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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.16 by root, Tue Nov 22 10:29:18 2011 UTC vs.
Revision 1.80 by root, Sat May 5 16:44:02 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
4 6
7#include "ecb.h"//D
8
9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
10#include "xthread.h"
11#include "schmorp.h"
12
13#ifdef I_DLFCN
14 #include <dlfcn.h>
15#endif
16
17// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
18// but then fail to define the api functions unless you ALSO define
19// this. This breaks 100% of the opencl 1.1 apps, for what reason?
20// after all, the functions are deprecated, not removed.
21// in addition, you cannot test for this in any future-proof way.
22// each time a new opencl version comes out, you need to make a new
23// release.
24#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
25
26#ifndef PREFER_1_1
27 #define PREFER_1_1 1
28#endif
29
30#if PREFER_1_1
31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
32#endif
33
34#ifdef __APPLE__
35 #define CLHDR(name) <OpenCL/name>
36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
5#include <CL/opencl.h> 40#include CLHDR(opencl.h)
41
42#ifndef CL_VERSION_1_2
43 #include CLHDR(cl_d3d9.h)
44#endif
45
46#if _WIN32
47 #include CLHDR(cl_d3d10.h)
48 #if CL_VERSION_1_2
49 #include CLHDR<cl_d3d11.h>
50 #endif
51 #include CLHDR<cl_dx9_media_sharing.h.h>
52#endif
53
54#ifndef CL_VERSION_1_2
55 #undef PREFER_1_1
56 #define PREFER_1_1 1
57#endif
58
59// make sure all constants we might use are actually defined
60#include "default.h"
6 61
7typedef cl_platform_id OpenCL__Platform; 62typedef cl_platform_id OpenCL__Platform;
8typedef cl_device_id OpenCL__Device; 63typedef cl_device_id OpenCL__Device;
64typedef cl_device_id OpenCL__SubDevice;
9typedef cl_context OpenCL__Context; 65typedef cl_context OpenCL__Context;
10typedef cl_command_queue OpenCL__Queue; 66typedef cl_command_queue OpenCL__Queue;
11typedef cl_mem OpenCL__Memory; 67typedef cl_mem OpenCL__Memory;
12typedef cl_mem OpenCL__Buffer; 68typedef cl_mem OpenCL__Buffer;
69typedef cl_mem OpenCL__BufferObj;
13typedef cl_mem OpenCL__Image; 70typedef cl_mem OpenCL__Image;
14typedef cl_mem OpenCL__Image2D;
15typedef cl_mem OpenCL__Image3D;
16typedef cl_mem OpenCL__Memory_ornull; 71typedef cl_mem OpenCL__Memory_ornull;
17typedef cl_mem OpenCL__Buffer_ornull; 72typedef cl_mem OpenCL__Buffer_ornull;
18typedef cl_mem OpenCL__Image_ornull; 73typedef cl_mem OpenCL__Image_ornull;
19typedef cl_mem OpenCL__Image2D_ornull;
20typedef cl_mem OpenCL__Image3D_ornull;
21typedef cl_sampler OpenCL__Sampler; 74typedef cl_sampler OpenCL__Sampler;
22typedef cl_program OpenCL__Program; 75typedef cl_program OpenCL__Program;
23typedef cl_kernel OpenCL__Kernel; 76typedef cl_kernel OpenCL__Kernel;
24typedef cl_event OpenCL__Event; 77typedef cl_event OpenCL__Event;
25typedef cl_event OpenCL__UserEvent; 78typedef cl_event OpenCL__UserEvent;
26 79
27typedef SV *FUTURE; 80typedef struct mapped * OpenCL__Mapped;
81
82static HV
83 *stash_platform,
84 *stash_device,
85 *stash_subdevice,
86 *stash_context,
87 *stash_queue,
88 *stash_program,
89 *stash_kernel,
90 *stash_sampler,
91 *stash_event,
92 *stash_userevent,
93 *stash_memory,
94 *stash_buffer,
95 *stash_bufferobj,
96 *stash_image,
97 *stash_image1d,
98 *stash_image1darray,
99 *stash_image1dbuffer,
100 *stash_image2d,
101 *stash_image2darray,
102 *stash_image3d,
103 *stash_mapped,
104 *stash_mappedbuffer,
105 *stash_mappedimage;
28 106
29/*****************************************************************************/ 107/*****************************************************************************/
30 108
109// name must include a leading underscore
110// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
111// for perl. doh.
112static void *
113glsym (const char *name)
114{
115 void *fun = 0;
116
117 #if defined I_DLFCN && defined RTLD_DEFAULT
118 fun = dlsym (RTLD_DEFAULT, name + 1);
119 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
120
121 if (!fun)
122 {
123 static void *libgl;
124 static const char *glso[] = {
125 "libGL.so.1",
126 "libGL.so.3",
127 "libGL.so.4.0",
128 "libGL.so",
129 "/usr/lib/libGL.so",
130 "/usr/X11R6/lib/libGL.1.dylib"
131 };
132 int i;
133
134 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
135 {
136 libgl = dlopen (glso [i], RTLD_LAZY);
137 if (libgl)
138 break;
139 }
140
141 if (libgl)
142 {
143 fun = dlsym (libgl, name + 1);
144 if (!fun) fun = dlsym (libgl, name);
145 }
146 }
147 #endif
148
149 return fun;
150}
151
152/*****************************************************************************/
153
31/* up to two temporary buffers */ 154/* up to two temporary buffers */
32static void * 155static void * ecb_noinline
33tmpbuf (size_t size) 156tmpbuf (size_t size)
34{ 157{
158 enum { buffers = 4 };
35 static int idx; 159 static int idx;
36 static void *buf [2]; 160 static void *buf [buffers];
37 static size_t len [2]; 161 static size_t len [buffers];
38 162
39 idx ^= 1; 163 idx = (idx + 1) % buffers;
40 164
41 if (len [idx] < size) 165 if (len [idx] < size)
42 { 166 {
43 free (buf [idx]); 167 free (buf [idx]);
44 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 168 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
46 } 170 }
47 171
48 return buf [idx]; 172 return buf [idx];
49} 173}
50 174
175static const char * ecb_noinline
176cv_get_name (CV *cv)
177{
178 static char fullname [256];
179
180 GV *gv = CvGV (cv); // gv better be nonzero
181
182 HV *stash = GvSTASH (gv);
183 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
184 const char *gvname = GvNAME (gv);
185
186 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
187 return fullname;
188}
189
51/*****************************************************************************/ 190/*****************************************************************************/
52 191
53typedef struct 192typedef struct
54{ 193{
55 IV iv; 194 IV iv;
56 const char *name; 195 const char *name;
57 #define const_iv(name) { (IV)CL_ ## name, # name }, 196 #define const_iv(name) { (IV)CL_ ## name, # name },
58} ivstr; 197} ivstr;
198
199typedef struct
200{
201 NV nv;
202 const char *name;
203 #define const_nv(name) { (NV)CL_ ## name, # name },
204} nvstr;
59 205
60static const char * 206static const char *
61iv2str (IV value, const ivstr *base, int count, const char *fallback) 207iv2str (IV value, const ivstr *base, int count, const char *fallback)
62{ 208{
63 int i; 209 int i;
112 if (res) \ 258 if (res) \
113 FAIL (name); 259 FAIL (name);
114 260
115/*****************************************************************************/ 261/*****************************************************************************/
116 262
117#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
118#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
119
120static void * 263static SV *
264new_clobj (HV *stash, IV id)
265{
266 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
267}
268
269#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
270#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
271
272/* cl objects are either \$iv, or [$iv, ...] */
273/* they can be upgraded at runtime to the array form */
274static void * ecb_noinline
121SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 275SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
122{ 276{
277 // sv_derived_from is quite slow :(
123 if (SvROK (sv) && sv_derived_from (sv, pkg)) 278 if (SvROK (sv) && sv_derived_from (sv, pkg))
124 return (void *)SvIV (SvRV (sv)); 279 return (void *)SvIV (SvRV (sv));
125 280
126 croak ("%s: %s is not of type %s", func, svname, pkg); 281 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
282}
283
284// the "no-inherit" version of the above
285static void * ecb_noinline
286SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
287{
288 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
289 return (void *)SvIV (SvRV (sv));
290
291 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
127} 292}
128 293
129/*****************************************************************************/ 294/*****************************************************************************/
295
296static cl_context_properties * ecb_noinline
297SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
298{
299 if (!sv || !SvOK (sv))
300 if (extra)
301 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
302 else
303 return 0;
304
305 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
306 {
307 AV *av = (AV *)SvRV (sv);
308 int i, len = av_len (av) + 1;
309 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
310 cl_context_properties *l = p;
311
312 if (len & 1)
313 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
314
315 while (extracount--)
316 *l++ = *extra++;
317
318 for (i = 0; i < len; i += 2)
319 {
320 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
321 SV *p_sv = *av_fetch (av, i + 1, 0);
322 cl_context_properties v = SvIV (p_sv); // code below can override
323
324 switch (t)
325 {
326 case CL_CONTEXT_PLATFORM:
327 if (SvROK (p_sv))
328 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
329 break;
330
331 case CL_GLX_DISPLAY_KHR:
332 if (!SvOK (p_sv))
333 {
334 void *func = glsym ("_glXGetCurrentDisplay");
335 if (func)
336 v = (cl_context_properties)((void *(*)(void))func)();
337 }
338 break;
339
340 case CL_GL_CONTEXT_KHR:
341 if (!SvOK (p_sv))
342 {
343 void *func = glsym ("_glXGetCurrentContext");
344 if (func)
345 v = (cl_context_properties)((void *(*)(void))func)();
346 }
347 break;
348
349 default:
350 /* unknown property, treat as int */
351 break;
352 }
353
354 *l++ = t;
355 *l++ = v;
356 }
357
358 *l = 0;
359
360 return p;
361 }
362
363 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
364}
365
366// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
367// is a pointer (and only on well-behaved systems).
368static void * ecb_noinline
369object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
370{
371 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
372 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
373 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
374
375 AV *av = (AV *)SvRV (arg);
376 void **list = 0;
377 cl_uint count = av_len (av) + 1;
378
379 if (count)
380 {
381 list = tmpbuf (sizeof (*list) * count);
382 int i;
383 for (i = 0; i < count; ++i)
384 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
385 }
386
387 if (!count && !or_undef)
388 croak ("%s: '%s' must contain at least one %s object",
389 cv_get_name (cv), argname, klass);
390
391 *rcount = count;
392 return (void *)list;
393}
394
395/*****************************************************************************/
396/* callback stuff */
397
398/* default context callback, log to stderr */
399static void CL_CALLBACK
400context_default_notify (const char *msg, const void *info, size_t cb, void *data)
401{
402 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
403}
404
405typedef struct
406{
407 int free_cb;
408 void (*push)(void *data1, void *data2, void *data3);
409} eq_vtbl;
410
411typedef struct eq_item
412{
413 struct eq_item *next;
414 eq_vtbl *vtbl;
415 SV *cb;
416 void *data1, *data2, *data3;
417} eq_item;
418
419static void (*eq_signal_func)(void *signal_arg, int value);
420static void *eq_signal_arg;
421static xmutex_t eq_lock = X_MUTEX_INIT;
422static eq_item *eq_head, *eq_tail;
423
424static void ecb_noinline
425eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
426{
427 eq_item *item = malloc (sizeof (eq_item));
428
429 item->next = 0;
430 item->vtbl = vtbl;
431 item->cb = cb;
432 item->data1 = data1;
433 item->data2 = data2;
434 item->data3 = data3;
435
436 X_LOCK (eq_lock);
437
438 *(eq_head ? &eq_tail->next : &eq_head) = item;
439 eq_tail = item;
440
441 X_UNLOCK (eq_lock);
442
443 eq_signal_func (eq_signal_arg, 0);
444}
445
446static eq_item *
447eq_dec (void)
448{
449 eq_item *res;
450
451 X_LOCK (eq_lock);
452
453 res = eq_head;
454 if (res)
455 eq_head = res->next;
456
457 X_UNLOCK (eq_lock);
458
459 return res;
460}
461
462static void
463eq_poll (void)
464{
465 eq_item *item;
466
467 while ((item = eq_dec ()))
468 {
469 ENTER;
470 SAVETMPS;
471
472 dSP;
473 PUSHMARK (SP);
474 EXTEND (SP, 2);
475
476 if (item->vtbl->free_cb)
477 sv_2mortal (item->cb);
478
479 PUTBACK;
480 item->vtbl->push (item->data1, item->data2, item->data3);
481
482 SV *cb = item->cb;
483 free (item);
484
485 call_sv (cb, G_DISCARD | G_VOID);
486
487 FREETMPS;
488 LEAVE;
489 }
490}
491
492static void
493eq_poll_interrupt (pTHX_ void *c_arg, int value)
494{
495 eq_poll ();
496}
497
498/*****************************************************************************/
499/* context notify */
500
501static void ecb_noinline
502eq_context_push (void *data1, void *data2, void *data3)
503{
504 dSP;
505 PUSHs (sv_2mortal (newSVpv (data1, 0)));
506 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
507 PUTBACK;
508
509 free (data1);
510 free (data2);
511}
512
513static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
514
515static void CL_CALLBACK
516eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
517{
518 void *pvt_copy = malloc (cb);
519 memcpy (pvt_copy, pvt, cb);
520 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
521}
522
523#define CONTEXT_NOTIFY_CALLBACK \
524 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
525 void *user_data = 0; \
526 \
527 if (SvOK (notify)) \
528 { \
529 pfn_notify = eq_context_notify; \
530 user_data = s_get_cv (notify); \
531 }
532
533static SV * ecb_noinline
534new_clobj_context (cl_context ctx, void *user_data)
535{
536 SV *sv = new_clobj (stash_context, (IV)ctx);
537
538 if (user_data)
539 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
540
541 return sv;
542}
543
544#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
545
546/*****************************************************************************/
547/* build/compile/link notify */
548
549static void
550eq_program_push (void *data1, void *data2, void *data3)
551{
552 dSP;
553 PUSH_CLOBJ (stash_program, data1);
554 PUTBACK;
555}
556
557static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
558
559static void CL_CALLBACK
560eq_program_notify (cl_program program, void *user_data)
561{
562 clRetainProgram (program);
563
564 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
565}
566
567typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
568
569static program_callback ecb_noinline
570make_program_callback (SV *notify, void **ruser_data)
571{
572 if (SvOK (notify))
573 {
574 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
575 return eq_program_notify;
576 }
577 else
578 {
579 *ruser_data = 0;
580 return 0;
581 }
582}
583
584struct build_args
585{
586 cl_program program;
587 char *options;
588 void *user_data;
589 cl_uint num_devices;
590};
591
592X_THREAD_PROC (build_program_thread)
593{
594 struct build_args *arg = thr_arg;
595
596 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
597
598 if (arg->user_data)
599 eq_program_notify (arg->program, arg->user_data);
600 else
601 clReleaseProgram (arg->program);
602
603 free (arg->options);
604 free (arg);
605
606 return 0;
607}
608
609static void
610build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
611{
612 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
613
614 arg->program = program;
615 arg->options = strdup (options);
616 arg->user_data = user_data;
617 arg->num_devices = num_devices;
618 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
619
620 xthread_t id;
621 thread_create (&id, build_program_thread, arg);
622}
623
624/*****************************************************************************/
625/* mem object destructor notify */
626
627static void ecb_noinline
628eq_destructor_push (void *data1, void *data2, void *data3)
629{
630}
631
632static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
633
634static void CL_CALLBACK
635eq_destructor_notify (cl_mem memobj, void *user_data)
636{
637 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
638}
639
640/*****************************************************************************/
641/* event objects */
642
643static void
644eq_event_push (void *data1, void *data2, void *data3)
645{
646 dSP;
647 PUSH_CLOBJ (stash_event, data1);
648 PUSHs (sv_2mortal (newSViv ((IV)data2)));
649 PUTBACK;
650}
651
652static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
653
654static void CL_CALLBACK
655eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
656{
657 clRetainEvent (event);
658 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
659}
660
661/*****************************************************************************/
662/* utilities for XS code */
130 663
131static size_t 664static size_t
132img_row_pitch (cl_mem img) 665img_row_pitch (cl_mem img)
133{ 666{
134 size_t res; 667 size_t res;
135 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 668 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
136 return res; 669 return res;
137} 670}
138 671
139static cl_event * 672static cl_event * ecb_noinline
140event_list (SV **items, int count) 673event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
141{ 674{
675 cl_uint count = *rcount;
676
677 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
678 *rcount = count = 0;
679
680 if (!count && !extra)
681 return 0;
682
142 cl_event *list = tmpbuf (sizeof (cl_event) * count); 683 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
684 int i = 0;
143 685
144 while (count--) 686 while (count--)
687 if (SvOK (items [count]))
145 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 688 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
146 689
690 if (extra)
691 list [i++] = extra;
692
693 *rcount = i;
694
147 return list; 695 return i ? list : 0;
148} 696}
149 697
150#define EVENT_LIST(items,count) \ 698#define EVENT_LIST(skip) \
151 cl_uint event_list_count = (count); \ 699 cl_uint event_list_count = items - (skip); \
152 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 700 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
153 701
154#define INFO(class) \ 702#define INFO(class) \
155{ \ 703{ \
156 size_t size; \ 704 size_t size; \
157 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 705 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
158 SV *sv = sv_2mortal (newSV (size)); \ 706 SV *sv = sv_2mortal (newSV (size)); \
159 SvUPGRADE (sv, SVt_PV); \ 707 SvUPGRADE (sv, SVt_PV); \
160 SvPOK_only (sv); \ 708 SvPOK_only (sv); \
161 SvCUR_set (sv, size); \ 709 SvCUR_set (sv, size); \
162 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 710 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
163 XPUSHs (sv); \ 711 XPUSHs (sv); \
164} 712}
165 713
714/*****************************************************************************/
715/* mapped_xxx */
716
717static OpenCL__Mapped
718SvMAPPED (SV *sv)
719{
720 // no typechecking atm., keep your fingers crossed
721 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
722}
723
724struct mapped
725{
726 cl_command_queue queue;
727 cl_mem memobj;
728 void *ptr;
729 size_t cb;
730 cl_event event;
731 size_t row_pitch;
732 size_t slice_pitch;
733
734 size_t element_size;
735 size_t width, height, depth;
736};
737
738static SV *
739mapped_new (
740 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
741 void *ptr, size_t cb, cl_event ev,
742 size_t row_pitch, size_t slice_pitch, size_t element_size,
743 size_t width, size_t height, size_t depth
744)
745{
746 SV *data = newSV (0);
747 SvUPGRADE (data, SVt_PVMG);
748
749 OpenCL__Mapped mapped;
750 New (0, mapped, 1, struct mapped);
751
752 clRetainCommandQueue (queue);
753
754 mapped->queue = queue;
755 mapped->memobj = memobj;
756 mapped->ptr = ptr;
757 mapped->cb = cb;
758 mapped->event = ev;
759 mapped->row_pitch = row_pitch;
760 mapped->slice_pitch = slice_pitch;
761
762 mapped->element_size = element_size;
763 mapped->width = width;
764 mapped->height = height;
765 mapped->depth = depth;
766
767 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
768
769 if (SvLEN (data))
770 Safefree (data);
771
772 SvPVX (data) = (char *)ptr;
773 SvCUR_set (data, cb);
774 SvLEN_set (data, 0);
775 SvPOK_only (data);
776
777 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
778
779 if (!(flags & CL_MAP_WRITE))
780 SvREADONLY_on (data);
781
782 return obj;
783}
784
785static void
786mapped_detach (SV *sv, OpenCL__Mapped mapped)
787{
788 SV *data = SvRV (sv);
789
790 // the next check checks both before AND after detach, where SvPVX should STILL be 0
791 if (SvPVX (data) != (char *)mapped->ptr)
792 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
793 else
794 {
795 SvREADONLY_off (data);
796 SvCUR_set (data, 0);
797 SvPVX (data) = 0;
798 SvOK_off (data);
799 }
800
801 mapped->ptr = 0;
802}
803
804static void
805mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
806{
807 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
808 cl_event ev;
809
810 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
811
812 clReleaseEvent (mapped->event);
813 mapped->event = ev;
814
815 mapped_detach (self, mapped);
816}
817
818static size_t
819mapped_element_size (OpenCL__Mapped self)
820{
821 if (!self->element_size)
822 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
823
824 return self->element_size;
825}
826
827/*****************************************************************************/
828
166MODULE = OpenCL PACKAGE = OpenCL 829MODULE = OpenCL PACKAGE = OpenCL
167 830
168PROTOTYPES: ENABLE 831PROTOTYPES: ENABLE
169 832
833void
834poll ()
835 CODE:
836 eq_poll ();
837
838void
839_eq_initialise (IV func, IV arg)
840 CODE:
841 eq_signal_func = (void (*)(void *, int))func;
842 eq_signal_arg = (void*)arg;
843
170BOOT: 844BOOT:
171{ 845{
172 HV *stash = gv_stashpv ("OpenCL", 1); 846 HV *stash = gv_stashpv ("OpenCL", 1);
847
173 static const ivstr *civ, const_iv[] = { 848 static const ivstr *civ, const_iv[] = {
174 { sizeof (cl_char ), "SIZEOF_CHAR" }, 849 { sizeof (cl_char ), "SIZEOF_CHAR" },
175 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 850 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
176 { sizeof (cl_short ), "SIZEOF_SHORT" }, 851 { sizeof (cl_short ), "SIZEOF_SHORT" },
177 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 852 { sizeof (cl_ushort), "SIZEOF_USHORT" },
178 { sizeof (cl_int ), "SIZEOF_INT" }, 853 { sizeof (cl_int ), "SIZEOF_INT" },
179 { sizeof (cl_uint ), "SIZEOF_UINT" }, 854 { sizeof (cl_uint ), "SIZEOF_UINT" },
180 { sizeof (cl_long ), "SIZEOF_LONG" }, 855 { sizeof (cl_long ), "SIZEOF_LONG" },
181 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 856 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
182 { sizeof (cl_half ), "SIZEOF_HALF" }, 857 { sizeof (cl_half ), "SIZEOF_HALF" },
183 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 858 { sizeof (cl_float ), "SIZEOF_FLOAT" },
184 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 859 { sizeof (cl_double), "SIZEOF_DOUBLE" },
860 { PREFER_1_1 , "PREFER_1_1" },
185#include "constiv.h" 861#include "constiv.h"
186 }; 862 };
863
187 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 864 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
188 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 865 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
866
867 static const nvstr *cnv, const_nv[] = {
868#include "constnv.h"
869 };
870
871 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
872 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
873
874 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
875
876 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
877 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
878 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
879 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
880 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
881 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
882 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
883 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
884 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
885 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
886 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
887 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
888 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
889 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
890 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
891 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
892 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
893 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
894 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
895 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
896 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
897 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
898 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
899
900 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
189} 901}
190 902
191cl_int 903cl_int
192errno () 904errno ()
193 CODE: 905 CODE:
194 errno = res; 906 RETVAL = res;
907 OUTPUT:
908 RETVAL
195 909
196const char * 910const char *
197err2str (cl_int err) 911err2str (cl_int err = res)
198 912
199const char * 913const char *
200enum2str (cl_uint value) 914enum2str (cl_uint value)
201 915
202void 916void
210 list = tmpbuf (sizeof (*list) * count); 924 list = tmpbuf (sizeof (*list) * count);
211 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 925 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
212 926
213 EXTEND (SP, count); 927 EXTEND (SP, count);
214 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
215 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 929 PUSH_CLOBJ (stash_platform, list [i]);
216 930
217void 931void
218context_from_type (FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 932context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
219 PPCODE: 933 PPCODE:
934 CONTEXT_NOTIFY_CALLBACK;
220 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (0, type, 0, 0, &res)); 935 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
221 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 936 XPUSH_CLOBJ_CONTEXT;
222 937
223void 938void
224context (FUTURE properties, FUTURE devices, FUTURE notify = 0) 939context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
225 PPCODE: 940 PPCODE:
226 /* der Gipfel der Kunst */ 941 cl_uint device_count;
942 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
943
944 CONTEXT_NOTIFY_CALLBACK;
945 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
946 XPUSH_CLOBJ_CONTEXT;
227 947
228void 948void
229wait_for_events (...) 949wait_for_events (...)
230 CODE: 950 CODE:
231 EVENT_LIST (0, items); 951 EVENT_LIST (0);
232 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 952 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
233 953
234PROTOTYPES: DISABLE 954PROTOTYPES: DISABLE
235 955
236MODULE = OpenCL PACKAGE = OpenCL::Platform 956MODULE = OpenCL PACKAGE = OpenCL::Platform
237 957
238void 958void
239info (OpenCL::Platform this, cl_platform_info name) 959info (OpenCL::Platform self, cl_platform_info name)
240 PPCODE: 960 PPCODE:
241 INFO (Platform) 961 INFO (Platform)
242 962
963void
964unload_compiler (OpenCL::Platform self)
965 CODE:
966#if CL_VERSION_1_2
967 clUnloadPlatformCompiler (self);
968#endif
969
243#BEGIN:platform 970#BEGIN:platform
244 971
245void 972void
246profile (OpenCL::Platform this) 973profile (OpenCL::Platform self)
247 ALIAS: 974 ALIAS:
248 profile = CL_PLATFORM_PROFILE 975 profile = CL_PLATFORM_PROFILE
249 version = CL_PLATFORM_VERSION 976 version = CL_PLATFORM_VERSION
250 name = CL_PLATFORM_NAME 977 name = CL_PLATFORM_NAME
251 vendor = CL_PLATFORM_VENDOR 978 vendor = CL_PLATFORM_VENDOR
252 extensions = CL_PLATFORM_EXTENSIONS 979 extensions = CL_PLATFORM_EXTENSIONS
253 PPCODE: 980 PPCODE:
254 size_t size; 981 size_t size;
255 NEED_SUCCESS (GetPlatformInfo, (this, ix, 0, 0, &size)); 982 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
256 char *value = tmpbuf (size); 983 char *value = tmpbuf (size);
257 NEED_SUCCESS (GetPlatformInfo, (this, ix, size, value, 0)); 984 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
258 EXTEND (SP, 1); 985 EXTEND (SP, 1);
259 const int i = 0; 986 const int i = 0;
260 PUSHs (sv_2mortal (newSVpv (value, 0))); 987 PUSHs (sv_2mortal (newSVpv (value, 0)));
261 988
262#END:platform 989#END:platform
263 990
264void 991void
265devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 992devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
266 PPCODE: 993 PPCODE:
267 cl_device_id *list; 994 cl_device_id *list;
268 cl_uint count; 995 cl_uint count;
269 int i; 996 int i;
270 997
271 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 998 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
272 list = tmpbuf (sizeof (*list) * count); 999 list = tmpbuf (sizeof (*list) * count);
273 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 1000 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
274 1001
275 EXTEND (SP, count); 1002 EXTEND (SP, count);
276 for (i = 0; i < count; ++i) 1003 for (i = 0; i < count; ++i)
277 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1004 PUSH_CLOBJ (stash_device, list [i]);
278 1005
279void 1006void
280context (OpenCL::Platform this, FUTURE properties, SV *devices, FUTURE notify = 0) 1007context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
281 PPCODE: 1008 PPCODE:
1009 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1010 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1011
1012 cl_uint device_count;
1013 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1014
1015 CONTEXT_NOTIFY_CALLBACK;
1016 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1017 XPUSH_CLOBJ_CONTEXT;
1018
1019void
1020context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1021 PPCODE:
1022 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1023 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1024
1025 CONTEXT_NOTIFY_CALLBACK;
1026 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1027 XPUSH_CLOBJ_CONTEXT;
1028
1029MODULE = OpenCL PACKAGE = OpenCL::Device
1030
1031void
1032info (OpenCL::Device self, cl_device_info name)
1033 PPCODE:
1034 INFO (Device)
1035
1036#if CL_VERSION_1_2
1037
1038void
1039sub_devices (OpenCL::Device self, SV *properties)
1040 PPCODE:
282 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1041 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
283 croak ("OpenCL::Platform argument 'device' must be an arrayref with device objects, in call"); 1042 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
284 1043
285 AV *av = (AV *)SvRV (devices); 1044 properties = SvRV (properties);
286 cl_uint num_devices = av_len (av) + 1; 1045
287 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1046 cl_uint count = av_len ((AV *)properties) + 1;
1047 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1048
288 int i; 1049 int i;
1050 for (i = 0; i < count; ++i)
1051 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
289 1052
290 for (i = num_devices; i--; ) 1053 props [count] = 0;
291 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
292 1054
293 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (0, num_devices, device_list, 0, 0, &res)); 1055 cl_uint num_devices;
294 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1056 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1057 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1058 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
295 1059
296void 1060 EXTEND (SP, num_devices);
297context_from_type (OpenCL::Platform this, FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1061 for (i = 0; i < count; ++i)
298 PPCODE: 1062 PUSH_CLOBJ (stash_subdevice, list [i]);
299 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 };
300 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
301 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
302 1063
303MODULE = OpenCL PACKAGE = OpenCL::Device 1064#endif
304
305void
306info (OpenCL::Device this, cl_device_info name)
307 PPCODE:
308 INFO (Device)
309 1065
310#BEGIN:device 1066#BEGIN:device
311 1067
312void 1068void
313type (OpenCL::Device this) 1069type (OpenCL::Device self)
1070 ALIAS:
1071 type = CL_DEVICE_TYPE
1072 address_bits = CL_DEVICE_ADDRESS_BITS
1073 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1074 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1075 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1076 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1077 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1078 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1079 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1080 properties = CL_DEVICE_QUEUE_PROPERTIES
1081 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1082 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
314 PPCODE: 1083 PPCODE:
315 cl_device_type value [1]; 1084 cl_ulong value [1];
316 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1085 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
317 EXTEND (SP, 1); 1086 EXTEND (SP, 1);
318 const int i = 0; 1087 const int i = 0;
319 PUSHs (sv_2mortal (newSViv (value [i]))); 1088 PUSHs (sv_2mortal (newSVuv (value [i])));
320 1089
321void 1090void
322vendor_id (OpenCL::Device this) 1091vendor_id (OpenCL::Device self)
323 ALIAS: 1092 ALIAS:
324 vendor_id = CL_DEVICE_VENDOR_ID 1093 vendor_id = CL_DEVICE_VENDOR_ID
325 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS 1094 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
326 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS 1095 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
327 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR 1096 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
335 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1104 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
336 image_support = CL_DEVICE_IMAGE_SUPPORT 1105 image_support = CL_DEVICE_IMAGE_SUPPORT
337 max_samplers = CL_DEVICE_MAX_SAMPLERS 1106 max_samplers = CL_DEVICE_MAX_SAMPLERS
338 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1107 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
339 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1108 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1109 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
340 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1110 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
341 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1111 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1112 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
342 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1113 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
343 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1114 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
344 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1115 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
345 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1116 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
346 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1117 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
347 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1118 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
348 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1119 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
349 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1120 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
350 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1121 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
351 PPCODE: 1122 PPCODE:
352 cl_uint value [1]; 1123 cl_uint value [1];
353 NEED_SUCCESS (GetDeviceInfo, (this, ix, sizeof (value), value, 0)); 1124 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
354 EXTEND (SP, 1); 1125 EXTEND (SP, 1);
355 const int i = 0; 1126 const int i = 0;
356 PUSHs (sv_2mortal (newSVuv (value [i]))); 1127 PUSHs (sv_2mortal (newSVuv (value [i])));
357 1128
358void 1129void
359max_work_group_size (OpenCL::Device this) 1130max_work_group_size (OpenCL::Device self)
360 ALIAS: 1131 ALIAS:
361 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE 1132 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
362 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH 1133 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
363 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT 1134 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
364 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH 1135 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
366 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH 1137 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
367 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE 1138 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
368 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION 1139 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
369 PPCODE: 1140 PPCODE:
370 size_t value [1]; 1141 size_t value [1];
371 NEED_SUCCESS (GetDeviceInfo, (this, ix, sizeof (value), value, 0)); 1142 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
372 EXTEND (SP, 1); 1143 EXTEND (SP, 1);
373 const int i = 0; 1144 const int i = 0;
374 PUSHs (sv_2mortal (newSVuv (value [i]))); 1145 PUSHs (sv_2mortal (newSVuv (value [i])));
375 1146
376void 1147void
377max_work_item_sizes (OpenCL::Device this) 1148max_work_item_sizes (OpenCL::Device self)
378 PPCODE: 1149 PPCODE:
379 size_t size; 1150 size_t size;
380 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size)); 1151 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
381 size_t *value = tmpbuf (size); 1152 size_t *value = tmpbuf (size);
382 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0)); 1153 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
383 int i, n = size / sizeof (*value); 1154 int i, n = size / sizeof (*value);
384 EXTEND (SP, n); 1155 EXTEND (SP, n);
385 for (i = 0; i < n; ++i) 1156 for (i = 0; i < n; ++i)
386 PUSHs (sv_2mortal (newSVuv (value [i]))); 1157 PUSHs (sv_2mortal (newSVuv (value [i])));
387 1158
388void 1159void
389address_bits (OpenCL::Device this)
390 PPCODE:
391 cl_bitfield value [1];
392 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
393 EXTEND (SP, 1);
394 const int i = 0;
395 PUSHs (sv_2mortal (newSVuv (value [i])));
396
397void
398max_mem_alloc_size (OpenCL::Device this)
399 ALIAS:
400 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
401 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
402 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
403 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
404 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
405 PPCODE:
406 cl_ulong value [1];
407 NEED_SUCCESS (GetDeviceInfo, (this, ix, sizeof (value), value, 0));
408 EXTEND (SP, 1);
409 const int i = 0;
410 PUSHs (sv_2mortal (newSVuv (value [i])));
411
412void
413single_fp_config (OpenCL::Device this)
414 ALIAS:
415 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
416 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
417 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
418 PPCODE:
419 cl_device_fp_config value [1];
420 NEED_SUCCESS (GetDeviceInfo, (this, ix, sizeof (value), value, 0));
421 EXTEND (SP, 1);
422 const int i = 0;
423 PUSHs (sv_2mortal (newSVuv (value [i])));
424
425void
426global_mem_cache_type (OpenCL::Device this)
427 PPCODE:
428 cl_device_mem_cache_type value [1];
429 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
430 EXTEND (SP, 1);
431 const int i = 0;
432 PUSHs (sv_2mortal (newSVuv (value [i])));
433
434void
435local_mem_type (OpenCL::Device this)
436 PPCODE:
437 cl_device_local_mem_type value [1];
438 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
439 EXTEND (SP, 1);
440 const int i = 0;
441 PUSHs (sv_2mortal (newSVuv (value [i])));
442
443void
444error_correction_support (OpenCL::Device this) 1160error_correction_support (OpenCL::Device self)
445 ALIAS: 1161 ALIAS:
446 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1162 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
447 endian_little = CL_DEVICE_ENDIAN_LITTLE 1163 endian_little = CL_DEVICE_ENDIAN_LITTLE
448 available = CL_DEVICE_AVAILABLE 1164 available = CL_DEVICE_AVAILABLE
449 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1165 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
450 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1166 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
451 PPCODE: 1167 PPCODE:
452 cl_bool value [1]; 1168 cl_uint value [1];
453 NEED_SUCCESS (GetDeviceInfo, (this, ix, sizeof (value), value, 0)); 1169 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
454 EXTEND (SP, 1); 1170 EXTEND (SP, 1);
455 const int i = 0; 1171 const int i = 0;
456 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1172 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
457 1173
458void 1174void
459execution_capabilities (OpenCL::Device this)
460 PPCODE:
461 cl_device_exec_capabilities value [1];
462 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
463 EXTEND (SP, 1);
464 const int i = 0;
465 PUSHs (sv_2mortal (newSVuv (value [i])));
466
467void
468properties (OpenCL::Device this)
469 PPCODE:
470 cl_command_queue_properties value [1];
471 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
472 EXTEND (SP, 1);
473 const int i = 0;
474 PUSHs (sv_2mortal (newSViv (value [i])));
475
476void
477platform (OpenCL::Device this) 1175platform (OpenCL::Device self)
478 PPCODE: 1176 PPCODE:
479 cl_platform_id value [1]; 1177 cl_platform_id value [1];
480 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1178 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
481 EXTEND (SP, 1); 1179 EXTEND (SP, 1);
482 const int i = 0; 1180 const int i = 0;
483 { 1181 PUSH_CLOBJ (stash_platform, value [i]);
484 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i]));
485 }
486 1182
487void 1183void
488name (OpenCL::Device this) 1184name (OpenCL::Device self)
489 ALIAS: 1185 ALIAS:
490 name = CL_DEVICE_NAME 1186 name = CL_DEVICE_NAME
491 vendor = CL_DEVICE_VENDOR 1187 vendor = CL_DEVICE_VENDOR
492 driver_version = CL_DRIVER_VERSION 1188 driver_version = CL_DRIVER_VERSION
493 profile = CL_DEVICE_PROFILE 1189 profile = CL_DEVICE_PROFILE
494 version = CL_DEVICE_VERSION 1190 version = CL_DEVICE_VERSION
495 extensions = CL_DEVICE_EXTENSIONS 1191 extensions = CL_DEVICE_EXTENSIONS
496 PPCODE: 1192 PPCODE:
497 size_t size; 1193 size_t size;
498 NEED_SUCCESS (GetDeviceInfo, (this, ix, 0, 0, &size)); 1194 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
499 char *value = tmpbuf (size); 1195 char *value = tmpbuf (size);
500 NEED_SUCCESS (GetDeviceInfo, (this, ix, size, value, 0)); 1196 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
501 EXTEND (SP, 1); 1197 EXTEND (SP, 1);
502 const int i = 0; 1198 const int i = 0;
503 PUSHs (sv_2mortal (newSVpv (value, 0))); 1199 PUSHs (sv_2mortal (newSVpv (value, 0)));
504 1200
505void 1201void
506parent_device_ext (OpenCL::Device this) 1202parent_device_ext (OpenCL::Device self)
507 PPCODE: 1203 PPCODE:
508 cl_device_id value [1]; 1204 cl_device_id value [1];
509 NEED_SUCCESS (GetDeviceInfo, (this, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1205 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
510 EXTEND (SP, 1); 1206 EXTEND (SP, 1);
511 const int i = 0; 1207 const int i = 0;
512 { 1208 PUSH_CLOBJ (stash_device, value [i]);
513 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
514 }
515 1209
516void 1210void
517partition_types_ext (OpenCL::Device this) 1211partition_types_ext (OpenCL::Device self)
518 ALIAS: 1212 ALIAS:
519 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1213 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
520 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT 1214 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
521 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT 1215 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
522 PPCODE: 1216 PPCODE:
523 size_t size; 1217 size_t size;
524 NEED_SUCCESS (GetDeviceInfo, (this, ix, 0, 0, &size)); 1218 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
525 cl_device_partition_property_ext *value = tmpbuf (size); 1219 cl_device_partition_property_ext *value = tmpbuf (size);
526 NEED_SUCCESS (GetDeviceInfo, (this, ix, size, value, 0)); 1220 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
527 int i, n = size / sizeof (*value); 1221 int i, n = size / sizeof (*value);
528 EXTEND (SP, n); 1222 EXTEND (SP, n);
529 for (i = 0; i < n; ++i) 1223 for (i = 0; i < n; ++i)
530 PUSHs (sv_2mortal (newSVuv (value [i]))); 1224 PUSHs (sv_2mortal (newSVuv (value [i])));
531 1225
532#END:device 1226#END:device
533 1227
1228MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1229
1230#if CL_VERSION_1_2
1231
1232void
1233DESTROY (OpenCL::SubDevice self)
1234 CODE:
1235 clReleaseDevice (self);
1236
1237#endif
1238
534MODULE = OpenCL PACKAGE = OpenCL::Context 1239MODULE = OpenCL PACKAGE = OpenCL::Context
535 1240
536void 1241void
537DESTROY (OpenCL::Context context) 1242DESTROY (OpenCL::Context self)
538 CODE: 1243 CODE:
539 clReleaseContext (context); 1244 clReleaseContext (self);
540 1245
541void 1246void
542info (OpenCL::Context this, cl_context_info name) 1247info (OpenCL::Context self, cl_context_info name)
543 PPCODE: 1248 PPCODE:
544 INFO (Context) 1249 INFO (Context)
545 1250
546void 1251void
547queue (OpenCL::Context this, OpenCL::Device device, cl_command_queue_properties properties = 0) 1252queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
548 PPCODE: 1253 PPCODE:
549 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (this, device, properties, &res)); 1254 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
550 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1255 XPUSH_CLOBJ (stash_queue, queue);
551 1256
552void 1257void
553user_event (OpenCL::Context this) 1258user_event (OpenCL::Context self)
554 PPCODE: 1259 PPCODE:
555 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (this, &res)); 1260 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
556 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1261 XPUSH_CLOBJ (stash_userevent, ev);
557 1262
558void 1263void
559buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1264buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
560 PPCODE: 1265 PPCODE:
561 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1266 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
562 croak ("clCreateBuffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1267 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
563 1268
564 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (this, flags, len, 0, &res)); 1269 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
565 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1270 XPUSH_CLOBJ (stash_bufferobj, mem);
566 1271
567void 1272void
568buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1273buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
569 PPCODE: 1274 PPCODE:
570 STRLEN len; 1275 STRLEN len;
571 char *ptr = SvPVbyte (data, len); 1276 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
572
573 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1277 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
574 croak ("clCreateBuffer: have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1278 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
575
576 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (this, flags, len, ptr, &res)); 1279 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
577 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 1280 XPUSH_CLOBJ (stash_bufferobj, mem);
578 1281
1282#if CL_VERSION_1_2
1283
579void 1284void
580image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1285image (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)
581 PPCODE: 1286 PPCODE:
582 STRLEN len; 1287 STRLEN len;
583 char *ptr = SvPVbyte (data, len); 1288 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
584 const cl_image_format format = { channel_order, channel_type }; 1289 const cl_image_format format = { channel_order, channel_type };
1290 const cl_image_desc desc = {
1291 type,
1292 width, height, depth,
1293 array_size, row_pitch, slice_pitch,
1294 num_mip_level, num_samples,
1295 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1296 };
585 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (this, flags, &format, width, height, row_pitch, ptr, &res)); 1297 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
586 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1298 HV *stash = stash_image;
1299 switch (type)
1300 {
1301 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1302 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1303 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1304 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1305 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1306 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1307 }
1308 XPUSH_CLOBJ (stash, mem);
587 1309
1310#endif
1311
588void 1312void
589image3d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1313image2d (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)
590 PPCODE: 1314 PPCODE:
591 STRLEN len; 1315 STRLEN len;
592 char *ptr = SvPVbyte (data, len); 1316 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
593 const cl_image_format format = { channel_order, channel_type }; 1317 const cl_image_format format = { channel_order, channel_type };
1318#if PREFER_1_1
1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1320#else
1321 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1322 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1323#endif
1324 XPUSH_CLOBJ (stash_image2d, mem);
1325
1326void
1327image3d (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)
1328 PPCODE:
1329 STRLEN len;
1330 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1331 const cl_image_format format = { channel_order, channel_type };
1332#if PREFER_1_1
594 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (this, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1333 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
595 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1334#else
1335 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1336 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1337#endif
1338 XPUSH_CLOBJ (stash_image3d, mem);
596 1339
1340#if cl_apple_gl_sharing || cl_khr_gl_sharing
1341
597void 1342void
1343gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1344 PPCODE:
1345 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1346 XPUSH_CLOBJ (stash_bufferobj, mem);
1347
1348void
1349gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1350 PPCODE:
1351 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1352 XPUSH_CLOBJ (stash_image2d, mem);
1353
1354#if CL_VERSION_1_2
1355
1356void
1357gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1358 ALIAS:
1359 PPCODE:
1360 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1361 cl_gl_object_type type;
1362 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1363 HV *stash = stash_memory;
1364 switch (type)
1365 {
1366 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1367 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1368 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1369 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1370 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1371 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1372 }
1373 XPUSH_CLOBJ (stash, mem);
1374
1375#endif
1376
1377void
1378gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1379 PPCODE:
1380#if PREFER_1_1
1381 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1382#else
1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1384#endif
1385 XPUSH_CLOBJ (stash_image2d, mem);
1386
1387void
1388gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1389 PPCODE:
1390#if PREFER_1_1
1391 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1392#else
1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1394#endif
1395 XPUSH_CLOBJ (stash_image3d, mem);
1396
1397#endif
1398
1399void
598supported_image_formats (OpenCL::Context this, cl_mem_flags flags, cl_mem_object_type image_type) 1400supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
599 PPCODE: 1401 PPCODE:
600{ 1402{
601 cl_uint count; 1403 cl_uint count;
602 cl_image_format *list; 1404 cl_image_format *list;
603 int i; 1405 int i;
604 1406
605 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1407 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
606 Newx (list, count, cl_image_format); 1408 Newx (list, count, cl_image_format);
607 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1409 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
608 1410
609 EXTEND (SP, count); 1411 EXTEND (SP, count);
610 for (i = 0; i < count; ++i) 1412 for (i = 0; i < count; ++i)
611 { 1413 {
612 AV *av = newAV (); 1414 AV *av = newAV ();
615 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1417 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
616 } 1418 }
617} 1419}
618 1420
619void 1421void
620sampler (OpenCL::Context this, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1422sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
621 PPCODE: 1423 PPCODE:
622 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (this, normalized_coords, addressing_mode, filter_mode, &res)); 1424 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
623 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1425 XPUSH_CLOBJ (stash_sampler, sampler);
624 1426
625void 1427void
626program_with_source (OpenCL::Context this, SV *program) 1428program_with_source (OpenCL::Context self, SV *program)
627 PPCODE: 1429 PPCODE:
628 STRLEN len; 1430 STRLEN len;
629 size_t len2; 1431 size_t len2;
630 const char *ptr = SvPVbyte (program, len); 1432 const char *ptr = SvPVbyte (program, len);
631 1433
632 len2 = len; 1434 len2 = len;
633 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (this, 1, &ptr, &len2, &res)); 1435 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
634 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1436 XPUSH_CLOBJ (stash_program, prog);
1437
1438void
1439program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1440 PPCODE:
1441 cl_uint device_count;
1442 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1443
1444 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1445 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1446
1447 binaries = SvRV (binaries);
1448
1449 if (device_count != av_len ((AV *)binaries) + 1)
1450 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1451
1452 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1453 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1454 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1455
1456 int i;
1457 for (i = 0; i < device_count; ++i)
1458 {
1459 STRLEN len;
1460 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1461 length_list [i] = len;
1462 }
1463
1464 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1465 length_list, binary_list,
1466 GIMME_V == G_ARRAY ? status_list : 0, &res));
1467
1468 EXTEND (SP, 2);
1469 PUSH_CLOBJ (stash_program, prog);
1470
1471 if (GIMME_V == G_ARRAY)
1472 {
1473 AV *av = newAV ();
1474 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1475
1476 for (i = device_count; i--; )
1477 av_store (av, i, newSViv (status_list [i]));
1478 }
1479
1480#if CL_VERSION_1_2
1481
1482void
1483program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1484 PPCODE:
1485 cl_uint device_count;
1486 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1489
1490 XPUSH_CLOBJ (stash_program, prog);
1491
1492void
1493link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1494 CODE:
1495 cl_uint device_count = 0;
1496 cl_device_id *device_list = 0;
1497
1498 if (SvOK (devices))
1499 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1500
1501 cl_uint program_count;
1502 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1503
1504 void *user_data;
1505 program_callback pfn_notify = make_program_callback (notify, &user_data);
1506
1507 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1508 program_count, program_list, pfn_notify, user_data, &res));
1509
1510 XPUSH_CLOBJ (stash_program, prog);
1511
1512#endif
635 1513
636#BEGIN:context 1514#BEGIN:context
637 1515
638void 1516void
639reference_count (OpenCL::Context this) 1517reference_count (OpenCL::Context self)
640 ALIAS: 1518 ALIAS:
641 reference_count = CL_CONTEXT_REFERENCE_COUNT 1519 reference_count = CL_CONTEXT_REFERENCE_COUNT
642 num_devices = CL_CONTEXT_NUM_DEVICES 1520 num_devices = CL_CONTEXT_NUM_DEVICES
643 PPCODE: 1521 PPCODE:
644 cl_uint value [1]; 1522 cl_uint value [1];
645 NEED_SUCCESS (GetContextInfo, (this, ix, sizeof (value), value, 0)); 1523 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
646 EXTEND (SP, 1); 1524 EXTEND (SP, 1);
647 const int i = 0; 1525 const int i = 0;
648 PUSHs (sv_2mortal (newSVuv (value [i]))); 1526 PUSHs (sv_2mortal (newSVuv (value [i])));
649 1527
650void 1528void
651devices (OpenCL::Context this) 1529devices (OpenCL::Context self)
652 PPCODE: 1530 PPCODE:
653 size_t size; 1531 size_t size;
654 NEED_SUCCESS (GetContextInfo, (this, CL_CONTEXT_DEVICES, 0, 0, &size)); 1532 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
655 cl_device_id *value = tmpbuf (size); 1533 cl_device_id *value = tmpbuf (size);
656 NEED_SUCCESS (GetContextInfo, (this, CL_CONTEXT_DEVICES, size, value, 0)); 1534 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
657 int i, n = size / sizeof (*value); 1535 int i, n = size / sizeof (*value);
658 EXTEND (SP, n); 1536 EXTEND (SP, n);
659 for (i = 0; i < n; ++i) 1537 for (i = 0; i < n; ++i)
660 { 1538 PUSH_CLOBJ (stash_device, value [i]);
661 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
662 }
663 1539
664void 1540void
665properties (OpenCL::Context this) 1541properties (OpenCL::Context self)
666 PPCODE: 1542 PPCODE:
667 size_t size; 1543 size_t size;
668 NEED_SUCCESS (GetContextInfo, (this, CL_CONTEXT_PROPERTIES, 0, 0, &size)); 1544 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
669 cl_context_properties *value = tmpbuf (size); 1545 cl_context_properties *value = tmpbuf (size);
670 NEED_SUCCESS (GetContextInfo, (this, CL_CONTEXT_PROPERTIES, size, value, 0)); 1546 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
671 int i, n = size / sizeof (*value); 1547 int i, n = size / sizeof (*value);
672 EXTEND (SP, n); 1548 EXTEND (SP, n);
673 for (i = 0; i < n; ++i) 1549 for (i = 0; i < n; ++i)
674 PUSHs (sv_2mortal (newSVuv ((UV)value [i]))); 1550 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
675 1551
676#END:context 1552#END:context
677 1553
678MODULE = OpenCL PACKAGE = OpenCL::Queue 1554MODULE = OpenCL PACKAGE = OpenCL::Queue
679 1555
680void 1556void
681DESTROY (OpenCL::Queue this) 1557DESTROY (OpenCL::Queue self)
682 CODE: 1558 CODE:
683 clReleaseCommandQueue (this); 1559 clReleaseCommandQueue (self);
684 1560
685void 1561void
686enqueue_read_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1562read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1563 ALIAS:
1564 enqueue_read_buffer = 0
687 PPCODE: 1565 PPCODE:
688 cl_event ev = 0;
689 EVENT_LIST (6, items - 6); 1566 EVENT_LIST (6);
690 1567
691 SvUPGRADE (data, SVt_PV); 1568 SvUPGRADE (data, SVt_PV);
692 SvGROW (data, len); 1569 SvGROW (data, len);
693 SvPOK_only (data); 1570 SvPOK_only (data);
694 SvCUR_set (data, len); 1571 SvCUR_set (data, len);
1572
1573 cl_event ev = 0;
695 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1574 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
696 1575
697 if (ev) 1576 if (ev)
698 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1577 XPUSH_CLOBJ (stash_event, ev);
699 1578
700void 1579void
701enqueue_write_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1580write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1581 ALIAS:
1582 enqueue_write_buffer = 0
702 PPCODE: 1583 PPCODE:
703 cl_event ev = 0; 1584 EVENT_LIST (5);
1585
704 STRLEN len; 1586 STRLEN len;
705 char *ptr = SvPVbyte (data, len); 1587 char *ptr = SvPVbyte (data, len);
706 EVENT_LIST (5, items - 5);
707 1588
1589 cl_event ev = 0;
708 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1590 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
709 1591
710 if (ev) 1592 if (ev)
711 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1593 XPUSH_CLOBJ (stash_event, ev);
712 1594
1595#if CL_VERSION_1_2
1596
713void 1597void
714enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1598fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1599 ALIAS:
1600 enqueue_fill_buffer = 0
715 PPCODE: 1601 PPCODE:
1602 EVENT_LIST (5);
1603
1604 STRLEN len;
1605 char *ptr = SvPVbyte (data, len);
1606
716 cl_event ev = 0; 1607 cl_event ev = 0;
717 EVENT_LIST (6, items - 6);
718
719 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1608 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
720 1609
721 if (ev) 1610 if (ev)
722 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1611 XPUSH_CLOBJ (stash_event, ev);
723 1612
724void 1613void
725enqueue_read_image (OpenCL::Queue this, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1614fill_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, ...)
1615 ALIAS:
1616 enqueue_fill_image = 0
726 PPCODE: 1617 PPCODE:
1618 EVENT_LIST (12);
1619
1620 const size_t origin [3] = { x, y, z };
1621 const size_t region [3] = { width, height, depth };
1622
1623 const cl_float c_f [4] = { r, g, b, a };
1624 const cl_uint c_u [4] = { r, g, b, a };
1625 const cl_int c_s [4] = { r, g, b, a };
1626 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1627 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1628 cl_image_format format;
1629 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1630 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1631 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1632 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1633
727 cl_event ev = 0; 1634 cl_event ev = 0;
1635 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1636 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1637
1638 if (ev)
1639 XPUSH_CLOBJ (stash_event, ev);
1640
1641#endif
1642
1643void
1644copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1645 ALIAS:
1646 enqueue_copy_buffer = 0
1647 PPCODE:
1648 EVENT_LIST (6);
1649
1650 cl_event ev = 0;
1651 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1652
1653 if (ev)
1654 XPUSH_CLOBJ (stash_event, ev);
1655
1656void
1657read_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, ...)
1658 ALIAS:
1659 enqueue_read_buffer_rect = 0
1660 PPCODE:
1661 EVENT_LIST (17);
1662
1663 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
728 const size_t src_origin[3] = { src_x, src_y, src_z }; 1664 const size_t host_origin[3] = { host_x, host_y, host_z };
729 const size_t region[3] = { width, height, depth }; 1665 const size_t region[3] = { width, height, depth };
730 EVENT_LIST (12, items - 12);
731 1666
732 if (!row_pitch) 1667 if (!buf_row_pitch)
733 row_pitch = img_row_pitch (src); 1668 buf_row_pitch = region [0];
734 1669
735 if (depth > 1 && !slice_pitch) 1670 if (!buf_slice_pitch)
736 slice_pitch = row_pitch * height; 1671 buf_slice_pitch = region [1] * buf_row_pitch;
737 1672
738 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1673 if (!host_row_pitch)
1674 host_row_pitch = region [0];
1675
1676 if (!host_slice_pitch)
1677 host_slice_pitch = region [1] * host_row_pitch;
1678
1679 size_t len = host_row_pitch * host_slice_pitch * region [2];
739 1680
740 SvUPGRADE (data, SVt_PV); 1681 SvUPGRADE (data, SVt_PV);
741 SvGROW (data, len); 1682 SvGROW (data, len);
742 SvPOK_only (data); 1683 SvPOK_only (data);
743 SvCUR_set (data, len); 1684 SvCUR_set (data, len);
744 NEED_SUCCESS (EnqueueReadImage, (this, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1685
1686 cl_event ev = 0;
1687 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));
745 1688
746 if (ev) 1689 if (ev)
747 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1690 XPUSH_CLOBJ (stash_event, ev);
748 1691
749void 1692void
750enqueue_write_image (OpenCL::Queue this, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1693write_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, ...)
1694 ALIAS:
1695 enqueue_write_buffer_rect = 0
751 PPCODE: 1696 PPCODE:
1697 EVENT_LIST (17);
1698
1699 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1700 const size_t host_origin[3] = { host_x, host_y, host_z };
1701 const size_t region[3] = { width, height, depth };
1702 STRLEN len;
1703 char *ptr = SvPVbyte (data, len);
1704
1705 if (!buf_row_pitch)
1706 buf_row_pitch = region [0];
1707
1708 if (!buf_slice_pitch)
1709 buf_slice_pitch = region [1] * buf_row_pitch;
1710
1711 if (!host_row_pitch)
1712 host_row_pitch = region [0];
1713
1714 if (!host_slice_pitch)
1715 host_slice_pitch = region [1] * host_row_pitch;
1716
1717 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1718
1719 if (len < min_len)
1720 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1721
752 cl_event ev = 0; 1722 cl_event ev = 0;
1723 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));
1724
1725 if (ev)
1726 XPUSH_CLOBJ (stash_event, ev);
1727
1728void
1729copy_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, ...)
1730 ALIAS:
1731 enqueue_copy_buffer_rect = 0
1732 PPCODE:
1733 EVENT_LIST (16);
1734
1735 const size_t src_origin[3] = { src_x, src_y, src_z };
1736 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1737 const size_t region[3] = { width, height, depth };
1738
1739 cl_event ev = 0;
1740 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));
1741
1742 if (ev)
1743 XPUSH_CLOBJ (stash_event, ev);
1744
1745void
1746read_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, ...)
1747 ALIAS:
1748 enqueue_read_image = 0
1749 PPCODE:
1750 EVENT_LIST (12);
1751
1752 const size_t src_origin[3] = { src_x, src_y, src_z };
1753 const size_t region[3] = { width, height, depth };
1754
1755 if (!row_pitch)
1756 row_pitch = img_row_pitch (src);
1757
1758 if (depth > 1 && !slice_pitch)
1759 slice_pitch = row_pitch * height;
1760
1761 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1762
1763 SvUPGRADE (data, SVt_PV);
1764 SvGROW (data, len);
1765 SvPOK_only (data);
1766 SvCUR_set (data, len);
1767
1768 cl_event ev = 0;
1769 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));
1770
1771 if (ev)
1772 XPUSH_CLOBJ (stash_event, ev);
1773
1774void
1775write_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, ...)
1776 ALIAS:
1777 enqueue_write_image = 0
1778 PPCODE:
1779 EVENT_LIST (12);
1780
753 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1781 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
754 const size_t region[3] = { width, height, depth }; 1782 const size_t region[3] = { width, height, depth };
755 STRLEN len; 1783 STRLEN len;
756 char *ptr = SvPVbyte (data, len); 1784 char *ptr = SvPVbyte (data, len);
757 EVENT_LIST (12, items - 12);
758 1785
759 if (!row_pitch) 1786 if (!row_pitch)
760 row_pitch = img_row_pitch (dst); 1787 row_pitch = img_row_pitch (dst);
761 1788
762 if (depth > 1 && !slice_pitch) 1789 if (depth > 1 && !slice_pitch)
765 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1792 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
766 1793
767 if (len < min_len) 1794 if (len < min_len)
768 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1795 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
769 1796
1797 cl_event ev = 0;
770 NEED_SUCCESS (EnqueueWriteImage, (this, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1798 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));
771 1799
772 if (ev) 1800 if (ev)
773 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1801 XPUSH_CLOBJ (stash_event, ev);
774 1802
775void 1803void
776enqueue_copy_buffer_rect (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1804copy_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, ...)
1805 ALIAS:
1806 enqueue_copy_image = 0
777 PPCODE: 1807 PPCODE:
778 cl_event ev = 0; 1808 EVENT_LIST (12);
1809
779 const size_t src_origin[3] = { src_x, src_y, src_z }; 1810 const size_t src_origin[3] = { src_x, src_y, src_z };
780 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1811 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
781 const size_t region[3] = { width, height, depth }; 1812 const size_t region[3] = { width, height, depth };
782 EVENT_LIST (16, items - 16);
783 1813
1814 cl_event ev = 0;
784 NEED_SUCCESS (EnqueueCopyBufferRect, (this, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1815 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
785 1816
786 if (ev) 1817 if (ev)
787 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1818 XPUSH_CLOBJ (stash_event, ev);
788 1819
789void 1820void
790enqueue_copy_buffer_to_image (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1821copy_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, ...)
1822 ALIAS:
1823 enqueue_copy_image_to_buffer = 0
791 PPCODE: 1824 PPCODE:
1825 EVENT_LIST (10);
1826
1827 const size_t src_origin[3] = { src_x, src_y, src_z };
1828 const size_t region [3] = { width, height, depth };
1829
792 cl_event ev = 0; 1830 cl_event ev = 0;
1831 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1832
1833 if (ev)
1834 XPUSH_CLOBJ (stash_event, ev);
1835
1836void
1837copy_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, ...)
1838 ALIAS:
1839 enqueue_copy_buffer_to_image = 0
1840 PPCODE:
1841 EVENT_LIST (10);
1842
793 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1843 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1844 const size_t region [3] = { width, height, depth };
1845
1846 cl_event ev = 0;
1847 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1848
1849 if (ev)
1850 XPUSH_CLOBJ (stash_event, ev);
1851
1852void
1853map_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, ...)
1854 ALIAS:
1855 enqueue_map_buffer = 0
1856 PPCODE:
1857 EVENT_LIST (6);
1858
1859 size_t cb = SvIV (cb_);
1860
1861 if (!SvOK (cb_))
1862 {
1863 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1864 cb -= offset;
1865 }
1866
1867 cl_event ev;
1868 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1869 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1870
1871void
1872map_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, ...)
1873 ALIAS:
1874 enqueue_map_image = 0
1875 PPCODE:
1876 size_t width = SvIV (width_);
1877 if (!SvOK (width_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1880 width -= x;
1881 }
1882
1883 size_t height = SvIV (width_);
1884 if (!SvOK (height_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1887 height -= y;
1888
1889 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1890 if (!height)
1891 height = 1;
1892 }
1893
1894 size_t depth = SvIV (width_);
1895 if (!SvOK (depth_))
1896 {
1897 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1898 depth -= z;
1899
1900 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1901 if (!depth)
1902 depth = 1;
1903 }
1904
1905 const size_t origin[3] = { x, y, z };
794 const size_t region[3] = { width, height, depth }; 1906 const size_t region[3] = { width, height, depth };
1907 size_t row_pitch, slice_pitch;
795 EVENT_LIST (10, items - 10); 1908 EVENT_LIST (10);
796 1909
797 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1910 cl_event ev;
1911 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));
1912
1913 size_t cb = slice_pitch ? slice_pitch * region [2]
1914 : row_pitch ? row_pitch * region [1]
1915 : region [0];
1916
1917 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1918
1919void
1920unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1921 PPCODE:
1922 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1923 if (GIMME_V != G_VOID)
1924 {
1925 clRetainEvent (mapped->event);
1926 XPUSH_CLOBJ (stash_event, mapped->event);
1927 }
1928
1929void
1930task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1931 ALIAS:
1932 enqueue_task = 0
1933 PPCODE:
1934 EVENT_LIST (2);
1935
1936 cl_event ev = 0;
1937 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
798 1938
799 if (ev) 1939 if (ev)
800 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1940 XPUSH_CLOBJ (stash_event, ev);
801 1941
802void 1942void
803enqueue_copy_image (OpenCL::Queue this, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
804 PPCODE:
805 cl_event ev = 0;
806 const size_t src_origin[3] = { src_x, src_y, src_z };
807 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
808 const size_t region[3] = { width, height, depth };
809 EVENT_LIST (12, items - 12);
810
811 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
812
813 if (ev)
814 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
815
816void
817enqueue_copy_image_to_buffer (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
818 PPCODE:
819 cl_event ev = 0;
820 const size_t src_origin[3] = { src_x, src_y, src_z };
821 const size_t region[3] = { width, height, depth };
822 EVENT_LIST (10, items - 10);
823
824 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
825
826 if (ev)
827 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
828
829void
830enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
831 PPCODE:
832 cl_event ev = 0;
833 EVENT_LIST (2, items - 2);
834
835 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
836
837 if (ev)
838 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
839
840void
841enqueue_nd_range_kernel (OpenCL::Queue this, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1943nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1944 ALIAS:
1945 enqueue_nd_range_kernel = 0
842 PPCODE: 1946 PPCODE:
843 cl_event ev = 0; 1947 EVENT_LIST (5);
1948
844 size_t *gwo = 0, *gws, *lws = 0; 1949 size_t *gwo = 0, *gws, *lws = 0;
845 int gws_len; 1950 int gws_len;
846 size_t *lists; 1951 size_t *lists;
847 int i; 1952 int i;
848 EVENT_LIST (5, items - 5);
849 1953
850 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1954 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
851 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
852 1956
853 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1957 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
854 1958
855 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1959 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
856 1960
857 gws = lists + gws_len * 0; 1961 gws = lists + gws_len * 0;
858 for (i = 0; i < gws_len; ++i) 1962 for (i = 0; i < gws_len; ++i)
1963 {
859 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1964 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1965 // at least nvidia crashes for 0-sized work group sizes, work around
1966 if (!gws [i])
1967 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1968 }
860 1969
861 if (SvOK (global_work_offset)) 1970 if (SvOK (global_work_offset))
862 { 1971 {
863 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1972 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
864 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1973 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
871 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1980 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
872 } 1981 }
873 1982
874 if (SvOK (local_work_size)) 1983 if (SvOK (local_work_size))
875 { 1984 {
876 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1985 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
877 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1986 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
878 1987
879 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1988 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
880 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1989 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
881 1990
882 lws = lists + gws_len * 2; 1991 lws = lists + gws_len * 2;
883 for (i = 0; i < gws_len; ++i) 1992 for (i = 0; i < gws_len; ++i)
1993 {
884 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1994 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1995 // at least nvidia crashes for 0-sized work group sizes, work around
1996 if (!lws [i])
1997 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1998 }
885 } 1999 }
886 2000
2001 cl_event ev = 0;
887 NEED_SUCCESS (EnqueueNDRangeKernel, (this, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2002 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
888 2003
889 if (ev) 2004 if (ev)
890 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2005 XPUSH_CLOBJ (stash_event, ev);
891 2006
2007#if CL_VERSION_1_2
2008
892void 2009void
893enqueue_marker (OpenCL::Queue this) 2010migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2011 ALIAS:
2012 enqueue_migrate_mem_objects = 0
894 PPCODE: 2013 PPCODE:
2014 EVENT_LIST (3);
2015
2016 cl_uint obj_count;
2017 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2018
2019 cl_event ev = 0;
2020 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2021
2022 if (ev)
2023 XPUSH_CLOBJ (stash_event, ev);
2024
2025#endif
2026
2027#if cl_apple_gl_sharing || cl_khr_gl_sharing
2028
2029void
2030acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2031 ALIAS:
2032 release_gl_objects = 1
2033 enqueue_acquire_gl_objects = 0
2034 enqueue_release_gl_objects = 1
2035 PPCODE:
2036 EVENT_LIST (2);
2037
2038 cl_uint obj_count;
2039 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2040
895 cl_event ev; 2041 cl_event ev = 0;
896 NEED_SUCCESS (EnqueueMarker, (this, &ev));
897 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
898 2042
2043 if (ix)
2044 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2045 else
2046 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2047
2048 if (ev)
2049 XPUSH_CLOBJ (stash_event, ev);
2050
2051#endif
2052
899void 2053void
900enqueue_wait_for_events (OpenCL::Queue this, ...) 2054wait_for_events (OpenCL::Queue self, ...)
2055 ALIAS:
2056 enqueue_wait_for_events = 0
901 CODE: 2057 CODE:
902 EVENT_LIST (1, items - 1); 2058 EVENT_LIST (1);
2059#if PREFER_1_1
903 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 2060 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2061#else
2062 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2063#endif
904 2064
905void 2065void
906enqueue_barrier (OpenCL::Queue this) 2066marker (OpenCL::Queue self, ...)
2067 ALIAS:
2068 enqueue_marker = 0
907 CODE: 2069 PPCODE:
2070 EVENT_LIST (1);
2071 cl_event ev = 0;
2072#if PREFER_1_1
2073 if (!event_list_count)
2074 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2075 else
2076#if CL_VERSION_1_2
2077 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2078#else
2079 {
2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2081 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2082 }
2083#endif
2084#else
2085 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2086#endif
2087 if (ev)
2088 XPUSH_CLOBJ (stash_event, ev);
2089
2090void
2091barrier (OpenCL::Queue self, ...)
2092 ALIAS:
2093 enqueue_barrier = 0
2094 PPCODE:
2095 EVENT_LIST (1);
2096 cl_event ev = 0;
2097#if PREFER_1_1
2098 if (!event_list_count && GIMME_V == G_VOID)
908 NEED_SUCCESS (EnqueueBarrier, (this)); 2099 NEED_SUCCESS (EnqueueBarrier, (self));
2100 else
2101#if CL_VERSION_1_2
2102 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2103#else
2104 {
2105 if (event_list_count)
2106 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
909 2107
2108 if (GIMME_V != G_VOID)
2109 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2110 }
2111#endif
2112#else
2113 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2114#endif
2115 if (ev)
2116 XPUSH_CLOBJ (stash_event, ev);
2117
910void 2118void
911flush (OpenCL::Queue this) 2119flush (OpenCL::Queue self)
912 CODE: 2120 CODE:
913 NEED_SUCCESS (Flush, (this)); 2121 NEED_SUCCESS (Flush, (self));
914 2122
915void 2123void
916finish (OpenCL::Queue this) 2124finish (OpenCL::Queue self)
917 CODE: 2125 CODE:
918 NEED_SUCCESS (Finish, (this)); 2126 NEED_SUCCESS (Finish, (self));
919 2127
920void 2128void
921info (OpenCL::Queue this, cl_command_queue_info name) 2129info (OpenCL::Queue self, cl_command_queue_info name)
922 PPCODE: 2130 PPCODE:
923 INFO (CommandQueue) 2131 INFO (CommandQueue)
924 2132
925#BEGIN:command_queue 2133#BEGIN:command_queue
926 2134
927void 2135void
928context (OpenCL::Queue this) 2136context (OpenCL::Queue self)
929 PPCODE: 2137 PPCODE:
930 cl_context value [1]; 2138 cl_context value [1];
931 NEED_SUCCESS (GetCommandQueueInfo, (this, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2139 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
932 EXTEND (SP, 1); 2140 EXTEND (SP, 1);
933 const int i = 0; 2141 const int i = 0;
934 {
935 NEED_SUCCESS (RetainContext, (value [i])); 2142 NEED_SUCCESS (RetainContext, (value [i]));
936 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2143 PUSH_CLOBJ (stash_context, value [i]);
937 }
938 2144
939void 2145void
940device (OpenCL::Queue this) 2146device (OpenCL::Queue self)
941 PPCODE: 2147 PPCODE:
942 cl_device_id value [1]; 2148 cl_device_id value [1];
943 NEED_SUCCESS (GetCommandQueueInfo, (this, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2149 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
944 EXTEND (SP, 1); 2150 EXTEND (SP, 1);
945 const int i = 0; 2151 const int i = 0;
946 { 2152 PUSH_CLOBJ (stash_device, value [i]);
947 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
948 }
949 2153
950void 2154void
951reference_count (OpenCL::Queue this) 2155reference_count (OpenCL::Queue self)
952 PPCODE: 2156 PPCODE:
953 cl_uint value [1]; 2157 cl_uint value [1];
954 NEED_SUCCESS (GetCommandQueueInfo, (this, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0)); 2158 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
955 EXTEND (SP, 1); 2159 EXTEND (SP, 1);
956 const int i = 0; 2160 const int i = 0;
957 PUSHs (sv_2mortal (newSVuv (value [i]))); 2161 PUSHs (sv_2mortal (newSVuv (value [i])));
958 2162
959void 2163void
960properties (OpenCL::Queue this) 2164properties (OpenCL::Queue self)
961 PPCODE: 2165 PPCODE:
962 cl_command_queue_properties value [1]; 2166 cl_ulong value [1];
963 NEED_SUCCESS (GetCommandQueueInfo, (this, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2167 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
964 EXTEND (SP, 1); 2168 EXTEND (SP, 1);
965 const int i = 0; 2169 const int i = 0;
966 PUSHs (sv_2mortal (newSViv (value [i]))); 2170 PUSHs (sv_2mortal (newSVuv (value [i])));
967 2171
968#END:command_queue 2172#END:command_queue
969 2173
970MODULE = OpenCL PACKAGE = OpenCL::Memory 2174MODULE = OpenCL PACKAGE = OpenCL::Memory
971 2175
972void 2176void
973DESTROY (OpenCL::Memory this) 2177DESTROY (OpenCL::Memory self)
974 CODE: 2178 CODE:
975 clReleaseMemObject (this); 2179 clReleaseMemObject (self);
976 2180
977void 2181void
978info (OpenCL::Memory this, cl_mem_info name) 2182info (OpenCL::Memory self, cl_mem_info name)
979 PPCODE: 2183 PPCODE:
980 INFO (MemObject) 2184 INFO (MemObject)
981 2185
2186void
2187destructor_callback (OpenCL::Memory self, SV *notify)
2188 PPCODE:
2189 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2190
982#BEGIN:mem 2191#BEGIN:mem
983 2192
984void 2193void
985type (OpenCL::Memory this) 2194type (OpenCL::Memory self)
2195 ALIAS:
2196 type = CL_MEM_TYPE
2197 map_count = CL_MEM_MAP_COUNT
2198 reference_count = CL_MEM_REFERENCE_COUNT
986 PPCODE: 2199 PPCODE:
987 cl_mem_object_type value [1]; 2200 cl_uint value [1];
988 NEED_SUCCESS (GetMemObjectInfo, (this, CL_MEM_TYPE, sizeof (value), value, 0)); 2201 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
989 EXTEND (SP, 1); 2202 EXTEND (SP, 1);
990 const int i = 0; 2203 const int i = 0;
991 PUSHs (sv_2mortal (newSViv (value [i]))); 2204 PUSHs (sv_2mortal (newSVuv (value [i])));
992 2205
993void 2206void
994flags (OpenCL::Memory this) 2207flags (OpenCL::Memory self)
995 PPCODE: 2208 PPCODE:
996 cl_mem_flags value [1]; 2209 cl_ulong value [1];
997 NEED_SUCCESS (GetMemObjectInfo, (this, CL_MEM_FLAGS, sizeof (value), value, 0)); 2210 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
998 EXTEND (SP, 1); 2211 EXTEND (SP, 1);
999 const int i = 0; 2212 const int i = 0;
1000 PUSHs (sv_2mortal (newSViv (value [i]))); 2213 PUSHs (sv_2mortal (newSVuv (value [i])));
1001 2214
1002void 2215void
1003size (OpenCL::Memory this) 2216size (OpenCL::Memory self)
1004 ALIAS: 2217 ALIAS:
1005 size = CL_MEM_SIZE 2218 size = CL_MEM_SIZE
1006 offset = CL_MEM_OFFSET 2219 offset = CL_MEM_OFFSET
1007 PPCODE: 2220 PPCODE:
1008 size_t value [1]; 2221 size_t value [1];
1009 NEED_SUCCESS (GetMemObjectInfo, (this, ix, sizeof (value), value, 0)); 2222 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1010 EXTEND (SP, 1); 2223 EXTEND (SP, 1);
1011 const int i = 0; 2224 const int i = 0;
1012 PUSHs (sv_2mortal (newSVuv (value [i]))); 2225 PUSHs (sv_2mortal (newSVuv (value [i])));
1013 2226
1014void 2227void
1015host_ptr (OpenCL::Memory this) 2228host_ptr (OpenCL::Memory self)
1016 PPCODE: 2229 PPCODE:
1017 void * value [1]; 2230 void * value [1];
1018 NEED_SUCCESS (GetMemObjectInfo, (this, CL_MEM_HOST_PTR, sizeof (value), value, 0)); 2231 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
1019 EXTEND (SP, 1); 2232 EXTEND (SP, 1);
1020 const int i = 0; 2233 const int i = 0;
1021 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2234 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1022 2235
1023void 2236void
1024map_count (OpenCL::Memory this)
1025 ALIAS:
1026 map_count = CL_MEM_MAP_COUNT
1027 reference_count = CL_MEM_REFERENCE_COUNT
1028 PPCODE:
1029 cl_uint value [1];
1030 NEED_SUCCESS (GetMemObjectInfo, (this, ix, sizeof (value), value, 0));
1031 EXTEND (SP, 1);
1032 const int i = 0;
1033 PUSHs (sv_2mortal (newSVuv (value [i])));
1034
1035void
1036context (OpenCL::Memory this) 2237context (OpenCL::Memory self)
1037 PPCODE: 2238 PPCODE:
1038 cl_context value [1]; 2239 cl_context value [1];
1039 NEED_SUCCESS (GetMemObjectInfo, (this, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1040 EXTEND (SP, 1); 2241 EXTEND (SP, 1);
1041 const int i = 0; 2242 const int i = 0;
1042 {
1043 NEED_SUCCESS (RetainContext, (value [i])); 2243 NEED_SUCCESS (RetainContext, (value [i]));
1044 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2244 PUSH_CLOBJ (stash_context, value [i]);
1045 }
1046 2245
1047void 2246void
1048associated_memobject (OpenCL::Memory this) 2247associated_memobject (OpenCL::Memory self)
1049 PPCODE: 2248 PPCODE:
1050 cl_mem value [1]; 2249 cl_mem value [1];
1051 NEED_SUCCESS (GetMemObjectInfo, (this, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1052 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1053 const int i = 0; 2252 const int i = 0;
1054 {
1055 NEED_SUCCESS (RetainMemObject, (value [i])); 2253 NEED_SUCCESS (RetainMemObject, (value [i]));
1056 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2254 PUSH_CLOBJ (stash_memory, value [i]);
1057 }
1058 2255
1059#END:mem 2256#END:mem
1060 2257
2258#if cl_apple_gl_sharing || cl_khr_gl_sharing
2259
2260void
2261gl_object_info (OpenCL::Memory self)
2262 PPCODE:
2263 cl_gl_object_type type;
2264 cl_GLuint name;
2265 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2266 EXTEND (SP, 2);
2267 PUSHs (sv_2mortal (newSVuv (type)));
2268 PUSHs (sv_2mortal (newSVuv (name)));
2269
2270#endif
2271
2272MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2273
2274void
2275sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2276 PPCODE:
2277 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2278 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2279
2280 cl_buffer_region crdata = { origin, size };
2281
2282 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2283 XPUSH_CLOBJ (stash_buffer, mem);
2284
1061MODULE = OpenCL PACKAGE = OpenCL::Image 2285MODULE = OpenCL PACKAGE = OpenCL::Image
1062 2286
1063void 2287void
1064image_info (OpenCL::Image this, cl_image_info name) 2288image_info (OpenCL::Image self, cl_image_info name)
1065 PPCODE: 2289 PPCODE:
1066 INFO (Image) 2290 INFO (Image)
1067 2291
2292void
2293format (OpenCL::Image self)
2294 PPCODE:
2295 cl_image_format format;
2296 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2297 EXTEND (SP, 2);
2298 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2299 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2300
1068#BEGIN:image 2301#BEGIN:image
1069 2302
1070void 2303void
1071element_size (OpenCL::Image this) 2304element_size (OpenCL::Image self)
1072 ALIAS: 2305 ALIAS:
1073 element_size = CL_IMAGE_ELEMENT_SIZE 2306 element_size = CL_IMAGE_ELEMENT_SIZE
1074 row_pitch = CL_IMAGE_ROW_PITCH 2307 row_pitch = CL_IMAGE_ROW_PITCH
1075 slice_pitch = CL_IMAGE_SLICE_PITCH 2308 slice_pitch = CL_IMAGE_SLICE_PITCH
1076 width = CL_IMAGE_WIDTH 2309 width = CL_IMAGE_WIDTH
1077 height = CL_IMAGE_HEIGHT 2310 height = CL_IMAGE_HEIGHT
1078 depth = CL_IMAGE_DEPTH 2311 depth = CL_IMAGE_DEPTH
1079 PPCODE: 2312 PPCODE:
1080 size_t value [1]; 2313 size_t value [1];
1081 NEED_SUCCESS (GetImageInfo, (this, ix, sizeof (value), value, 0)); 2314 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
1082 EXTEND (SP, 1); 2315 EXTEND (SP, 1);
1083 const int i = 0; 2316 const int i = 0;
1084 PUSHs (sv_2mortal (newSVuv (value [i]))); 2317 PUSHs (sv_2mortal (newSVuv (value [i])));
1085 2318
1086#END:image 2319#END:image
1087 2320
2321#if cl_apple_gl_sharing || cl_khr_gl_sharing
2322
2323#BEGIN:gl_texture
2324
2325void
2326target (OpenCL::Image self)
2327 PPCODE:
2328 cl_GLenum value [1];
2329 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2330 EXTEND (SP, 1);
2331 const int i = 0;
2332 PUSHs (sv_2mortal (newSVuv (value [i])));
2333
2334void
2335gl_mipmap_level (OpenCL::Image self)
2336 PPCODE:
2337 cl_GLint value [1];
2338 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2339 EXTEND (SP, 1);
2340 const int i = 0;
2341 PUSHs (sv_2mortal (newSViv (value [i])));
2342
2343#END:gl_texture
2344
2345#endif
2346
1088MODULE = OpenCL PACKAGE = OpenCL::Sampler 2347MODULE = OpenCL PACKAGE = OpenCL::Sampler
1089 2348
1090void 2349void
1091DESTROY (OpenCL::Sampler this) 2350DESTROY (OpenCL::Sampler self)
1092 CODE: 2351 CODE:
1093 clReleaseSampler (this); 2352 clReleaseSampler (self);
1094 2353
1095void 2354void
1096info (OpenCL::Sampler this, cl_sampler_info name) 2355info (OpenCL::Sampler self, cl_sampler_info name)
1097 PPCODE: 2356 PPCODE:
1098 INFO (Sampler) 2357 INFO (Sampler)
1099 2358
1100#BEGIN:sampler 2359#BEGIN:sampler
1101 2360
1102void 2361void
1103reference_count (OpenCL::Sampler this) 2362reference_count (OpenCL::Sampler self)
2363 ALIAS:
2364 reference_count = CL_SAMPLER_REFERENCE_COUNT
2365 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2366 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1104 PPCODE: 2367 PPCODE:
1105 cl_uint value [1]; 2368 cl_uint value [1];
1106 NEED_SUCCESS (GetSamplerInfo, (this, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2369 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1107 EXTEND (SP, 1); 2370 EXTEND (SP, 1);
1108 const int i = 0; 2371 const int i = 0;
1109 PUSHs (sv_2mortal (newSVuv (value [i]))); 2372 PUSHs (sv_2mortal (newSVuv (value [i])));
1110 2373
1111void 2374void
1112context (OpenCL::Sampler this) 2375context (OpenCL::Sampler self)
1113 PPCODE: 2376 PPCODE:
1114 cl_context value [1]; 2377 cl_context value [1];
1115 NEED_SUCCESS (GetSamplerInfo, (this, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2378 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1116 EXTEND (SP, 1); 2379 EXTEND (SP, 1);
1117 const int i = 0; 2380 const int i = 0;
1118 {
1119 NEED_SUCCESS (RetainContext, (value [i])); 2381 NEED_SUCCESS (RetainContext, (value [i]));
1120 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2382 PUSH_CLOBJ (stash_context, value [i]);
1121 }
1122 2383
1123void 2384void
1124normalized_coords (OpenCL::Sampler this)
1125 PPCODE:
1126 cl_addressing_mode value [1];
1127 NEED_SUCCESS (GetSamplerInfo, (this, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1128 EXTEND (SP, 1);
1129 const int i = 0;
1130 PUSHs (sv_2mortal (newSViv (value [i])));
1131
1132void
1133addressing_mode (OpenCL::Sampler this)
1134 PPCODE:
1135 cl_filter_mode value [1];
1136 NEED_SUCCESS (GetSamplerInfo, (this, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1137 EXTEND (SP, 1);
1138 const int i = 0;
1139 PUSHs (sv_2mortal (newSViv (value [i])));
1140
1141void
1142filter_mode (OpenCL::Sampler this) 2385filter_mode (OpenCL::Sampler self)
1143 PPCODE: 2386 PPCODE:
1144 cl_bool value [1]; 2387 cl_uint value [1];
1145 NEED_SUCCESS (GetSamplerInfo, (this, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1146 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1147 const int i = 0; 2390 const int i = 0;
1148 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2391 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1149 2392
1150#END:sampler 2393#END:sampler
1151 2394
1152MODULE = OpenCL PACKAGE = OpenCL::Program 2395MODULE = OpenCL PACKAGE = OpenCL::Program
1153 2396
1154void 2397void
1155DESTROY (OpenCL::Program this) 2398DESTROY (OpenCL::Program self)
1156 CODE: 2399 CODE:
1157 clReleaseProgram (this); 2400 clReleaseProgram (self);
1158 2401
1159void 2402void
1160build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2403build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2404 ALIAS:
2405 build_async = 1
1161 CODE: 2406 CODE:
2407 cl_uint device_count = 0;
2408 cl_device_id *device_list = 0;
2409
2410 if (SvOK (devices))
2411 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2412
2413 void *user_data;
2414 program_callback pfn_notify = make_program_callback (notify, &user_data);
2415
2416 if (ix)
2417 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2418 else
2419 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2420
2421#if CL_VERSION_1_2
2422
2423void
2424compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2425 CODE:
2426 cl_uint device_count = 0;
2427 cl_device_id *device_list = 0;
2428
2429 if (SvOK (devices))
2430 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2431
2432 cl_uint header_count = 0;
2433 cl_program *header_list = 0;
2434 const char **header_name = 0;
2435
2436 if (SvOK (headers))
2437 {
2438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2439 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2440
2441 HV *hv = (HV *)SvRV (devices);
2442
2443 header_count = hv_iterinit (hv);
2444 header_list = tmpbuf (sizeof (*header_list) * header_count);
2445 header_name = tmpbuf (sizeof (*header_name) * header_count);
2446
2447 HE *he;
2448 int i = 0;
2449 while (he = hv_iternext (hv))
2450 {
2451 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2452 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2453 ++i;
2454 }
2455 }
2456
2457 void *user_data;
2458 program_callback pfn_notify = make_program_callback (notify, &user_data);
2459
1162 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2460 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2461 header_count, header_list, header_name, pfn_notify, user_data));
1163 2462
2463#endif
2464
1164void 2465void
1165build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2466build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1166 PPCODE: 2467 PPCODE:
1167 size_t size; 2468 size_t size;
1168 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2469 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
1169 SV *sv = sv_2mortal (newSV (size)); 2470 SV *sv = sv_2mortal (newSV (size));
1170 SvUPGRADE (sv, SVt_PV); 2471 SvUPGRADE (sv, SVt_PV);
1171 SvPOK_only (sv); 2472 SvPOK_only (sv);
1172 SvCUR_set (sv, size); 2473 SvCUR_set (sv, size);
1173 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2474 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
1174 XPUSHs (sv); 2475 XPUSHs (sv);
1175 2476
1176#BEGIN:program_build 2477#BEGIN:program_build
1177 2478
1178void 2479void
1179build_status (OpenCL::Program this, OpenCL::Device device) 2480build_status (OpenCL::Program self, OpenCL::Device device)
1180 PPCODE: 2481 PPCODE:
1181 cl_build_status value [1]; 2482 cl_int value [1];
1182 NEED_SUCCESS (GetProgramBuildInfo, (this, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2483 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1183 EXTEND (SP, 1); 2484 EXTEND (SP, 1);
1184 const int i = 0; 2485 const int i = 0;
1185 PUSHs (sv_2mortal (newSViv (value [i]))); 2486 PUSHs (sv_2mortal (newSViv (value [i])));
1186 2487
1187void 2488void
1188build_options (OpenCL::Program this, OpenCL::Device device) 2489build_options (OpenCL::Program self, OpenCL::Device device)
1189 ALIAS: 2490 ALIAS:
1190 build_options = CL_PROGRAM_BUILD_OPTIONS 2491 build_options = CL_PROGRAM_BUILD_OPTIONS
1191 build_log = CL_PROGRAM_BUILD_LOG 2492 build_log = CL_PROGRAM_BUILD_LOG
1192 PPCODE: 2493 PPCODE:
1193 size_t size; 2494 size_t size;
1194 NEED_SUCCESS (GetProgramBuildInfo, (this, device, ix, 0, 0, &size)); 2495 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
1195 char *value = tmpbuf (size); 2496 char *value = tmpbuf (size);
1196 NEED_SUCCESS (GetProgramBuildInfo, (this, device, ix, size, value, 0)); 2497 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1197 EXTEND (SP, 1); 2498 EXTEND (SP, 1);
1198 const int i = 0; 2499 const int i = 0;
1199 PUSHs (sv_2mortal (newSVpv (value, 0))); 2500 PUSHs (sv_2mortal (newSVpv (value, 0)));
1200 2501
2502void
2503binary_type (OpenCL::Program self, OpenCL::Device device)
2504 PPCODE:
2505 cl_uint value [1];
2506 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2507 EXTEND (SP, 1);
2508 const int i = 0;
2509 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2510
1201#END:program_build 2511#END:program_build
1202 2512
1203void 2513void
1204kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
1205 PPCODE: 2515 PPCODE:
1206 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1207 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
1208 2518
1209void 2519void
2520kernels_in_program (OpenCL::Program program)
2521 PPCODE:
2522 cl_uint num_kernels;
2523 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2524 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2526
2527 int i;
2528 EXTEND (SP, num_kernels);
2529 for (i = 0; i < num_kernels; ++i)
2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
2531
2532void
1210info (OpenCL::Program this, cl_program_info name) 2533info (OpenCL::Program self, cl_program_info name)
1211 PPCODE: 2534 PPCODE:
1212 INFO (Program) 2535 INFO (Program)
1213 2536
1214void 2537void
1215binaries (OpenCL::Program this) 2538binaries (OpenCL::Program self)
1216 PPCODE: 2539 PPCODE:
1217 cl_uint n, i; 2540 cl_uint n, i;
1218 size_t size; 2541 size_t size;
1219 2542
1220 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0)); 2543 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
1221 if (!n) XSRETURN_EMPTY; 2544 if (!n) XSRETURN_EMPTY;
1222 2545
1223 size_t *sizes = tmpbuf (sizeof (*sizes) * n); 2546 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
1224 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size)); 2547 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
1225 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY; 2548 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
1226 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n); 2549 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
1227 2550
1228 EXTEND (SP, n); 2551 EXTEND (SP, n);
1229 for (i = 0; i < n; ++i) 2552 for (i = 0; i < n; ++i)
1230 { 2553 {
1231 SV *sv = sv_2mortal (newSV (sizes [i])); 2554 SV *sv = sv_2mortal (newSV (sizes [i]));
1232 SvUPGRADE (sv, SVt_PV); 2555 SvUPGRADE (sv, SVt_PV);
1233 SvPOK_only (sv); 2556 SvPOK_only (sv);
1234 SvCUR_set (sv, sizes [i]); 2557 SvCUR_set (sv, sizes [i]);
1235 ptrs [i] = SvPVX (sv); 2558 ptrs [i] = (void *)SvPVX (sv);
1236 PUSHs (sv); 2559 PUSHs (sv);
1237 } 2560 }
1238 2561
1239 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2562 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1240 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2563 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1241 2564
1242#BEGIN:program 2565#BEGIN:program
1243 2566
1244void 2567void
1245reference_count (OpenCL::Program this) 2568reference_count (OpenCL::Program self)
1246 ALIAS: 2569 ALIAS:
1247 reference_count = CL_PROGRAM_REFERENCE_COUNT 2570 reference_count = CL_PROGRAM_REFERENCE_COUNT
1248 num_devices = CL_PROGRAM_NUM_DEVICES 2571 num_devices = CL_PROGRAM_NUM_DEVICES
1249 PPCODE: 2572 PPCODE:
1250 cl_uint value [1]; 2573 cl_uint value [1];
1251 NEED_SUCCESS (GetProgramInfo, (this, ix, sizeof (value), value, 0)); 2574 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
1252 EXTEND (SP, 1); 2575 EXTEND (SP, 1);
1253 const int i = 0; 2576 const int i = 0;
1254 PUSHs (sv_2mortal (newSVuv (value [i]))); 2577 PUSHs (sv_2mortal (newSVuv (value [i])));
1255 2578
1256void 2579void
1257context (OpenCL::Program this) 2580context (OpenCL::Program self)
1258 PPCODE: 2581 PPCODE:
1259 cl_context value [1]; 2582 cl_context value [1];
1260 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1261 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
1262 const int i = 0; 2585 const int i = 0;
1263 {
1264 NEED_SUCCESS (RetainContext, (value [i])); 2586 NEED_SUCCESS (RetainContext, (value [i]));
1265 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2587 PUSH_CLOBJ (stash_context, value [i]);
1266 }
1267 2588
1268void 2589void
1269devices (OpenCL::Program this) 2590devices (OpenCL::Program self)
1270 PPCODE: 2591 PPCODE:
1271 size_t size; 2592 size_t size;
1272 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_DEVICES, 0, 0, &size)); 2593 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
1273 cl_device_id *value = tmpbuf (size); 2594 cl_device_id *value = tmpbuf (size);
1274 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_DEVICES, size, value, 0)); 2595 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1275 int i, n = size / sizeof (*value); 2596 int i, n = size / sizeof (*value);
1276 EXTEND (SP, n); 2597 EXTEND (SP, n);
1277 for (i = 0; i < n; ++i) 2598 for (i = 0; i < n; ++i)
1278 { 2599 PUSH_CLOBJ (stash_device, value [i]);
1279 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1280 }
1281 2600
1282void 2601void
1283source (OpenCL::Program this) 2602source (OpenCL::Program self)
1284 PPCODE: 2603 PPCODE:
1285 size_t size; 2604 size_t size;
1286 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_SOURCE, 0, 0, &size)); 2605 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
1287 char *value = tmpbuf (size); 2606 char *value = tmpbuf (size);
1288 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_SOURCE, size, value, 0)); 2607 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
1289 EXTEND (SP, 1); 2608 EXTEND (SP, 1);
1290 const int i = 0; 2609 const int i = 0;
1291 PUSHs (sv_2mortal (newSVpv (value, 0))); 2610 PUSHs (sv_2mortal (newSVpv (value, 0)));
1292 2611
1293void 2612void
1294binary_sizes (OpenCL::Program this) 2613binary_sizes (OpenCL::Program self)
1295 PPCODE: 2614 PPCODE:
1296 size_t size; 2615 size_t size;
1297 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_BINARY_SIZES, 0, 0, &size)); 2616 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
1298 size_t *value = tmpbuf (size); 2617 size_t *value = tmpbuf (size);
1299 NEED_SUCCESS (GetProgramInfo, (this, CL_PROGRAM_BINARY_SIZES, size, value, 0)); 2618 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
1300 int i, n = size / sizeof (*value); 2619 int i, n = size / sizeof (*value);
1301 EXTEND (SP, n); 2620 EXTEND (SP, n);
1302 for (i = 0; i < n; ++i) 2621 for (i = 0; i < n; ++i)
1303 PUSHs (sv_2mortal (newSVuv (value [i]))); 2622 PUSHs (sv_2mortal (newSVuv (value [i])));
1304 2623
1305#END:program 2624#END:program
1306 2625
1307MODULE = OpenCL PACKAGE = OpenCL::Kernel 2626MODULE = OpenCL PACKAGE = OpenCL::Kernel
1308 2627
1309void 2628void
1310DESTROY (OpenCL::Kernel this) 2629DESTROY (OpenCL::Kernel self)
1311 CODE: 2630 CODE:
1312 clReleaseKernel (this); 2631 clReleaseKernel (self);
1313 2632
1314void 2633void
2634setf (OpenCL::Kernel self, const char *format, ...)
2635 CODE:
2636 int i;
2637 for (i = 2; ; ++i)
2638 {
2639 while (*format == ' ')
2640 ++format;
2641
2642 char type = *format++;
2643
2644 if (!type)
2645 break;
2646
2647 if (i >= items)
2648 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2649
2650 SV *sv = ST (i);
2651
2652 union
2653 {
2654 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2655 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2656 cl_half ch; cl_float cf; cl_double cd;
2657 cl_mem cm;
2658 cl_sampler ca;
2659 size_t cz;
2660 cl_event ce;
2661 } arg;
2662 size_t size;
2663 int nullarg = 0;
2664
2665 switch (type)
2666 {
2667 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2668 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2669 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2670 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2671 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2672 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2673 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2674 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2675
2676 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2677 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2678 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2679
2680 case 'z': nullarg = 1; size = SvIV (sv); break;
2681
2682 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2683 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2684 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2685
2686 default:
2687 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2688 }
2689
2690 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2691 if (res)
2692 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2693 }
2694
2695 if (i != items)
2696 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2697
2698void
1315set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2699set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1316 CODE: 2700 CODE:
1317 clSetKernelArg (this, idx, sizeof (value), &value); 2701 clSetKernelArg (self, idx, sizeof (value), &value);
1318 2702
1319void 2703void
1320set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value) 2704set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
1321 CODE: 2705 CODE:
1322 clSetKernelArg (this, idx, sizeof (value), &value); 2706 clSetKernelArg (self, idx, sizeof (value), &value);
1323 2707
1324void 2708void
1325set_short (OpenCL::Kernel this, cl_uint idx, cl_short value) 2709set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
1326 CODE: 2710 CODE:
1327 clSetKernelArg (this, idx, sizeof (value), &value); 2711 clSetKernelArg (self, idx, sizeof (value), &value);
1328 2712
1329void 2713void
1330set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2714set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
1331 CODE: 2715 CODE:
1332 clSetKernelArg (this, idx, sizeof (value), &value); 2716 clSetKernelArg (self, idx, sizeof (value), &value);
1333 2717
1334void 2718void
1335set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2719set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
1336 CODE: 2720 CODE:
1337 clSetKernelArg (this, idx, sizeof (value), &value); 2721 clSetKernelArg (self, idx, sizeof (value), &value);
1338 2722
1339void 2723void
1340set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value) 2724set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
1341 CODE: 2725 CODE:
1342 clSetKernelArg (this, idx, sizeof (value), &value); 2726 clSetKernelArg (self, idx, sizeof (value), &value);
1343 2727
1344void 2728void
1345set_long (OpenCL::Kernel this, cl_uint idx, cl_long value) 2729set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
1346 CODE: 2730 CODE:
1347 clSetKernelArg (this, idx, sizeof (value), &value); 2731 clSetKernelArg (self, idx, sizeof (value), &value);
1348 2732
1349void 2733void
1350set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value) 2734set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
1351 CODE: 2735 CODE:
1352 clSetKernelArg (this, idx, sizeof (value), &value); 2736 clSetKernelArg (self, idx, sizeof (value), &value);
1353 2737
1354void 2738void
1355set_half (OpenCL::Kernel this, cl_uint idx, cl_half value) 2739set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
1356 CODE: 2740 CODE:
1357 clSetKernelArg (this, idx, sizeof (value), &value); 2741 clSetKernelArg (self, idx, sizeof (value), &value);
1358 2742
1359void 2743void
1360set_float (OpenCL::Kernel this, cl_uint idx, cl_float value) 2744set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
1361 CODE: 2745 CODE:
1362 clSetKernelArg (this, idx, sizeof (value), &value); 2746 clSetKernelArg (self, idx, sizeof (value), &value);
1363 2747
1364void 2748void
1365set_double (OpenCL::Kernel this, cl_uint idx, cl_double value) 2749set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
1366 CODE: 2750 CODE:
1367 clSetKernelArg (this, idx, sizeof (value), &value); 2751 clSetKernelArg (self, idx, sizeof (value), &value);
1368 2752
1369void 2753void
1370set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value) 2754set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1371 CODE: 2755 CODE:
1372 clSetKernelArg (this, idx, sizeof (value), &value); 2756 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1373 2757
1374void 2758void
1375set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value) 2759set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1376 CODE: 2760 CODE:
1377 clSetKernelArg (this, idx, sizeof (value), &value); 2761 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1378 2762
1379void 2763void
1380set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value) 2764set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
1381 CODE: 2765 CODE:
1382 clSetKernelArg (this, idx, sizeof (value), &value); 2766 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1383 2767
1384void 2768void
1385set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
1386 CODE:
1387 clSetKernelArg (this, idx, sizeof (value), &value);
1388
1389void
1390set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value) 2769set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1391 CODE: 2770 CODE:
1392 clSetKernelArg (this, idx, sizeof (value), &value); 2771 clSetKernelArg (self, idx, sizeof (value), &value);
1393 2772
1394void 2773void
2774set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2775 CODE:
2776 clSetKernelArg (self, idx, size, 0);
2777
2778void
1395set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value) 2779set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1396 CODE: 2780 CODE:
1397 clSetKernelArg (this, idx, sizeof (value), &value); 2781 clSetKernelArg (self, idx, sizeof (value), &value);
1398 2782
1399void 2783void
1400info (OpenCL::Kernel this, cl_kernel_info name) 2784info (OpenCL::Kernel self, cl_kernel_info name)
1401 PPCODE: 2785 PPCODE:
1402 INFO (Kernel) 2786 INFO (Kernel)
1403 2787
1404#BEGIN:kernel 2788#BEGIN:kernel
1405 2789
1406void 2790void
1407function_name (OpenCL::Kernel this) 2791function_name (OpenCL::Kernel self)
1408 PPCODE: 2792 PPCODE:
1409 size_t size; 2793 size_t size;
1410 NEED_SUCCESS (GetKernelInfo, (this, CL_KERNEL_FUNCTION_NAME, 0, 0, &size)); 2794 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
1411 char *value = tmpbuf (size); 2795 char *value = tmpbuf (size);
1412 NEED_SUCCESS (GetKernelInfo, (this, CL_KERNEL_FUNCTION_NAME, size, value, 0)); 2796 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
1413 EXTEND (SP, 1); 2797 EXTEND (SP, 1);
1414 const int i = 0; 2798 const int i = 0;
1415 PUSHs (sv_2mortal (newSVpv (value, 0))); 2799 PUSHs (sv_2mortal (newSVpv (value, 0)));
1416 2800
1417void 2801void
1418num_args (OpenCL::Kernel this) 2802num_args (OpenCL::Kernel self)
1419 ALIAS: 2803 ALIAS:
1420 num_args = CL_KERNEL_NUM_ARGS 2804 num_args = CL_KERNEL_NUM_ARGS
1421 reference_count = CL_KERNEL_REFERENCE_COUNT 2805 reference_count = CL_KERNEL_REFERENCE_COUNT
1422 PPCODE: 2806 PPCODE:
1423 cl_uint value [1]; 2807 cl_uint value [1];
1424 NEED_SUCCESS (GetKernelInfo, (this, ix, sizeof (value), value, 0)); 2808 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
1425 EXTEND (SP, 1); 2809 EXTEND (SP, 1);
1426 const int i = 0; 2810 const int i = 0;
1427 PUSHs (sv_2mortal (newSVuv (value [i]))); 2811 PUSHs (sv_2mortal (newSVuv (value [i])));
1428 2812
1429void 2813void
1430context (OpenCL::Kernel this) 2814context (OpenCL::Kernel self)
1431 PPCODE: 2815 PPCODE:
1432 cl_context value [1]; 2816 cl_context value [1];
1433 NEED_SUCCESS (GetKernelInfo, (this, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2817 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1434 EXTEND (SP, 1); 2818 EXTEND (SP, 1);
1435 const int i = 0; 2819 const int i = 0;
1436 {
1437 NEED_SUCCESS (RetainContext, (value [i])); 2820 NEED_SUCCESS (RetainContext, (value [i]));
1438 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2821 PUSH_CLOBJ (stash_context, value [i]);
1439 }
1440 2822
1441void 2823void
1442program (OpenCL::Kernel this) 2824program (OpenCL::Kernel self)
1443 PPCODE: 2825 PPCODE:
1444 cl_program value [1]; 2826 cl_program value [1];
1445 NEED_SUCCESS (GetKernelInfo, (this, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2827 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1446 EXTEND (SP, 1); 2828 EXTEND (SP, 1);
1447 const int i = 0; 2829 const int i = 0;
1448 {
1449 NEED_SUCCESS (RetainProgram, (value [i])); 2830 NEED_SUCCESS (RetainProgram, (value [i]));
1450 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2831 PUSH_CLOBJ (stash_program, value [i]);
1451 }
1452 2832
1453#END:kernel 2833#END:kernel
1454 2834
1455void 2835void
1456work_group_info (OpenCL::Kernel this, OpenCL::Device device, cl_kernel_work_group_info name) 2836work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
1457 PPCODE: 2837 PPCODE:
1458 size_t size; 2838 size_t size;
1459 NEED_SUCCESS (GetKernelWorkGroupInfo, (this, device, name, 0, 0, &size)); 2839 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
1460 SV *sv = sv_2mortal (newSV (size)); 2840 SV *sv = sv_2mortal (newSV (size));
1461 SvUPGRADE (sv, SVt_PV); 2841 SvUPGRADE (sv, SVt_PV);
1462 SvPOK_only (sv); 2842 SvPOK_only (sv);
1463 SvCUR_set (sv, size); 2843 SvCUR_set (sv, size);
1464 NEED_SUCCESS (GetKernelWorkGroupInfo, (this, device, name, size, SvPVX (sv), 0)); 2844 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
1465 XPUSHs (sv); 2845 XPUSHs (sv);
1466 2846
1467#BEGIN:kernel_work_group 2847#BEGIN:kernel_work_group
1468 2848
1469void 2849void
1470work_group_size (OpenCL::Kernel this, OpenCL::Device device) 2850work_group_size (OpenCL::Kernel self, OpenCL::Device device)
1471 ALIAS: 2851 ALIAS:
1472 work_group_size = CL_KERNEL_WORK_GROUP_SIZE 2852 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
1473 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE 2853 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
1474 PPCODE: 2854 PPCODE:
1475 size_t value [1]; 2855 size_t value [1];
1476 NEED_SUCCESS (GetKernelWorkGroupInfo, (this, device, ix, sizeof (value), value, 0)); 2856 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
1477 EXTEND (SP, 1); 2857 EXTEND (SP, 1);
1478 const int i = 0; 2858 const int i = 0;
1479 PUSHs (sv_2mortal (newSVuv (value [i]))); 2859 PUSHs (sv_2mortal (newSVuv (value [i])));
1480 2860
1481void 2861void
1482compile_work_group_size (OpenCL::Kernel this, OpenCL::Device device) 2862compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
1483 PPCODE: 2863 PPCODE:
1484 size_t size; 2864 size_t size;
1485 NEED_SUCCESS (GetKernelWorkGroupInfo, (this, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size)); 2865 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
1486 size_t *value = tmpbuf (size); 2866 size_t *value = tmpbuf (size);
1487 NEED_SUCCESS (GetKernelWorkGroupInfo, (this, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0)); 2867 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
1488 int i, n = size / sizeof (*value); 2868 int i, n = size / sizeof (*value);
1489 EXTEND (SP, n); 2869 EXTEND (SP, n);
1490 for (i = 0; i < n; ++i) 2870 for (i = 0; i < n; ++i)
1491 PUSHs (sv_2mortal (newSVuv (value [i]))); 2871 PUSHs (sv_2mortal (newSVuv (value [i])));
1492 2872
1493void 2873void
1494local_mem_size (OpenCL::Kernel this, OpenCL::Device device) 2874local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
1495 ALIAS: 2875 ALIAS:
1496 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE 2876 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
1497 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE 2877 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
1498 PPCODE: 2878 PPCODE:
1499 cl_ulong value [1]; 2879 cl_ulong value [1];
1500 NEED_SUCCESS (GetKernelWorkGroupInfo, (this, device, ix, sizeof (value), value, 0)); 2880 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
1501 EXTEND (SP, 1); 2881 EXTEND (SP, 1);
1502 const int i = 0; 2882 const int i = 0;
1503 PUSHs (sv_2mortal (newSVuv (value [i]))); 2883 PUSHs (sv_2mortal (newSVuv (value [i])));
1504 2884
1505#END:kernel_work_group 2885#END:kernel_work_group
1506 2886
2887#if CL_VERSION_1_2
2888
2889void
2890arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2891 PPCODE:
2892 size_t size;
2893 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2894 SV *sv = sv_2mortal (newSV (size));
2895 SvUPGRADE (sv, SVt_PV);
2896 SvPOK_only (sv);
2897 SvCUR_set (sv, size);
2898 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2899 XPUSHs (sv);
2900
2901#BEGIN:kernel_arg
2902
2903void
2904arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2905 ALIAS:
2906 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2907 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2908 PPCODE:
2909 cl_uint value [1];
2910 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2911 EXTEND (SP, 1);
2912 const int i = 0;
2913 PUSHs (sv_2mortal (newSVuv (value [i])));
2914
2915void
2916arg_type_name (OpenCL::Kernel self, cl_uint idx)
2917 ALIAS:
2918 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2919 arg_name = CL_KERNEL_ARG_NAME
2920 PPCODE:
2921 size_t size;
2922 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2923 char *value = tmpbuf (size);
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2925 EXTEND (SP, 1);
2926 const int i = 0;
2927 PUSHs (sv_2mortal (newSVpv (value, 0)));
2928
2929void
2930arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2931 PPCODE:
2932 cl_ulong value [1];
2933 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2934 EXTEND (SP, 1);
2935 const int i = 0;
2936 PUSHs (sv_2mortal (newSVuv (value [i])));
2937
2938#END:kernel_arg
2939
2940#endif
2941
1507MODULE = OpenCL PACKAGE = OpenCL::Event 2942MODULE = OpenCL PACKAGE = OpenCL::Event
1508 2943
1509void 2944void
1510DESTROY (OpenCL::Event this) 2945DESTROY (OpenCL::Event self)
1511 CODE: 2946 CODE:
1512 clReleaseEvent (this); 2947 clReleaseEvent (self);
1513 2948
1514void 2949void
1515wait (OpenCL::Event this) 2950wait (OpenCL::Event self)
1516 CODE: 2951 CODE:
1517 clWaitForEvents (1, &this); 2952 clWaitForEvents (1, &self);
1518 2953
1519void 2954void
2955cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2956 CODE:
2957 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
2958
2959void
1520info (OpenCL::Event this, cl_event_info name) 2960info (OpenCL::Event self, cl_event_info name)
1521 PPCODE: 2961 PPCODE:
1522 INFO (Event) 2962 INFO (Event)
1523 2963
1524#BEGIN:event 2964#BEGIN:event
1525 2965
1526void 2966void
1527command_queue (OpenCL::Event this) 2967command_queue (OpenCL::Event self)
1528 PPCODE: 2968 PPCODE:
1529 cl_command_queue value [1]; 2969 cl_command_queue value [1];
1530 NEED_SUCCESS (GetEventInfo, (this, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2970 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1531 EXTEND (SP, 1); 2971 EXTEND (SP, 1);
1532 const int i = 0; 2972 const int i = 0;
1533 {
1534 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2973 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1535 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2974 PUSH_CLOBJ (stash_queue, value [i]);
1536 }
1537 2975
1538void 2976void
1539command_type (OpenCL::Event this) 2977command_type (OpenCL::Event self)
1540 PPCODE:
1541 cl_command_type value [1];
1542 NEED_SUCCESS (GetEventInfo, (this, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
1543 EXTEND (SP, 1);
1544 const int i = 0;
1545 PUSHs (sv_2mortal (newSVuv (value [i])));
1546
1547void
1548reference_count (OpenCL::Event this)
1549 ALIAS: 2978 ALIAS:
2979 command_type = CL_EVENT_COMMAND_TYPE
1550 reference_count = CL_EVENT_REFERENCE_COUNT 2980 reference_count = CL_EVENT_REFERENCE_COUNT
1551 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2981 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
1552 PPCODE: 2982 PPCODE:
1553 cl_uint value [1]; 2983 cl_uint value [1];
1554 NEED_SUCCESS (GetEventInfo, (this, ix, sizeof (value), value, 0)); 2984 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
1555 EXTEND (SP, 1); 2985 EXTEND (SP, 1);
1556 const int i = 0; 2986 const int i = 0;
1557 PUSHs (sv_2mortal (newSVuv (value [i]))); 2987 PUSHs (sv_2mortal (newSVuv (value [i])));
1558 2988
1559void 2989void
1560context (OpenCL::Event this) 2990context (OpenCL::Event self)
1561 PPCODE: 2991 PPCODE:
1562 cl_context value [1]; 2992 cl_context value [1];
1563 NEED_SUCCESS (GetEventInfo, (this, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2993 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1564 EXTEND (SP, 1); 2994 EXTEND (SP, 1);
1565 const int i = 0; 2995 const int i = 0;
1566 {
1567 NEED_SUCCESS (RetainContext, (value [i])); 2996 NEED_SUCCESS (RetainContext, (value [i]));
1568 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2997 PUSH_CLOBJ (stash_context, value [i]);
1569 }
1570 2998
1571#END:event 2999#END:event
1572 3000
1573void 3001void
1574profiling_info (OpenCL::Event this, cl_profiling_info name) 3002profiling_info (OpenCL::Event self, cl_profiling_info name)
1575 PPCODE: 3003 PPCODE:
1576 INFO (EventProfiling) 3004 INFO (EventProfiling)
1577 3005
1578#BEGIN:profiling 3006#BEGIN:profiling
1579 3007
1580void 3008void
1581profiling_command_queued (OpenCL::Event this) 3009profiling_command_queued (OpenCL::Event self)
1582 ALIAS: 3010 ALIAS:
1583 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED 3011 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
1584 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT 3012 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
1585 profiling_command_start = CL_PROFILING_COMMAND_START 3013 profiling_command_start = CL_PROFILING_COMMAND_START
1586 profiling_command_end = CL_PROFILING_COMMAND_END 3014 profiling_command_end = CL_PROFILING_COMMAND_END
1587 PPCODE: 3015 PPCODE:
1588 cl_ulong value [1]; 3016 cl_ulong value [1];
1589 NEED_SUCCESS (GetEventProfilingInfo, (this, ix, sizeof (value), value, 0)); 3017 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
1590 EXTEND (SP, 1); 3018 EXTEND (SP, 1);
1591 const int i = 0; 3019 const int i = 0;
1592 PUSHs (sv_2mortal (newSVuv (value [i]))); 3020 PUSHs (sv_2mortal (newSVuv (value [i])));
1593 3021
1594#END:profiling 3022#END:profiling
1595 3023
1596MODULE = OpenCL PACKAGE = OpenCL::UserEvent 3024MODULE = OpenCL PACKAGE = OpenCL::UserEvent
1597 3025
1598void 3026void
1599set_status (OpenCL::UserEvent this, cl_int execution_status) 3027set_status (OpenCL::UserEvent self, cl_int execution_status)
1600 CODE: 3028 CODE:
1601 clSetUserEventStatus (this, execution_status); 3029 clSetUserEventStatus (self, execution_status);
1602 3030
3031MODULE = OpenCL PACKAGE = OpenCL::Mapped
3032
3033void
3034DESTROY (SV *self)
3035 CODE:
3036 OpenCL__Mapped mapped = SvMAPPED (self);
3037
3038 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3039 mapped_detach (self, mapped);
3040
3041 clReleaseCommandQueue (mapped->queue);
3042 clReleaseEvent (mapped->event);
3043 Safefree (mapped);
3044
3045void
3046unmap (OpenCL::Mapped self, ...)
3047 CODE:
3048 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3049
3050bool
3051mapped (OpenCL::Mapped self)
3052 CODE:
3053 RETVAL = !!self->ptr;
3054 OUTPUT:
3055 RETVAL
3056
3057void
3058wait (OpenCL::Mapped self)
3059 PPCODE:
3060 if (self->event)
3061 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3062
3063void
3064event (OpenCL::Mapped self)
3065 PPCODE:
3066 if (!self->event)
3067 XSRETURN_UNDEF;
3068
3069 clRetainEvent (self->event);
3070 XPUSH_CLOBJ (stash_event, self->event);
3071
3072#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3073#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3074#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3075#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3076#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3077#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3078
3079IV
3080size (OpenCL::Mapped self)
3081 ALIAS:
3082 size = MAPPED_OFFSET_CB
3083 row_pitch = MAPPED_OFFSET_ROW_PITCH
3084 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3085 width = MAPPED_OFFSET_WIDTH
3086 height = MAPPED_OFFSET_HEIGHT
3087 depth = MAPPED_OFFSET_DEPTH
3088 CODE:
3089 RETVAL = *(size_t *)((char *)self + ix);
3090 OUTPUT:
3091 RETVAL
3092
3093IV
3094ptr (OpenCL::Mapped self)
3095 CODE:
3096 RETVAL = PTR2IV (self->ptr);
3097 OUTPUT:
3098 RETVAL
3099
3100void
3101set (OpenCL::Mapped self, size_t offset, SV *data)
3102 CODE:
3103 STRLEN len;
3104 const char *ptr = SvPVbyte (data, len);
3105
3106 if (offset + len > self->cb)
3107 croak ("OpenCL::Mapped::set out of bound condition detected");
3108
3109 memcpy (offset + (char *)self->ptr, ptr, len);
3110
3111void
3112get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3113 PPCODE:
3114 if (!SvOK (ST (1)))
3115 count = self->width - x;
3116
3117 if (x + count > self->width)
3118 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3119
3120 if (y >= self->height)
3121 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3122
3123 if (z >= self->depth)
3124 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3125
3126 size_t element = mapped_element_size (self);
3127
3128 count *= element;
3129 x *= element;
3130
3131 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3132 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3133
3134void
3135set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3136 PPCODE:
3137 STRLEN count;
3138 char *dataptr = SvPVbyte (data, count);
3139 size_t element = mapped_element_size (self);
3140
3141 x *= element;
3142
3143 if (x + count > self->width * element)
3144 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3145
3146 if (y >= self->height)
3147 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3148
3149 if (z >= self->depth)
3150 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3151
3152 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3153 memcpy (ptr, dataptr, count);
3154
3155MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3158
3159IV
3160element_size (OpenCL::Mapped self)
3161 CODE:
3162 RETVAL = mapped_element_size (self);
3163 OUTPUT:
3164 RETVAL
3165

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines