sizeof (eq_item));
353
354 item->next = 0;
355 item->vtbl = vtbl;
356 item->cb = cb;
357 item->data1 = data1;
358 item->data2 = data2;
359 item->data3 = data3;
360
361 X_LOCK (eq_lock);
362
363 *(eq_head ? &eq_tail->next : &eq_head) = item;
364 eq_tail = item;
365
366 X_UNLOCK (eq_lock);
367
368 eq_signal_func (eq_signal_arg, 0);
369}
370
371static eq_item *
372eq_dec (void)
373{
374 eq_item *res;
375
376 X_LOCK (eq_lock);
377
378 res = eq_head;
379 if (res)
380 eq_head = res->next;
381
382 X_UNLOCK (eq_lock);
383
384 return res;
385}
386
387static void
388eq_poll (void)
389{
390 eq_item *item;
391
392 while ((item = eq_dec ()))
393 {
394 ENTER;
395 SAVETMPS;
396
397 dSP;
398 PUSHMARK (SP);
399 EXTEND (SP, 2);
400
401 if (item->vtbl->free_cb)
402 sv_2mortal (item->cb);
403
404 PUTBACK;
405 item->vtbl->push (item->data1, item->data2, item->data3);
406
407 SV *cb = item->cb;
408 free (item);
409
410 call_sv (cb, G_DISCARD | G_VOID);
411
412 FREETMPS;
413 LEAVE;
414 }
415}
416
417static void
418eq_poll_interrupt (pTHX_ void *c_arg, int value)
419{
420 eq_poll ();
421}
422
423/*****************************************************************************/
424/* context notify */
425
426static void ecb_noinline
427eq_context_push (void *data1, void *data2, void *data3)
428{
429 dSP;
430 PUSHs (sv_2mortal (newSVpv (data1, 0)));
431 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
432 PUTBACK;
433
434 free (data1);
435 free (data2);
436}
437
438static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
439
440static void CL_CALLBACK
441eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
442{
443 void *pvt_copy = malloc (cb);
444 memcpy (pvt_copy, pvt, cb);
445 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
446}
447
448#define CONTEXT_NOTIFY_CALLBACK \
449 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
450 void *user_data = 0; \
451 \
452 if (SvOK (notify)) \
453 { \
454 pfn_notify = eq_context_notify; \
455 user_data = s_get_cv (notify); \
456 }
457
458static SV * ecb_noinline
459new_clobj_context (cl_context ctx, void *user_data)
460{
461 SV *sv = new_clobj (stash_context, (IV)ctx);
462
463 if (user_data)
464 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
465
466 return sv;
467}
468
469#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
470
471/*****************************************************************************/
472/* build/compile/link notify */
473
474static void
475eq_program_push (void *data1, void *data2, void *data3)
476{
477 dSP;
478 PUSH_CLOBJ (stash_program, data1);
479 PUTBACK;
480}
481
482static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
483
484static void CL_CALLBACK
485eq_program_notify (cl_program program, void *user_data)
486{
487 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
488}
489
490struct build_args
491{
492 cl_program program;
493 char *options;
494 void *user_data;
495 cl_uint num_devices;
496};
497
498X_THREAD_PROC (build_program_thread)
499{
500 struct build_args *arg = thr_arg;
501
502 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
503
504 if (arg->user_data)
505 eq_program_notify (arg->program, arg->user_data);
506 else
507 clReleaseProgram (arg->program);
508
509 free (arg->options);
510 free (arg);
511
512 return 0;
513}
514
515static void
516build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
517{
518 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
519
520 arg->program = program;
521 arg->options = strdup (options);
522 arg->user_data = user_data;
523 arg->num_devices = num_devices;
524 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
525
526 xthread_t id;
527 thread_create (&id, build_program_thread, arg);
528}
529
530/*****************************************************************************/
531/* event objects */
532
533static void
534eq_event_push (void *data1, void *data2, void *data3)
535{
536 dSP;
537 PUSH_CLOBJ (stash_event, data1);
538 PUSHs (sv_2mortal (newSViv ((IV)data2)));
539 PUTBACK;
540}
541
542static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
543
544static void CL_CALLBACK
545eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
546{
547 clRetainEvent (event);
548 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
549}
550
551/*****************************************************************************/
552/* utilities for XS code */
553
207static size_t 554static size_t
208img_row_pitch (cl_mem img) 555img_row_pitch (cl_mem img)
209{ 556{
210 size_t res; 557 size_t res;
211 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 558 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
212 return res; 559 return res;
213} 560}
214 561
215static cl_event * 562static cl_event * ecb_noinline
216event_list (SV **items, int count) 563event_list (SV **items, cl_uint *rcount, cl_event extra)
217{ 564{
218 if (!count) 565 cl_uint count = *rcount;
566
567 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
568 *rcount = count = 0;
569
570 if (!count && !extra)
219 return 0; 571 return 0;
220 572
221 cl_event *list = tmpbuf (sizeof (cl_event) * count); 573 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
574 int i = 0;
222 575
223 while (count--) 576 while (count--)
577 if (SvOK (items [count]))
224 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 578 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
225 579
580 if (extra)
581 list [i++] = extra;
582
583 *rcount = i;
584
226 return list; 585 return i ? list : 0;
227} 586}
228 587
229#define EVENT_LIST(items,count) \ 588#define EVENT_LIST(skip) \
230 cl_uint event_list_count = (count); \ 589 cl_uint event_list_count = items - (skip); \
231 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 590 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
232 591
233#define INFO(class) \ 592#define INFO(class) \
234{ \ 593{ \
235 size_t size; \ 594 size_t size; \
236 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 595 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
240 SvCUR_set (sv, size); \ 599 SvCUR_set (sv, size); \
241 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 600 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
242 XPUSHs (sv); \ 601 XPUSHs (sv); \
243} 602}
244 603
604/*****************************************************************************/
605/* mapped_xxx */
606
607static OpenCL__Mapped
608SvMAPPED (SV *sv)
609{
610 // no typechecking atm., keep your fingers crossed
611 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
612}
613
614struct mapped
615{
616 cl_command_queue queue;
617 cl_mem memobj;
618 void *ptr;
619 size_t cb;
620 cl_event event;
621 size_t row_pitch;
622 size_t slice_pitch;
623};
624
625static SV *
626mapped_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)
627{
628 SV *data = newSV (0);
629 SvUPGRADE (data, SVt_PVMG);
630
631 OpenCL__Mapped mapped;
632 New (0, mapped, 1, struct mapped);
633
634 clRetainCommandQueue (queue);
635
636 mapped->queue = queue;
637 mapped->memobj = memobj;
638 mapped->ptr = ptr;
639 mapped->cb = cb;
640 mapped->event = ev;
641 mapped->row_pitch = row_pitch;
642 mapped->slice_pitch = slice_pitch;
643
644 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
645
646 if (SvLEN (data))
647 Safefree (data);
648
649 SvPVX (data) = (char *)ptr;
650 SvCUR_set (data, cb);
651 SvLEN_set (data, 0);
652 SvPOK_only (data);
653
654 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
655
656 if (!(flags & CL_MAP_WRITE))
657 SvREADONLY_on (data);
658
659 return obj;
660}
661
662static void
663mapped_detach (SV *sv, OpenCL__Mapped mapped)
664{
665 SV *data = SvRV (sv);
666
667 // the next check checks both before AND after detach, where SvPVX should STILL be 0
668 if (SvPVX (data) != (char *)mapped->ptr)
669 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
670 else
671 {
672 SvREADONLY_off (data);
673 SvCUR_set (data, 0);
674 SvPVX (data) = 0;
675 SvOK_off (data);
676 }
677
678 mapped->ptr = 0;
679}
680
681static void
682mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
683{
684 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
685 cl_event ev;
686
687 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
688
689 clReleaseEvent (mapped->event);
690 mapped->event = ev;
691
692 mapped_detach (self, mapped);
693}
694
695/*****************************************************************************/
696
245MODULE = OpenCL PACKAGE = OpenCL 697MODULE = OpenCL PACKAGE = OpenCL
246 698
247PROTOTYPES: ENABLE 699PROTOTYPES: ENABLE
700
701void
702poll ()
703 CODE:
704 eq_poll ();
705
706void
707_eq_initialise (IV func, IV arg)
708 CODE:
709 eq_signal_func = (void (*)(void *, int))func;
710 eq_signal_arg = (void*)arg;
248 711
249BOOT: 712BOOT:
250{ 713{
251 HV *stash = gv_stashpv ("OpenCL", 1); 714 HV *stash = gv_stashpv ("OpenCL", 1);
252 static const ivstr *civ, const_iv[] = { 715 static const ivstr *civ, const_iv[] = {
261 { sizeof (cl_half ), "SIZEOF_HALF" }, 724 { sizeof (cl_half ), "SIZEOF_HALF" },
262 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 725 { sizeof (cl_float ), "SIZEOF_FLOAT" },
263 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 726 { sizeof (cl_double), "SIZEOF_DOUBLE" },
264#include "constiv.h" 727#include "constiv.h"
265 }; 728 };
729
266 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 730 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
267 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 731 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
732
733 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
734 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
735 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
736 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
737 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
738 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
739 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
740 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
741 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
742 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
743 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
744 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
745 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
746 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
747 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
748 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
749 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
750 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
751 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
752 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
753 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
754 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
755
756 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
268} 757}
269 758
270cl_int 759cl_int
271errno () 760errno ()
272 CODE: 761 CODE:
273 errno = res; 762 RETVAL = res;
763 OUTPUT:
764 RETVAL
274 765
275const char * 766const char *
276err2str (cl_int err) 767err2str (cl_int err = res)
277 768
278const char * 769const char *
279enum2str (cl_uint value) 770enum2str (cl_uint value)
280 771
281void 772void
289 list = tmpbuf (sizeof (*list) * count); 780 list = tmpbuf (sizeof (*list) * count);
290 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 781 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
291 782
292 EXTEND (SP, count); 783 EXTEND (SP, count);
293 for (i = 0; i < count; ++i) 784 for (i = 0; i < count; ++i)
294 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 785 PUSH_CLOBJ (stash_platform, (IV)list [i]);
295 786
296void 787void
297context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 788context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
298 PPCODE: 789 PPCODE:
790 CONTEXT_NOTIFY_CALLBACK;
299 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 791 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
300 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 792 XPUSH_CLOBJ_CONTEXT;
301 793
302void 794void
303context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 795context (FUTURE properties, FUTURE devices, FUTURE notify)
304 PPCODE: 796 PPCODE:
305 /* der Gipfel der Kunst */ 797 /* der Gipfel der Kunst */
306 798
307void 799void
308wait_for_events (...) 800wait_for_events (...)
309 CODE: 801 CODE:
310 EVENT_LIST (0, items); 802 EVENT_LIST (0);
311 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 803 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
312 804
313PROTOTYPES: DISABLE 805PROTOTYPES: DISABLE
314 806
315MODULE = OpenCL PACKAGE = OpenCL::Platform 807MODULE = OpenCL PACKAGE = OpenCL::Platform
316 808
317void 809void
318info (OpenCL::Platform self, cl_platform_info name) 810info (OpenCL::Platform self, cl_platform_info name)
319 PPCODE: 811 PPCODE:
320 INFO (Platform) 812 INFO (Platform)
813
814void
815unload_compiler (OpenCL::Platform self)
816 CODE:
817#if CL_VERSION_1_2
818 clUnloadPlatformCompiler (self);
819#endif
321 820
322#BEGIN:platform 821#BEGIN:platform
323 822
324void 823void
325profile (OpenCL::Platform self) 824profile (OpenCL::Platform self)
351 list = tmpbuf (sizeof (*list) * count); 850 list = tmpbuf (sizeof (*list) * count);
352 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 851 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
353 852
354 EXTEND (SP, count); 853 EXTEND (SP, count);
355 for (i = 0; i < count; ++i) 854 for (i = 0; i < count; ++i)
356 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 855 PUSH_CLOBJ (stash_device, list [i]);
357 856
358void 857void
359context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 858context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
360 PPCODE: 859 PPCODE:
361 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 860 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
362 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 861 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
862
863 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
864 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
363 865
364 AV *av = (AV *)SvRV (devices); 866 AV *av = (AV *)SvRV (devices);
365 cl_uint num_devices = av_len (av) + 1; 867 cl_uint num_devices = av_len (av) + 1;
366 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 868 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
869
367 int i; 870 int i;
368
369 for (i = num_devices; i--; ) 871 for (i = num_devices; i--; )
370 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 872 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
371 873
874 CONTEXT_NOTIFY_CALLBACK;
372 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 875 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res));
373 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 876 XPUSH_CLOBJ_CONTEXT;
374 877
375void 878void
376context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 879context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
377 PPCODE: 880 PPCODE:
378 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 881 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
379 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 882 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
883
884 CONTEXT_NOTIFY_CALLBACK;
380 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 885 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
381 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 886 XPUSH_CLOBJ_CONTEXT;
382 887
383MODULE = OpenCL PACKAGE = OpenCL::Device 888MODULE = OpenCL PACKAGE = OpenCL::Device
384 889
385void 890void
386info (OpenCL::Device self, cl_device_info name) 891info (OpenCL::Device self, cl_device_info name)
394 PPCODE: 899 PPCODE:
395 cl_device_type value [1]; 900 cl_device_type value [1];
396 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
397 EXTEND (SP, 1); 902 EXTEND (SP, 1);
398 const int i = 0; 903 const int i = 0;
399 PUSHs (sv_2mortal (newSViv (value [i]))); 904 PUSHs (sv_2mortal (newSVuv (value [i])));
400 905
401void 906void
402vendor_id (OpenCL::Device self) 907vendor_id (OpenCL::Device self)
403 ALIAS: 908 ALIAS:
404 vendor_id = CL_DEVICE_VENDOR_ID 909 vendor_id = CL_DEVICE_VENDOR_ID
425 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 930 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
426 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 931 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
427 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 932 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
428 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 933 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
429 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 934 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
430 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 935 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
431 PPCODE: 936 PPCODE:
432 cl_uint value [1]; 937 cl_uint value [1];
433 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 938 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
434 EXTEND (SP, 1); 939 EXTEND (SP, 1);
435 const int i = 0; 940 const int i = 0;
549 PPCODE: 1054 PPCODE:
550 cl_command_queue_properties value [1]; 1055 cl_command_queue_properties value [1];
551 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1056 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
552 EXTEND (SP, 1); 1057 EXTEND (SP, 1);
553 const int i = 0; 1058 const int i = 0;
554 PUSHs (sv_2mortal (newSViv (value [i]))); 1059 PUSHs (sv_2mortal (newSVuv (value [i])));
555 1060
556void 1061void
557platform (OpenCL::Device self) 1062platform (OpenCL::Device self)
558 PPCODE: 1063 PPCODE:
559 cl_platform_id value [1]; 1064 cl_platform_id value [1];
560 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1065 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
561 EXTEND (SP, 1); 1066 EXTEND (SP, 1);
562 const int i = 0; 1067 const int i = 0;
563 { 1068 {
564 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1069 PUSH_CLOBJ (stash_platform, value [i]);
565 } 1070 }
566 1071
567void 1072void
568name (OpenCL::Device self) 1073name (OpenCL::Device self)
569 ALIAS: 1074 ALIAS:
588 cl_device_id value [1]; 1093 cl_device_id value [1];
589 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1094 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
590 EXTEND (SP, 1); 1095 EXTEND (SP, 1);
591 const int i = 0; 1096 const int i = 0;
592 { 1097 {
593 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1098 PUSH_CLOBJ (stash_device, value [i]);
594 } 1099 }
595 1100
596void 1101void
597partition_types_ext (OpenCL::Device self) 1102partition_types_ext (OpenCL::Device self)
598 ALIAS: 1103 ALIAS:
625 1130
626void 1131void
627queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1132queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
628 PPCODE: 1133 PPCODE:
629 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1134 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
630 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1135 XPUSH_CLOBJ (stash_queue, queue);
631 1136
632void 1137void
633user_event (OpenCL::Context self) 1138user_event (OpenCL::Context self)
634 PPCODE: 1139 PPCODE:
635 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1140 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
636 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1141 XPUSH_CLOBJ (stash_userevent, ev);
637 1142
638void 1143void
639buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1144buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
640 PPCODE: 1145 PPCODE:
641 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1146 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
642 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1147 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
643 1148
644 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1149 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
645 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1150 XPUSH_CLOBJ (stash_bufferobj, mem);
646 1151
647void 1152void
648buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1153buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
649 PPCODE: 1154 PPCODE:
650 STRLEN len; 1155 STRLEN len;
651 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1156 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
652 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1157 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
653 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1158 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
654 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1159 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
655 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1160 XPUSH_CLOBJ (stash_bufferobj, mem);
656 1161
1162#if CL_VERSION_1_2
1163
657void 1164void
658image2d (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) 1165image (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)
659 PPCODE: 1166 PPCODE:
660 STRLEN len; 1167 STRLEN len;
661 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1168 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
662 const cl_image_format format = { channel_order, channel_type }; 1169 const cl_image_format format = { channel_order, channel_type };
1170 const cl_image_desc desc = {
1171 type,
1172 width, height, depth,
1173 array_size, row_pitch, slice_pitch,
1174 num_mip_level, num_samples,
1175 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1176 };
663 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1177 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
664 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1178 HV *stash = stash_image;
1179 switch (type)
1180 {
1181 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1182 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1183 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1184 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1185 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1186 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1187 }
1188 XPUSH_CLOBJ (stash, mem);
665 1189
1190#endif
1191
666void 1192void
667image3d (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) 1193image2d (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)
668 PPCODE: 1194 PPCODE:
669 STRLEN len; 1195 STRLEN len;
670 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1196 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
671 const cl_image_format format = { channel_order, channel_type }; 1197 const cl_image_format format = { channel_order, channel_type };
1198#if PREFER_1_1
1199 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1200#else
1201 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1202 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1203#endif
1204 XPUSH_CLOBJ (stash_image2d, mem);
1205
1206void
1207image3d (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)
1208 PPCODE:
1209 STRLEN len;
1210 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1211 const cl_image_format format = { channel_order, channel_type };
1212#if PREFER_1_1
672 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1213 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
673 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1214#else
1215 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1216 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1217#endif
1218 XPUSH_CLOBJ (stash_image3d, mem);
674 1219
675#if cl_apple_gl_sharing || cl_khr_gl_sharing 1220#if cl_apple_gl_sharing || cl_khr_gl_sharing
676 1221
677void 1222void
678gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1223gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
679 PPCODE: 1224 PPCODE:
680 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1225 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
681 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1226 XPUSH_CLOBJ (stash_bufferobj, mem);
1227
1228void
1229gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1230 PPCODE:
1231 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1232 XPUSH_CLOBJ (stash_image2d, mem);
1233
1234#if CL_VERSION_1_2
1235
1236void
1237gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1238 ALIAS:
1239 PPCODE:
1240 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1241 cl_gl_object_type type;
1242 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1243 HV *stash = stash_memory;
1244 switch (type)
1245 {
1246 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1247 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1248 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1249 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1250 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1251 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1252 }
1253 XPUSH_CLOBJ (stash, mem);
1254
1255#endif
682 1256
683void 1257void
684gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1258gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
685 PPCODE: 1259 PPCODE:
1260#if PREFER_1_1
686 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1261 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
687 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1262#else
1263 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1264#endif
1265 XPUSH_CLOBJ (stash_image2d, mem);
688 1266
689void 1267void
690gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1268gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
691 PPCODE: 1269 PPCODE:
1270#if PREFER_1_1
692 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
693 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1272#else
694 1273 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
695void 1274#endif
696gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1275 XPUSH_CLOBJ (stash_image3d, mem);
697 PPCODE:
698 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
699 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
700 1276
701#endif 1277#endif
702 1278
703void 1279void
704supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1280supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
724 1300
725void 1301void
726sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1302sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
727 PPCODE: 1303 PPCODE:
728 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1304 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
729 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1305 XPUSH_CLOBJ (stash_sampler, sampler);
730 1306
731void 1307void
732program_with_source (OpenCL::Context self, SV *program) 1308program_with_source (OpenCL::Context self, SV *program)
733 PPCODE: 1309 PPCODE:
734 STRLEN len; 1310 STRLEN len;
735 size_t len2; 1311 size_t len2;
736 const char *ptr = SvPVbyte (program, len); 1312 const char *ptr = SvPVbyte (program, len);
737 1313
738 len2 = len; 1314 len2 = len;
739 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1315 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
740 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1316 XPUSH_CLOBJ (stash_program, prog);
1317
1318void
1319program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1320 PPCODE:
1321 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1322 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects");
1323
1324 devices = SvRV (devices);
1325
1326 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1327 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1328
1329 binaries = SvRV (binaries);
1330
1331 if (av_len ((AV *)devices) != av_len ((AV *)binaries))
1332 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1333
1334 int count = av_len ((AV *)devices) + 1;
1335 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1336 size_t *length_list = tmpbuf (sizeof (*length_list) * count);
1337 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count);
1338 cl_int *status_list = tmpbuf (sizeof (*status_list) * count);
1339
1340 int i;
1341 for (i = 0; i < count; ++i)
1342 {
1343 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1344 STRLEN len;
1345 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1346 length_list [i] = len;
1347 }
1348
1349 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res));
1350
1351 EXTEND (SP, 2);
1352 PUSH_CLOBJ (stash_program, prog);
1353
1354 if (GIMME_V == G_ARRAY)
1355 {
1356 AV *av = newAV ();
1357 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1358
1359 for (i = count; i--; )
1360 av_store (av, i, newSViv (status_list [i]));
1361 }
741 1362
742#BEGIN:context 1363#BEGIN:context
743 1364
744void 1365void
745reference_count (OpenCL::Context self) 1366reference_count (OpenCL::Context self)
762 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1383 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
763 int i, n = size / sizeof (*value); 1384 int i, n = size / sizeof (*value);
764 EXTEND (SP, n); 1385 EXTEND (SP, n);
765 for (i = 0; i < n; ++i) 1386 for (i = 0; i < n; ++i)
766 { 1387 {
767 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1388 PUSH_CLOBJ (stash_device, value [i]);
768 } 1389 }
769 1390
770void 1391void
771properties (OpenCL::Context self) 1392properties (OpenCL::Context self)
772 PPCODE: 1393 PPCODE:
787DESTROY (OpenCL::Queue self) 1408DESTROY (OpenCL::Queue self)
788 CODE: 1409 CODE:
789 clReleaseCommandQueue (self); 1410 clReleaseCommandQueue (self);
790 1411
791void 1412void
792enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1413read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1414 ALIAS:
1415 enqueue_read_buffer = 0
793 PPCODE: 1416 PPCODE:
794 cl_event ev = 0; 1417 cl_event ev = 0;
795 EVENT_LIST (6, items - 6); 1418 EVENT_LIST (6);
796 1419
797 SvUPGRADE (data, SVt_PV); 1420 SvUPGRADE (data, SVt_PV);
798 SvGROW (data, len); 1421 SvGROW (data, len);
799 SvPOK_only (data); 1422 SvPOK_only (data);
800 SvCUR_set (data, len); 1423 SvCUR_set (data, len);
801 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1424 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
802 1425
803 if (ev) 1426 if (ev)
804 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1427 XPUSH_CLOBJ (stash_event, ev);
805 1428
806void 1429void
807enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1430write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1431 ALIAS:
1432 enqueue_write_buffer = 0
808 PPCODE: 1433 PPCODE:
809 cl_event ev = 0; 1434 cl_event ev = 0;
810 STRLEN len; 1435 STRLEN len;
811 char *ptr = SvPVbyte (data, len); 1436 char *ptr = SvPVbyte (data, len);
812 EVENT_LIST (5, items - 5); 1437 EVENT_LIST (5);
813 1438
814 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1439 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
815 1440
816 if (ev) 1441 if (ev)
817 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1442 XPUSH_CLOBJ (stash_event, ev);
818 1443
1444#if CL_VERSION_1_2
1445
819void 1446void
820enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1447fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1448 ALIAS:
1449 enqueue_fill_buffer = 0
821 PPCODE: 1450 PPCODE:
822 cl_event ev = 0; 1451 cl_event ev = 0;
1452 STRLEN len;
1453 char *ptr = SvPVbyte (data, len);
1454 EVENT_LIST (5);
1455
1456 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1457
1458 if (ev)
1459 XPUSH_CLOBJ (stash_event, ev);
1460
1461void
1462fill_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, ...)
1463 ALIAS:
1464 enqueue_fill_image = 0
1465 PPCODE:
1466 cl_event ev = 0;
1467 const size_t origin [3] = { x, y, z };
1468 const size_t region [3] = { width, height, depth };
1469 EVENT_LIST (12);
1470
1471 const cl_float c_f [4] = { r, g, b, a };
1472 const cl_uint c_u [4] = { r, g, b, a };
1473 const cl_int c_s [4] = { r, g, b, a };
1474 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1475 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1476 cl_image_format format;
1477 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1478 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1479 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1480 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1481
1482 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1483 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1484
1485 if (ev)
1486 XPUSH_CLOBJ (stash_event, ev);
1487
1488#endif
1489
1490void
1491copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1492 ALIAS:
1493 enqueue_copy_buffer = 0
1494 PPCODE:
1495 cl_event ev = 0;
823 EVENT_LIST (6, items - 6); 1496 EVENT_LIST (6);
824 1497
825 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1498 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
826 1499
827 if (ev) 1500 if (ev)
828 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1501 XPUSH_CLOBJ (stash_event, ev);
829 1502
830void 1503void
831enqueue_read_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, ...) 1504read_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, ...)
1505 ALIAS:
1506 enqueue_read_buffer_rect = 0
832 PPCODE: 1507 PPCODE:
833 cl_event ev = 0; 1508 cl_event ev = 0;
834 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1509 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
835 const size_t host_origin[3] = { host_x, host_y, host_z }; 1510 const size_t host_origin[3] = { host_x, host_y, host_z };
836 const size_t region[3] = { width, height, depth }; 1511 const size_t region[3] = { width, height, depth };
837 EVENT_LIST (17, items - 17); 1512 EVENT_LIST (17);
838 1513
839 if (!buf_row_pitch) 1514 if (!buf_row_pitch)
840 buf_row_pitch = region [0]; 1515 buf_row_pitch = region [0];
841 1516
842 if (!buf_slice_pitch) 1517 if (!buf_slice_pitch)
855 SvPOK_only (data); 1530 SvPOK_only (data);
856 SvCUR_set (data, len); 1531 SvCUR_set (data, len);
857 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)); 1532 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));
858 1533
859 if (ev) 1534 if (ev)
860 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1535 XPUSH_CLOBJ (stash_event, ev);
861 1536
862void 1537void
863enqueue_write_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, ...) 1538write_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, ...)
1539 ALIAS:
1540 enqueue_write_buffer_rect = 0
864 PPCODE: 1541 PPCODE:
865 cl_event ev = 0; 1542 cl_event ev = 0;
866 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1543 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
867 const size_t host_origin[3] = { host_x, host_y, host_z }; 1544 const size_t host_origin[3] = { host_x, host_y, host_z };
868 const size_t region[3] = { width, height, depth }; 1545 const size_t region[3] = { width, height, depth };
869 STRLEN len; 1546 STRLEN len;
870 char *ptr = SvPVbyte (data, len); 1547 char *ptr = SvPVbyte (data, len);
871 EVENT_LIST (17, items - 17); 1548 EVENT_LIST (17);
872 1549
873 if (!buf_row_pitch) 1550 if (!buf_row_pitch)
874 buf_row_pitch = region [0]; 1551 buf_row_pitch = region [0];
875 1552
876 if (!buf_slice_pitch) 1553 if (!buf_slice_pitch)
885 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1562 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
886 1563
887 if (len < min_len) 1564 if (len < min_len)
888 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1565 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
889 1566
890 NEED_SUCCESS (EnqueueWriteBufferRect, (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)); 1567 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));
891 1568
892 if (ev) 1569 if (ev)
893 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1570 XPUSH_CLOBJ (stash_event, ev);
894 1571
895void 1572void
896enqueue_copy_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, ...) 1573copy_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, ...)
1574 ALIAS:
1575 enqueue_copy_buffer_rect = 0
897 PPCODE: 1576 PPCODE:
898 cl_event ev = 0; 1577 cl_event ev = 0;
899 const size_t src_origin[3] = { src_x, src_y, src_z }; 1578 const size_t src_origin[3] = { src_x, src_y, src_z };
900 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1579 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
901 const size_t region[3] = { width, height, depth }; 1580 const size_t region[3] = { width, height, depth };
902 EVENT_LIST (16, items - 16); 1581 EVENT_LIST (16);
903 1582
904 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)); 1583 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));
905 1584
906 if (ev) 1585 if (ev)
907 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1586 XPUSH_CLOBJ (stash_event, ev);
908 1587
909void 1588void
910enqueue_read_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, ...) 1589read_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, ...)
1590 ALIAS:
1591 enqueue_read_image = 0
911 PPCODE: 1592 PPCODE:
912 cl_event ev = 0; 1593 cl_event ev = 0;
913 const size_t src_origin[3] = { src_x, src_y, src_z }; 1594 const size_t src_origin[3] = { src_x, src_y, src_z };
914 const size_t region[3] = { width, height, depth }; 1595 const size_t region[3] = { width, height, depth };
915 EVENT_LIST (12, items - 12); 1596 EVENT_LIST (12);
916 1597
917 if (!row_pitch) 1598 if (!row_pitch)
918 row_pitch = img_row_pitch (src); 1599 row_pitch = img_row_pitch (src);
919 1600
920 if (depth > 1 && !slice_pitch) 1601 if (depth > 1 && !slice_pitch)
927 SvPOK_only (data); 1608 SvPOK_only (data);
928 SvCUR_set (data, len); 1609 SvCUR_set (data, len);
929 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)); 1610 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));
930 1611
931 if (ev) 1612 if (ev)
932 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1613 XPUSH_CLOBJ (stash_event, ev);
933 1614
934void 1615void
935enqueue_write_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, ...) 1616write_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, ...)
1617 ALIAS:
1618 enqueue_write_image = 0
936 PPCODE: 1619 PPCODE:
937 cl_event ev = 0; 1620 cl_event ev = 0;
938 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1621 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
939 const size_t region[3] = { width, height, depth }; 1622 const size_t region[3] = { width, height, depth };
940 STRLEN len; 1623 STRLEN len;
941 char *ptr = SvPVbyte (data, len); 1624 char *ptr = SvPVbyte (data, len);
942 EVENT_LIST (12, items - 12); 1625 EVENT_LIST (12);
943 1626
944 if (!row_pitch) 1627 if (!row_pitch)
945 row_pitch = img_row_pitch (dst); 1628 row_pitch = img_row_pitch (dst);
946 1629
947 if (depth > 1 && !slice_pitch) 1630 if (depth > 1 && !slice_pitch)
950 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1633 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
951 1634
952 if (len < min_len) 1635 if (len < min_len)
953 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1636 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
954 1637
955 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1638 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));
956 1639
957 if (ev) 1640 if (ev)
958 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1641 XPUSH_CLOBJ (stash_event, ev);
959 1642
960void 1643void
961enqueue_copy_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, ...) 1644copy_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, ...)
1645 ALIAS:
1646 enqueue_copy_image = 0
962 PPCODE: 1647 PPCODE:
963 cl_event ev = 0; 1648 cl_event ev = 0;
964 const size_t src_origin[3] = { src_x, src_y, src_z }; 1649 const size_t src_origin[3] = { src_x, src_y, src_z };
965 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1650 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
966 const size_t region[3] = { width, height, depth }; 1651 const size_t region[3] = { width, height, depth };
967 EVENT_LIST (12, items - 12); 1652 EVENT_LIST (12);
968 1653
969 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1654 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
970 1655
971 if (ev) 1656 if (ev)
972 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1657 XPUSH_CLOBJ (stash_event, ev);
973 1658
974void 1659void
975enqueue_copy_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, ...) 1660copy_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, ...)
1661 ALIAS:
1662 enqueue_copy_image_to_buffer = 0
976 PPCODE: 1663 PPCODE:
977 cl_event ev = 0; 1664 cl_event ev = 0;
978 const size_t src_origin[3] = { src_x, src_y, src_z }; 1665 const size_t src_origin[3] = { src_x, src_y, src_z };
1666 const size_t region [3] = { width, height, depth };
1667 EVENT_LIST (10);
1668
1669 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1670
1671 if (ev)
1672 XPUSH_CLOBJ (stash_event, ev);
1673
1674void
1675copy_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, ...)
1676 ALIAS:
1677 enqueue_copy_buffer_to_image = 0
1678 PPCODE:
1679 cl_event ev = 0;
1680 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1681 const size_t region [3] = { width, height, depth };
1682 EVENT_LIST (10);
1683
1684 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1685
1686 if (ev)
1687 XPUSH_CLOBJ (stash_event, ev);
1688
1689void
1690map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1691 ALIAS:
1692 enqueue_map_buffer = 0
1693 PPCODE:
1694 cl_event ev;
1695 EVENT_LIST (6);
1696 size_t cb = SvIV (cb_);
1697
1698 if (!SvOK (cb_))
1699 {
1700 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1701 cb -= offset;
1702 }
1703
1704 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1705 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1706
1707void
1708map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1709 ALIAS:
1710 enqueue_map_image = 0
1711 PPCODE:
1712 size_t width = SvIV (width_);
1713 if (!SvOK (width_))
1714 {
1715 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1716 width -= x;
1717 }
1718
1719 size_t height = SvIV (width_);
1720 if (!SvOK (height_))
1721 {
1722 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1723 height -= y;
1724 }
1725
1726 size_t depth = SvIV (width_);
1727 if (!SvOK (depth_))
1728 {
1729 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1730 depth -= z;
1731
1732 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1733 if (!depth)
1734 depth = 1;
1735 }
1736
1737 const size_t origin[3] = { x, y, z };
979 const size_t region[3] = { width, height, depth }; 1738 const size_t region[3] = { width, height, depth };
1739 size_t row_pitch, slice_pitch;
980 EVENT_LIST (10, items - 10); 1740 EVENT_LIST (10);
981 1741
1742 cl_event ev;
1743 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));
1744
1745 size_t cb = slice_pitch ? slice_pitch * region [2]
1746 : row_pitch ? row_pitch * region [1]
1747 : region [0];
1748
1749 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1750
1751void
1752unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1753 PPCODE:
1754 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1755 if (GIMME_V != G_VOID)
1756 {
1757 clRetainEvent (mapped->event);
1758 XPUSH_CLOBJ (stash_event, mapped->event);
1759 }
1760
1761void
1762task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1763 ALIAS:
1764 enqueue_task = 0
1765 PPCODE:
1766 cl_event ev = 0;
1767 EVENT_LIST (2);
1768
982 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1769 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
983 1770
984 if (ev) 1771 if (ev)
985 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1772 XPUSH_CLOBJ (stash_event, ev);
986 1773
987void 1774void
988enqueue_copy_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, ...)
989 PPCODE:
990 cl_event ev = 0;
991 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
992 const size_t region[3] = { width, height, depth };
993 EVENT_LIST (10, items - 10);
994
995 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
996
997 if (ev)
998 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
999
1000void
1001enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1002 PPCODE:
1003 cl_event ev = 0;
1004 EVENT_LIST (2, items - 2);
1005
1006 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1007
1008 if (ev)
1009 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1010
1011void
1012enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1775nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1776 ALIAS:
1777 enqueue_nd_range_kernel = 0
1013 PPCODE: 1778 PPCODE:
1014 cl_event ev = 0; 1779 cl_event ev = 0;
1015 size_t *gwo = 0, *gws, *lws = 0; 1780 size_t *gwo = 0, *gws, *lws = 0;
1016 int gws_len; 1781 int gws_len;
1017 size_t *lists; 1782 size_t *lists;
1018 int i; 1783 int i;
1019 EVENT_LIST (5, items - 5); 1784 EVENT_LIST (5);
1020 1785
1021 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1786 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1022 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1787 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1023 1788
1024 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1789 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1025 1790
1026 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1791 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1027 1792
1028 gws = lists + gws_len * 0; 1793 gws = lists + gws_len * 0;
1029 for (i = 0; i < gws_len; ++i) 1794 for (i = 0; i < gws_len; ++i)
1795 {
1030 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1796 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1797 // at least nvidia crashes for 0-sized work group sizes, work around
1798 if (!gws [i])
1799 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1800 }
1031 1801
1032 if (SvOK (global_work_offset)) 1802 if (SvOK (global_work_offset))
1033 { 1803 {
1034 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1804 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1035 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1805 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1042 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1812 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1043 } 1813 }
1044 1814
1045 if (SvOK (local_work_size)) 1815 if (SvOK (local_work_size))
1046 { 1816 {
1047 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1817 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1048 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1818 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1049 1819
1050 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1820 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1051 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1821 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1052 1822
1053 lws = lists + gws_len * 2; 1823 lws = lists + gws_len * 2;
1054 for (i = 0; i < gws_len; ++i) 1824 for (i = 0; i < gws_len; ++i)
1825 {
1055 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1826 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1827 // at least nvidia crashes for 0-sized work group sizes, work around
1828 if (!lws [i])
1829 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1830 }
1056 } 1831 }
1057 1832
1058 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1833 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1059 1834
1060 if (ev) 1835 if (ev)
1061 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1836 XPUSH_CLOBJ (stash_event, ev);
1062 1837
1063#if cl_apple_gl_sharing || cl_khr_gl_sharing 1838#if cl_apple_gl_sharing || cl_khr_gl_sharing
1064 1839
1065void 1840void
1066enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1841acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1067 ALIAS: 1842 ALIAS:
1843 release_gl_objects = 1
1844 enqueue_acquire_gl_objects = 0
1068 enqueue_release_gl_objects = 1 1845 enqueue_release_gl_objects = 1
1069 CODE: 1846 PPCODE:
1070 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1847 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1071 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1848 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1072 1849
1073 cl_event ev = 0; 1850 cl_event ev = 0;
1074 EVENT_LIST (2, items - 2); 1851 EVENT_LIST (2);
1075 AV *av = (AV *)SvRV (objects); 1852 AV *av = (AV *)SvRV (objects);
1076 cl_uint num_objects = av_len (av) + 1; 1853 cl_uint num_objects = av_len (av) + 1;
1077 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1854 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1078 int i; 1855 int i;
1079 1856
1080 for (i = num_objects; i--; ) 1857 for (i = num_objects; i--; )
1081 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory"); 1858 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1082 1859
1083 if (ix) 1860 if (ix)
1084 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1861 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1085 else 1862 else
1086 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1863 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1087 1864
1088 if (ev) 1865 if (ev)
1089 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1866 XPUSH_CLOBJ (stash_event, ev);
1090 1867
1091#endif 1868#endif
1092 1869
1093void 1870void
1094enqueue_marker (OpenCL::Queue self)
1095 PPCODE:
1096 cl_event ev;
1097 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1098 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1099
1100void
1101enqueue_wait_for_events (OpenCL::Queue self, ...) 1871wait_for_events (OpenCL::Queue self, ...)
1872 ALIAS:
1873 enqueue_wait_for_events = 0
1102 CODE: 1874 CODE:
1103 EVENT_LIST (1, items - 1); 1875 EVENT_LIST (1);
1876#if PREFER_1_1
1104 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1877 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1878#else
1879 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1880#endif
1105 1881
1106void 1882void
1107enqueue_barrier (OpenCL::Queue self) 1883marker (OpenCL::Queue self, ...)
1884 ALIAS:
1885 enqueue_marker = 0
1108 CODE: 1886 PPCODE:
1887 cl_event ev = 0;
1888 EVENT_LIST (1);
1889#if PREFER_1_1
1890 if (!event_list_count)
1891 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1892 else
1893#if CL_VERSION_1_2
1894 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1895#else
1896 {
1897 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
1898 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1899 }
1900#endif
1901#else
1902 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1903#endif
1904 if (ev)
1905 XPUSH_CLOBJ (stash_event, ev);
1906
1907void
1908barrier (OpenCL::Queue self, ...)
1909 ALIAS:
1910 enqueue_barrier = 0
1911 PPCODE:
1912 cl_event ev = 0;
1913 EVENT_LIST (1);
1914#if PREFER_1_1
1915 if (!event_list_count && GIMME_V == G_VOID)
1109 NEED_SUCCESS (EnqueueBarrier, (self)); 1916 NEED_SUCCESS (EnqueueBarrier, (self));
1917 else
1918#if CL_VERSION_1_2
1919 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1920#else
1921 {
1922 if (event_list_count)
1923 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1924
1925 if (GIMME_V != G_VOID)
1926 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1927 }
1928#endif
1929#else
1930 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1931#endif
1932 if (ev)
1933 XPUSH_CLOBJ (stash_event, ev);
1110 1934
1111void 1935void
1112flush (OpenCL::Queue self) 1936flush (OpenCL::Queue self)
1113 CODE: 1937 CODE:
1114 NEED_SUCCESS (Flush, (self)); 1938 NEED_SUCCESS (Flush, (self));
1132 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1956 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1133 EXTEND (SP, 1); 1957 EXTEND (SP, 1);
1134 const int i = 0; 1958 const int i = 0;
1135 { 1959 {
1136 NEED_SUCCESS (RetainContext, (value [i])); 1960 NEED_SUCCESS (RetainContext, (value [i]));
1137 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 1961 PUSH_CLOBJ (stash_context, value [i]);
1138 } 1962 }
1139 1963
1140void 1964void
1141device (OpenCL::Queue self) 1965device (OpenCL::Queue self)
1142 PPCODE: 1966 PPCODE:
1143 cl_device_id value [1]; 1967 cl_device_id value [1];
1144 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1968 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1145 EXTEND (SP, 1); 1969 EXTEND (SP, 1);
1146 const int i = 0; 1970 const int i = 0;
1147 { 1971 {
1148 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1972 PUSH_CLOBJ (stash_device, value [i]);
1149 } 1973 }
1150 1974
1151void 1975void
1152reference_count (OpenCL::Queue self) 1976reference_count (OpenCL::Queue self)
1153 PPCODE: 1977 PPCODE:
1162 PPCODE: 1986 PPCODE:
1163 cl_command_queue_properties value [1]; 1987 cl_command_queue_properties value [1];
1164 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1988 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1165 EXTEND (SP, 1); 1989 EXTEND (SP, 1);
1166 const int i = 0; 1990 const int i = 0;
1167 PUSHs (sv_2mortal (newSViv (value [i]))); 1991 PUSHs (sv_2mortal (newSVuv (value [i])));
1168 1992
1169#END:command_queue 1993#END:command_queue
1170 1994
1171MODULE = OpenCL PACKAGE = OpenCL::Memory 1995MODULE = OpenCL PACKAGE = OpenCL::Memory
1172 1996
1187 PPCODE: 2011 PPCODE:
1188 cl_mem_object_type value [1]; 2012 cl_mem_object_type value [1];
1189 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2013 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1190 EXTEND (SP, 1); 2014 EXTEND (SP, 1);
1191 const int i = 0; 2015 const int i = 0;
1192 PUSHs (sv_2mortal (newSViv (value [i]))); 2016 PUSHs (sv_2mortal (newSVuv (value [i])));
1193 2017
1194void 2018void
1195flags (OpenCL::Memory self) 2019flags (OpenCL::Memory self)
1196 PPCODE: 2020 PPCODE:
1197 cl_mem_flags value [1]; 2021 cl_mem_flags value [1];
1198 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2022 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1199 EXTEND (SP, 1); 2023 EXTEND (SP, 1);
1200 const int i = 0; 2024 const int i = 0;
1201 PUSHs (sv_2mortal (newSViv (value [i]))); 2025 PUSHs (sv_2mortal (newSVuv (value [i])));
1202 2026
1203void 2027void
1204size (OpenCL::Memory self) 2028size (OpenCL::Memory self)
1205 ALIAS: 2029 ALIAS:
1206 size = CL_MEM_SIZE 2030 size = CL_MEM_SIZE
1240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2064 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1241 EXTEND (SP, 1); 2065 EXTEND (SP, 1);
1242 const int i = 0; 2066 const int i = 0;
1243 { 2067 {
1244 NEED_SUCCESS (RetainContext, (value [i])); 2068 NEED_SUCCESS (RetainContext, (value [i]));
1245 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2069 PUSH_CLOBJ (stash_context, value [i]);
1246 } 2070 }
1247 2071
1248void 2072void
1249associated_memobject (OpenCL::Memory self) 2073associated_memobject (OpenCL::Memory self)
1250 PPCODE: 2074 PPCODE:
1252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2076 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1253 EXTEND (SP, 1); 2077 EXTEND (SP, 1);
1254 const int i = 0; 2078 const int i = 0;
1255 { 2079 {
1256 NEED_SUCCESS (RetainMemObject, (value [i])); 2080 NEED_SUCCESS (RetainMemObject, (value [i]));
1257 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2081 PUSH_CLOBJ (stash_memory, value [i]);
1258 } 2082 }
1259 2083
1260#END:mem 2084#END:mem
1261 2085
1262#if cl_apple_gl_sharing || cl_khr_gl_sharing 2086#if cl_apple_gl_sharing || cl_khr_gl_sharing
1264void 2088void
1265gl_object_info (OpenCL::Memory self) 2089gl_object_info (OpenCL::Memory self)
1266 PPCODE: 2090 PPCODE:
1267 cl_gl_object_type type; 2091 cl_gl_object_type type;
1268 cl_GLuint name; 2092 cl_GLuint name;
1269 NEED_SUCCESS (clGetGLObjectInfo, (self, &type, &name)); 2093 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
1270 EXTEND (SP, 2); 2094 EXTEND (SP, 2);
1271 PUSHs (sv_2mortal (newSVuv (type))); 2095 PUSHs (sv_2mortal (newSVuv (type)));
1272 PUSHs (sv_2mortal (newSVuv (name))); 2096 PUSHs (sv_2mortal (newSVuv (name)));
1273 2097
1274#endif 2098#endif
1282 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2106 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1283 2107
1284 cl_buffer_region crdata = { origin, size }; 2108 cl_buffer_region crdata = { origin, size };
1285 2109
1286 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2110 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1287 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2111 XPUSH_CLOBJ (stash_buffer, mem);
1288 2112
1289MODULE = OpenCL PACKAGE = OpenCL::Image 2113MODULE = OpenCL PACKAGE = OpenCL::Image
1290 2114
1291void 2115void
1292image_info (OpenCL::Image self, cl_image_info name) 2116image_info (OpenCL::Image self, cl_image_info name)
1293 PPCODE: 2117 PPCODE:
1294 INFO (Image) 2118 INFO (Image)
2119
2120void
2121format (OpenCL::Image self)
2122 PPCODE:
2123 cl_image_format format;
2124 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2125 EXTEND (SP, 2);
2126 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2127 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1295 2128
1296#BEGIN:image 2129#BEGIN:image
1297 2130
1298void 2131void
1299element_size (OpenCL::Image self) 2132element_size (OpenCL::Image self)
1319 2152
1320void 2153void
1321target (OpenCL::Image self) 2154target (OpenCL::Image self)
1322 PPCODE: 2155 PPCODE:
1323 cl_GLenum value [1]; 2156 cl_GLenum value [1];
1324 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0)); 2157 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
1325 EXTEND (SP, 1); 2158 EXTEND (SP, 1);
1326 const int i = 0; 2159 const int i = 0;
1327 PUSHs (sv_2mortal (newSVuv (value [i]))); 2160 PUSHs (sv_2mortal (newSVuv (value [i])));
1328 2161
1329void 2162void
1330gl_mipmap_level (OpenCL::Image self) 2163gl_mipmap_level (OpenCL::Image self)
1331 PPCODE: 2164 PPCODE:
1332 cl_GLint value [1]; 2165 cl_GLint value [1];
1333 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0)); 2166 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
1334 EXTEND (SP, 1); 2167 EXTEND (SP, 1);
1335 const int i = 0; 2168 const int i = 0;
1336 PUSHs (sv_2mortal (newSViv (value [i]))); 2169 PUSHs (sv_2mortal (newSViv (value [i])));
1337 2170
1338#END:gl_texture 2171#END:gl_texture
1369 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2202 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1370 EXTEND (SP, 1); 2203 EXTEND (SP, 1);
1371 const int i = 0; 2204 const int i = 0;
1372 { 2205 {
1373 NEED_SUCCESS (RetainContext, (value [i])); 2206 NEED_SUCCESS (RetainContext, (value [i]));
1374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2207 PUSH_CLOBJ (stash_context, value [i]);
1375 } 2208 }
1376 2209
1377void 2210void
1378normalized_coords (OpenCL::Sampler self) 2211normalized_coords (OpenCL::Sampler self)
1379 PPCODE: 2212 PPCODE:
1380 cl_addressing_mode value [1]; 2213 cl_addressing_mode value [1];
1381 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2214 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1382 EXTEND (SP, 1); 2215 EXTEND (SP, 1);
1383 const int i = 0; 2216 const int i = 0;
1384 PUSHs (sv_2mortal (newSViv (value [i]))); 2217 PUSHs (sv_2mortal (newSVuv (value [i])));
1385 2218
1386void 2219void
1387addressing_mode (OpenCL::Sampler self) 2220addressing_mode (OpenCL::Sampler self)
1388 PPCODE: 2221 PPCODE:
1389 cl_filter_mode value [1]; 2222 cl_filter_mode value [1];
1390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2223 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1391 EXTEND (SP, 1); 2224 EXTEND (SP, 1);
1392 const int i = 0; 2225 const int i = 0;
1393 PUSHs (sv_2mortal (newSViv (value [i]))); 2226 PUSHs (sv_2mortal (newSVuv (value [i])));
1394 2227
1395void 2228void
1396filter_mode (OpenCL::Sampler self) 2229filter_mode (OpenCL::Sampler self)
1397 PPCODE: 2230 PPCODE:
1398 cl_bool value [1]; 2231 cl_bool value [1];
1409DESTROY (OpenCL::Program self) 2242DESTROY (OpenCL::Program self)
1410 CODE: 2243 CODE:
1411 clReleaseProgram (self); 2244 clReleaseProgram (self);
1412 2245
1413void 2246void
1414build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2247build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2248 ALIAS:
2249 build_async = 1
1415 CODE: 2250 CODE:
2251 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2252 void *user_data = 0;
2253 cl_uint num_devices = 0;
2254 cl_device_id *device_list = 0;
2255
2256 if (SvOK (devices))
2257 {
2258 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
2259 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects.");
2260
2261 AV *av = (AV *)SvRV (devices);
2262 num_devices = av_len (av) + 1;
2263
2264 if (num_devices)
2265 {
2266 device_list = tmpbuf (sizeof (*device_list) * num_devices);
2267 int count;
2268 for (count = 0; count < num_devices; ++count)
2269 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2270 }
2271 }
2272
2273 if (SvOK (notify))
2274 {
2275 NEED_SUCCESS (RetainProgram, (self));
2276 pfn_notify = eq_program_notify;
2277 user_data = SvREFCNT_inc (s_get_cv (notify));
2278 }
2279
2280 if (ix)
2281 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2282 else
1416 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2283 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
1417 2284
1418void 2285void
1419build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2286build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1420 PPCODE: 2287 PPCODE:
1421 size_t size; 2288 size_t size;
1456 2323
1457void 2324void
1458kernel (OpenCL::Program program, SV *function) 2325kernel (OpenCL::Program program, SV *function)
1459 PPCODE: 2326 PPCODE:
1460 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2327 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1461 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2328 XPUSH_CLOBJ (stash_kernel, kernel);
2329
2330void
2331kernels_in_program (OpenCL::Program program)
2332 PPCODE:
2333 cl_uint num_kernels;
2334 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2335 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2336 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2337
2338 int i;
2339 EXTEND (SP, num_kernels);
2340 for (i = 0; i < num_kernels; ++i)
2341 PUSH_CLOBJ (stash_kernel, kernels [i]);
1462 2342
1463void 2343void
1464info (OpenCL::Program self, cl_program_info name) 2344info (OpenCL::Program self, cl_program_info name)
1465 PPCODE: 2345 PPCODE:
1466 INFO (Program) 2346 INFO (Program)
1484 { 2364 {
1485 SV *sv = sv_2mortal (newSV (sizes [i])); 2365 SV *sv = sv_2mortal (newSV (sizes [i]));
1486 SvUPGRADE (sv, SVt_PV); 2366 SvUPGRADE (sv, SVt_PV);
1487 SvPOK_only (sv); 2367 SvPOK_only (sv);
1488 SvCUR_set (sv, sizes [i]); 2368 SvCUR_set (sv, sizes [i]);
1489 ptrs [i] = SvPVX (sv); 2369 ptrs [i] = (void *)SvPVX (sv);
1490 PUSHs (sv); 2370 PUSHs (sv);
1491 } 2371 }
1492 2372
1493 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2373 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1494 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2374 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1514 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2394 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1515 EXTEND (SP, 1); 2395 EXTEND (SP, 1);
1516 const int i = 0; 2396 const int i = 0;
1517 { 2397 {
1518 NEED_SUCCESS (RetainContext, (value [i])); 2398 NEED_SUCCESS (RetainContext, (value [i]));
1519 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2399 PUSH_CLOBJ (stash_context, value [i]);
1520 } 2400 }
1521 2401
1522void 2402void
1523devices (OpenCL::Program self) 2403devices (OpenCL::Program self)
1524 PPCODE: 2404 PPCODE:
1528 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2408 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1529 int i, n = size / sizeof (*value); 2409 int i, n = size / sizeof (*value);
1530 EXTEND (SP, n); 2410 EXTEND (SP, n);
1531 for (i = 0; i < n; ++i) 2411 for (i = 0; i < n; ++i)
1532 { 2412 {
1533 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2413 PUSH_CLOBJ (stash_device, value [i]);
1534 } 2414 }
1535 2415
1536void 2416void
1537source (OpenCL::Program self) 2417source (OpenCL::Program self)
1538 PPCODE: 2418 PPCODE:
1564DESTROY (OpenCL::Kernel self) 2444DESTROY (OpenCL::Kernel self)
1565 CODE: 2445 CODE:
1566 clReleaseKernel (self); 2446 clReleaseKernel (self);
1567 2447
1568void 2448void
2449setf (OpenCL::Kernel self, const char *format, ...)
2450 CODE:
2451 int i;
2452 for (i = 2; ; ++i)
2453 {
2454 while (*format == ' ')
2455 ++format;
2456
2457 char type = *format++;
2458
2459 if (!type)
2460 break;
2461
2462 if (i >= items)
2463 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2464
2465 SV *sv = ST (i);
2466
2467 union
2468 {
2469 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2470 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2471 cl_half ch; cl_float cf; cl_double cd;
2472 cl_mem cm;
2473 cl_sampler ca;
2474 size_t cz;
2475 cl_event ce;
2476 } arg;
2477 size_t size;
2478 int nullarg = 0;
2479
2480 switch (type)
2481 {
2482 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2483 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2484 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2485 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2486 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2487 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2488 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2489 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2490
2491 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2492 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2493 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2494
2495 case 'z': nullarg = 1; size = SvIV (sv); break;
2496
2497 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2498 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2499 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2500
2501 default:
2502 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2503 }
2504
2505 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2506 if (res)
2507 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2508 }
2509
2510 if (i != items)
2511 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2512
2513void
1569set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2514set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1570 CODE: 2515 CODE:
1571 clSetKernelArg (self, idx, sizeof (value), &value); 2516 clSetKernelArg (self, idx, sizeof (value), &value);
1572 2517
1573void 2518void
1621 clSetKernelArg (self, idx, sizeof (value), &value); 2566 clSetKernelArg (self, idx, sizeof (value), &value);
1622 2567
1623void 2568void
1624set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2569set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1625 CODE: 2570 CODE:
2571 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2572
2573void
2574set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2575 CODE:
2576 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2577
2578void
2579set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2580 ALIAS:
2581 set_image2d = 0
2582 set_image3d = 0
2583 CODE:
2584 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2585
2586void
2587set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2588 CODE:
1626 clSetKernelArg (self, idx, sizeof (value), &value); 2589 clSetKernelArg (self, idx, sizeof (value), &value);
1627 2590
1628void 2591void
1629set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2592set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1630 CODE: 2593 CODE:
1631 clSetKernelArg (self, idx, sizeof (value), &value); 2594 clSetKernelArg (self, idx, size, 0);
1632
1633void
1634set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1635 CODE:
1636 clSetKernelArg (self, idx, sizeof (value), &value);
1637
1638void
1639set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1640 CODE:
1641 clSetKernelArg (self, idx, sizeof (value), &value);
1642
1643void
1644set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1645 CODE:
1646 clSetKernelArg (self, idx, sizeof (value), &value);
1647 2595
1648void 2596void
1649set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2597set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1650 CODE: 2598 CODE:
1651 clSetKernelArg (self, idx, sizeof (value), &value); 2599 clSetKernelArg (self, idx, sizeof (value), &value);
1687 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2635 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1688 EXTEND (SP, 1); 2636 EXTEND (SP, 1);
1689 const int i = 0; 2637 const int i = 0;
1690 { 2638 {
1691 NEED_SUCCESS (RetainContext, (value [i])); 2639 NEED_SUCCESS (RetainContext, (value [i]));
1692 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2640 PUSH_CLOBJ (stash_context, value [i]);
1693 } 2641 }
1694 2642
1695void 2643void
1696program (OpenCL::Kernel self) 2644program (OpenCL::Kernel self)
1697 PPCODE: 2645 PPCODE:
1699 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2647 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1700 EXTEND (SP, 1); 2648 EXTEND (SP, 1);
1701 const int i = 0; 2649 const int i = 0;
1702 { 2650 {
1703 NEED_SUCCESS (RetainProgram, (value [i])); 2651 NEED_SUCCESS (RetainProgram, (value [i]));
1704 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2652 PUSH_CLOBJ (stash_program, value [i]);
1705 } 2653 }
1706 2654
1707#END:kernel 2655#END:kernel
1708 2656
1709void 2657void
1769wait (OpenCL::Event self) 2717wait (OpenCL::Event self)
1770 CODE: 2718 CODE:
1771 clWaitForEvents (1, &self); 2719 clWaitForEvents (1, &self);
1772 2720
1773void 2721void
2722cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2723 CODE:
2724 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2725
2726void
1774info (OpenCL::Event self, cl_event_info name) 2727info (OpenCL::Event self, cl_event_info name)
1775 PPCODE: 2728 PPCODE:
1776 INFO (Event) 2729 INFO (Event)
1777 2730
1778#BEGIN:event 2731#BEGIN:event
1784 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2737 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1785 EXTEND (SP, 1); 2738 EXTEND (SP, 1);
1786 const int i = 0; 2739 const int i = 0;
1787 { 2740 {
1788 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2741 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1789 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2742 PUSH_CLOBJ (stash_queue, value [i]);
1790 } 2743 }
1791 2744
1792void 2745void
1793command_type (OpenCL::Event self) 2746command_type (OpenCL::Event self)
1794 PPCODE: 2747 PPCODE:
1817 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2770 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1818 EXTEND (SP, 1); 2771 EXTEND (SP, 1);
1819 const int i = 0; 2772 const int i = 0;
1820 { 2773 {
1821 NEED_SUCCESS (RetainContext, (value [i])); 2774 NEED_SUCCESS (RetainContext, (value [i]));
1822 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2775 PUSH_CLOBJ (stash_context, value [i]);
1823 } 2776 }
1824 2777
1825#END:event 2778#END:event
1826 2779
1827void 2780void
1852void 2805void
1853set_status (OpenCL::UserEvent self, cl_int execution_status) 2806set_status (OpenCL::UserEvent self, cl_int execution_status)
1854 CODE: 2807 CODE:
1855 clSetUserEventStatus (self, execution_status); 2808 clSetUserEventStatus (self, execution_status);
1856 2809
2810MODULE = OpenCL PACKAGE = OpenCL::Mapped
2811
2812void
2813DESTROY (SV *self)
2814 CODE:
2815 OpenCL__Mapped mapped = SvMAPPED (self);
2816
2817 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2818 mapped_detach (self, mapped);
2819
2820 clReleaseCommandQueue (mapped->queue);
2821 clReleaseEvent (mapped->event);
2822 Safefree (mapped);
2823
2824void
2825unmap (OpenCL::Mapped self, ...)
2826 CODE:
2827 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2828
2829bool
2830mapped (OpenCL::Mapped self)
2831 CODE:
2832 RETVAL = !!self->ptr;
2833 OUTPUT:
2834 RETVAL
2835
2836void
2837wait (OpenCL::Mapped self)
2838 PPCODE:
2839 if (self->event)
2840 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2841
2842void
2843event (OpenCL::Mapped self)
2844 PPCODE:
2845 if (!self->event)
2846 XSRETURN_UNDEF;
2847
2848 clRetainEvent (self->event);
2849 XPUSH_CLOBJ (stash_event, self->event);
2850
2851size_t
2852size (OpenCL::Mapped self)
2853 CODE:
2854 RETVAL = self->cb;
2855 OUTPUT:
2856 RETVAL
2857
2858IV
2859ptr (OpenCL::Mapped self)
2860 CODE:
2861 RETVAL = PTR2IV (self->ptr);
2862 OUTPUT:
2863 RETVAL
2864
2865void
2866set (OpenCL::Mapped self, size_t offset, SV *data)
2867 CODE:
2868 STRLEN len;
2869 const char *ptr = SvPVbyte (data, len);
2870
2871 if (offset + len > self->cb)
2872 croak ("OpenCL::Mapped::set out of bound condition detected");
2873
2874 memcpy (offset + (char *)self->ptr, ptr, len);
2875
2876MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2877
2878MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2879
2880IV
2881row_pitch (OpenCL::Mapped self)
2882 ALIAS:
2883 slice_pitch = 1
2884 CODE:
2885 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2886 OUTPUT:
2887 RETVAL
2888
2889
2890

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines