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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.5 by root, Wed Nov 16 00:35:30 2011 UTC vs.
Revision 1.76 by root, Sat May 5 14:49:52 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
10
11#ifdef I_DLFCN
12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
30#endif
31
32#ifdef __APPLE__
33 #define CLHDR(name) <OpenCL/name>
34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
5#include <CL/opencl.h> 38#include CLHDR(opencl.h)
6 39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
50#endif
51
52#ifndef CL_VERSION_1_2
53 #undef PREFER_1_1
54 #define PREFER_1_1 1
55#endif
56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
7typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
8typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
9typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
10typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
11typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
12typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
67typedef cl_mem OpenCL__BufferObj;
13typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
14typedef cl_mem OpenCL__Image2D;
15typedef cl_mem OpenCL__Image3D;
16typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
17typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
18typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
19typedef cl_mem OpenCL__Image2D_ornull;
20typedef cl_mem OpenCL__Image3D_ornull;
21typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
22typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
23typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
24typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
25typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
77
78typedef struct mapped * OpenCL__Mapped;
79
80static HV
81 *stash_platform,
82 *stash_device,
83 *stash_subdevice,
84 *stash_context,
85 *stash_queue,
86 *stash_program,
87 *stash_kernel,
88 *stash_sampler,
89 *stash_event,
90 *stash_userevent,
91 *stash_memory,
92 *stash_buffer,
93 *stash_bufferobj,
94 *stash_image,
95 *stash_image1d,
96 *stash_image1darray,
97 *stash_image1dbuffer,
98 *stash_image2d,
99 *stash_image2darray,
100 *stash_image3d,
101 *stash_mapped,
102 *stash_mappedbuffer,
103 *stash_mappedimage;
26 104
27/*****************************************************************************/ 105/*****************************************************************************/
28 106
107// name must include a leading underscore
108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
109// for perl. doh.
110static void *
111glsym (const char *name)
112{
113 void *fun = 0;
114
115 #if defined I_DLFCN && defined RTLD_DEFAULT
116 fun = dlsym (RTLD_DEFAULT, name + 1);
117 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
118
119 if (!fun)
120 {
121 static void *libgl;
122 static const char *glso[] = {
123 "libGL.so.1",
124 "libGL.so.3",
125 "libGL.so.4.0",
126 "libGL.so",
127 "/usr/lib/libGL.so",
128 "/usr/X11R6/lib/libGL.1.dylib"
129 };
130 int i;
131
132 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
133 {
134 libgl = dlopen (glso [i], RTLD_LAZY);
135 if (libgl)
136 break;
137 }
138
139 if (libgl)
140 {
141 fun = dlsym (libgl, name + 1);
142 if (!fun) fun = dlsym (libgl, name);
143 }
144 }
145 #endif
146
147 return fun;
148}
149
150/*****************************************************************************/
151
29/* up to two temporary buffers */ 152/* up to two temporary buffers */
30static void * 153static void * ecb_noinline
31tmpbuf (size_t size) 154tmpbuf (size_t size)
32{ 155{
156 enum { buffers = 4 };
33 static int idx; 157 static int idx;
34 static void *buf [2]; 158 static void *buf [buffers];
35 static size_t len [2]; 159 static size_t len [buffers];
36 160
37 idx ^= 1; 161 idx = (idx + 1) % buffers;
38 162
39 if (len [idx] < size) 163 if (len [idx] < size)
40 { 164 {
41 free (buf [idx]); 165 free (buf [idx]);
42 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
44 } 168 }
45 169
46 return buf [idx]; 170 return buf [idx];
47} 171}
48 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
49/*****************************************************************************/ 188/*****************************************************************************/
50 189
51typedef struct 190typedef struct
52{ 191{
53 IV iv; 192 IV iv;
54 const char *name; 193 const char *name;
55 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
56} ivstr; 195} ivstr;
57 196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
203
58static const char * 204static const char *
59iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
60{ 206{
61 int i; 207 int i;
62 static char strbuf [32]; 208 static char strbuf [32];
90 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)"); 236 return iv2str (err, errstr, sizeof (errstr) / sizeof (errstr [0]), "ERROR(%d)");
91} 237}
92 238
93/*****************************************************************************/ 239/*****************************************************************************/
94 240
95static cl_int last_error; 241static cl_int res;
96 242
97#define FAIL(name,err) \ 243#define FAIL(name) \
98 croak ("cl" # name ": %s", err2str (last_error = err)); 244 croak ("cl" # name ": %s", err2str (res));
99 245
100#define NEED_SUCCESS(name,args) \ 246#define NEED_SUCCESS(name,args) \
101 do { \ 247 do { \
102 cl_int res = cl ## name args; \ 248 res = cl ## name args; \
103 \ 249 \
104 if (res) \ 250 if (res) \
105 FAIL (name, res); \ 251 FAIL (name); \
106 } while (0) 252 } while (0)
107 253
254#define NEED_SUCCESS_ARG(retdecl, name, args) \
255 retdecl = cl ## name args; \
256 if (res) \
257 FAIL (name);
258
108/*****************************************************************************/ 259/*****************************************************************************/
109 260
110#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
111#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
112
113static void * 261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
114SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
115{ 274{
275 // sv_derived_from is quite slow :(
116 if (SvROK (sv) && sv_derived_from (sv, pkg)) 276 if (SvROK (sv) && sv_derived_from (sv, pkg))
117 return (void *)SvIV (SvRV (sv)); 277 return (void *)SvIV (SvRV (sv));
118 278
119 croak ("%s: %s is not of type %s", func, svname, pkg); 279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
120} 290}
121 291
122/*****************************************************************************/ 292/*****************************************************************************/
123 293
124static cl_event * 294static cl_context_properties * ecb_noinline
125event_list (SV **items, int count) 295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
126{ 296{
297 if (!sv || !SvOK (sv))
298 if (extra)
299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
300 else
301 return 0;
302
303 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
304 {
305 AV *av = (AV *)SvRV (sv);
306 int i, len = av_len (av) + 1;
307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
308 cl_context_properties *l = p;
309
310 if (len & 1)
311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
312
313 while (extracount--)
314 *l++ = *extra++;
315
316 for (i = 0; i < len; i += 2)
317 {
318 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
319 SV *p_sv = *av_fetch (av, i + 1, 0);
320 cl_context_properties v = SvIV (p_sv); // code below can override
321
322 switch (t)
323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
329 case CL_GLX_DISPLAY_KHR:
330 if (!SvOK (p_sv))
331 {
332 void *func = glsym ("_glXGetCurrentDisplay");
333 if (func)
334 v = (cl_context_properties)((void *(*)(void))func)();
335 }
336 break;
337
338 case CL_GL_CONTEXT_KHR:
339 if (!SvOK (p_sv))
340 {
341 void *func = glsym ("_glXGetCurrentContext");
342 if (func)
343 v = (cl_context_properties)((void *(*)(void))func)();
344 }
345 break;
346
347 default:
348 /* unknown property, treat as int */
349 break;
350 }
351
352 *l++ = t;
353 *l++ = v;
354 }
355
356 *l = 0;
357
358 return p;
359 }
360
361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
362}
363
364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
365// is a pointer (and only on well-behaved systems).
366static void * ecb_noinline
367object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
368{
369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
372
373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
378 {
379 list = tmpbuf (sizeof (*list) * count);
380 int i;
381 for (i = 0; i < count; ++i)
382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
383 }
384
385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
391}
392
393/*****************************************************************************/
394/* callback stuff */
395
396/* default context callback, log to stderr */
397static void CL_CALLBACK
398context_default_notify (const char *msg, const void *info, size_t cb, void *data)
399{
400 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
401}
402
403typedef struct
404{
405 int free_cb;
406 void (*push)(void *data1, void *data2, void *data3);
407} eq_vtbl;
408
409typedef struct eq_item
410{
411 struct eq_item *next;
412 eq_vtbl *vtbl;
413 SV *cb;
414 void *data1, *data2, *data3;
415} eq_item;
416
417static void (*eq_signal_func)(void *signal_arg, int value);
418static void *eq_signal_arg;
419static xmutex_t eq_lock = X_MUTEX_INIT;
420static eq_item *eq_head, *eq_tail;
421
422static void ecb_noinline
423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
424{
425 eq_item *item = malloc (sizeof (eq_item));
426
427 item->next = 0;
428 item->vtbl = vtbl;
429 item->cb = cb;
430 item->data1 = data1;
431 item->data2 = data2;
432 item->data3 = data3;
433
434 X_LOCK (eq_lock);
435
436 *(eq_head ? &eq_tail->next : &eq_head) = item;
437 eq_tail = item;
438
439 X_UNLOCK (eq_lock);
440
441 eq_signal_func (eq_signal_arg, 0);
442}
443
444static eq_item *
445eq_dec (void)
446{
447 eq_item *res;
448
449 X_LOCK (eq_lock);
450
451 res = eq_head;
452 if (res)
453 eq_head = res->next;
454
455 X_UNLOCK (eq_lock);
456
457 return res;
458}
459
460static void
461eq_poll (void)
462{
463 eq_item *item;
464
465 while ((item = eq_dec ()))
466 {
467 ENTER;
468 SAVETMPS;
469
470 dSP;
471 PUSHMARK (SP);
472 EXTEND (SP, 2);
473
474 if (item->vtbl->free_cb)
475 sv_2mortal (item->cb);
476
477 PUTBACK;
478 item->vtbl->push (item->data1, item->data2, item->data3);
479
480 SV *cb = item->cb;
481 free (item);
482
483 call_sv (cb, G_DISCARD | G_VOID);
484
485 FREETMPS;
486 LEAVE;
487 }
488}
489
490static void
491eq_poll_interrupt (pTHX_ void *c_arg, int value)
492{
493 eq_poll ();
494}
495
496/*****************************************************************************/
497/* context notify */
498
499static void ecb_noinline
500eq_context_push (void *data1, void *data2, void *data3)
501{
502 dSP;
503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
505 PUTBACK;
506
507 free (data1);
508 free (data2);
509}
510
511static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
512
513static void CL_CALLBACK
514eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
515{
516 void *pvt_copy = malloc (cb);
517 memcpy (pvt_copy, pvt, cb);
518 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
519}
520
521#define CONTEXT_NOTIFY_CALLBACK \
522 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
523 void *user_data = 0; \
524 \
525 if (SvOK (notify)) \
526 { \
527 pfn_notify = eq_context_notify; \
528 user_data = s_get_cv (notify); \
529 }
530
531static SV * ecb_noinline
532new_clobj_context (cl_context ctx, void *user_data)
533{
534 SV *sv = new_clobj (stash_context, (IV)ctx);
535
536 if (user_data)
537 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
538
539 return sv;
540}
541
542#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
543
544/*****************************************************************************/
545/* build/compile/link notify */
546
547static void
548eq_program_push (void *data1, void *data2, void *data3)
549{
550 dSP;
551 PUSH_CLOBJ (stash_program, data1);
552 PUTBACK;
553}
554
555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
556
557static void CL_CALLBACK
558eq_program_notify (cl_program program, void *user_data)
559{
560 clRetainProgram (program);
561
562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
580}
581
582struct build_args
583{
584 cl_program program;
585 char *options;
586 void *user_data;
587 cl_uint num_devices;
588};
589
590X_THREAD_PROC (build_program_thread)
591{
592 struct build_args *arg = thr_arg;
593
594 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
595
596 if (arg->user_data)
597 eq_program_notify (arg->program, arg->user_data);
598 else
599 clReleaseProgram (arg->program);
600
601 free (arg->options);
602 free (arg);
603
604 return 0;
605}
606
607static void
608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
609{
610 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
611
612 arg->program = program;
613 arg->options = strdup (options);
614 arg->user_data = user_data;
615 arg->num_devices = num_devices;
616 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
617
618 xthread_t id;
619 thread_create (&id, build_program_thread, arg);
620}
621
622/*****************************************************************************/
623/* event objects */
624
625static void
626eq_event_push (void *data1, void *data2, void *data3)
627{
628 dSP;
629 PUSH_CLOBJ (stash_event, data1);
630 PUSHs (sv_2mortal (newSViv ((IV)data2)));
631 PUTBACK;
632}
633
634static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
635
636static void CL_CALLBACK
637eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
638{
639 clRetainEvent (event);
640 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
641}
642
643/*****************************************************************************/
644/* utilities for XS code */
645
646static size_t
647img_row_pitch (cl_mem img)
648{
649 size_t res;
650 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
651 return res;
652}
653
654static cl_event * ecb_noinline
655event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
656{
657 cl_uint count = *rcount;
658
659 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
660 *rcount = count = 0;
661
662 if (!count && !extra)
663 return 0;
664
127 cl_event *list = tmpbuf (sizeof (cl_event) * count); 665 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
666 int i = 0;
128 667
129 while (count--) 668 while (count--)
669 if (SvOK (items [count]))
130 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 670 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
131 671
672 if (extra)
673 list [i++] = extra;
674
675 *rcount = i;
676
132 return list; 677 return i ? list : 0;
133} 678}
134 679
135#define EVENT_LIST(items,count) \ 680#define EVENT_LIST(skip) \
136 cl_uint event_list_count = (count); \ 681 cl_uint event_list_count = items - (skip); \
137 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 682 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
138 683
139#define INFO(class) \ 684#define INFO(class) \
140{ \ 685{ \
141 size_t size; \ 686 size_t size; \
142 SV *sv; \
143 \
144 NEED_SUCCESS (Get ## class ## Info, (this, name, 0, 0, &size)); \ 687 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
145 sv = sv_2mortal (newSV (size)); \ 688 SV *sv = sv_2mortal (newSV (size)); \
146 SvUPGRADE (sv, SVt_PV); \ 689 SvUPGRADE (sv, SVt_PV); \
147 SvPOK_only (sv); \ 690 SvPOK_only (sv); \
148 SvCUR_set (sv, size); \ 691 SvCUR_set (sv, size); \
149 NEED_SUCCESS (Get ## class ## Info, (this, name, size, SvPVX (sv), 0)); \ 692 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
150 XPUSHs (sv); \ 693 XPUSHs (sv); \
151} 694}
152 695
696/*****************************************************************************/
697/* mapped_xxx */
698
699static OpenCL__Mapped
700SvMAPPED (SV *sv)
701{
702 // no typechecking atm., keep your fingers crossed
703 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
704}
705
706struct mapped
707{
708 cl_command_queue queue;
709 cl_mem memobj;
710 void *ptr;
711 size_t cb;
712 cl_event event;
713 size_t row_pitch;
714 size_t slice_pitch;
715};
716
717static SV *
718mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
719{
720 SV *data = newSV (0);
721 SvUPGRADE (data, SVt_PVMG);
722
723 OpenCL__Mapped mapped;
724 New (0, mapped, 1, struct mapped);
725
726 clRetainCommandQueue (queue);
727
728 mapped->queue = queue;
729 mapped->memobj = memobj;
730 mapped->ptr = ptr;
731 mapped->cb = cb;
732 mapped->event = ev;
733 mapped->row_pitch = row_pitch;
734 mapped->slice_pitch = slice_pitch;
735
736 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
737
738 if (SvLEN (data))
739 Safefree (data);
740
741 SvPVX (data) = (char *)ptr;
742 SvCUR_set (data, cb);
743 SvLEN_set (data, 0);
744 SvPOK_only (data);
745
746 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
747
748 if (!(flags & CL_MAP_WRITE))
749 SvREADONLY_on (data);
750
751 return obj;
752}
753
754static void
755mapped_detach (SV *sv, OpenCL__Mapped mapped)
756{
757 SV *data = SvRV (sv);
758
759 // the next check checks both before AND after detach, where SvPVX should STILL be 0
760 if (SvPVX (data) != (char *)mapped->ptr)
761 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
762 else
763 {
764 SvREADONLY_off (data);
765 SvCUR_set (data, 0);
766 SvPVX (data) = 0;
767 SvOK_off (data);
768 }
769
770 mapped->ptr = 0;
771}
772
773static void
774mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
775{
776 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
777 cl_event ev;
778
779 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
780
781 clReleaseEvent (mapped->event);
782 mapped->event = ev;
783
784 mapped_detach (self, mapped);
785}
786
787/*****************************************************************************/
788
153MODULE = OpenCL PACKAGE = OpenCL 789MODULE = OpenCL PACKAGE = OpenCL
154 790
155PROTOTYPES: ENABLE 791PROTOTYPES: ENABLE
156 792
793void
794poll ()
795 CODE:
796 eq_poll ();
797
798void
799_eq_initialise (IV func, IV arg)
800 CODE:
801 eq_signal_func = (void (*)(void *, int))func;
802 eq_signal_arg = (void*)arg;
803
157BOOT: 804BOOT:
158{ 805{
159 HV *stash = gv_stashpv ("OpenCL", 1); 806 HV *stash = gv_stashpv ("OpenCL", 1);
807
160 static const ivstr *civ, const_iv[] = { 808 static const ivstr *civ, const_iv[] = {
161 { sizeof (cl_char ), "SIZEOF_CHAR" }, 809 { sizeof (cl_char ), "SIZEOF_CHAR" },
162 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 810 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
163 { sizeof (cl_short ), "SIZEOF_SHORT" }, 811 { sizeof (cl_short ), "SIZEOF_SHORT" },
164 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 812 { sizeof (cl_ushort), "SIZEOF_USHORT" },
165 { sizeof (cl_int ), "SIZEOF_INT" }, 813 { sizeof (cl_int ), "SIZEOF_INT" },
166 { sizeof (cl_uint ), "SIZEOF_UINT" }, 814 { sizeof (cl_uint ), "SIZEOF_UINT" },
167 { sizeof (cl_long ), "SIZEOF_LONG" }, 815 { sizeof (cl_long ), "SIZEOF_LONG" },
168 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 816 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
169 { sizeof (cl_half ), "SIZEOF_HALF" }, 817 { sizeof (cl_half ), "SIZEOF_HALF" },
170 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 818 { sizeof (cl_float ), "SIZEOF_FLOAT" },
171 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 819 { sizeof (cl_double), "SIZEOF_DOUBLE" },
820 { PREFER_1_1 , "PREFER_1_1" },
172#include "constiv.h" 821#include "constiv.h"
173 }; 822 };
823
174 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 824 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
175 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 825 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
826
827 static const nvstr *cnv, const_nv[] = {
828#include "constnv.h"
829 };
830
831 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
832 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
833
834 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
835
836 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
837 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
838 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
839 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
840 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
841 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
842 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
843 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
844 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
845 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
846 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
847 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
848 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
849 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
850 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
851 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
852 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
853 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
854 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
855 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
856 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
857 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
858 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
859
860 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
176} 861}
177 862
178cl_int 863cl_int
179errno () 864errno ()
180 CODE: 865 CODE:
181 errno = last_error; 866 RETVAL = res;
867 OUTPUT:
868 RETVAL
182 869
183const char * 870const char *
184err2str (cl_int err) 871err2str (cl_int err = res)
185 872
186const char * 873const char *
187enum2str (cl_uint value) 874enum2str (cl_uint value)
188 875
189void 876void
190platforms () 877platforms ()
191 PPCODE: 878 PPCODE:
192{
193 cl_platform_id *list; 879 cl_platform_id *list;
194 cl_uint count; 880 cl_uint count;
195 int i; 881 int i;
196 882
197 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count)); 883 NEED_SUCCESS (GetPlatformIDs, (0, 0, &count));
198 list = tmpbuf (sizeof (*list) * count); 884 list = tmpbuf (sizeof (*list) * count);
199 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 885 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
200 886
201 EXTEND (SP, count); 887 EXTEND (SP, count);
202 for (i = 0; i < count; ++i) 888 for (i = 0; i < count; ++i)
203 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 889 PUSH_CLOBJ (stash_platform, list [i]);
204}
205 890
206void 891void
207context_from_type_simple (cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 892context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
208 PPCODE: 893 PPCODE:
209{ 894 CONTEXT_NOTIFY_CALLBACK;
210 cl_int res; 895 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
211 cl_context ctx = clCreateContextFromType (0, type, 0, 0, &res); 896 XPUSH_CLOBJ_CONTEXT;
212 897
213 if (res) 898void
214 FAIL (CreateContextFromType, res); 899context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
900 PPCODE:
901 cl_uint device_count;
902 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
215 903
216 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 904 CONTEXT_NOTIFY_CALLBACK;
217} 905 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
906 XPUSH_CLOBJ_CONTEXT;
218 907
219void 908void
220wait_for_events (...) 909wait_for_events (...)
221 CODE: 910 CODE:
222{
223 EVENT_LIST (0, items); 911 EVENT_LIST (0);
224 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 912 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
225}
226 913
227PROTOTYPES: DISABLE 914PROTOTYPES: DISABLE
228 915
229MODULE = OpenCL PACKAGE = OpenCL::Platform 916MODULE = OpenCL PACKAGE = OpenCL::Platform
230 917
231void 918void
232info (OpenCL::Platform this, cl_platform_info name) 919info (OpenCL::Platform self, cl_platform_info name)
233 PPCODE: 920 PPCODE:
234 INFO (Platform) 921 INFO (Platform)
235 922
236void 923void
924unload_compiler (OpenCL::Platform self)
925 CODE:
926#if CL_VERSION_1_2
927 clUnloadPlatformCompiler (self);
928#endif
929
930#BEGIN:platform
931
932void
933profile (OpenCL::Platform self)
934 ALIAS:
935 profile = CL_PLATFORM_PROFILE
936 version = CL_PLATFORM_VERSION
937 name = CL_PLATFORM_NAME
938 vendor = CL_PLATFORM_VENDOR
939 extensions = CL_PLATFORM_EXTENSIONS
940 PPCODE:
941 size_t size;
942 NEED_SUCCESS (GetPlatformInfo, (self, ix, 0, 0, &size));
943 char *value = tmpbuf (size);
944 NEED_SUCCESS (GetPlatformInfo, (self, ix, size, value, 0));
945 EXTEND (SP, 1);
946 const int i = 0;
947 PUSHs (sv_2mortal (newSVpv (value, 0)));
948
949#END:platform
950
951void
237devices (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_ALL) 952devices (OpenCL::Platform self, cl_device_type type = CL_DEVICE_TYPE_ALL)
238 PPCODE: 953 PPCODE:
239{
240 cl_device_id *list; 954 cl_device_id *list;
241 cl_uint count; 955 cl_uint count;
242 int i; 956 int i;
243 957
244 NEED_SUCCESS (GetDeviceIDs, (this, type, 0, 0, &count)); 958 NEED_SUCCESS (GetDeviceIDs, (self, type, 0, 0, &count));
245 list = tmpbuf (sizeof (*list) * count); 959 list = tmpbuf (sizeof (*list) * count);
246 NEED_SUCCESS (GetDeviceIDs, (this, type, count, list, 0)); 960 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
247 961
248 EXTEND (SP, count); 962 EXTEND (SP, count);
249 for (i = 0; i < count; ++i) 963 for (i = 0; i < count; ++i)
250 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 964 PUSH_CLOBJ (stash_device, list [i]);
251}
252 965
253void 966void
254context_from_type_simple (OpenCL::Platform this, cl_device_type type = CL_DEVICE_TYPE_DEFAULT) 967context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
255 PPCODE: 968 PPCODE:
256{
257 cl_int res;
258 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)this, 0 }; 969 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
259 cl_context ctx = clCreateContextFromType (props, type, 0, 0, &res); 970 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
260 971
261 if (res) 972 cl_uint device_count;
262 FAIL (CreateContextFromType, res); 973 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
263 974
264 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 975 CONTEXT_NOTIFY_CALLBACK;
265} 976 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
977 XPUSH_CLOBJ_CONTEXT;
978
979void
980context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
981 PPCODE:
982 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
983 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
984
985 CONTEXT_NOTIFY_CALLBACK;
986 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
987 XPUSH_CLOBJ_CONTEXT;
266 988
267MODULE = OpenCL PACKAGE = OpenCL::Device 989MODULE = OpenCL PACKAGE = OpenCL::Device
268 990
269void 991void
270info (OpenCL::Device this, cl_device_info name) 992info (OpenCL::Device self, cl_device_info name)
271 PPCODE: 993 PPCODE:
272 INFO (Device) 994 INFO (Device)
273 995
996#if CL_VERSION_1_2
997
274void 998void
275context_simple (OpenCL::Device this) 999sub_devices (OpenCL::Device self, SV *properties)
1000 PPCODE:
1001 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
1002 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
1003
1004 properties = SvRV (properties);
1005
1006 cl_uint count = av_len ((AV *)properties) + 1;
1007 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1008
1009 int i;
1010 for (i = 0; i < count; ++i)
1011 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
1012
1013 props [count] = 0;
1014
1015 cl_uint num_devices;
1016 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1017 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1018 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
1019
1020 EXTEND (SP, num_devices);
1021 for (i = 0; i < count; ++i)
1022 PUSH_CLOBJ (stash_subdevice, list [i]);
1023
1024#endif
1025
1026#BEGIN:device
1027
1028void
1029type (OpenCL::Device self)
276 PPCODE: 1030 PPCODE:
277{ 1031 cl_device_type value [1];
278 cl_int res; 1032 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
279 cl_context ctx = clCreateContext (0, 1, &this, 0, 0, &res); 1033 EXTEND (SP, 1);
1034 const int i = 0;
1035 PUSHs (sv_2mortal (newSVuv (value [i])));
280 1036
281 if (res) 1037void
282 FAIL (CreateContext, res); 1038vendor_id (OpenCL::Device self)
1039 ALIAS:
1040 vendor_id = CL_DEVICE_VENDOR_ID
1041 max_compute_units = CL_DEVICE_MAX_COMPUTE_UNITS
1042 max_work_item_dimensions = CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS
1043 preferred_vector_width_char = CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR
1044 preferred_vector_width_short = CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT
1045 preferred_vector_width_int = CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT
1046 preferred_vector_width_long = CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG
1047 preferred_vector_width_float = CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT
1048 preferred_vector_width_double = CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE
1049 max_clock_frequency = CL_DEVICE_MAX_CLOCK_FREQUENCY
1050 max_read_image_args = CL_DEVICE_MAX_READ_IMAGE_ARGS
1051 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
1052 image_support = CL_DEVICE_IMAGE_SUPPORT
1053 max_samplers = CL_DEVICE_MAX_SAMPLERS
1054 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
1055 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1056 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
1057 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1058 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
1059 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
1060 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
1061 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
1062 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
1063 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
1064 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
1065 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
1066 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
1067 PPCODE:
1068 cl_uint value [1];
1069 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1070 EXTEND (SP, 1);
1071 const int i = 0;
1072 PUSHs (sv_2mortal (newSVuv (value [i])));
283 1073
284 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1074void
285} 1075max_work_group_size (OpenCL::Device self)
1076 ALIAS:
1077 max_work_group_size = CL_DEVICE_MAX_WORK_GROUP_SIZE
1078 image2d_max_width = CL_DEVICE_IMAGE2D_MAX_WIDTH
1079 image2d_max_height = CL_DEVICE_IMAGE2D_MAX_HEIGHT
1080 image3d_max_width = CL_DEVICE_IMAGE3D_MAX_WIDTH
1081 image3d_max_height = CL_DEVICE_IMAGE3D_MAX_HEIGHT
1082 image3d_max_depth = CL_DEVICE_IMAGE3D_MAX_DEPTH
1083 max_parameter_size = CL_DEVICE_MAX_PARAMETER_SIZE
1084 profiling_timer_resolution = CL_DEVICE_PROFILING_TIMER_RESOLUTION
1085 PPCODE:
1086 size_t value [1];
1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1088 EXTEND (SP, 1);
1089 const int i = 0;
1090 PUSHs (sv_2mortal (newSVuv (value [i])));
1091
1092void
1093max_work_item_sizes (OpenCL::Device self)
1094 PPCODE:
1095 size_t size;
1096 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, 0, 0, &size));
1097 size_t *value = tmpbuf (size);
1098 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_MAX_WORK_ITEM_SIZES, size, value, 0));
1099 int i, n = size / sizeof (*value);
1100 EXTEND (SP, n);
1101 for (i = 0; i < n; ++i)
1102 PUSHs (sv_2mortal (newSVuv (value [i])));
1103
1104void
1105address_bits (OpenCL::Device self)
1106 PPCODE:
1107 cl_bitfield value [1];
1108 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
1109 EXTEND (SP, 1);
1110 const int i = 0;
1111 PUSHs (sv_2mortal (newSVuv (value [i])));
1112
1113void
1114max_mem_alloc_size (OpenCL::Device self)
1115 ALIAS:
1116 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1117 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1118 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1119 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1120 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1121 PPCODE:
1122 cl_ulong value [1];
1123 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1124 EXTEND (SP, 1);
1125 const int i = 0;
1126 PUSHs (sv_2mortal (newSVuv (value [i])));
1127
1128void
1129single_fp_config (OpenCL::Device self)
1130 ALIAS:
1131 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1132 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1133 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
1134 PPCODE:
1135 cl_device_fp_config value [1];
1136 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1137 EXTEND (SP, 1);
1138 const int i = 0;
1139 PUSHs (sv_2mortal (newSVuv (value [i])));
1140
1141void
1142global_mem_cache_type (OpenCL::Device self)
1143 PPCODE:
1144 cl_device_mem_cache_type value [1];
1145 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
1146 EXTEND (SP, 1);
1147 const int i = 0;
1148 PUSHs (sv_2mortal (newSVuv (value [i])));
1149
1150void
1151local_mem_type (OpenCL::Device self)
1152 PPCODE:
1153 cl_device_local_mem_type value [1];
1154 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
1155 EXTEND (SP, 1);
1156 const int i = 0;
1157 PUSHs (sv_2mortal (newSVuv (value [i])));
1158
1159void
1160error_correction_support (OpenCL::Device self)
1161 ALIAS:
1162 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1163 endian_little = CL_DEVICE_ENDIAN_LITTLE
1164 available = CL_DEVICE_AVAILABLE
1165 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1166 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1167 PPCODE:
1168 cl_bool value [1];
1169 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1170 EXTEND (SP, 1);
1171 const int i = 0;
1172 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1173
1174void
1175execution_capabilities (OpenCL::Device self)
1176 PPCODE:
1177 cl_device_exec_capabilities value [1];
1178 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1179 EXTEND (SP, 1);
1180 const int i = 0;
1181 PUSHs (sv_2mortal (newSVuv (value [i])));
1182
1183void
1184properties (OpenCL::Device self)
1185 PPCODE:
1186 cl_command_queue_properties value [1];
1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1188 EXTEND (SP, 1);
1189 const int i = 0;
1190 PUSHs (sv_2mortal (newSVuv (value [i])));
1191
1192void
1193platform (OpenCL::Device self)
1194 PPCODE:
1195 cl_platform_id value [1];
1196 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1197 EXTEND (SP, 1);
1198 const int i = 0;
1199 {
1200 PUSH_CLOBJ (stash_platform, value [i]);
1201 }
1202
1203void
1204name (OpenCL::Device self)
1205 ALIAS:
1206 name = CL_DEVICE_NAME
1207 vendor = CL_DEVICE_VENDOR
1208 driver_version = CL_DRIVER_VERSION
1209 profile = CL_DEVICE_PROFILE
1210 version = CL_DEVICE_VERSION
1211 extensions = CL_DEVICE_EXTENSIONS
1212 PPCODE:
1213 size_t size;
1214 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1215 char *value = tmpbuf (size);
1216 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1217 EXTEND (SP, 1);
1218 const int i = 0;
1219 PUSHs (sv_2mortal (newSVpv (value, 0)));
1220
1221void
1222parent_device_ext (OpenCL::Device self)
1223 PPCODE:
1224 cl_device_id value [1];
1225 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1226 EXTEND (SP, 1);
1227 const int i = 0;
1228 {
1229 PUSH_CLOBJ (stash_device, value [i]);
1230 }
1231
1232void
1233partition_types_ext (OpenCL::Device self)
1234 ALIAS:
1235 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1236 affinity_domains_ext = CL_DEVICE_AFFINITY_DOMAINS_EXT
1237 partition_style_ext = CL_DEVICE_PARTITION_STYLE_EXT
1238 PPCODE:
1239 size_t size;
1240 NEED_SUCCESS (GetDeviceInfo, (self, ix, 0, 0, &size));
1241 cl_device_partition_property_ext *value = tmpbuf (size);
1242 NEED_SUCCESS (GetDeviceInfo, (self, ix, size, value, 0));
1243 int i, n = size / sizeof (*value);
1244 EXTEND (SP, n);
1245 for (i = 0; i < n; ++i)
1246 PUSHs (sv_2mortal (newSVuv (value [i])));
1247
1248#END:device
1249
1250MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1251
1252#if CL_VERSION_1_2
1253
1254void
1255DESTROY (OpenCL::SubDevice self)
1256 CODE:
1257 clReleaseDevice (self);
1258
1259#endif
286 1260
287MODULE = OpenCL PACKAGE = OpenCL::Context 1261MODULE = OpenCL PACKAGE = OpenCL::Context
288 1262
289void 1263void
290DESTROY (OpenCL::Context context) 1264DESTROY (OpenCL::Context self)
291 CODE: 1265 CODE:
292 clReleaseContext (context); 1266 clReleaseContext (self);
293 1267
294void 1268void
295info (OpenCL::Context this, cl_context_info name) 1269info (OpenCL::Context self, cl_context_info name)
296 PPCODE: 1270 PPCODE:
297 INFO (Context) 1271 INFO (Context)
298 1272
299void 1273void
300command_queue_simple (OpenCL::Context this, OpenCL::Device device) 1274queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
301 PPCODE: 1275 PPCODE:
302{ 1276 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
303 cl_int res; 1277 XPUSH_CLOBJ (stash_queue, queue);
304 cl_command_queue queue = clCreateCommandQueue (this, device, 0, &res);
305 1278
306 if (res)
307 FAIL (CreateCommandQueue, res);
308
309 XPUSH_NEW_OBJ ("OpenCL::Queue", queue);
310}
311
312void 1279void
313user_event (OpenCL::Context this) 1280user_event (OpenCL::Context self)
314 PPCODE: 1281 PPCODE:
315{ 1282 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
316 cl_int res; 1283 XPUSH_CLOBJ (stash_userevent, ev);
317 cl_event ev = clCreateUserEvent (this, &res);
318 1284
319 if (res)
320 FAIL (CreateUserevent, res);
321
322 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev);
323}
324
325void 1285void
326buffer (OpenCL::Context this, cl_mem_flags flags, size_t len) 1286buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
327 PPCODE: 1287 PPCODE:
328{
329 cl_int res;
330 cl_mem mem;
331
332 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1288 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
333 croak ("clCreateBuffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1289 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
334 1290
335 mem = clCreateBuffer (this, flags, len, 0, &res); 1291 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1292 XPUSH_CLOBJ (stash_bufferobj, mem);
336 1293
337 if (res)
338 FAIL (CreateBuffer, res);
339
340 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem);
341}
342
343void 1294void
344buffer_sv (OpenCL::Context this, cl_mem_flags flags, SV *data) 1295buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
345 PPCODE: 1296 PPCODE:
346{
347 STRLEN len; 1297 STRLEN len;
348 char *ptr = SvPVbyte (data, len); 1298 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
349 cl_int res;
350 cl_mem mem;
351
352 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1299 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
353 croak ("clCreateBuffer: have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1300 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
354 1301 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
355 mem = clCreateBuffer (this, flags, len, ptr, &res); 1302 XPUSH_CLOBJ (stash_bufferobj, mem);
356 1303
357 if (res) 1304#if CL_VERSION_1_2
358 FAIL (CreateBuffer, res);
359 1305
360 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem);
361}
362
363void 1306void
364image2d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, SV *data) 1307image (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)
365 PPCODE: 1308 PPCODE:
366{
367 STRLEN len; 1309 STRLEN len;
368 char *ptr = SvPVbyte (data, len); 1310 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
369 const cl_image_format format = { channel_order, channel_type }; 1311 const cl_image_format format = { channel_order, channel_type };
370 cl_int res; 1312 const cl_image_desc desc = {
371 cl_mem mem = clCreateImage2D (this, flags, &format, width, height, len / height, ptr, &res); 1313 type,
1314 width, height, depth,
1315 array_size, row_pitch, slice_pitch,
1316 num_mip_level, num_samples,
1317 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1318 };
1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1320 HV *stash = stash_image;
1321 switch (type)
1322 {
1323 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1324 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1325 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1326 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1327 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1328 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1329 }
1330 XPUSH_CLOBJ (stash, mem);
372 1331
373 if (res) 1332#endif
374 FAIL (CreateImage2D, res);
375 1333
376 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
377}
378
379void 1334void
380image3d (OpenCL::Context this, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t slice_pitch, SV *data) 1335image2d (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)
381 PPCODE: 1336 PPCODE:
382{
383 STRLEN len; 1337 STRLEN len;
384 char *ptr = SvPVbyte (data, len); 1338 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
385 const cl_image_format format = { channel_order, channel_type }; 1339 const cl_image_format format = { channel_order, channel_type };
386 cl_int res; 1340#if PREFER_1_1
387 cl_mem mem = clCreateImage3D (this, flags, &format, width, height, 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
388 depth, len / (height * slice_pitch), slice_pitch, ptr, &res); 1342#else
1343 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1344 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1345#endif
1346 XPUSH_CLOBJ (stash_image2d, mem);
389 1347
390 if (res)
391 FAIL (CreateImage3D, res);
392
393 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem);
394}
395
396void 1348void
1349image3d (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)
1350 PPCODE:
1351 STRLEN len;
1352 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1353 const cl_image_format format = { channel_order, channel_type };
1354#if PREFER_1_1
1355 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1356#else
1357 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1358 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1359#endif
1360 XPUSH_CLOBJ (stash_image3d, mem);
1361
1362#if cl_apple_gl_sharing || cl_khr_gl_sharing
1363
1364void
1365gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1366 PPCODE:
1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1368 XPUSH_CLOBJ (stash_bufferobj, mem);
1369
1370void
1371gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1372 PPCODE:
1373 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1374 XPUSH_CLOBJ (stash_image2d, mem);
1375
1376#if CL_VERSION_1_2
1377
1378void
1379gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1380 ALIAS:
1381 PPCODE:
1382 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1383 cl_gl_object_type type;
1384 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1385 HV *stash = stash_memory;
1386 switch (type)
1387 {
1388 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1389 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1390 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1391 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1392 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1393 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1394 }
1395 XPUSH_CLOBJ (stash, mem);
1396
1397#endif
1398
1399void
1400gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1401 PPCODE:
1402#if PREFER_1_1
1403 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1404#else
1405 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1406#endif
1407 XPUSH_CLOBJ (stash_image2d, mem);
1408
1409void
1410gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1411 PPCODE:
1412#if PREFER_1_1
1413 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1414#else
1415 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1416#endif
1417 XPUSH_CLOBJ (stash_image3d, mem);
1418
1419#endif
1420
1421void
397supported_image_formats (OpenCL::Context this, cl_mem_flags flags, cl_mem_object_type image_type) 1422supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
398 PPCODE: 1423 PPCODE:
399{ 1424{
400 cl_uint count; 1425 cl_uint count;
401 cl_image_format *list; 1426 cl_image_format *list;
402 int i; 1427 int i;
403 1428
404 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, 0, 0, &count)); 1429 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
405 Newx (list, count, cl_image_format); 1430 Newx (list, count, cl_image_format);
406 NEED_SUCCESS (GetSupportedImageFormats, (this, flags, image_type, count, list, 0)); 1431 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
407 1432
408 EXTEND (SP, count); 1433 EXTEND (SP, count);
409 for (i = 0; i < count; ++i) 1434 for (i = 0; i < count; ++i)
410 { 1435 {
411 AV *av = newAV (); 1436 AV *av = newAV ();
414 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1439 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
415 } 1440 }
416} 1441}
417 1442
418void 1443void
419sampler (OpenCL::Context this, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1444sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
420 PPCODE: 1445 PPCODE:
421{
422 cl_int res;
423 cl_sampler sampler = clCreateSampler (this, normalized_coords, addressing_mode, filter_mode, &res); 1446 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
424
425 if (res)
426 FAIL (CreateSampler, res);
427
428 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1447 XPUSH_CLOBJ (stash_sampler, sampler);
429}
430 1448
431void 1449void
432program_with_source (OpenCL::Context this, SV *program) 1450program_with_source (OpenCL::Context self, SV *program)
433 PPCODE: 1451 PPCODE:
434{
435 STRLEN len; 1452 STRLEN len;
436 size_t len2; 1453 size_t len2;
437 const char *ptr = SvPVbyte (program, len); 1454 const char *ptr = SvPVbyte (program, len);
438 cl_int res;
439 cl_program prog;
440 1455
441 len2 = len; 1456 len2 = len;
442 prog = clCreateProgramWithSource (this, 1, &ptr, &len2, &res); 1457 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
443
444 if (res)
445 FAIL (CreateProgramWithSource, res);
446
447 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1458 XPUSH_CLOBJ (stash_program, prog);
448} 1459
1460void
1461program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1462 PPCODE:
1463 cl_uint device_count;
1464 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1465
1466 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1467 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1468
1469 binaries = SvRV (binaries);
1470
1471 if (device_count != av_len ((AV *)binaries) + 1)
1472 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1473
1474 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1475 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1476 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1477
1478 int i;
1479 for (i = 0; i < device_count; ++i)
1480 {
1481 STRLEN len;
1482 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1483 length_list [i] = len;
1484 }
1485
1486 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1487 length_list, binary_list,
1488 GIMME_V == G_ARRAY ? status_list : 0, &res));
1489
1490 EXTEND (SP, 2);
1491 PUSH_CLOBJ (stash_program, prog);
1492
1493 if (GIMME_V == G_ARRAY)
1494 {
1495 AV *av = newAV ();
1496 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1497
1498 for (i = device_count; i--; )
1499 av_store (av, i, newSViv (status_list [i]));
1500 }
1501
1502#if CL_VERSION_1_2
1503
1504void
1505program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1506 PPCODE:
1507 cl_uint device_count;
1508 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1509
1510 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514void
1515link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1516 CODE:
1517 cl_uint device_count = 0;
1518 cl_device_id *device_list = 0;
1519
1520 if (SvOK (devices))
1521 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1522
1523 cl_uint program_count;
1524 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1525
1526 void *user_data;
1527 program_callback pfn_notify = make_program_callback (notify, &user_data);
1528
1529 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1530 program_count, program_list, pfn_notify, user_data, &res));
1531
1532 XPUSH_CLOBJ (stash_program, prog);
1533
1534#endif
1535
1536#BEGIN:context
1537
1538void
1539reference_count (OpenCL::Context self)
1540 ALIAS:
1541 reference_count = CL_CONTEXT_REFERENCE_COUNT
1542 num_devices = CL_CONTEXT_NUM_DEVICES
1543 PPCODE:
1544 cl_uint value [1];
1545 NEED_SUCCESS (GetContextInfo, (self, ix, sizeof (value), value, 0));
1546 EXTEND (SP, 1);
1547 const int i = 0;
1548 PUSHs (sv_2mortal (newSVuv (value [i])));
1549
1550void
1551devices (OpenCL::Context self)
1552 PPCODE:
1553 size_t size;
1554 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, 0, 0, &size));
1555 cl_device_id *value = tmpbuf (size);
1556 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1557 int i, n = size / sizeof (*value);
1558 EXTEND (SP, n);
1559 for (i = 0; i < n; ++i)
1560 {
1561 PUSH_CLOBJ (stash_device, value [i]);
1562 }
1563
1564void
1565properties (OpenCL::Context self)
1566 PPCODE:
1567 size_t size;
1568 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, 0, 0, &size));
1569 cl_context_properties *value = tmpbuf (size);
1570 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_PROPERTIES, size, value, 0));
1571 int i, n = size / sizeof (*value);
1572 EXTEND (SP, n);
1573 for (i = 0; i < n; ++i)
1574 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
1575
1576#END:context
449 1577
450MODULE = OpenCL PACKAGE = OpenCL::Queue 1578MODULE = OpenCL PACKAGE = OpenCL::Queue
451 1579
452void 1580void
453DESTROY (OpenCL::Queue this) 1581DESTROY (OpenCL::Queue self)
454 CODE: 1582 CODE:
455 clReleaseCommandQueue (this); 1583 clReleaseCommandQueue (self);
456 1584
457void 1585void
458info (OpenCL::Queue this, cl_command_queue_info name)
459 PPCODE:
460 INFO (CommandQueue)
461
462void
463enqueue_read_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1586read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1587 ALIAS:
1588 enqueue_read_buffer = 0
464 PPCODE: 1589 PPCODE:
465{
466 cl_event ev = 0;
467 EVENT_LIST (6, items - 6); 1590 EVENT_LIST (6);
468 1591
469 SvUPGRADE (data, SVt_PV); 1592 SvUPGRADE (data, SVt_PV);
470 SvGROW (data, len); 1593 SvGROW (data, len);
471 SvPOK_only (data); 1594 SvPOK_only (data);
472 SvCUR_set (data, len); 1595 SvCUR_set (data, len);
1596
1597 cl_event ev = 0;
473 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1598 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
474 1599
475 if (ev) 1600 if (ev)
476 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1601 XPUSH_CLOBJ (stash_event, ev);
477}
478 1602
479void 1603void
480enqueue_write_buffer (OpenCL::Queue this, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1604write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1605 ALIAS:
1606 enqueue_write_buffer = 0
481 PPCODE: 1607 PPCODE:
482{ 1608 EVENT_LIST (5);
483 cl_event ev = 0; 1609
484 STRLEN len; 1610 STRLEN len;
485 char *ptr = SvPVbyte (data, len); 1611 char *ptr = SvPVbyte (data, len);
486 EVENT_LIST (5, items - 5);
487 1612
1613 cl_event ev = 0;
488 NEED_SUCCESS (EnqueueReadBuffer, (this, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
489 1615
490 if (ev) 1616 if (ev)
491 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1617 XPUSH_CLOBJ (stash_event, ev);
492}
493 1618
1619#if CL_VERSION_1_2
1620
494void 1621void
495enqueue_copy_buffer (OpenCL::Queue this, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1622fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1623 ALIAS:
1624 enqueue_fill_buffer = 0
496 PPCODE: 1625 PPCODE:
497{ 1626 EVENT_LIST (5);
1627
1628 STRLEN len;
1629 char *ptr = SvPVbyte (data, len);
1630
498 cl_event ev = 0; 1631 cl_event ev = 0;
499 EVENT_LIST (6, items - 6);
500
501 NEED_SUCCESS (EnqueueCopyBuffer, (this, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1632 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
502 1633
503 if (ev) 1634 if (ev)
504 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1635 XPUSH_CLOBJ (stash_event, ev);
505}
506 1636
507 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueReadBufferRect.html */
508 /*TODO http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clEnqueueWriteBufferRect.html */
509
510void 1637void
511enqueue_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, ...) 1638fill_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, ...)
1639 ALIAS:
1640 enqueue_fill_image = 0
512 PPCODE: 1641 PPCODE:
513{ 1642 EVENT_LIST (12);
1643
1644 const size_t origin [3] = { x, y, z };
1645 const size_t region [3] = { width, height, depth };
1646
1647 const cl_float c_f [4] = { r, g, b, a };
1648 const cl_uint c_u [4] = { r, g, b, a };
1649 const cl_int c_s [4] = { r, g, b, a };
1650 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1651 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1652 cl_image_format format;
1653 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1654 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1655 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1656 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1657
514 cl_event ev = 0; 1658 cl_event ev = 0;
1659 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1660 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1661
1662 if (ev)
1663 XPUSH_CLOBJ (stash_event, ev);
1664
1665#endif
1666
1667void
1668copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1669 ALIAS:
1670 enqueue_copy_buffer = 0
1671 PPCODE:
1672 EVENT_LIST (6);
1673
1674 cl_event ev = 0;
1675 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1676
1677 if (ev)
1678 XPUSH_CLOBJ (stash_event, ev);
1679
1680void
1681read_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, ...)
1682 ALIAS:
1683 enqueue_read_buffer_rect = 0
1684 PPCODE:
1685 EVENT_LIST (17);
1686
1687 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
515 const size_t src_origin[3] = { src_x, src_y, src_z }; 1688 const size_t host_origin[3] = { host_x, host_y, host_z };
516 const size_t region[3] = { width, height, depth }; 1689 const size_t region[3] = { width, height, depth };
1690
1691 if (!buf_row_pitch)
1692 buf_row_pitch = region [0];
1693
1694 if (!buf_slice_pitch)
1695 buf_slice_pitch = region [1] * buf_row_pitch;
1696
1697 if (!host_row_pitch)
1698 host_row_pitch = region [0];
1699
1700 if (!host_slice_pitch)
1701 host_slice_pitch = region [1] * host_row_pitch;
1702
517 size_t len = row_pitch * slice_pitch * depth; 1703 size_t len = host_row_pitch * host_slice_pitch * region [2];
518 EVENT_LIST (11, items - 11);
519 1704
520 SvUPGRADE (data, SVt_PV); 1705 SvUPGRADE (data, SVt_PV);
521 SvGROW (data, len); 1706 SvGROW (data, len);
522 SvPOK_only (data); 1707 SvPOK_only (data);
523 SvCUR_set (data, len); 1708 SvCUR_set (data, len);
524 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)); 1709
1710 cl_event ev = 0;
1711 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));
525 1712
526 if (ev) 1713 if (ev)
527 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1714 XPUSH_CLOBJ (stash_event, ev);
528}
529 1715
530void 1716void
531enqueue_write_image (OpenCL::Queue this, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, SV *data, ...) 1717write_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, ...)
1718 ALIAS:
1719 enqueue_write_buffer_rect = 0
532 PPCODE: 1720 PPCODE:
533{ 1721 EVENT_LIST (17);
1722
1723 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1724 const size_t host_origin[3] = { host_x, host_y, host_z };
1725 const size_t region[3] = { width, height, depth };
1726 STRLEN len;
1727 char *ptr = SvPVbyte (data, len);
1728
1729 if (!buf_row_pitch)
1730 buf_row_pitch = region [0];
1731
1732 if (!buf_slice_pitch)
1733 buf_slice_pitch = region [1] * buf_row_pitch;
1734
1735 if (!host_row_pitch)
1736 host_row_pitch = region [0];
1737
1738 if (!host_slice_pitch)
1739 host_slice_pitch = region [1] * host_row_pitch;
1740
1741 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1742
1743 if (len < min_len)
1744 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1745
534 cl_event ev = 0; 1746 cl_event ev = 0;
1747 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));
1748
1749 if (ev)
1750 XPUSH_CLOBJ (stash_event, ev);
1751
1752void
1753copy_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, ...)
1754 ALIAS:
1755 enqueue_copy_buffer_rect = 0
1756 PPCODE:
1757 EVENT_LIST (16);
1758
1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1760 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1761 const size_t region[3] = { width, height, depth };
1762
1763 cl_event ev = 0;
1764 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));
1765
1766 if (ev)
1767 XPUSH_CLOBJ (stash_event, ev);
1768
1769void
1770read_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, ...)
1771 ALIAS:
1772 enqueue_read_image = 0
1773 PPCODE:
1774 EVENT_LIST (12);
1775
1776 const size_t src_origin[3] = { src_x, src_y, src_z };
1777 const size_t region[3] = { width, height, depth };
1778
1779 if (!row_pitch)
1780 row_pitch = img_row_pitch (src);
1781
1782 if (depth > 1 && !slice_pitch)
1783 slice_pitch = row_pitch * height;
1784
1785 size_t len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1786
1787 SvUPGRADE (data, SVt_PV);
1788 SvGROW (data, len);
1789 SvPOK_only (data);
1790 SvCUR_set (data, len);
1791
1792 cl_event ev = 0;
1793 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));
1794
1795 if (ev)
1796 XPUSH_CLOBJ (stash_event, ev);
1797
1798void
1799write_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, ...)
1800 ALIAS:
1801 enqueue_write_image = 0
1802 PPCODE:
1803 EVENT_LIST (12);
1804
535 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1805 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
536 const size_t region[3] = { width, height, depth }; 1806 const size_t region[3] = { width, height, depth };
537 STRLEN len; 1807 STRLEN len;
538 char *ptr = SvPVbyte (data, len); 1808 char *ptr = SvPVbyte (data, len);
1809
1810 if (!row_pitch)
1811 row_pitch = img_row_pitch (dst);
1812
1813 if (depth > 1 && !slice_pitch)
539 size_t slice_pitch = len / (row_pitch * height); 1814 slice_pitch = row_pitch * height;
540 EVENT_LIST (11, items - 11);
541 1815
1816 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1817
1818 if (len < min_len)
1819 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1820
1821 cl_event ev = 0;
542 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)); 1822 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));
543 1823
544 if (ev) 1824 if (ev)
545 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
546}
547 1826
548void 1827void
549enqueue_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, ...) 1828copy_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, ...)
1829 ALIAS:
1830 enqueue_copy_image = 0
550 PPCODE: 1831 PPCODE:
551{ 1832 EVENT_LIST (12);
552 cl_event ev = 0; 1833
553 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
554 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1835 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
555 const size_t region[3] = { width, height, depth }; 1836 const size_t region[3] = { width, height, depth };
556 EVENT_LIST (16, items - 16);
557 1837
1838 cl_event ev = 0;
558 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)); 1839 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
559 1840
560 if (ev) 1841 if (ev)
561 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1842 XPUSH_CLOBJ (stash_event, ev);
562}
563 1843
564void 1844void
565enqueue_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, ...) 1845copy_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, ...)
1846 ALIAS:
1847 enqueue_copy_image_to_buffer = 0
566 PPCODE: 1848 PPCODE:
567{ 1849 EVENT_LIST (10);
1850
1851 const size_t src_origin[3] = { src_x, src_y, src_z };
1852 const size_t region [3] = { width, height, depth };
1853
568 cl_event ev = 0; 1854 cl_event ev = 0;
1855 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1856
1857 if (ev)
1858 XPUSH_CLOBJ (stash_event, ev);
1859
1860void
1861copy_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, ...)
1862 ALIAS:
1863 enqueue_copy_buffer_to_image = 0
1864 PPCODE:
1865 EVENT_LIST (10);
1866
569 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1867 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1868 const size_t region [3] = { width, height, depth };
1869
1870 cl_event ev = 0;
1871 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1872
1873 if (ev)
1874 XPUSH_CLOBJ (stash_event, ev);
1875
1876void
1877map_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, ...)
1878 ALIAS:
1879 enqueue_map_buffer = 0
1880 PPCODE:
1881 EVENT_LIST (6);
1882
1883 size_t cb = SvIV (cb_);
1884
1885 if (!SvOK (cb_))
1886 {
1887 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1888 cb -= offset;
1889 }
1890
1891 cl_event ev;
1892 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1893 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1894
1895void
1896map_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, ...)
1897 ALIAS:
1898 enqueue_map_image = 0
1899 PPCODE:
1900 size_t width = SvIV (width_);
1901 if (!SvOK (width_))
1902 {
1903 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1904 width -= x;
1905 }
1906
1907 size_t height = SvIV (width_);
1908 if (!SvOK (height_))
1909 {
1910 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1911 height -= y;
1912 }
1913
1914 size_t depth = SvIV (width_);
1915 if (!SvOK (depth_))
1916 {
1917 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1918 depth -= z;
1919
1920 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1921 if (!depth)
1922 depth = 1;
1923 }
1924
1925 const size_t origin[3] = { x, y, z };
570 const size_t region[3] = { width, height, depth }; 1926 const size_t region[3] = { width, height, depth };
1927 size_t row_pitch, slice_pitch;
571 EVENT_LIST (10, items - 10); 1928 EVENT_LIST (10);
572 1929
573 NEED_SUCCESS (EnqueueCopyBufferToImage, (this, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1930 cl_event ev;
1931 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));
1932
1933 size_t cb = slice_pitch ? slice_pitch * region [2]
1934 : row_pitch ? row_pitch * region [1]
1935 : region [0];
1936
1937 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1938
1939void
1940unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1941 PPCODE:
1942 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1943 if (GIMME_V != G_VOID)
1944 {
1945 clRetainEvent (mapped->event);
1946 XPUSH_CLOBJ (stash_event, mapped->event);
1947 }
1948
1949void
1950task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1951 ALIAS:
1952 enqueue_task = 0
1953 PPCODE:
1954 EVENT_LIST (2);
1955
1956 cl_event ev = 0;
1957 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
574 1958
575 if (ev) 1959 if (ev)
576 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1960 XPUSH_CLOBJ (stash_event, ev);
577}
578 1961
579void 1962void
580enqueue_copy_image (OpenCL::Queue this, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
581 PPCODE:
582{
583 cl_event ev = 0;
584 const size_t src_origin[3] = { src_x, src_y, src_z };
585 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
586 const size_t region[3] = { width, height, depth };
587 EVENT_LIST (12, items - 12);
588
589 NEED_SUCCESS (EnqueueCopyImage, (this, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
590
591 if (ev)
592 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
593}
594
595void
596enqueue_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, ...)
597 PPCODE:
598{
599 cl_event ev = 0;
600 const size_t src_origin[3] = { src_x, src_y, src_z };
601 const size_t region[3] = { width, height, depth };
602 EVENT_LIST (10, items - 10);
603
604 NEED_SUCCESS (EnqueueCopyImageToBuffer, (this, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
605
606 if (ev)
607 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
608}
609
610void
611enqueue_task (OpenCL::Queue this, OpenCL::Kernel kernel, ...)
612 PPCODE:
613{
614 cl_event ev = 0;
615 EVENT_LIST (2, items - 2);
616
617 NEED_SUCCESS (EnqueueTask, (this, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
618
619 if (ev)
620 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
621}
622
623void
624enqueue_nd_range_kernel (OpenCL::Queue this, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1963nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1964 ALIAS:
1965 enqueue_nd_range_kernel = 0
625 PPCODE: 1966 PPCODE:
626{ 1967 EVENT_LIST (5);
627 cl_event ev = 0; 1968
628 size_t *gwo = 0, *gws, *lws = 0; 1969 size_t *gwo = 0, *gws, *lws = 0;
629 int gws_len; 1970 int gws_len;
630 size_t *lists; 1971 size_t *lists;
631 int i; 1972 int i;
632 EVENT_LIST (5, items - 5);
633 1973
634 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1974 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
635 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
636 1976
637 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1977 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
638 1978
639 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1979 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
640 1980
641 gws = lists + gws_len * 0; 1981 gws = lists + gws_len * 0;
642 for (i = 0; i < gws_len; ++i) 1982 for (i = 0; i < gws_len; ++i)
1983 {
643 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1984 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1985 // at least nvidia crashes for 0-sized work group sizes, work around
1986 if (!gws [i])
1987 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1988 }
644 1989
645 if (SvOK (global_work_offset)) 1990 if (SvOK (global_work_offset))
646 { 1991 {
647 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1992 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
648 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1993 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
655 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 2000 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
656 } 2001 }
657 2002
658 if (SvOK (local_work_size)) 2003 if (SvOK (local_work_size))
659 { 2004 {
660 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 2005 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
661 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 2006 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
662 2007
663 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 2008 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
664 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 2009 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
665 2010
666 lws = lists + gws_len * 2; 2011 lws = lists + gws_len * 2;
667 for (i = 0; i < gws_len; ++i) 2012 for (i = 0; i < gws_len; ++i)
2013 {
668 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 2014 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
2015 // at least nvidia crashes for 0-sized work group sizes, work around
2016 if (!lws [i])
2017 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2018 }
669 } 2019 }
670 2020
2021 cl_event ev = 0;
671 NEED_SUCCESS (EnqueueNDRangeKernel, (this, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2022 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
672 2023
673 if (ev) 2024 if (ev)
674 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2025 XPUSH_CLOBJ (stash_event, ev);
675}
676 2026
2027#if CL_VERSION_1_2
2028
677void 2029void
678enqueue_marker (OpenCL::Queue this) 2030migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2031 ALIAS:
2032 enqueue_migrate_mem_objects = 0
679 PPCODE: 2033 PPCODE:
680{ 2034 EVENT_LIST (3);
2035
2036 cl_uint obj_count;
2037 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2038
2039 cl_event ev = 0;
2040 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2041
2042 if (ev)
2043 XPUSH_CLOBJ (stash_event, ev);
2044
2045#endif
2046
2047#if cl_apple_gl_sharing || cl_khr_gl_sharing
2048
2049void
2050acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2051 ALIAS:
2052 release_gl_objects = 1
2053 enqueue_acquire_gl_objects = 0
2054 enqueue_release_gl_objects = 1
2055 PPCODE:
2056 EVENT_LIST (2);
2057
2058 cl_uint obj_count;
2059 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2060
681 cl_event ev; 2061 cl_event ev = 0;
682 NEED_SUCCESS (EnqueueMarker, (this, &ev));
683 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
684}
685 2062
2063 if (ix)
2064 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2065 else
2066 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2067
2068 if (ev)
2069 XPUSH_CLOBJ (stash_event, ev);
2070
2071#endif
2072
686void 2073void
687enqueue_wait_for_events (OpenCL::Queue this, ...) 2074wait_for_events (OpenCL::Queue self, ...)
2075 ALIAS:
2076 enqueue_wait_for_events = 0
688 CODE: 2077 CODE:
689{
690 EVENT_LIST (1, items - 1); 2078 EVENT_LIST (1);
2079#if PREFER_1_1
691 NEED_SUCCESS (EnqueueWaitForEvents, (this, event_list_count, event_list_ptr)); 2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
692} 2081#else
2082 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2083#endif
693 2084
694void 2085void
695enqueue_barrier (OpenCL::Queue this) 2086marker (OpenCL::Queue self, ...)
2087 ALIAS:
2088 enqueue_marker = 0
696 CODE: 2089 PPCODE:
2090 EVENT_LIST (1);
2091 cl_event ev = 0;
2092#if PREFER_1_1
2093 if (!event_list_count)
2094 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2095 else
2096#if CL_VERSION_1_2
2097 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2098#else
2099 {
2100 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2101 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2102 }
2103#endif
2104#else
2105 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2106#endif
2107 if (ev)
2108 XPUSH_CLOBJ (stash_event, ev);
2109
2110void
2111barrier (OpenCL::Queue self, ...)
2112 ALIAS:
2113 enqueue_barrier = 0
2114 PPCODE:
2115 EVENT_LIST (1);
2116 cl_event ev = 0;
2117#if PREFER_1_1
2118 if (!event_list_count && GIMME_V == G_VOID)
697 NEED_SUCCESS (EnqueueBarrier, (this)); 2119 NEED_SUCCESS (EnqueueBarrier, (self));
2120 else
2121#if CL_VERSION_1_2
2122 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2123#else
2124 {
2125 if (event_list_count)
2126 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
698 2127
2128 if (GIMME_V != G_VOID)
2129 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2130 }
2131#endif
2132#else
2133 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2134#endif
2135 if (ev)
2136 XPUSH_CLOBJ (stash_event, ev);
2137
699void 2138void
700flush (OpenCL::Queue this) 2139flush (OpenCL::Queue self)
701 CODE: 2140 CODE:
702 NEED_SUCCESS (Flush, (this)); 2141 NEED_SUCCESS (Flush, (self));
703 2142
704void 2143void
705finish (OpenCL::Queue this) 2144finish (OpenCL::Queue self)
706 CODE: 2145 CODE:
707 NEED_SUCCESS (Finish, (this)); 2146 NEED_SUCCESS (Finish, (self));
2147
2148void
2149info (OpenCL::Queue self, cl_command_queue_info name)
2150 PPCODE:
2151 INFO (CommandQueue)
2152
2153#BEGIN:command_queue
2154
2155void
2156context (OpenCL::Queue self)
2157 PPCODE:
2158 cl_context value [1];
2159 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
2160 EXTEND (SP, 1);
2161 const int i = 0;
2162 {
2163 NEED_SUCCESS (RetainContext, (value [i]));
2164 PUSH_CLOBJ (stash_context, value [i]);
2165 }
2166
2167void
2168device (OpenCL::Queue self)
2169 PPCODE:
2170 cl_device_id value [1];
2171 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
2172 EXTEND (SP, 1);
2173 const int i = 0;
2174 {
2175 PUSH_CLOBJ (stash_device, value [i]);
2176 }
2177
2178void
2179reference_count (OpenCL::Queue self)
2180 PPCODE:
2181 cl_uint value [1];
2182 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_REFERENCE_COUNT, sizeof (value), value, 0));
2183 EXTEND (SP, 1);
2184 const int i = 0;
2185 PUSHs (sv_2mortal (newSVuv (value [i])));
2186
2187void
2188properties (OpenCL::Queue self)
2189 PPCODE:
2190 cl_command_queue_properties value [1];
2191 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
2192 EXTEND (SP, 1);
2193 const int i = 0;
2194 PUSHs (sv_2mortal (newSVuv (value [i])));
2195
2196#END:command_queue
708 2197
709MODULE = OpenCL PACKAGE = OpenCL::Memory 2198MODULE = OpenCL PACKAGE = OpenCL::Memory
710 2199
711void 2200void
712DESTROY (OpenCL::Memory this) 2201DESTROY (OpenCL::Memory self)
713 CODE: 2202 CODE:
714 clReleaseMemObject (this); 2203 clReleaseMemObject (self);
715 2204
716void 2205void
717info (OpenCL::Memory this, cl_mem_info name) 2206info (OpenCL::Memory self, cl_mem_info name)
718 PPCODE: 2207 PPCODE:
719 INFO (MemObject) 2208 INFO (MemObject)
720 2209
2210#BEGIN:mem
2211
2212void
2213type (OpenCL::Memory self)
2214 PPCODE:
2215 cl_mem_object_type value [1];
2216 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
2217 EXTEND (SP, 1);
2218 const int i = 0;
2219 PUSHs (sv_2mortal (newSVuv (value [i])));
2220
2221void
2222flags (OpenCL::Memory self)
2223 PPCODE:
2224 cl_mem_flags value [1];
2225 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
2226 EXTEND (SP, 1);
2227 const int i = 0;
2228 PUSHs (sv_2mortal (newSVuv (value [i])));
2229
2230void
2231size (OpenCL::Memory self)
2232 ALIAS:
2233 size = CL_MEM_SIZE
2234 offset = CL_MEM_OFFSET
2235 PPCODE:
2236 size_t value [1];
2237 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2238 EXTEND (SP, 1);
2239 const int i = 0;
2240 PUSHs (sv_2mortal (newSVuv (value [i])));
2241
2242void
2243host_ptr (OpenCL::Memory self)
2244 PPCODE:
2245 void * value [1];
2246 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_HOST_PTR, sizeof (value), value, 0));
2247 EXTEND (SP, 1);
2248 const int i = 0;
2249 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
2250
2251void
2252map_count (OpenCL::Memory self)
2253 ALIAS:
2254 map_count = CL_MEM_MAP_COUNT
2255 reference_count = CL_MEM_REFERENCE_COUNT
2256 PPCODE:
2257 cl_uint value [1];
2258 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
2259 EXTEND (SP, 1);
2260 const int i = 0;
2261 PUSHs (sv_2mortal (newSVuv (value [i])));
2262
2263void
2264context (OpenCL::Memory self)
2265 PPCODE:
2266 cl_context value [1];
2267 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
2268 EXTEND (SP, 1);
2269 const int i = 0;
2270 {
2271 NEED_SUCCESS (RetainContext, (value [i]));
2272 PUSH_CLOBJ (stash_context, value [i]);
2273 }
2274
2275void
2276associated_memobject (OpenCL::Memory self)
2277 PPCODE:
2278 cl_mem value [1];
2279 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
2280 EXTEND (SP, 1);
2281 const int i = 0;
2282 {
2283 NEED_SUCCESS (RetainMemObject, (value [i]));
2284 PUSH_CLOBJ (stash_memory, value [i]);
2285 }
2286
2287#END:mem
2288
2289#if cl_apple_gl_sharing || cl_khr_gl_sharing
2290
2291void
2292gl_object_info (OpenCL::Memory self)
2293 PPCODE:
2294 cl_gl_object_type type;
2295 cl_GLuint name;
2296 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2297 EXTEND (SP, 2);
2298 PUSHs (sv_2mortal (newSVuv (type)));
2299 PUSHs (sv_2mortal (newSVuv (name)));
2300
2301#endif
2302
2303MODULE = OpenCL PACKAGE = OpenCL::BufferObj
2304
2305void
2306sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
2307 PPCODE:
2308 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR))
2309 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
2310
2311 cl_buffer_region crdata = { origin, size };
2312
2313 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
2314 XPUSH_CLOBJ (stash_buffer, mem);
2315
2316MODULE = OpenCL PACKAGE = OpenCL::Image
2317
2318void
2319image_info (OpenCL::Image self, cl_image_info name)
2320 PPCODE:
2321 INFO (Image)
2322
2323void
2324format (OpenCL::Image self)
2325 PPCODE:
2326 cl_image_format format;
2327 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2328 EXTEND (SP, 2);
2329 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2330 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
2331
2332#BEGIN:image
2333
2334void
2335element_size (OpenCL::Image self)
2336 ALIAS:
2337 element_size = CL_IMAGE_ELEMENT_SIZE
2338 row_pitch = CL_IMAGE_ROW_PITCH
2339 slice_pitch = CL_IMAGE_SLICE_PITCH
2340 width = CL_IMAGE_WIDTH
2341 height = CL_IMAGE_HEIGHT
2342 depth = CL_IMAGE_DEPTH
2343 PPCODE:
2344 size_t value [1];
2345 NEED_SUCCESS (GetImageInfo, (self, ix, sizeof (value), value, 0));
2346 EXTEND (SP, 1);
2347 const int i = 0;
2348 PUSHs (sv_2mortal (newSVuv (value [i])));
2349
2350#END:image
2351
2352#if cl_apple_gl_sharing || cl_khr_gl_sharing
2353
2354#BEGIN:gl_texture
2355
2356void
2357target (OpenCL::Image self)
2358 PPCODE:
2359 cl_GLenum value [1];
2360 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2361 EXTEND (SP, 1);
2362 const int i = 0;
2363 PUSHs (sv_2mortal (newSVuv (value [i])));
2364
2365void
2366gl_mipmap_level (OpenCL::Image self)
2367 PPCODE:
2368 cl_GLint value [1];
2369 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2370 EXTEND (SP, 1);
2371 const int i = 0;
2372 PUSHs (sv_2mortal (newSViv (value [i])));
2373
2374#END:gl_texture
2375
2376#endif
2377
721MODULE = OpenCL PACKAGE = OpenCL::Sampler 2378MODULE = OpenCL PACKAGE = OpenCL::Sampler
722 2379
723void 2380void
724DESTROY (OpenCL::Sampler this) 2381DESTROY (OpenCL::Sampler self)
725 CODE: 2382 CODE:
726 clReleaseSampler (this); 2383 clReleaseSampler (self);
727 2384
728void 2385void
729info (OpenCL::Sampler this, cl_sampler_info name) 2386info (OpenCL::Sampler self, cl_sampler_info name)
730 PPCODE: 2387 PPCODE:
731 INFO (Sampler) 2388 INFO (Sampler)
732 2389
2390#BEGIN:sampler
2391
2392void
2393reference_count (OpenCL::Sampler self)
2394 PPCODE:
2395 cl_uint value [1];
2396 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0));
2397 EXTEND (SP, 1);
2398 const int i = 0;
2399 PUSHs (sv_2mortal (newSVuv (value [i])));
2400
2401void
2402context (OpenCL::Sampler self)
2403 PPCODE:
2404 cl_context value [1];
2405 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2406 EXTEND (SP, 1);
2407 const int i = 0;
2408 {
2409 NEED_SUCCESS (RetainContext, (value [i]));
2410 PUSH_CLOBJ (stash_context, value [i]);
2411 }
2412
2413void
2414normalized_coords (OpenCL::Sampler self)
2415 PPCODE:
2416 cl_addressing_mode value [1];
2417 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2418 EXTEND (SP, 1);
2419 const int i = 0;
2420 PUSHs (sv_2mortal (newSVuv (value [i])));
2421
2422void
2423addressing_mode (OpenCL::Sampler self)
2424 PPCODE:
2425 cl_filter_mode value [1];
2426 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2427 EXTEND (SP, 1);
2428 const int i = 0;
2429 PUSHs (sv_2mortal (newSVuv (value [i])));
2430
2431void
2432filter_mode (OpenCL::Sampler self)
2433 PPCODE:
2434 cl_bool value [1];
2435 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2436 EXTEND (SP, 1);
2437 const int i = 0;
2438 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2439
2440#END:sampler
2441
733MODULE = OpenCL PACKAGE = OpenCL::Program 2442MODULE = OpenCL PACKAGE = OpenCL::Program
734 2443
735void 2444void
736DESTROY (OpenCL::Program this) 2445DESTROY (OpenCL::Program self)
737 CODE: 2446 CODE:
738 clReleaseProgram (this); 2447 clReleaseProgram (self);
739 2448
740void 2449void
741info (OpenCL::Program this, cl_program_info name) 2450build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2451 ALIAS:
2452 build_async = 1
742 PPCODE: 2453 CODE:
743 INFO (Program) 2454 cl_uint device_count = 0;
2455 cl_device_id *device_list = 0;
744 2456
2457 if (SvOK (devices))
2458 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2459
2460 void *user_data;
2461 program_callback pfn_notify = make_program_callback (notify, &user_data);
2462
2463 if (ix)
2464 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2465 else
2466 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2467
2468#if CL_VERSION_1_2
2469
745void 2470void
746build (OpenCL::Program this, OpenCL::Device device, SV *options = &PL_sv_undef) 2471compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
747 CODE: 2472 CODE:
2473 cl_uint device_count = 0;
2474 cl_device_id *device_list = 0;
2475
2476 if (SvOK (devices))
2477 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2478
2479 cl_uint header_count = 0;
2480 cl_program *header_list = 0;
2481 const char **header_name = 0;
2482
2483 if (SvOK (headers))
2484 {
2485 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2486 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2487
2488 HV *hv = (HV *)SvRV (devices);
2489
2490 header_count = hv_iterinit (hv);
2491 header_list = tmpbuf (sizeof (*header_list) * header_count);
2492 header_name = tmpbuf (sizeof (*header_name) * header_count);
2493
2494 HE *he;
2495 int i = 0;
2496 while (he = hv_iternext (hv))
2497 {
2498 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2499 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2500 ++i;
2501 }
2502 }
2503
2504 void *user_data;
2505 program_callback pfn_notify = make_program_callback (notify, &user_data);
2506
748 NEED_SUCCESS (BuildProgram, (this, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2507 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2508 header_count, header_list, header_name, pfn_notify, user_data));
749 2509
2510#endif
2511
750void 2512void
751build_info (OpenCL::Program this, OpenCL::Device device, cl_program_build_info name) 2513build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
752 PPCODE: 2514 PPCODE:
753{
754 size_t size; 2515 size_t size;
755 SV *sv;
756
757 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, 0, 0, &size)); 2516 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
758 sv = sv_2mortal (newSV (size)); 2517 SV *sv = sv_2mortal (newSV (size));
759 SvUPGRADE (sv, SVt_PV); 2518 SvUPGRADE (sv, SVt_PV);
760 SvPOK_only (sv); 2519 SvPOK_only (sv);
761 SvCUR_set (sv, size); 2520 SvCUR_set (sv, size);
762 NEED_SUCCESS (GetProgramBuildInfo, (this, device, name, size, SvPVX (sv), 0)); 2521 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
763 XPUSHs (sv); 2522 XPUSHs (sv);
764} 2523
2524#BEGIN:program_build
2525
2526void
2527build_status (OpenCL::Program self, OpenCL::Device device)
2528 PPCODE:
2529 cl_build_status value [1];
2530 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2531 EXTEND (SP, 1);
2532 const int i = 0;
2533 PUSHs (sv_2mortal (newSViv (value [i])));
2534
2535void
2536build_options (OpenCL::Program self, OpenCL::Device device)
2537 ALIAS:
2538 build_options = CL_PROGRAM_BUILD_OPTIONS
2539 build_log = CL_PROGRAM_BUILD_LOG
2540 PPCODE:
2541 size_t size;
2542 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, 0, 0, &size));
2543 char *value = tmpbuf (size);
2544 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2545 EXTEND (SP, 1);
2546 const int i = 0;
2547 PUSHs (sv_2mortal (newSVpv (value, 0)));
2548
2549void
2550binary_type (OpenCL::Program self, OpenCL::Device device)
2551 PPCODE:
2552 cl_program_binary_type value [1];
2553 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2554 EXTEND (SP, 1);
2555 const int i = 0;
2556 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2557
2558#END:program_build
765 2559
766void 2560void
767kernel (OpenCL::Program program, SV *function) 2561kernel (OpenCL::Program program, SV *function)
768 PPCODE: 2562 PPCODE:
769{
770 cl_int res;
771 cl_kernel kernel = clCreateKernel (program, SvPVbyte_nolen (function), &res); 2563 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2564 XPUSH_CLOBJ (stash_kernel, kernel);
772 2565
773 if (res) 2566void
774 FAIL (CreateKernel, res); 2567kernels_in_program (OpenCL::Program program)
2568 PPCODE:
2569 cl_uint num_kernels;
2570 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2571 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2572 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
775 2573
776 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2574 int i;
777} 2575 EXTEND (SP, num_kernels);
2576 for (i = 0; i < num_kernels; ++i)
2577 PUSH_CLOBJ (stash_kernel, kernels [i]);
2578
2579void
2580info (OpenCL::Program self, cl_program_info name)
2581 PPCODE:
2582 INFO (Program)
2583
2584void
2585binaries (OpenCL::Program self)
2586 PPCODE:
2587 cl_uint n, i;
2588 size_t size;
2589
2590 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_NUM_DEVICES , sizeof (n) , &n , 0));
2591 if (!n) XSRETURN_EMPTY;
2592
2593 size_t *sizes = tmpbuf (sizeof (*sizes) * n);
2594 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, sizeof (*sizes) * n, sizes, &size));
2595 if (size != sizeof (*sizes) * n) XSRETURN_EMPTY;
2596 unsigned char **ptrs = tmpbuf (sizeof (*ptrs) * n);
2597
2598 EXTEND (SP, n);
2599 for (i = 0; i < n; ++i)
2600 {
2601 SV *sv = sv_2mortal (newSV (sizes [i]));
2602 SvUPGRADE (sv, SVt_PV);
2603 SvPOK_only (sv);
2604 SvCUR_set (sv, sizes [i]);
2605 ptrs [i] = (void *)SvPVX (sv);
2606 PUSHs (sv);
2607 }
2608
2609 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
2610 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
2611
2612#BEGIN:program
2613
2614void
2615reference_count (OpenCL::Program self)
2616 ALIAS:
2617 reference_count = CL_PROGRAM_REFERENCE_COUNT
2618 num_devices = CL_PROGRAM_NUM_DEVICES
2619 PPCODE:
2620 cl_uint value [1];
2621 NEED_SUCCESS (GetProgramInfo, (self, ix, sizeof (value), value, 0));
2622 EXTEND (SP, 1);
2623 const int i = 0;
2624 PUSHs (sv_2mortal (newSVuv (value [i])));
2625
2626void
2627context (OpenCL::Program self)
2628 PPCODE:
2629 cl_context value [1];
2630 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2631 EXTEND (SP, 1);
2632 const int i = 0;
2633 {
2634 NEED_SUCCESS (RetainContext, (value [i]));
2635 PUSH_CLOBJ (stash_context, value [i]);
2636 }
2637
2638void
2639devices (OpenCL::Program self)
2640 PPCODE:
2641 size_t size;
2642 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, 0, 0, &size));
2643 cl_device_id *value = tmpbuf (size);
2644 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2645 int i, n = size / sizeof (*value);
2646 EXTEND (SP, n);
2647 for (i = 0; i < n; ++i)
2648 {
2649 PUSH_CLOBJ (stash_device, value [i]);
2650 }
2651
2652void
2653source (OpenCL::Program self)
2654 PPCODE:
2655 size_t size;
2656 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, 0, 0, &size));
2657 char *value = tmpbuf (size);
2658 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_SOURCE, size, value, 0));
2659 EXTEND (SP, 1);
2660 const int i = 0;
2661 PUSHs (sv_2mortal (newSVpv (value, 0)));
2662
2663void
2664binary_sizes (OpenCL::Program self)
2665 PPCODE:
2666 size_t size;
2667 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, 0, 0, &size));
2668 size_t *value = tmpbuf (size);
2669 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARY_SIZES, size, value, 0));
2670 int i, n = size / sizeof (*value);
2671 EXTEND (SP, n);
2672 for (i = 0; i < n; ++i)
2673 PUSHs (sv_2mortal (newSVuv (value [i])));
2674
2675#END:program
778 2676
779MODULE = OpenCL PACKAGE = OpenCL::Kernel 2677MODULE = OpenCL PACKAGE = OpenCL::Kernel
780 2678
781void 2679void
782DESTROY (OpenCL::Kernel this) 2680DESTROY (OpenCL::Kernel self)
783 CODE: 2681 CODE:
784 clReleaseKernel (this); 2682 clReleaseKernel (self);
785 2683
786void 2684void
2685setf (OpenCL::Kernel self, const char *format, ...)
2686 CODE:
2687 int i;
2688 for (i = 2; ; ++i)
2689 {
2690 while (*format == ' ')
2691 ++format;
2692
2693 char type = *format++;
2694
2695 if (!type)
2696 break;
2697
2698 if (i >= items)
2699 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2700
2701 SV *sv = ST (i);
2702
2703 union
2704 {
2705 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2706 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2707 cl_half ch; cl_float cf; cl_double cd;
2708 cl_mem cm;
2709 cl_sampler ca;
2710 size_t cz;
2711 cl_event ce;
2712 } arg;
2713 size_t size;
2714 int nullarg = 0;
2715
2716 switch (type)
2717 {
2718 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2719 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2720 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2721 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2722 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2723 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2724 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2725 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2726
2727 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2728 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2729 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2730
2731 case 'z': nullarg = 1; size = SvIV (sv); break;
2732
2733 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2734 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2735 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2736
2737 default:
2738 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2739 }
2740
2741 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2742 if (res)
2743 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2744 }
2745
2746 if (i != items)
2747 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2748
2749void
2750set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2751 CODE:
2752 clSetKernelArg (self, idx, sizeof (value), &value);
2753
2754void
2755set_uchar (OpenCL::Kernel self, cl_uint idx, cl_uchar value)
2756 CODE:
2757 clSetKernelArg (self, idx, sizeof (value), &value);
2758
2759void
2760set_short (OpenCL::Kernel self, cl_uint idx, cl_short value)
2761 CODE:
2762 clSetKernelArg (self, idx, sizeof (value), &value);
2763
2764void
2765set_ushort (OpenCL::Kernel self, cl_uint idx, cl_ushort value)
2766 CODE:
2767 clSetKernelArg (self, idx, sizeof (value), &value);
2768
2769void
2770set_int (OpenCL::Kernel self, cl_uint idx, cl_int value)
2771 CODE:
2772 clSetKernelArg (self, idx, sizeof (value), &value);
2773
2774void
2775set_uint (OpenCL::Kernel self, cl_uint idx, cl_uint value)
2776 CODE:
2777 clSetKernelArg (self, idx, sizeof (value), &value);
2778
2779void
2780set_long (OpenCL::Kernel self, cl_uint idx, cl_long value)
2781 CODE:
2782 clSetKernelArg (self, idx, sizeof (value), &value);
2783
2784void
2785set_ulong (OpenCL::Kernel self, cl_uint idx, cl_ulong value)
2786 CODE:
2787 clSetKernelArg (self, idx, sizeof (value), &value);
2788
2789void
2790set_half (OpenCL::Kernel self, cl_uint idx, cl_half value)
2791 CODE:
2792 clSetKernelArg (self, idx, sizeof (value), &value);
2793
2794void
2795set_float (OpenCL::Kernel self, cl_uint idx, cl_float value)
2796 CODE:
2797 clSetKernelArg (self, idx, sizeof (value), &value);
2798
2799void
2800set_double (OpenCL::Kernel self, cl_uint idx, cl_double value)
2801 CODE:
2802 clSetKernelArg (self, idx, sizeof (value), &value);
2803
2804void
2805set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2806 CODE:
2807 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2808
2809void
2810set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2811 CODE:
2812 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2813
2814void
2815set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2816 CODE:
2817 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2818
2819void
2820set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2821 CODE:
2822 clSetKernelArg (self, idx, sizeof (value), &value);
2823
2824void
2825set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
2826 CODE:
2827 clSetKernelArg (self, idx, size, 0);
2828
2829void
2830set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
2831 CODE:
2832 clSetKernelArg (self, idx, sizeof (value), &value);
2833
2834void
787info (OpenCL::Kernel this, cl_kernel_info name) 2835info (OpenCL::Kernel self, cl_kernel_info name)
788 PPCODE: 2836 PPCODE:
789 INFO (Kernel) 2837 INFO (Kernel)
790 2838
2839#BEGIN:kernel
2840
791void 2841void
792set_char (OpenCL::Kernel this, cl_uint idx, cl_char value) 2842function_name (OpenCL::Kernel self)
2843 PPCODE:
2844 size_t size;
2845 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, 0, 0, &size));
2846 char *value = tmpbuf (size);
2847 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_FUNCTION_NAME, size, value, 0));
2848 EXTEND (SP, 1);
2849 const int i = 0;
2850 PUSHs (sv_2mortal (newSVpv (value, 0)));
2851
2852void
2853num_args (OpenCL::Kernel self)
2854 ALIAS:
2855 num_args = CL_KERNEL_NUM_ARGS
2856 reference_count = CL_KERNEL_REFERENCE_COUNT
2857 PPCODE:
2858 cl_uint value [1];
2859 NEED_SUCCESS (GetKernelInfo, (self, ix, sizeof (value), value, 0));
2860 EXTEND (SP, 1);
2861 const int i = 0;
2862 PUSHs (sv_2mortal (newSVuv (value [i])));
2863
2864void
2865context (OpenCL::Kernel self)
2866 PPCODE:
2867 cl_context value [1];
2868 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2869 EXTEND (SP, 1);
2870 const int i = 0;
2871 {
2872 NEED_SUCCESS (RetainContext, (value [i]));
2873 PUSH_CLOBJ (stash_context, value [i]);
2874 }
2875
2876void
2877program (OpenCL::Kernel self)
2878 PPCODE:
2879 cl_program value [1];
2880 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2881 EXTEND (SP, 1);
2882 const int i = 0;
2883 {
2884 NEED_SUCCESS (RetainProgram, (value [i]));
2885 PUSH_CLOBJ (stash_program, value [i]);
2886 }
2887
2888#END:kernel
2889
2890void
2891work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
793 CODE: 2892 PPCODE:
794 clSetKernelArg (this, idx, sizeof (value), &value); 2893 size_t size;
2894 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, 0, 0, &size));
2895 SV *sv = sv_2mortal (newSV (size));
2896 SvUPGRADE (sv, SVt_PV);
2897 SvPOK_only (sv);
2898 SvCUR_set (sv, size);
2899 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, name, size, SvPVX (sv), 0));
2900 XPUSHs (sv);
795 2901
2902#BEGIN:kernel_work_group
2903
796void 2904void
797set_uchar (OpenCL::Kernel this, cl_uint idx, cl_uchar value) 2905work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2906 ALIAS:
2907 work_group_size = CL_KERNEL_WORK_GROUP_SIZE
2908 preferred_work_group_size_multiple = CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
2909 PPCODE:
2910 size_t value [1];
2911 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2912 EXTEND (SP, 1);
2913 const int i = 0;
2914 PUSHs (sv_2mortal (newSVuv (value [i])));
2915
2916void
2917compile_work_group_size (OpenCL::Kernel self, OpenCL::Device device)
2918 PPCODE:
2919 size_t size;
2920 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, 0, 0, &size));
2921 size_t *value = tmpbuf (size);
2922 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, CL_KERNEL_COMPILE_WORK_GROUP_SIZE, size, value, 0));
2923 int i, n = size / sizeof (*value);
2924 EXTEND (SP, n);
2925 for (i = 0; i < n; ++i)
2926 PUSHs (sv_2mortal (newSVuv (value [i])));
2927
2928void
2929local_mem_size (OpenCL::Kernel self, OpenCL::Device device)
2930 ALIAS:
2931 local_mem_size = CL_KERNEL_LOCAL_MEM_SIZE
2932 private_mem_size = CL_KERNEL_PRIVATE_MEM_SIZE
2933 PPCODE:
2934 cl_ulong value [1];
2935 NEED_SUCCESS (GetKernelWorkGroupInfo, (self, device, ix, sizeof (value), value, 0));
2936 EXTEND (SP, 1);
2937 const int i = 0;
2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2939
2940#END:kernel_work_group
2941
2942#if CL_VERSION_1_2
2943
2944void
2945arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
798 CODE: 2946 PPCODE:
799 clSetKernelArg (this, idx, sizeof (value), &value); 2947 size_t size;
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2949 SV *sv = sv_2mortal (newSV (size));
2950 SvUPGRADE (sv, SVt_PV);
2951 SvPOK_only (sv);
2952 SvCUR_set (sv, size);
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2954 XPUSHs (sv);
800 2955
801void 2956#BEGIN:kernel_arg
802set_short (OpenCL::Kernel this, cl_uint idx, cl_short value)
803 CODE:
804 clSetKernelArg (this, idx, sizeof (value), &value);
805 2957
806void 2958void
807set_ushort (OpenCL::Kernel this, cl_uint idx, cl_ushort value) 2959arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
808 CODE: 2960 PPCODE:
809 clSetKernelArg (this, idx, sizeof (value), &value); 2961 cl_kernel_arg_address_qualifier value [1];
2962 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2963 EXTEND (SP, 1);
2964 const int i = 0;
2965 PUSHs (sv_2mortal (newSVuv (value [i])));
810 2966
811void 2967void
812set_int (OpenCL::Kernel this, cl_uint idx, cl_int value) 2968arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
813 CODE: 2969 PPCODE:
814 clSetKernelArg (this, idx, sizeof (value), &value); 2970 cl_kernel_arg_access_qualifier value [1];
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVuv (value [i])));
815 2975
816void 2976void
817set_uint (OpenCL::Kernel this, cl_uint idx, cl_uint value) 2977arg_type_name (OpenCL::Kernel self, cl_uint idx)
818 CODE: 2978 ALIAS:
819 clSetKernelArg (this, idx, sizeof (value), &value); 2979 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2980 arg_name = CL_KERNEL_ARG_NAME
2981 PPCODE:
2982 size_t size;
2983 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2984 char *value = tmpbuf (size);
2985 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2986 EXTEND (SP, 1);
2987 const int i = 0;
2988 PUSHs (sv_2mortal (newSVpv (value, 0)));
820 2989
821void 2990void
822set_long (OpenCL::Kernel this, cl_uint idx, cl_long value) 2991arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
823 CODE: 2992 PPCODE:
824 clSetKernelArg (this, idx, sizeof (value), &value); 2993 cl_kernel_arg_type_qualifier value [1];
2994 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2995 EXTEND (SP, 1);
2996 const int i = 0;
2997 PUSHs (sv_2mortal (newSVuv (value [i])));
825 2998
826void 2999#END:kernel_arg
827set_ulong (OpenCL::Kernel this, cl_uint idx, cl_ulong value)
828 CODE:
829 clSetKernelArg (this, idx, sizeof (value), &value);
830 3000
831void 3001#endif
832set_half (OpenCL::Kernel this, cl_uint idx, cl_half value)
833 CODE:
834 clSetKernelArg (this, idx, sizeof (value), &value);
835
836void
837set_float (OpenCL::Kernel this, cl_uint idx, cl_float value)
838 CODE:
839 clSetKernelArg (this, idx, sizeof (value), &value);
840
841void
842set_double (OpenCL::Kernel this, cl_uint idx, cl_double value)
843 CODE:
844 clSetKernelArg (this, idx, sizeof (value), &value);
845
846void
847set_memory (OpenCL::Kernel this, cl_uint idx, OpenCL::Memory_ornull value)
848 CODE:
849 clSetKernelArg (this, idx, sizeof (value), &value);
850
851void
852set_buffer (OpenCL::Kernel this, cl_uint idx, OpenCL::Buffer_ornull value)
853 CODE:
854 clSetKernelArg (this, idx, sizeof (value), &value);
855
856void
857set_image2d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image2D_ornull value)
858 CODE:
859 clSetKernelArg (this, idx, sizeof (value), &value);
860
861void
862set_image3d (OpenCL::Kernel this, cl_uint idx, OpenCL::Image3D_ornull value)
863 CODE:
864 clSetKernelArg (this, idx, sizeof (value), &value);
865
866void
867set_sampler (OpenCL::Kernel this, cl_uint idx, OpenCL::Sampler value)
868 CODE:
869 clSetKernelArg (this, idx, sizeof (value), &value);
870
871void
872set_event (OpenCL::Kernel this, cl_uint idx, OpenCL::Event value)
873 CODE:
874 clSetKernelArg (this, idx, sizeof (value), &value);
875 3002
876MODULE = OpenCL PACKAGE = OpenCL::Event 3003MODULE = OpenCL PACKAGE = OpenCL::Event
877 3004
878void 3005void
879DESTROY (OpenCL::Event this) 3006DESTROY (OpenCL::Event self)
880 CODE: 3007 CODE:
881 clReleaseEvent (this); 3008 clReleaseEvent (self);
882 3009
883void 3010void
3011wait (OpenCL::Event self)
3012 CODE:
3013 clWaitForEvents (1, &self);
3014
3015void
3016cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
3017 CODE:
3018 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
3019
3020void
884info (OpenCL::Event this, cl_event_info name) 3021info (OpenCL::Event self, cl_event_info name)
885 PPCODE: 3022 PPCODE:
886 INFO (Event) 3023 INFO (Event)
887 3024
3025#BEGIN:event
3026
888void 3027void
889wait (OpenCL::Event this) 3028command_queue (OpenCL::Event self)
3029 PPCODE:
3030 cl_command_queue value [1];
3031 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
3032 EXTEND (SP, 1);
3033 const int i = 0;
3034 {
3035 NEED_SUCCESS (RetainCommandQueue, (value [i]));
3036 PUSH_CLOBJ (stash_queue, value [i]);
3037 }
3038
3039void
3040command_type (OpenCL::Event self)
3041 PPCODE:
3042 cl_command_type value [1];
3043 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
3044 EXTEND (SP, 1);
3045 const int i = 0;
3046 PUSHs (sv_2mortal (newSVuv (value [i])));
3047
3048void
3049reference_count (OpenCL::Event self)
3050 ALIAS:
3051 reference_count = CL_EVENT_REFERENCE_COUNT
3052 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
3053 PPCODE:
3054 cl_uint value [1];
3055 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
3056 EXTEND (SP, 1);
3057 const int i = 0;
3058 PUSHs (sv_2mortal (newSVuv (value [i])));
3059
3060void
3061context (OpenCL::Event self)
3062 PPCODE:
3063 cl_context value [1];
3064 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
3065 EXTEND (SP, 1);
3066 const int i = 0;
3067 {
3068 NEED_SUCCESS (RetainContext, (value [i]));
3069 PUSH_CLOBJ (stash_context, value [i]);
3070 }
3071
3072#END:event
3073
3074void
3075profiling_info (OpenCL::Event self, cl_profiling_info name)
890 CODE: 3076 PPCODE:
891 clWaitForEvents (1, &this); 3077 INFO (EventProfiling)
3078
3079#BEGIN:profiling
3080
3081void
3082profiling_command_queued (OpenCL::Event self)
3083 ALIAS:
3084 profiling_command_queued = CL_PROFILING_COMMAND_QUEUED
3085 profiling_command_submit = CL_PROFILING_COMMAND_SUBMIT
3086 profiling_command_start = CL_PROFILING_COMMAND_START
3087 profiling_command_end = CL_PROFILING_COMMAND_END
3088 PPCODE:
3089 cl_ulong value [1];
3090 NEED_SUCCESS (GetEventProfilingInfo, (self, ix, sizeof (value), value, 0));
3091 EXTEND (SP, 1);
3092 const int i = 0;
3093 PUSHs (sv_2mortal (newSVuv (value [i])));
3094
3095#END:profiling
892 3096
893MODULE = OpenCL PACKAGE = OpenCL::UserEvent 3097MODULE = OpenCL PACKAGE = OpenCL::UserEvent
894 3098
895void 3099void
896set_status (OpenCL::UserEvent this, cl_int execution_status) 3100set_status (OpenCL::UserEvent self, cl_int execution_status)
897 CODE: 3101 CODE:
898 clSetUserEventStatus (this, execution_status); 3102 clSetUserEventStatus (self, execution_status);
899 3103
3104MODULE = OpenCL PACKAGE = OpenCL::Mapped
3105
3106void
3107DESTROY (SV *self)
3108 CODE:
3109 OpenCL__Mapped mapped = SvMAPPED (self);
3110
3111 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3112 mapped_detach (self, mapped);
3113
3114 clReleaseCommandQueue (mapped->queue);
3115 clReleaseEvent (mapped->event);
3116 Safefree (mapped);
3117
3118void
3119unmap (OpenCL::Mapped self, ...)
3120 CODE:
3121 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3122
3123bool
3124mapped (OpenCL::Mapped self)
3125 CODE:
3126 RETVAL = !!self->ptr;
3127 OUTPUT:
3128 RETVAL
3129
3130void
3131wait (OpenCL::Mapped self)
3132 PPCODE:
3133 if (self->event)
3134 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3135
3136void
3137event (OpenCL::Mapped self)
3138 PPCODE:
3139 if (!self->event)
3140 XSRETURN_UNDEF;
3141
3142 clRetainEvent (self->event);
3143 XPUSH_CLOBJ (stash_event, self->event);
3144
3145size_t
3146size (OpenCL::Mapped self)
3147 CODE:
3148 RETVAL = self->cb;
3149 OUTPUT:
3150 RETVAL
3151
3152IV
3153ptr (OpenCL::Mapped self)
3154 CODE:
3155 RETVAL = PTR2IV (self->ptr);
3156 OUTPUT:
3157 RETVAL
3158
3159void
3160set (OpenCL::Mapped self, size_t offset, SV *data)
3161 CODE:
3162 STRLEN len;
3163 const char *ptr = SvPVbyte (data, len);
3164
3165 if (offset + len > self->cb)
3166 croak ("OpenCL::Mapped::set out of bound condition detected");
3167
3168 memcpy (offset + (char *)self->ptr, ptr, len);
3169
3170MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3171
3172MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3173
3174IV
3175row_pitch (OpenCL::Mapped self)
3176 ALIAS:
3177 slice_pitch = 1
3178 CODE:
3179 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3180 OUTPUT:
3181 RETVAL
3182
3183
3184

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines