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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.31 by root, Thu Apr 19 14:36:46 2012 UTC vs.
Revision 1.80 by root, Sat May 5 16:44:02 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
4 6
7#include "ecb.h"//D
8
9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
10#include "xthread.h"
11#include "schmorp.h"
12
5#ifdef I_DLFCN 13#ifdef I_DLFCN
6 #include <dlfcn.h> 14 #include <dlfcn.h>
7#endif 15#endif
8 16
17// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
18// but then fail to define the api functions unless you ALSO define
19// this. This breaks 100% of the opencl 1.1 apps, for what reason?
20// after all, the functions are deprecated, not removed.
21// in addition, you cannot test for this in any future-proof way.
22// each time a new opencl version comes out, you need to make a new
23// release.
24#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
25
26#ifndef PREFER_1_1
27 #define PREFER_1_1 1
28#endif
29
30#if PREFER_1_1
31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
32#endif
33
9#ifdef __APPLE__ 34#ifdef __APPLE__
10 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
11#else 36#else
12 #include <CL/opencl.h> 37 #define CLHDR(name) <CL/name>
13#endif 38#endif
39
40#include CLHDR(opencl.h)
41
42#ifndef CL_VERSION_1_2
43 #include CLHDR(cl_d3d9.h)
44#endif
45
46#if _WIN32
47 #include CLHDR(cl_d3d10.h)
48 #if CL_VERSION_1_2
49 #include CLHDR<cl_d3d11.h>
50 #endif
51 #include CLHDR<cl_dx9_media_sharing.h.h>
52#endif
53
54#ifndef CL_VERSION_1_2
55 #undef PREFER_1_1
56 #define PREFER_1_1 1
57#endif
58
59// make sure all constants we might use are actually defined
60#include "default.h"
14 61
15typedef cl_platform_id OpenCL__Platform; 62typedef cl_platform_id OpenCL__Platform;
16typedef cl_device_id OpenCL__Device; 63typedef cl_device_id OpenCL__Device;
64typedef cl_device_id OpenCL__SubDevice;
17typedef cl_context OpenCL__Context; 65typedef cl_context OpenCL__Context;
18typedef cl_command_queue OpenCL__Queue; 66typedef cl_command_queue OpenCL__Queue;
19typedef cl_mem OpenCL__Memory; 67typedef cl_mem OpenCL__Memory;
20typedef cl_mem OpenCL__Buffer; 68typedef cl_mem OpenCL__Buffer;
21typedef cl_mem OpenCL__BufferObj; 69typedef cl_mem OpenCL__BufferObj;
22typedef cl_mem OpenCL__Image; 70typedef cl_mem OpenCL__Image;
23typedef cl_mem OpenCL__Image2D;
24typedef cl_mem OpenCL__Image3D;
25typedef cl_mem OpenCL__Memory_ornull; 71typedef cl_mem OpenCL__Memory_ornull;
26typedef cl_mem OpenCL__Buffer_ornull; 72typedef cl_mem OpenCL__Buffer_ornull;
27typedef cl_mem OpenCL__Image_ornull; 73typedef cl_mem OpenCL__Image_ornull;
28typedef cl_mem OpenCL__Image2D_ornull;
29typedef cl_mem OpenCL__Image3D_ornull;
30typedef cl_sampler OpenCL__Sampler; 74typedef cl_sampler OpenCL__Sampler;
31typedef cl_program OpenCL__Program; 75typedef cl_program OpenCL__Program;
32typedef cl_kernel OpenCL__Kernel; 76typedef cl_kernel OpenCL__Kernel;
33typedef cl_event OpenCL__Event; 77typedef cl_event OpenCL__Event;
34typedef cl_event OpenCL__UserEvent; 78typedef cl_event OpenCL__UserEvent;
35 79
36typedef SV *FUTURE; 80typedef struct mapped * OpenCL__Mapped;
81
82static HV
83 *stash_platform,
84 *stash_device,
85 *stash_subdevice,
86 *stash_context,
87 *stash_queue,
88 *stash_program,
89 *stash_kernel,
90 *stash_sampler,
91 *stash_event,
92 *stash_userevent,
93 *stash_memory,
94 *stash_buffer,
95 *stash_bufferobj,
96 *stash_image,
97 *stash_image1d,
98 *stash_image1darray,
99 *stash_image1dbuffer,
100 *stash_image2d,
101 *stash_image2darray,
102 *stash_image3d,
103 *stash_mapped,
104 *stash_mappedbuffer,
105 *stash_mappedimage;
37 106
38/*****************************************************************************/ 107/*****************************************************************************/
39 108
40// name must include a leading underscore 109// name must include a leading underscore
110// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
111// for perl. doh.
41static void * 112static void *
42getsym (const char *name) 113glsym (const char *name)
43{ 114{
115 void *fun = 0;
116
44 #if defined I_DLFCN && defined RTLD_DEFAULT 117 #if defined I_DLFCN && defined RTLD_DEFAULT
45 #if !DLSYM_NEEDS_UNDERSCORE 118 fun = dlsym (RTLD_DEFAULT, name + 1);
46 ++name; // skip _
47 #endif
48 return dlsym (RTLD_DEFAULT, name); 119 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
49 #else 120
50 return 0; 121 if (!fun)
122 {
123 static void *libgl;
124 static const char *glso[] = {
125 "libGL.so.1",
126 "libGL.so.3",
127 "libGL.so.4.0",
128 "libGL.so",
129 "/usr/lib/libGL.so",
130 "/usr/X11R6/lib/libGL.1.dylib"
131 };
132 int i;
133
134 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
135 {
136 libgl = dlopen (glso [i], RTLD_LAZY);
137 if (libgl)
138 break;
139 }
140
141 if (libgl)
142 {
143 fun = dlsym (libgl, name + 1);
144 if (!fun) fun = dlsym (libgl, name);
145 }
146 }
51 #endif 147 #endif
148
149 return fun;
52} 150}
53 151
54/*****************************************************************************/ 152/*****************************************************************************/
55 153
56/* up to two temporary buffers */ 154/* up to two temporary buffers */
57static void * 155static void * ecb_noinline
58tmpbuf (size_t size) 156tmpbuf (size_t size)
59{ 157{
60 enum { buffers = 3 }; 158 enum { buffers = 4 };
61 static int idx; 159 static int idx;
62 static void *buf [buffers]; 160 static void *buf [buffers];
63 static size_t len [buffers]; 161 static size_t len [buffers];
64 162
65 idx = ++idx % buffers; 163 idx = (idx + 1) % buffers;
66 164
67 if (len [idx] < size) 165 if (len [idx] < size)
68 { 166 {
69 free (buf [idx]); 167 free (buf [idx]);
70 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 168 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
72 } 170 }
73 171
74 return buf [idx]; 172 return buf [idx];
75} 173}
76 174
175static const char * ecb_noinline
176cv_get_name (CV *cv)
177{
178 static char fullname [256];
179
180 GV *gv = CvGV (cv); // gv better be nonzero
181
182 HV *stash = GvSTASH (gv);
183 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
184 const char *gvname = GvNAME (gv);
185
186 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
187 return fullname;
188}
189
77/*****************************************************************************/ 190/*****************************************************************************/
78 191
79typedef struct 192typedef struct
80{ 193{
81 IV iv; 194 IV iv;
82 const char *name; 195 const char *name;
83 #define const_iv(name) { (IV)CL_ ## name, # name }, 196 #define const_iv(name) { (IV)CL_ ## name, # name },
84} ivstr; 197} ivstr;
198
199typedef struct
200{
201 NV nv;
202 const char *name;
203 #define const_nv(name) { (NV)CL_ ## name, # name },
204} nvstr;
85 205
86static const char * 206static const char *
87iv2str (IV value, const ivstr *base, int count, const char *fallback) 207iv2str (IV value, const ivstr *base, int count, const char *fallback)
88{ 208{
89 int i; 209 int i;
138 if (res) \ 258 if (res) \
139 FAIL (name); 259 FAIL (name);
140 260
141/*****************************************************************************/ 261/*****************************************************************************/
142 262
143#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
144#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
145
146static void * 263static SV *
264new_clobj (HV *stash, IV id)
265{
266 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
267}
268
269#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
270#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
271
272/* cl objects are either \$iv, or [$iv, ...] */
273/* they can be upgraded at runtime to the array form */
274static void * ecb_noinline
147SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 275SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
148{ 276{
277 // sv_derived_from is quite slow :(
149 if (SvROK (sv) && sv_derived_from (sv, pkg)) 278 if (SvROK (sv) && sv_derived_from (sv, pkg))
150 return (void *)SvIV (SvRV (sv)); 279 return (void *)SvIV (SvRV (sv));
151 280
152 croak ("%s: %s is not of type %s", func, svname, pkg); 281 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
282}
283
284// the "no-inherit" version of the above
285static void * ecb_noinline
286SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
287{
288 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
289 return (void *)SvIV (SvRV (sv));
290
291 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
153} 292}
154 293
155/*****************************************************************************/ 294/*****************************************************************************/
156 295
157static cl_context_properties * 296static cl_context_properties * ecb_noinline
158SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 297SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
159{ 298{
160 if (!sv || !SvOK (sv)) 299 if (!sv || !SvOK (sv))
161 if (extra) 300 if (extra)
162 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 301 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
163 else 302 else
169 int i, len = av_len (av) + 1; 308 int i, len = av_len (av) + 1;
170 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 309 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
171 cl_context_properties *l = p; 310 cl_context_properties *l = p;
172 311
173 if (len & 1) 312 if (len & 1)
174 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 313 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
175 314
176 while (extracount--) 315 while (extracount--)
177 *l++ = *extra++; 316 *l++ = *extra++;
178 317
179 for (i = 0; i < len; i += 2) 318 for (i = 0; i < len; i += 2)
180 { 319 {
181 cl_context_properties t = SvIV (*av_fetch (av, i , 0)); 320 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
182 SV *p_sv = *av_fetch (av, i + 1, 0); 321 SV *p_sv = *av_fetch (av, i + 1, 0);
183 cl_context_properties v; 322 cl_context_properties v = SvIV (p_sv); // code below can override
184 323
185 switch (t) 324 switch (t)
186 { 325 {
326 case CL_CONTEXT_PLATFORM:
327 if (SvROK (p_sv))
328 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
329 break;
330
331 case CL_GLX_DISPLAY_KHR:
332 if (!SvOK (p_sv))
333 {
334 void *func = glsym ("_glXGetCurrentDisplay");
335 if (func)
336 v = (cl_context_properties)((void *(*)(void))func)();
337 }
338 break;
339
340 case CL_GL_CONTEXT_KHR:
341 if (!SvOK (p_sv))
342 {
343 void *func = glsym ("_glXGetCurrentContext");
344 if (func)
345 v = (cl_context_properties)((void *(*)(void))func)();
346 }
347 break;
348
187 default: 349 default:
188 /* unknown property, treat as int */ 350 /* unknown property, treat as int */
189 v = SvIV (p_sv);
190 break; 351 break;
191 } 352 }
192 353
193 *l++ = t; 354 *l++ = t;
194 *l++ = v; 355 *l++ = v;
197 *l = 0; 358 *l = 0;
198 359
199 return p; 360 return p;
200 } 361 }
201 362
202 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 363 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
364}
365
366// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
367// is a pointer (and only on well-behaved systems).
368static void * ecb_noinline
369object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
370{
371 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
372 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
373 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
374
375 AV *av = (AV *)SvRV (arg);
376 void **list = 0;
377 cl_uint count = av_len (av) + 1;
378
379 if (count)
380 {
381 list = tmpbuf (sizeof (*list) * count);
382 int i;
383 for (i = 0; i < count; ++i)
384 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
385 }
386
387 if (!count && !or_undef)
388 croak ("%s: '%s' must contain at least one %s object",
389 cv_get_name (cv), argname, klass);
390
391 *rcount = count;
392 return (void *)list;
203} 393}
204 394
205/*****************************************************************************/ 395/*****************************************************************************/
396/* callback stuff */
397
398/* default context callback, log to stderr */
399static void CL_CALLBACK
400context_default_notify (const char *msg, const void *info, size_t cb, void *data)
401{
402 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
403}
404
405typedef struct
406{
407 int free_cb;
408 void (*push)(void *data1, void *data2, void *data3);
409} eq_vtbl;
410
411typedef struct eq_item
412{
413 struct eq_item *next;
414 eq_vtbl *vtbl;
415 SV *cb;
416 void *data1, *data2, *data3;
417} eq_item;
418
419static void (*eq_signal_func)(void *signal_arg, int value);
420static void *eq_signal_arg;
421static xmutex_t eq_lock = X_MUTEX_INIT;
422static eq_item *eq_head, *eq_tail;
423
424static void ecb_noinline
425eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
426{
427 eq_item *item = malloc (sizeof (eq_item));
428
429 item->next = 0;
430 item->vtbl = vtbl;
431 item->cb = cb;
432 item->data1 = data1;
433 item->data2 = data2;
434 item->data3 = data3;
435
436 X_LOCK (eq_lock);
437
438 *(eq_head ? &eq_tail->next : &eq_head) = item;
439 eq_tail = item;
440
441 X_UNLOCK (eq_lock);
442
443 eq_signal_func (eq_signal_arg, 0);
444}
445
446static eq_item *
447eq_dec (void)
448{
449 eq_item *res;
450
451 X_LOCK (eq_lock);
452
453 res = eq_head;
454 if (res)
455 eq_head = res->next;
456
457 X_UNLOCK (eq_lock);
458
459 return res;
460}
461
462static void
463eq_poll (void)
464{
465 eq_item *item;
466
467 while ((item = eq_dec ()))
468 {
469 ENTER;
470 SAVETMPS;
471
472 dSP;
473 PUSHMARK (SP);
474 EXTEND (SP, 2);
475
476 if (item->vtbl->free_cb)
477 sv_2mortal (item->cb);
478
479 PUTBACK;
480 item->vtbl->push (item->data1, item->data2, item->data3);
481
482 SV *cb = item->cb;
483 free (item);
484
485 call_sv (cb, G_DISCARD | G_VOID);
486
487 FREETMPS;
488 LEAVE;
489 }
490}
491
492static void
493eq_poll_interrupt (pTHX_ void *c_arg, int value)
494{
495 eq_poll ();
496}
497
498/*****************************************************************************/
499/* context notify */
500
501static void ecb_noinline
502eq_context_push (void *data1, void *data2, void *data3)
503{
504 dSP;
505 PUSHs (sv_2mortal (newSVpv (data1, 0)));
506 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
507 PUTBACK;
508
509 free (data1);
510 free (data2);
511}
512
513static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
514
515static void CL_CALLBACK
516eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
517{
518 void *pvt_copy = malloc (cb);
519 memcpy (pvt_copy, pvt, cb);
520 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
521}
522
523#define CONTEXT_NOTIFY_CALLBACK \
524 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
525 void *user_data = 0; \
526 \
527 if (SvOK (notify)) \
528 { \
529 pfn_notify = eq_context_notify; \
530 user_data = s_get_cv (notify); \
531 }
532
533static SV * ecb_noinline
534new_clobj_context (cl_context ctx, void *user_data)
535{
536 SV *sv = new_clobj (stash_context, (IV)ctx);
537
538 if (user_data)
539 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
540
541 return sv;
542}
543
544#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
545
546/*****************************************************************************/
547/* build/compile/link notify */
548
549static void
550eq_program_push (void *data1, void *data2, void *data3)
551{
552 dSP;
553 PUSH_CLOBJ (stash_program, data1);
554 PUTBACK;
555}
556
557static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
558
559static void CL_CALLBACK
560eq_program_notify (cl_program program, void *user_data)
561{
562 clRetainProgram (program);
563
564 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
565}
566
567typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
568
569static program_callback ecb_noinline
570make_program_callback (SV *notify, void **ruser_data)
571{
572 if (SvOK (notify))
573 {
574 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
575 return eq_program_notify;
576 }
577 else
578 {
579 *ruser_data = 0;
580 return 0;
581 }
582}
583
584struct build_args
585{
586 cl_program program;
587 char *options;
588 void *user_data;
589 cl_uint num_devices;
590};
591
592X_THREAD_PROC (build_program_thread)
593{
594 struct build_args *arg = thr_arg;
595
596 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
597
598 if (arg->user_data)
599 eq_program_notify (arg->program, arg->user_data);
600 else
601 clReleaseProgram (arg->program);
602
603 free (arg->options);
604 free (arg);
605
606 return 0;
607}
608
609static void
610build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
611{
612 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
613
614 arg->program = program;
615 arg->options = strdup (options);
616 arg->user_data = user_data;
617 arg->num_devices = num_devices;
618 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
619
620 xthread_t id;
621 thread_create (&id, build_program_thread, arg);
622}
623
624/*****************************************************************************/
625/* mem object destructor notify */
626
627static void ecb_noinline
628eq_destructor_push (void *data1, void *data2, void *data3)
629{
630}
631
632static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
633
634static void CL_CALLBACK
635eq_destructor_notify (cl_mem memobj, void *user_data)
636{
637 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
638}
639
640/*****************************************************************************/
641/* event objects */
642
643static void
644eq_event_push (void *data1, void *data2, void *data3)
645{
646 dSP;
647 PUSH_CLOBJ (stash_event, data1);
648 PUSHs (sv_2mortal (newSViv ((IV)data2)));
649 PUTBACK;
650}
651
652static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
653
654static void CL_CALLBACK
655eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
656{
657 clRetainEvent (event);
658 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
659}
660
661/*****************************************************************************/
662/* utilities for XS code */
206 663
207static size_t 664static size_t
208img_row_pitch (cl_mem img) 665img_row_pitch (cl_mem img)
209{ 666{
210 size_t res; 667 size_t res;
211 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 668 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
212 return res; 669 return res;
213} 670}
214 671
215static cl_event * 672static cl_event * ecb_noinline
216event_list (SV **items, int count) 673event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
217{ 674{
218 if (!count) 675 cl_uint count = *rcount;
676
677 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
678 *rcount = count = 0;
679
680 if (!count && !extra)
219 return 0; 681 return 0;
220 682
221 cl_event *list = tmpbuf (sizeof (cl_event) * count); 683 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
684 int i = 0;
222 685
223 while (count--) 686 while (count--)
687 if (SvOK (items [count]))
224 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 688 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
225 689
690 if (extra)
691 list [i++] = extra;
692
693 *rcount = i;
694
226 return list; 695 return i ? list : 0;
227} 696}
228 697
229#define EVENT_LIST(items,count) \ 698#define EVENT_LIST(skip) \
230 cl_uint event_list_count = (count); \ 699 cl_uint event_list_count = items - (skip); \
231 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 700 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
232 701
233#define INFO(class) \ 702#define INFO(class) \
234{ \ 703{ \
235 size_t size; \ 704 size_t size; \
236 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 705 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
240 SvCUR_set (sv, size); \ 709 SvCUR_set (sv, size); \
241 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 710 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
242 XPUSHs (sv); \ 711 XPUSHs (sv); \
243} 712}
244 713
714/*****************************************************************************/
715/* mapped_xxx */
716
717static OpenCL__Mapped
718SvMAPPED (SV *sv)
719{
720 // no typechecking atm., keep your fingers crossed
721 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
722}
723
724struct mapped
725{
726 cl_command_queue queue;
727 cl_mem memobj;
728 void *ptr;
729 size_t cb;
730 cl_event event;
731 size_t row_pitch;
732 size_t slice_pitch;
733
734 size_t element_size;
735 size_t width, height, depth;
736};
737
738static SV *
739mapped_new (
740 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
741 void *ptr, size_t cb, cl_event ev,
742 size_t row_pitch, size_t slice_pitch, size_t element_size,
743 size_t width, size_t height, size_t depth
744)
745{
746 SV *data = newSV (0);
747 SvUPGRADE (data, SVt_PVMG);
748
749 OpenCL__Mapped mapped;
750 New (0, mapped, 1, struct mapped);
751
752 clRetainCommandQueue (queue);
753
754 mapped->queue = queue;
755 mapped->memobj = memobj;
756 mapped->ptr = ptr;
757 mapped->cb = cb;
758 mapped->event = ev;
759 mapped->row_pitch = row_pitch;
760 mapped->slice_pitch = slice_pitch;
761
762 mapped->element_size = element_size;
763 mapped->width = width;
764 mapped->height = height;
765 mapped->depth = depth;
766
767 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
768
769 if (SvLEN (data))
770 Safefree (data);
771
772 SvPVX (data) = (char *)ptr;
773 SvCUR_set (data, cb);
774 SvLEN_set (data, 0);
775 SvPOK_only (data);
776
777 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
778
779 if (!(flags & CL_MAP_WRITE))
780 SvREADONLY_on (data);
781
782 return obj;
783}
784
785static void
786mapped_detach (SV *sv, OpenCL__Mapped mapped)
787{
788 SV *data = SvRV (sv);
789
790 // the next check checks both before AND after detach, where SvPVX should STILL be 0
791 if (SvPVX (data) != (char *)mapped->ptr)
792 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
793 else
794 {
795 SvREADONLY_off (data);
796 SvCUR_set (data, 0);
797 SvPVX (data) = 0;
798 SvOK_off (data);
799 }
800
801 mapped->ptr = 0;
802}
803
804static void
805mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
806{
807 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
808 cl_event ev;
809
810 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
811
812 clReleaseEvent (mapped->event);
813 mapped->event = ev;
814
815 mapped_detach (self, mapped);
816}
817
818static size_t
819mapped_element_size (OpenCL__Mapped self)
820{
821 if (!self->element_size)
822 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
823
824 return self->element_size;
825}
826
827/*****************************************************************************/
828
245MODULE = OpenCL PACKAGE = OpenCL 829MODULE = OpenCL PACKAGE = OpenCL
246 830
247PROTOTYPES: ENABLE 831PROTOTYPES: ENABLE
248 832
833void
834poll ()
835 CODE:
836 eq_poll ();
837
838void
839_eq_initialise (IV func, IV arg)
840 CODE:
841 eq_signal_func = (void (*)(void *, int))func;
842 eq_signal_arg = (void*)arg;
843
249BOOT: 844BOOT:
250{ 845{
251 HV *stash = gv_stashpv ("OpenCL", 1); 846 HV *stash = gv_stashpv ("OpenCL", 1);
847
252 static const ivstr *civ, const_iv[] = { 848 static const ivstr *civ, const_iv[] = {
253 { sizeof (cl_char ), "SIZEOF_CHAR" }, 849 { sizeof (cl_char ), "SIZEOF_CHAR" },
254 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 850 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
255 { sizeof (cl_short ), "SIZEOF_SHORT" }, 851 { sizeof (cl_short ), "SIZEOF_SHORT" },
256 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 852 { sizeof (cl_ushort), "SIZEOF_USHORT" },
259 { sizeof (cl_long ), "SIZEOF_LONG" }, 855 { sizeof (cl_long ), "SIZEOF_LONG" },
260 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 856 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
261 { sizeof (cl_half ), "SIZEOF_HALF" }, 857 { sizeof (cl_half ), "SIZEOF_HALF" },
262 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 858 { sizeof (cl_float ), "SIZEOF_FLOAT" },
263 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 859 { sizeof (cl_double), "SIZEOF_DOUBLE" },
860 { PREFER_1_1 , "PREFER_1_1" },
264#include "constiv.h" 861#include "constiv.h"
265 }; 862 };
863
266 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 864 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
267 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 865 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
866
867 static const nvstr *cnv, const_nv[] = {
868#include "constnv.h"
869 };
870
871 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
872 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
873
874 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
875
876 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
877 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
878 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
879 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
880 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
881 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
882 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
883 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
884 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
885 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
886 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
887 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
888 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
889 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
890 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
891 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
892 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
893 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
894 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
895 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
896 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
897 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
898 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
899
900 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
268} 901}
269 902
270cl_int 903cl_int
271errno () 904errno ()
272 CODE: 905 CODE:
273 errno = res; 906 RETVAL = res;
907 OUTPUT:
908 RETVAL
274 909
275const char * 910const char *
276err2str (cl_int err) 911err2str (cl_int err = res)
277 912
278const char * 913const char *
279enum2str (cl_uint value) 914enum2str (cl_uint value)
280 915
281void 916void
289 list = tmpbuf (sizeof (*list) * count); 924 list = tmpbuf (sizeof (*list) * count);
290 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 925 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
291 926
292 EXTEND (SP, count); 927 EXTEND (SP, count);
293 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
294 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 929 PUSH_CLOBJ (stash_platform, list [i]);
295 930
296void 931void
297context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 932context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
298 PPCODE: 933 PPCODE:
934 CONTEXT_NOTIFY_CALLBACK;
299 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 935 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
300 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 936 XPUSH_CLOBJ_CONTEXT;
301 937
302void 938void
303context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 939context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
304 PPCODE: 940 PPCODE:
305 /* der Gipfel der Kunst */ 941 cl_uint device_count;
942 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
943
944 CONTEXT_NOTIFY_CALLBACK;
945 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
946 XPUSH_CLOBJ_CONTEXT;
306 947
307void 948void
308wait_for_events (...) 949wait_for_events (...)
309 CODE: 950 CODE:
310 EVENT_LIST (0, items); 951 EVENT_LIST (0);
311 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 952 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
312 953
313PROTOTYPES: DISABLE 954PROTOTYPES: DISABLE
314 955
315MODULE = OpenCL PACKAGE = OpenCL::Platform 956MODULE = OpenCL PACKAGE = OpenCL::Platform
316 957
317void 958void
318info (OpenCL::Platform self, cl_platform_info name) 959info (OpenCL::Platform self, cl_platform_info name)
319 PPCODE: 960 PPCODE:
320 INFO (Platform) 961 INFO (Platform)
962
963void
964unload_compiler (OpenCL::Platform self)
965 CODE:
966#if CL_VERSION_1_2
967 clUnloadPlatformCompiler (self);
968#endif
321 969
322#BEGIN:platform 970#BEGIN:platform
323 971
324void 972void
325profile (OpenCL::Platform self) 973profile (OpenCL::Platform self)
351 list = tmpbuf (sizeof (*list) * count); 999 list = tmpbuf (sizeof (*list) * count);
352 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 1000 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
353 1001
354 EXTEND (SP, count); 1002 EXTEND (SP, count);
355 for (i = 0; i < count; ++i) 1003 for (i = 0; i < count; ++i)
356 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1004 PUSH_CLOBJ (stash_device, list [i]);
357 1005
358void 1006void
359context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 1007context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
360 PPCODE: 1008 PPCODE:
1009 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1010 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1011
1012 cl_uint device_count;
1013 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1014
1015 CONTEXT_NOTIFY_CALLBACK;
1016 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1017 XPUSH_CLOBJ_CONTEXT;
1018
1019void
1020context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1021 PPCODE:
1022 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1023 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1024
1025 CONTEXT_NOTIFY_CALLBACK;
1026 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1027 XPUSH_CLOBJ_CONTEXT;
1028
1029MODULE = OpenCL PACKAGE = OpenCL::Device
1030
1031void
1032info (OpenCL::Device self, cl_device_info name)
1033 PPCODE:
1034 INFO (Device)
1035
1036#if CL_VERSION_1_2
1037
1038void
1039sub_devices (OpenCL::Device self, SV *properties)
1040 PPCODE:
361 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1041 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
362 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1042 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
363 1043
364 AV *av = (AV *)SvRV (devices); 1044 properties = SvRV (properties);
365 cl_uint num_devices = av_len (av) + 1; 1045
366 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1046 cl_uint count = av_len ((AV *)properties) + 1;
1047 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1048
367 int i; 1049 int i;
1050 for (i = 0; i < count; ++i)
1051 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
368 1052
369 for (i = num_devices; i--; ) 1053 props [count] = 0;
370 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
371 1054
372 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 1055 cl_uint num_devices;
373 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1056 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1057 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1058 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
374 1059
375void 1060 EXTEND (SP, num_devices);
376context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1061 for (i = 0; i < count; ++i)
377 PPCODE: 1062 PUSH_CLOBJ (stash_subdevice, list [i]);
378 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
379 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
380 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
381 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
382 1063
383MODULE = OpenCL PACKAGE = OpenCL::Device 1064#endif
384
385void
386info (OpenCL::Device self, cl_device_info name)
387 PPCODE:
388 INFO (Device)
389 1065
390#BEGIN:device 1066#BEGIN:device
391 1067
392void 1068void
393type (OpenCL::Device self) 1069type (OpenCL::Device self)
1070 ALIAS:
1071 type = CL_DEVICE_TYPE
1072 address_bits = CL_DEVICE_ADDRESS_BITS
1073 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1074 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1075 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1076 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1077 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1078 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1079 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1080 properties = CL_DEVICE_QUEUE_PROPERTIES
1081 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1082 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
394 PPCODE: 1083 PPCODE:
395 cl_device_type value [1]; 1084 cl_ulong value [1];
396 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1085 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
397 EXTEND (SP, 1); 1086 EXTEND (SP, 1);
398 const int i = 0; 1087 const int i = 0;
399 PUSHs (sv_2mortal (newSViv (value [i]))); 1088 PUSHs (sv_2mortal (newSVuv (value [i])));
400 1089
401void 1090void
402vendor_id (OpenCL::Device self) 1091vendor_id (OpenCL::Device self)
403 ALIAS: 1092 ALIAS:
404 vendor_id = CL_DEVICE_VENDOR_ID 1093 vendor_id = CL_DEVICE_VENDOR_ID
415 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1104 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
416 image_support = CL_DEVICE_IMAGE_SUPPORT 1105 image_support = CL_DEVICE_IMAGE_SUPPORT
417 max_samplers = CL_DEVICE_MAX_SAMPLERS 1106 max_samplers = CL_DEVICE_MAX_SAMPLERS
418 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1107 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
419 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1108 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1109 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
420 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1110 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
421 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1111 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1112 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
422 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1113 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
423 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1114 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
424 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1115 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
425 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1116 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
426 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1117 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
427 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1118 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
428 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1119 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
429 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1120 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
430 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1121 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
431 PPCODE: 1122 PPCODE:
432 cl_uint value [1]; 1123 cl_uint value [1];
433 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1124 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
434 EXTEND (SP, 1); 1125 EXTEND (SP, 1);
435 const int i = 0; 1126 const int i = 0;
464 EXTEND (SP, n); 1155 EXTEND (SP, n);
465 for (i = 0; i < n; ++i) 1156 for (i = 0; i < n; ++i)
466 PUSHs (sv_2mortal (newSVuv (value [i]))); 1157 PUSHs (sv_2mortal (newSVuv (value [i])));
467 1158
468void 1159void
469address_bits (OpenCL::Device self)
470 PPCODE:
471 cl_bitfield value [1];
472 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
473 EXTEND (SP, 1);
474 const int i = 0;
475 PUSHs (sv_2mortal (newSVuv (value [i])));
476
477void
478max_mem_alloc_size (OpenCL::Device self)
479 ALIAS:
480 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
481 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
482 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
483 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
484 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
485 PPCODE:
486 cl_ulong value [1];
487 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
488 EXTEND (SP, 1);
489 const int i = 0;
490 PUSHs (sv_2mortal (newSVuv (value [i])));
491
492void
493single_fp_config (OpenCL::Device self)
494 ALIAS:
495 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
496 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
497 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
498 PPCODE:
499 cl_device_fp_config value [1];
500 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
501 EXTEND (SP, 1);
502 const int i = 0;
503 PUSHs (sv_2mortal (newSVuv (value [i])));
504
505void
506global_mem_cache_type (OpenCL::Device self)
507 PPCODE:
508 cl_device_mem_cache_type value [1];
509 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
510 EXTEND (SP, 1);
511 const int i = 0;
512 PUSHs (sv_2mortal (newSVuv (value [i])));
513
514void
515local_mem_type (OpenCL::Device self)
516 PPCODE:
517 cl_device_local_mem_type value [1];
518 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
519 EXTEND (SP, 1);
520 const int i = 0;
521 PUSHs (sv_2mortal (newSVuv (value [i])));
522
523void
524error_correction_support (OpenCL::Device self) 1160error_correction_support (OpenCL::Device self)
525 ALIAS: 1161 ALIAS:
526 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1162 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
527 endian_little = CL_DEVICE_ENDIAN_LITTLE 1163 endian_little = CL_DEVICE_ENDIAN_LITTLE
528 available = CL_DEVICE_AVAILABLE 1164 available = CL_DEVICE_AVAILABLE
529 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1165 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
530 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1166 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
531 PPCODE: 1167 PPCODE:
532 cl_bool value [1]; 1168 cl_uint value [1];
533 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1169 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
534 EXTEND (SP, 1); 1170 EXTEND (SP, 1);
535 const int i = 0; 1171 const int i = 0;
536 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1172 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
537
538void
539execution_capabilities (OpenCL::Device self)
540 PPCODE:
541 cl_device_exec_capabilities value [1];
542 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
543 EXTEND (SP, 1);
544 const int i = 0;
545 PUSHs (sv_2mortal (newSVuv (value [i])));
546
547void
548properties (OpenCL::Device self)
549 PPCODE:
550 cl_command_queue_properties value [1];
551 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
552 EXTEND (SP, 1);
553 const int i = 0;
554 PUSHs (sv_2mortal (newSViv (value [i])));
555 1173
556void 1174void
557platform (OpenCL::Device self) 1175platform (OpenCL::Device self)
558 PPCODE: 1176 PPCODE:
559 cl_platform_id value [1]; 1177 cl_platform_id value [1];
560 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1178 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
561 EXTEND (SP, 1); 1179 EXTEND (SP, 1);
562 const int i = 0; 1180 const int i = 0;
563 { 1181 PUSH_CLOBJ (stash_platform, value [i]);
564 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i]));
565 }
566 1182
567void 1183void
568name (OpenCL::Device self) 1184name (OpenCL::Device self)
569 ALIAS: 1185 ALIAS:
570 name = CL_DEVICE_NAME 1186 name = CL_DEVICE_NAME
587 PPCODE: 1203 PPCODE:
588 cl_device_id value [1]; 1204 cl_device_id value [1];
589 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1205 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
590 EXTEND (SP, 1); 1206 EXTEND (SP, 1);
591 const int i = 0; 1207 const int i = 0;
592 { 1208 PUSH_CLOBJ (stash_device, value [i]);
593 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
594 }
595 1209
596void 1210void
597partition_types_ext (OpenCL::Device self) 1211partition_types_ext (OpenCL::Device self)
598 ALIAS: 1212 ALIAS:
599 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1213 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
609 for (i = 0; i < n; ++i) 1223 for (i = 0; i < n; ++i)
610 PUSHs (sv_2mortal (newSVuv (value [i]))); 1224 PUSHs (sv_2mortal (newSVuv (value [i])));
611 1225
612#END:device 1226#END:device
613 1227
1228MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1229
1230#if CL_VERSION_1_2
1231
1232void
1233DESTROY (OpenCL::SubDevice self)
1234 CODE:
1235 clReleaseDevice (self);
1236
1237#endif
1238
614MODULE = OpenCL PACKAGE = OpenCL::Context 1239MODULE = OpenCL PACKAGE = OpenCL::Context
615 1240
616void 1241void
617DESTROY (OpenCL::Context context) 1242DESTROY (OpenCL::Context self)
618 CODE: 1243 CODE:
619 clReleaseContext (context); 1244 clReleaseContext (self);
620 1245
621void 1246void
622info (OpenCL::Context self, cl_context_info name) 1247info (OpenCL::Context self, cl_context_info name)
623 PPCODE: 1248 PPCODE:
624 INFO (Context) 1249 INFO (Context)
625 1250
626void 1251void
627queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1252queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
628 PPCODE: 1253 PPCODE:
629 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1254 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
630 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1255 XPUSH_CLOBJ (stash_queue, queue);
631 1256
632void 1257void
633user_event (OpenCL::Context self) 1258user_event (OpenCL::Context self)
634 PPCODE: 1259 PPCODE:
635 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1260 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
636 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1261 XPUSH_CLOBJ (stash_userevent, ev);
637 1262
638void 1263void
639buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1264buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
640 PPCODE: 1265 PPCODE:
641 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1266 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
642 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1267 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
643 1268
644 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1269 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
645 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1270 XPUSH_CLOBJ (stash_bufferobj, mem);
646 1271
647void 1272void
648buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1273buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
649 PPCODE: 1274 PPCODE:
650 STRLEN len; 1275 STRLEN len;
651 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1276 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
652 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1277 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
653 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1278 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
654 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1279 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
655 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1280 XPUSH_CLOBJ (stash_bufferobj, mem);
656 1281
1282#if CL_VERSION_1_2
1283
657void 1284void
658image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1285image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
659 PPCODE: 1286 PPCODE:
660 STRLEN len; 1287 STRLEN len;
661 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1288 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
662 const cl_image_format format = { channel_order, channel_type }; 1289 const cl_image_format format = { channel_order, channel_type };
1290 const cl_image_desc desc = {
1291 type,
1292 width, height, depth,
1293 array_size, row_pitch, slice_pitch,
1294 num_mip_level, num_samples,
1295 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1296 };
663 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1297 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
664 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1298 HV *stash = stash_image;
1299 switch (type)
1300 {
1301 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1302 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1303 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1304 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1305 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1306 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1307 }
1308 XPUSH_CLOBJ (stash, mem);
665 1309
1310#endif
1311
666void 1312void
667image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1313image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
668 PPCODE: 1314 PPCODE:
669 STRLEN len; 1315 STRLEN len;
670 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1316 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
671 const cl_image_format format = { channel_order, channel_type }; 1317 const cl_image_format format = { channel_order, channel_type };
1318#if PREFER_1_1
1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1320#else
1321 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1322 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1323#endif
1324 XPUSH_CLOBJ (stash_image2d, mem);
1325
1326void
1327image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1328 PPCODE:
1329 STRLEN len;
1330 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1331 const cl_image_format format = { channel_order, channel_type };
1332#if PREFER_1_1
672 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1333 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
673 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1334#else
1335 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1336 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1337#endif
1338 XPUSH_CLOBJ (stash_image3d, mem);
674 1339
675#if cl_apple_gl_sharing || cl_khr_gl_sharing 1340#if cl_apple_gl_sharing || cl_khr_gl_sharing
676 1341
677void 1342void
678gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1343gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
679 PPCODE: 1344 PPCODE:
680 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1345 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
681 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1346 XPUSH_CLOBJ (stash_bufferobj, mem);
1347
1348void
1349gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1350 PPCODE:
1351 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1352 XPUSH_CLOBJ (stash_image2d, mem);
1353
1354#if CL_VERSION_1_2
1355
1356void
1357gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1358 ALIAS:
1359 PPCODE:
1360 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1361 cl_gl_object_type type;
1362 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1363 HV *stash = stash_memory;
1364 switch (type)
1365 {
1366 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1367 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1368 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1369 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1370 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1371 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1372 }
1373 XPUSH_CLOBJ (stash, mem);
1374
1375#endif
682 1376
683void 1377void
684gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1378gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
685 PPCODE: 1379 PPCODE:
1380#if PREFER_1_1
686 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1381 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
687 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1382#else
1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1384#endif
1385 XPUSH_CLOBJ (stash_image2d, mem);
688 1386
689void 1387void
690gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1388gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
691 PPCODE: 1389 PPCODE:
1390#if PREFER_1_1
692 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1391 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
693 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1392#else
694 1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
695void 1394#endif
696gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1395 XPUSH_CLOBJ (stash_image3d, mem);
697 PPCODE:
698 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
699 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
700 1396
701#endif 1397#endif
702 1398
703void 1399void
704supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1400supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
724 1420
725void 1421void
726sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1422sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
727 PPCODE: 1423 PPCODE:
728 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1424 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
729 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1425 XPUSH_CLOBJ (stash_sampler, sampler);
730 1426
731void 1427void
732program_with_source (OpenCL::Context self, SV *program) 1428program_with_source (OpenCL::Context self, SV *program)
733 PPCODE: 1429 PPCODE:
734 STRLEN len; 1430 STRLEN len;
735 size_t len2; 1431 size_t len2;
736 const char *ptr = SvPVbyte (program, len); 1432 const char *ptr = SvPVbyte (program, len);
737 1433
738 len2 = len; 1434 len2 = len;
739 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1435 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
740 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1436 XPUSH_CLOBJ (stash_program, prog);
1437
1438void
1439program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1440 PPCODE:
1441 cl_uint device_count;
1442 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1443
1444 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1445 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1446
1447 binaries = SvRV (binaries);
1448
1449 if (device_count != av_len ((AV *)binaries) + 1)
1450 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1451
1452 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1453 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1454 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1455
1456 int i;
1457 for (i = 0; i < device_count; ++i)
1458 {
1459 STRLEN len;
1460 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1461 length_list [i] = len;
1462 }
1463
1464 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1465 length_list, binary_list,
1466 GIMME_V == G_ARRAY ? status_list : 0, &res));
1467
1468 EXTEND (SP, 2);
1469 PUSH_CLOBJ (stash_program, prog);
1470
1471 if (GIMME_V == G_ARRAY)
1472 {
1473 AV *av = newAV ();
1474 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1475
1476 for (i = device_count; i--; )
1477 av_store (av, i, newSViv (status_list [i]));
1478 }
1479
1480#if CL_VERSION_1_2
1481
1482void
1483program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1484 PPCODE:
1485 cl_uint device_count;
1486 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1489
1490 XPUSH_CLOBJ (stash_program, prog);
1491
1492void
1493link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1494 CODE:
1495 cl_uint device_count = 0;
1496 cl_device_id *device_list = 0;
1497
1498 if (SvOK (devices))
1499 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1500
1501 cl_uint program_count;
1502 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1503
1504 void *user_data;
1505 program_callback pfn_notify = make_program_callback (notify, &user_data);
1506
1507 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1508 program_count, program_list, pfn_notify, user_data, &res));
1509
1510 XPUSH_CLOBJ (stash_program, prog);
1511
1512#endif
741 1513
742#BEGIN:context 1514#BEGIN:context
743 1515
744void 1516void
745reference_count (OpenCL::Context self) 1517reference_count (OpenCL::Context self)
761 cl_device_id *value = tmpbuf (size); 1533 cl_device_id *value = tmpbuf (size);
762 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1534 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
763 int i, n = size / sizeof (*value); 1535 int i, n = size / sizeof (*value);
764 EXTEND (SP, n); 1536 EXTEND (SP, n);
765 for (i = 0; i < n; ++i) 1537 for (i = 0; i < n; ++i)
766 { 1538 PUSH_CLOBJ (stash_device, value [i]);
767 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
768 }
769 1539
770void 1540void
771properties (OpenCL::Context self) 1541properties (OpenCL::Context self)
772 PPCODE: 1542 PPCODE:
773 size_t size; 1543 size_t size;
787DESTROY (OpenCL::Queue self) 1557DESTROY (OpenCL::Queue self)
788 CODE: 1558 CODE:
789 clReleaseCommandQueue (self); 1559 clReleaseCommandQueue (self);
790 1560
791void 1561void
792enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1562read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1563 ALIAS:
1564 enqueue_read_buffer = 0
793 PPCODE: 1565 PPCODE:
794 cl_event ev = 0;
795 EVENT_LIST (6, items - 6); 1566 EVENT_LIST (6);
796 1567
797 SvUPGRADE (data, SVt_PV); 1568 SvUPGRADE (data, SVt_PV);
798 SvGROW (data, len); 1569 SvGROW (data, len);
799 SvPOK_only (data); 1570 SvPOK_only (data);
800 SvCUR_set (data, len); 1571 SvCUR_set (data, len);
1572
1573 cl_event ev = 0;
801 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1574 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
802 1575
803 if (ev) 1576 if (ev)
804 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1577 XPUSH_CLOBJ (stash_event, ev);
805 1578
806void 1579void
807enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1580write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1581 ALIAS:
1582 enqueue_write_buffer = 0
808 PPCODE: 1583 PPCODE:
809 cl_event ev = 0; 1584 EVENT_LIST (5);
1585
810 STRLEN len; 1586 STRLEN len;
811 char *ptr = SvPVbyte (data, len); 1587 char *ptr = SvPVbyte (data, len);
812 EVENT_LIST (5, items - 5);
813 1588
1589 cl_event ev = 0;
814 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1590 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
815 1591
816 if (ev) 1592 if (ev)
817 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1593 XPUSH_CLOBJ (stash_event, ev);
818 1594
1595#if CL_VERSION_1_2
1596
819void 1597void
820enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1598fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1599 ALIAS:
1600 enqueue_fill_buffer = 0
821 PPCODE: 1601 PPCODE:
1602 EVENT_LIST (5);
1603
1604 STRLEN len;
1605 char *ptr = SvPVbyte (data, len);
1606
822 cl_event ev = 0; 1607 cl_event ev = 0;
1608 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1609
1610 if (ev)
1611 XPUSH_CLOBJ (stash_event, ev);
1612
1613void
1614fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1615 ALIAS:
1616 enqueue_fill_image = 0
1617 PPCODE:
1618 EVENT_LIST (12);
1619
1620 const size_t origin [3] = { x, y, z };
1621 const size_t region [3] = { width, height, depth };
1622
1623 const cl_float c_f [4] = { r, g, b, a };
1624 const cl_uint c_u [4] = { r, g, b, a };
1625 const cl_int c_s [4] = { r, g, b, a };
1626 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1627 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1628 cl_image_format format;
1629 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1630 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1631 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1632 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1633
1634 cl_event ev = 0;
1635 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1636 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1637
1638 if (ev)
1639 XPUSH_CLOBJ (stash_event, ev);
1640
1641#endif
1642
1643void
1644copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1645 ALIAS:
1646 enqueue_copy_buffer = 0
1647 PPCODE:
823 EVENT_LIST (6, items - 6); 1648 EVENT_LIST (6);
824 1649
1650 cl_event ev = 0;
825 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1651 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
826 1652
827 if (ev) 1653 if (ev)
828 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1654 XPUSH_CLOBJ (stash_event, ev);
829 1655
830void 1656void
831enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1657read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1658 ALIAS:
1659 enqueue_read_buffer_rect = 0
832 PPCODE: 1660 PPCODE:
833 cl_event ev = 0; 1661 EVENT_LIST (17);
1662
834 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1663 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
835 const size_t host_origin[3] = { host_x, host_y, host_z }; 1664 const size_t host_origin[3] = { host_x, host_y, host_z };
836 const size_t region[3] = { width, height, depth }; 1665 const size_t region[3] = { width, height, depth };
837 EVENT_LIST (17, items - 17);
838 1666
839 if (!buf_row_pitch) 1667 if (!buf_row_pitch)
840 buf_row_pitch = region [0]; 1668 buf_row_pitch = region [0];
841 1669
842 if (!buf_slice_pitch) 1670 if (!buf_slice_pitch)
852 1680
853 SvUPGRADE (data, SVt_PV); 1681 SvUPGRADE (data, SVt_PV);
854 SvGROW (data, len); 1682 SvGROW (data, len);
855 SvPOK_only (data); 1683 SvPOK_only (data);
856 SvCUR_set (data, len); 1684 SvCUR_set (data, len);
1685
1686 cl_event ev = 0;
857 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1687 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
858 1688
859 if (ev) 1689 if (ev)
860 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1690 XPUSH_CLOBJ (stash_event, ev);
861 1691
862void 1692void
863enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1693write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1694 ALIAS:
1695 enqueue_write_buffer_rect = 0
864 PPCODE: 1696 PPCODE:
865 cl_event ev = 0; 1697 EVENT_LIST (17);
1698
866 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1699 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
867 const size_t host_origin[3] = { host_x, host_y, host_z }; 1700 const size_t host_origin[3] = { host_x, host_y, host_z };
868 const size_t region[3] = { width, height, depth }; 1701 const size_t region[3] = { width, height, depth };
869 STRLEN len; 1702 STRLEN len;
870 char *ptr = SvPVbyte (data, len); 1703 char *ptr = SvPVbyte (data, len);
871 EVENT_LIST (17, items - 17);
872 1704
873 if (!buf_row_pitch) 1705 if (!buf_row_pitch)
874 buf_row_pitch = region [0]; 1706 buf_row_pitch = region [0];
875 1707
876 if (!buf_slice_pitch) 1708 if (!buf_slice_pitch)
885 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1717 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
886 1718
887 if (len < min_len) 1719 if (len < min_len)
888 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1720 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
889 1721
1722 cl_event ev = 0;
890 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1723 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
891 1724
892 if (ev) 1725 if (ev)
893 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1726 XPUSH_CLOBJ (stash_event, ev);
894 1727
895void 1728void
896enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1729copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1730 ALIAS:
1731 enqueue_copy_buffer_rect = 0
897 PPCODE: 1732 PPCODE:
898 cl_event ev = 0; 1733 EVENT_LIST (16);
1734
899 const size_t src_origin[3] = { src_x, src_y, src_z }; 1735 const size_t src_origin[3] = { src_x, src_y, src_z };
900 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1736 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
901 const size_t region[3] = { width, height, depth }; 1737 const size_t region[3] = { width, height, depth };
902 EVENT_LIST (16, items - 16);
903 1738
1739 cl_event ev = 0;
904 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1740 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
905 1741
906 if (ev) 1742 if (ev)
907 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1743 XPUSH_CLOBJ (stash_event, ev);
908 1744
909void 1745void
910enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1746read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1747 ALIAS:
1748 enqueue_read_image = 0
911 PPCODE: 1749 PPCODE:
912 cl_event ev = 0; 1750 EVENT_LIST (12);
1751
913 const size_t src_origin[3] = { src_x, src_y, src_z }; 1752 const size_t src_origin[3] = { src_x, src_y, src_z };
914 const size_t region[3] = { width, height, depth }; 1753 const size_t region[3] = { width, height, depth };
915 EVENT_LIST (12, items - 12);
916 1754
917 if (!row_pitch) 1755 if (!row_pitch)
918 row_pitch = img_row_pitch (src); 1756 row_pitch = img_row_pitch (src);
919 1757
920 if (depth > 1 && !slice_pitch) 1758 if (depth > 1 && !slice_pitch)
924 1762
925 SvUPGRADE (data, SVt_PV); 1763 SvUPGRADE (data, SVt_PV);
926 SvGROW (data, len); 1764 SvGROW (data, len);
927 SvPOK_only (data); 1765 SvPOK_only (data);
928 SvCUR_set (data, len); 1766 SvCUR_set (data, len);
1767
1768 cl_event ev = 0;
929 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1769 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
930 1770
931 if (ev) 1771 if (ev)
932 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1772 XPUSH_CLOBJ (stash_event, ev);
933 1773
934void 1774void
935enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1775write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1776 ALIAS:
1777 enqueue_write_image = 0
936 PPCODE: 1778 PPCODE:
937 cl_event ev = 0; 1779 EVENT_LIST (12);
1780
938 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1781 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
939 const size_t region[3] = { width, height, depth }; 1782 const size_t region[3] = { width, height, depth };
940 STRLEN len; 1783 STRLEN len;
941 char *ptr = SvPVbyte (data, len); 1784 char *ptr = SvPVbyte (data, len);
942 EVENT_LIST (12, items - 12);
943 1785
944 if (!row_pitch) 1786 if (!row_pitch)
945 row_pitch = img_row_pitch (dst); 1787 row_pitch = img_row_pitch (dst);
946 1788
947 if (depth > 1 && !slice_pitch) 1789 if (depth > 1 && !slice_pitch)
950 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1792 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
951 1793
952 if (len < min_len) 1794 if (len < min_len)
953 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1795 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
954 1796
1797 cl_event ev = 0;
955 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1798 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
956 1799
957 if (ev) 1800 if (ev)
958 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1801 XPUSH_CLOBJ (stash_event, ev);
959 1802
960void 1803void
961enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1804copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1805 ALIAS:
1806 enqueue_copy_image = 0
962 PPCODE: 1807 PPCODE:
963 cl_event ev = 0; 1808 EVENT_LIST (12);
1809
964 const size_t src_origin[3] = { src_x, src_y, src_z }; 1810 const size_t src_origin[3] = { src_x, src_y, src_z };
965 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1811 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
966 const size_t region[3] = { width, height, depth }; 1812 const size_t region[3] = { width, height, depth };
967 EVENT_LIST (12, items - 12);
968 1813
1814 cl_event ev = 0;
969 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1815 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
970 1816
971 if (ev) 1817 if (ev)
972 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1818 XPUSH_CLOBJ (stash_event, ev);
973 1819
974void 1820void
975enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1821copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1822 ALIAS:
1823 enqueue_copy_image_to_buffer = 0
976 PPCODE: 1824 PPCODE:
1825 EVENT_LIST (10);
1826
1827 const size_t src_origin[3] = { src_x, src_y, src_z };
1828 const size_t region [3] = { width, height, depth };
1829
977 cl_event ev = 0; 1830 cl_event ev = 0;
1831 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1832
1833 if (ev)
1834 XPUSH_CLOBJ (stash_event, ev);
1835
1836void
1837copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1838 ALIAS:
1839 enqueue_copy_buffer_to_image = 0
1840 PPCODE:
1841 EVENT_LIST (10);
1842
978 const size_t src_origin[3] = { src_x, src_y, src_z }; 1843 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1844 const size_t region [3] = { width, height, depth };
1845
1846 cl_event ev = 0;
1847 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1848
1849 if (ev)
1850 XPUSH_CLOBJ (stash_event, ev);
1851
1852void
1853map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1854 ALIAS:
1855 enqueue_map_buffer = 0
1856 PPCODE:
1857 EVENT_LIST (6);
1858
1859 size_t cb = SvIV (cb_);
1860
1861 if (!SvOK (cb_))
1862 {
1863 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1864 cb -= offset;
1865 }
1866
1867 cl_event ev;
1868 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1869 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1870
1871void
1872map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1873 ALIAS:
1874 enqueue_map_image = 0
1875 PPCODE:
1876 size_t width = SvIV (width_);
1877 if (!SvOK (width_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1880 width -= x;
1881 }
1882
1883 size_t height = SvIV (width_);
1884 if (!SvOK (height_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1887 height -= y;
1888
1889 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1890 if (!height)
1891 height = 1;
1892 }
1893
1894 size_t depth = SvIV (width_);
1895 if (!SvOK (depth_))
1896 {
1897 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1898 depth -= z;
1899
1900 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1901 if (!depth)
1902 depth = 1;
1903 }
1904
1905 const size_t origin[3] = { x, y, z };
979 const size_t region[3] = { width, height, depth }; 1906 const size_t region[3] = { width, height, depth };
1907 size_t row_pitch, slice_pitch;
980 EVENT_LIST (10, items - 10); 1908 EVENT_LIST (10);
981 1909
1910 cl_event ev;
1911 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1912
1913 size_t cb = slice_pitch ? slice_pitch * region [2]
1914 : row_pitch ? row_pitch * region [1]
1915 : region [0];
1916
1917 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1918
1919void
1920unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1921 PPCODE:
1922 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1923 if (GIMME_V != G_VOID)
1924 {
1925 clRetainEvent (mapped->event);
1926 XPUSH_CLOBJ (stash_event, mapped->event);
1927 }
1928
1929void
1930task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1931 ALIAS:
1932 enqueue_task = 0
1933 PPCODE:
1934 EVENT_LIST (2);
1935
1936 cl_event ev = 0;
982 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1937 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
983 1938
984 if (ev) 1939 if (ev)
985 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1940 XPUSH_CLOBJ (stash_event, ev);
986 1941
987void 1942void
988enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
989 PPCODE:
990 cl_event ev = 0;
991 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
992 const size_t region[3] = { width, height, depth };
993 EVENT_LIST (10, items - 10);
994
995 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
996
997 if (ev)
998 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
999
1000void
1001enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1002 PPCODE:
1003 cl_event ev = 0;
1004 EVENT_LIST (2, items - 2);
1005
1006 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1007
1008 if (ev)
1009 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1010
1011void
1012enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1943nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1944 ALIAS:
1945 enqueue_nd_range_kernel = 0
1013 PPCODE: 1946 PPCODE:
1014 cl_event ev = 0; 1947 EVENT_LIST (5);
1948
1015 size_t *gwo = 0, *gws, *lws = 0; 1949 size_t *gwo = 0, *gws, *lws = 0;
1016 int gws_len; 1950 int gws_len;
1017 size_t *lists; 1951 size_t *lists;
1018 int i; 1952 int i;
1019 EVENT_LIST (5, items - 5);
1020 1953
1021 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1954 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1022 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1023 1956
1024 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1957 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1025 1958
1026 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1959 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1027 1960
1028 gws = lists + gws_len * 0; 1961 gws = lists + gws_len * 0;
1029 for (i = 0; i < gws_len; ++i) 1962 for (i = 0; i < gws_len; ++i)
1963 {
1030 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1964 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1965 // at least nvidia crashes for 0-sized work group sizes, work around
1966 if (!gws [i])
1967 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1968 }
1031 1969
1032 if (SvOK (global_work_offset)) 1970 if (SvOK (global_work_offset))
1033 { 1971 {
1034 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1972 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1035 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1973 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1042 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1980 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1043 } 1981 }
1044 1982
1045 if (SvOK (local_work_size)) 1983 if (SvOK (local_work_size))
1046 { 1984 {
1047 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1985 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1048 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1986 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1049 1987
1050 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1988 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1051 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1989 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1052 1990
1053 lws = lists + gws_len * 2; 1991 lws = lists + gws_len * 2;
1054 for (i = 0; i < gws_len; ++i) 1992 for (i = 0; i < gws_len; ++i)
1993 {
1055 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1994 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1995 // at least nvidia crashes for 0-sized work group sizes, work around
1996 if (!lws [i])
1997 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1998 }
1056 } 1999 }
1057 2000
2001 cl_event ev = 0;
1058 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2002 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1059 2003
1060 if (ev) 2004 if (ev)
1061 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2005 XPUSH_CLOBJ (stash_event, ev);
2006
2007#if CL_VERSION_1_2
2008
2009void
2010migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2011 ALIAS:
2012 enqueue_migrate_mem_objects = 0
2013 PPCODE:
2014 EVENT_LIST (3);
2015
2016 cl_uint obj_count;
2017 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2018
2019 cl_event ev = 0;
2020 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2021
2022 if (ev)
2023 XPUSH_CLOBJ (stash_event, ev);
2024
2025#endif
1062 2026
1063#if cl_apple_gl_sharing || cl_khr_gl_sharing 2027#if cl_apple_gl_sharing || cl_khr_gl_sharing
1064 2028
1065void 2029void
1066enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2030acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1067 ALIAS: 2031 ALIAS:
2032 release_gl_objects = 1
2033 enqueue_acquire_gl_objects = 0
1068 enqueue_release_gl_objects = 1 2034 enqueue_release_gl_objects = 1
1069 CODE: 2035 PPCODE:
1070 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2036 EVENT_LIST (2);
1071 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2037
2038 cl_uint obj_count;
2039 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1072 2040
1073 cl_event ev = 0; 2041 cl_event ev = 0;
1074 EVENT_LIST (2, items - 2);
1075 AV *av = (AV *)SvRV (objects);
1076 cl_uint num_objects = av_len (av) + 1;
1077 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1078 int i;
1079
1080 for (i = num_objects; i--; )
1081 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1082 2042
1083 if (ix) 2043 if (ix)
1084 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2044 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1085 else 2045 else
1086 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1087 2047
1088 if (ev) 2048 if (ev)
1089 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2049 XPUSH_CLOBJ (stash_event, ev);
1090 2050
1091#endif 2051#endif
1092 2052
1093void 2053void
1094enqueue_marker (OpenCL::Queue self)
1095 PPCODE:
1096 cl_event ev;
1097 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1098 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1099
1100void
1101enqueue_wait_for_events (OpenCL::Queue self, ...) 2054wait_for_events (OpenCL::Queue self, ...)
2055 ALIAS:
2056 enqueue_wait_for_events = 0
1102 CODE: 2057 CODE:
1103 EVENT_LIST (1, items - 1); 2058 EVENT_LIST (1);
2059#if PREFER_1_1
1104 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2060 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2061#else
2062 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2063#endif
1105 2064
1106void 2065void
1107enqueue_barrier (OpenCL::Queue self) 2066marker (OpenCL::Queue self, ...)
2067 ALIAS:
2068 enqueue_marker = 0
1108 CODE: 2069 PPCODE:
2070 EVENT_LIST (1);
2071 cl_event ev = 0;
2072#if PREFER_1_1
2073 if (!event_list_count)
2074 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2075 else
2076#if CL_VERSION_1_2
2077 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2078#else
2079 {
2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2081 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2082 }
2083#endif
2084#else
2085 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2086#endif
2087 if (ev)
2088 XPUSH_CLOBJ (stash_event, ev);
2089
2090void
2091barrier (OpenCL::Queue self, ...)
2092 ALIAS:
2093 enqueue_barrier = 0
2094 PPCODE:
2095 EVENT_LIST (1);
2096 cl_event ev = 0;
2097#if PREFER_1_1
2098 if (!event_list_count && GIMME_V == G_VOID)
1109 NEED_SUCCESS (EnqueueBarrier, (self)); 2099 NEED_SUCCESS (EnqueueBarrier, (self));
2100 else
2101#if CL_VERSION_1_2
2102 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2103#else
2104 {
2105 if (event_list_count)
2106 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2107
2108 if (GIMME_V != G_VOID)
2109 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2110 }
2111#endif
2112#else
2113 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2114#endif
2115 if (ev)
2116 XPUSH_CLOBJ (stash_event, ev);
1110 2117
1111void 2118void
1112flush (OpenCL::Queue self) 2119flush (OpenCL::Queue self)
1113 CODE: 2120 CODE:
1114 NEED_SUCCESS (Flush, (self)); 2121 NEED_SUCCESS (Flush, (self));
1130 PPCODE: 2137 PPCODE:
1131 cl_context value [1]; 2138 cl_context value [1];
1132 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2139 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1133 EXTEND (SP, 1); 2140 EXTEND (SP, 1);
1134 const int i = 0; 2141 const int i = 0;
1135 {
1136 NEED_SUCCESS (RetainContext, (value [i])); 2142 NEED_SUCCESS (RetainContext, (value [i]));
1137 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2143 PUSH_CLOBJ (stash_context, value [i]);
1138 }
1139 2144
1140void 2145void
1141device (OpenCL::Queue self) 2146device (OpenCL::Queue self)
1142 PPCODE: 2147 PPCODE:
1143 cl_device_id value [1]; 2148 cl_device_id value [1];
1144 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2149 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1145 EXTEND (SP, 1); 2150 EXTEND (SP, 1);
1146 const int i = 0; 2151 const int i = 0;
1147 { 2152 PUSH_CLOBJ (stash_device, value [i]);
1148 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1149 }
1150 2153
1151void 2154void
1152reference_count (OpenCL::Queue self) 2155reference_count (OpenCL::Queue self)
1153 PPCODE: 2156 PPCODE:
1154 cl_uint value [1]; 2157 cl_uint value [1];
1158 PUSHs (sv_2mortal (newSVuv (value [i]))); 2161 PUSHs (sv_2mortal (newSVuv (value [i])));
1159 2162
1160void 2163void
1161properties (OpenCL::Queue self) 2164properties (OpenCL::Queue self)
1162 PPCODE: 2165 PPCODE:
1163 cl_command_queue_properties value [1]; 2166 cl_ulong value [1];
1164 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2167 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1165 EXTEND (SP, 1); 2168 EXTEND (SP, 1);
1166 const int i = 0; 2169 const int i = 0;
1167 PUSHs (sv_2mortal (newSViv (value [i]))); 2170 PUSHs (sv_2mortal (newSVuv (value [i])));
1168 2171
1169#END:command_queue 2172#END:command_queue
1170 2173
1171MODULE = OpenCL PACKAGE = OpenCL::Memory 2174MODULE = OpenCL PACKAGE = OpenCL::Memory
1172 2175
1178void 2181void
1179info (OpenCL::Memory self, cl_mem_info name) 2182info (OpenCL::Memory self, cl_mem_info name)
1180 PPCODE: 2183 PPCODE:
1181 INFO (MemObject) 2184 INFO (MemObject)
1182 2185
2186void
2187destructor_callback (OpenCL::Memory self, SV *notify)
2188 PPCODE:
2189 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2190
1183#BEGIN:mem 2191#BEGIN:mem
1184 2192
1185void 2193void
1186type (OpenCL::Memory self) 2194type (OpenCL::Memory self)
2195 ALIAS:
2196 type = CL_MEM_TYPE
2197 map_count = CL_MEM_MAP_COUNT
2198 reference_count = CL_MEM_REFERENCE_COUNT
1187 PPCODE: 2199 PPCODE:
1188 cl_mem_object_type value [1]; 2200 cl_uint value [1];
1189 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2201 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1190 EXTEND (SP, 1); 2202 EXTEND (SP, 1);
1191 const int i = 0; 2203 const int i = 0;
1192 PUSHs (sv_2mortal (newSViv (value [i]))); 2204 PUSHs (sv_2mortal (newSVuv (value [i])));
1193 2205
1194void 2206void
1195flags (OpenCL::Memory self) 2207flags (OpenCL::Memory self)
1196 PPCODE: 2208 PPCODE:
1197 cl_mem_flags value [1]; 2209 cl_ulong value [1];
1198 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2210 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1199 EXTEND (SP, 1); 2211 EXTEND (SP, 1);
1200 const int i = 0; 2212 const int i = 0;
1201 PUSHs (sv_2mortal (newSViv (value [i]))); 2213 PUSHs (sv_2mortal (newSVuv (value [i])));
1202 2214
1203void 2215void
1204size (OpenCL::Memory self) 2216size (OpenCL::Memory self)
1205 ALIAS: 2217 ALIAS:
1206 size = CL_MEM_SIZE 2218 size = CL_MEM_SIZE
1220 EXTEND (SP, 1); 2232 EXTEND (SP, 1);
1221 const int i = 0; 2233 const int i = 0;
1222 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2234 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1223 2235
1224void 2236void
1225map_count (OpenCL::Memory self)
1226 ALIAS:
1227 map_count = CL_MEM_MAP_COUNT
1228 reference_count = CL_MEM_REFERENCE_COUNT
1229 PPCODE:
1230 cl_uint value [1];
1231 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1232 EXTEND (SP, 1);
1233 const int i = 0;
1234 PUSHs (sv_2mortal (newSVuv (value [i])));
1235
1236void
1237context (OpenCL::Memory self) 2237context (OpenCL::Memory self)
1238 PPCODE: 2238 PPCODE:
1239 cl_context value [1]; 2239 cl_context value [1];
1240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2240 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1241 EXTEND (SP, 1); 2241 EXTEND (SP, 1);
1242 const int i = 0; 2242 const int i = 0;
1243 {
1244 NEED_SUCCESS (RetainContext, (value [i])); 2243 NEED_SUCCESS (RetainContext, (value [i]));
1245 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2244 PUSH_CLOBJ (stash_context, value [i]);
1246 }
1247 2245
1248void 2246void
1249associated_memobject (OpenCL::Memory self) 2247associated_memobject (OpenCL::Memory self)
1250 PPCODE: 2248 PPCODE:
1251 cl_mem value [1]; 2249 cl_mem value [1];
1252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1253 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1254 const int i = 0; 2252 const int i = 0;
1255 {
1256 NEED_SUCCESS (RetainMemObject, (value [i])); 2253 NEED_SUCCESS (RetainMemObject, (value [i]));
1257 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2254 PUSH_CLOBJ (stash_memory, value [i]);
1258 }
1259 2255
1260#END:mem 2256#END:mem
1261 2257
1262#if cl_apple_gl_sharing || cl_khr_gl_sharing 2258#if cl_apple_gl_sharing || cl_khr_gl_sharing
1263 2259
1282 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2278 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1283 2279
1284 cl_buffer_region crdata = { origin, size }; 2280 cl_buffer_region crdata = { origin, size };
1285 2281
1286 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2282 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1287 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2283 XPUSH_CLOBJ (stash_buffer, mem);
1288 2284
1289MODULE = OpenCL PACKAGE = OpenCL::Image 2285MODULE = OpenCL PACKAGE = OpenCL::Image
1290 2286
1291void 2287void
1292image_info (OpenCL::Image self, cl_image_info name) 2288image_info (OpenCL::Image self, cl_image_info name)
1293 PPCODE: 2289 PPCODE:
1294 INFO (Image) 2290 INFO (Image)
2291
2292void
2293format (OpenCL::Image self)
2294 PPCODE:
2295 cl_image_format format;
2296 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2297 EXTEND (SP, 2);
2298 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2299 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1295 2300
1296#BEGIN:image 2301#BEGIN:image
1297 2302
1298void 2303void
1299element_size (OpenCL::Image self) 2304element_size (OpenCL::Image self)
1353 2358
1354#BEGIN:sampler 2359#BEGIN:sampler
1355 2360
1356void 2361void
1357reference_count (OpenCL::Sampler self) 2362reference_count (OpenCL::Sampler self)
2363 ALIAS:
2364 reference_count = CL_SAMPLER_REFERENCE_COUNT
2365 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2366 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1358 PPCODE: 2367 PPCODE:
1359 cl_uint value [1]; 2368 cl_uint value [1];
1360 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2369 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1361 EXTEND (SP, 1); 2370 EXTEND (SP, 1);
1362 const int i = 0; 2371 const int i = 0;
1363 PUSHs (sv_2mortal (newSVuv (value [i]))); 2372 PUSHs (sv_2mortal (newSVuv (value [i])));
1364 2373
1365void 2374void
1367 PPCODE: 2376 PPCODE:
1368 cl_context value [1]; 2377 cl_context value [1];
1369 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2378 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1370 EXTEND (SP, 1); 2379 EXTEND (SP, 1);
1371 const int i = 0; 2380 const int i = 0;
1372 {
1373 NEED_SUCCESS (RetainContext, (value [i])); 2381 NEED_SUCCESS (RetainContext, (value [i]));
1374 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2382 PUSH_CLOBJ (stash_context, value [i]);
1375 }
1376
1377void
1378normalized_coords (OpenCL::Sampler self)
1379 PPCODE:
1380 cl_addressing_mode value [1];
1381 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1382 EXTEND (SP, 1);
1383 const int i = 0;
1384 PUSHs (sv_2mortal (newSViv (value [i])));
1385
1386void
1387addressing_mode (OpenCL::Sampler self)
1388 PPCODE:
1389 cl_filter_mode value [1];
1390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1391 EXTEND (SP, 1);
1392 const int i = 0;
1393 PUSHs (sv_2mortal (newSViv (value [i])));
1394 2383
1395void 2384void
1396filter_mode (OpenCL::Sampler self) 2385filter_mode (OpenCL::Sampler self)
1397 PPCODE: 2386 PPCODE:
1398 cl_bool value [1]; 2387 cl_uint value [1];
1399 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1400 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1401 const int i = 0; 2390 const int i = 0;
1402 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2391 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1403 2392
1409DESTROY (OpenCL::Program self) 2398DESTROY (OpenCL::Program self)
1410 CODE: 2399 CODE:
1411 clReleaseProgram (self); 2400 clReleaseProgram (self);
1412 2401
1413void 2402void
1414build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2403build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2404 ALIAS:
2405 build_async = 1
1415 CODE: 2406 CODE:
2407 cl_uint device_count = 0;
2408 cl_device_id *device_list = 0;
2409
2410 if (SvOK (devices))
2411 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2412
2413 void *user_data;
2414 program_callback pfn_notify = make_program_callback (notify, &user_data);
2415
2416 if (ix)
2417 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2418 else
2419 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2420
2421#if CL_VERSION_1_2
2422
2423void
2424compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2425 CODE:
2426 cl_uint device_count = 0;
2427 cl_device_id *device_list = 0;
2428
2429 if (SvOK (devices))
2430 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2431
2432 cl_uint header_count = 0;
2433 cl_program *header_list = 0;
2434 const char **header_name = 0;
2435
2436 if (SvOK (headers))
2437 {
2438 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2439 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2440
2441 HV *hv = (HV *)SvRV (devices);
2442
2443 header_count = hv_iterinit (hv);
2444 header_list = tmpbuf (sizeof (*header_list) * header_count);
2445 header_name = tmpbuf (sizeof (*header_name) * header_count);
2446
2447 HE *he;
2448 int i = 0;
2449 while (he = hv_iternext (hv))
2450 {
2451 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2452 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2453 ++i;
2454 }
2455 }
2456
2457 void *user_data;
2458 program_callback pfn_notify = make_program_callback (notify, &user_data);
2459
1416 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2460 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2461 header_count, header_list, header_name, pfn_notify, user_data));
2462
2463#endif
1417 2464
1418void 2465void
1419build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2466build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1420 PPCODE: 2467 PPCODE:
1421 size_t size; 2468 size_t size;
1430#BEGIN:program_build 2477#BEGIN:program_build
1431 2478
1432void 2479void
1433build_status (OpenCL::Program self, OpenCL::Device device) 2480build_status (OpenCL::Program self, OpenCL::Device device)
1434 PPCODE: 2481 PPCODE:
1435 cl_build_status value [1]; 2482 cl_int value [1];
1436 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2483 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1437 EXTEND (SP, 1); 2484 EXTEND (SP, 1);
1438 const int i = 0; 2485 const int i = 0;
1439 PUSHs (sv_2mortal (newSViv (value [i]))); 2486 PUSHs (sv_2mortal (newSViv (value [i])));
1440 2487
1450 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2497 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1451 EXTEND (SP, 1); 2498 EXTEND (SP, 1);
1452 const int i = 0; 2499 const int i = 0;
1453 PUSHs (sv_2mortal (newSVpv (value, 0))); 2500 PUSHs (sv_2mortal (newSVpv (value, 0)));
1454 2501
2502void
2503binary_type (OpenCL::Program self, OpenCL::Device device)
2504 PPCODE:
2505 cl_uint value [1];
2506 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2507 EXTEND (SP, 1);
2508 const int i = 0;
2509 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2510
1455#END:program_build 2511#END:program_build
1456 2512
1457void 2513void
1458kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
1459 PPCODE: 2515 PPCODE:
1460 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1461 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
2518
2519void
2520kernels_in_program (OpenCL::Program program)
2521 PPCODE:
2522 cl_uint num_kernels;
2523 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2524 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2526
2527 int i;
2528 EXTEND (SP, num_kernels);
2529 for (i = 0; i < num_kernels; ++i)
2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
1462 2531
1463void 2532void
1464info (OpenCL::Program self, cl_program_info name) 2533info (OpenCL::Program self, cl_program_info name)
1465 PPCODE: 2534 PPCODE:
1466 INFO (Program) 2535 INFO (Program)
1484 { 2553 {
1485 SV *sv = sv_2mortal (newSV (sizes [i])); 2554 SV *sv = sv_2mortal (newSV (sizes [i]));
1486 SvUPGRADE (sv, SVt_PV); 2555 SvUPGRADE (sv, SVt_PV);
1487 SvPOK_only (sv); 2556 SvPOK_only (sv);
1488 SvCUR_set (sv, sizes [i]); 2557 SvCUR_set (sv, sizes [i]);
1489 ptrs [i] = SvPVX (sv); 2558 ptrs [i] = (void *)SvPVX (sv);
1490 PUSHs (sv); 2559 PUSHs (sv);
1491 } 2560 }
1492 2561
1493 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2562 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1494 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2563 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1512 PPCODE: 2581 PPCODE:
1513 cl_context value [1]; 2582 cl_context value [1];
1514 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1515 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
1516 const int i = 0; 2585 const int i = 0;
1517 {
1518 NEED_SUCCESS (RetainContext, (value [i])); 2586 NEED_SUCCESS (RetainContext, (value [i]));
1519 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2587 PUSH_CLOBJ (stash_context, value [i]);
1520 }
1521 2588
1522void 2589void
1523devices (OpenCL::Program self) 2590devices (OpenCL::Program self)
1524 PPCODE: 2591 PPCODE:
1525 size_t size; 2592 size_t size;
1527 cl_device_id *value = tmpbuf (size); 2594 cl_device_id *value = tmpbuf (size);
1528 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2595 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1529 int i, n = size / sizeof (*value); 2596 int i, n = size / sizeof (*value);
1530 EXTEND (SP, n); 2597 EXTEND (SP, n);
1531 for (i = 0; i < n; ++i) 2598 for (i = 0; i < n; ++i)
1532 { 2599 PUSH_CLOBJ (stash_device, value [i]);
1533 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1534 }
1535 2600
1536void 2601void
1537source (OpenCL::Program self) 2602source (OpenCL::Program self)
1538 PPCODE: 2603 PPCODE:
1539 size_t size; 2604 size_t size;
1564DESTROY (OpenCL::Kernel self) 2629DESTROY (OpenCL::Kernel self)
1565 CODE: 2630 CODE:
1566 clReleaseKernel (self); 2631 clReleaseKernel (self);
1567 2632
1568void 2633void
2634setf (OpenCL::Kernel self, const char *format, ...)
2635 CODE:
2636 int i;
2637 for (i = 2; ; ++i)
2638 {
2639 while (*format == ' ')
2640 ++format;
2641
2642 char type = *format++;
2643
2644 if (!type)
2645 break;
2646
2647 if (i >= items)
2648 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2649
2650 SV *sv = ST (i);
2651
2652 union
2653 {
2654 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2655 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2656 cl_half ch; cl_float cf; cl_double cd;
2657 cl_mem cm;
2658 cl_sampler ca;
2659 size_t cz;
2660 cl_event ce;
2661 } arg;
2662 size_t size;
2663 int nullarg = 0;
2664
2665 switch (type)
2666 {
2667 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2668 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2669 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2670 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2671 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2672 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2673 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2674 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2675
2676 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2677 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2678 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2679
2680 case 'z': nullarg = 1; size = SvIV (sv); break;
2681
2682 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2683 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2684 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2685
2686 default:
2687 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2688 }
2689
2690 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2691 if (res)
2692 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2693 }
2694
2695 if (i != items)
2696 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2697
2698void
1569set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2699set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1570 CODE: 2700 CODE:
1571 clSetKernelArg (self, idx, sizeof (value), &value); 2701 clSetKernelArg (self, idx, sizeof (value), &value);
1572 2702
1573void 2703void
1621 clSetKernelArg (self, idx, sizeof (value), &value); 2751 clSetKernelArg (self, idx, sizeof (value), &value);
1622 2752
1623void 2753void
1624set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2754set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1625 CODE: 2755 CODE:
2756 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2757
2758void
2759set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2760 CODE:
2761 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2762
2763void
2764set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2765 CODE:
2766 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2767
2768void
2769set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2770 CODE:
1626 clSetKernelArg (self, idx, sizeof (value), &value); 2771 clSetKernelArg (self, idx, sizeof (value), &value);
1627 2772
1628void 2773void
1629set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2774set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1630 CODE: 2775 CODE:
1631 clSetKernelArg (self, idx, sizeof (value), &value); 2776 clSetKernelArg (self, idx, size, 0);
1632
1633void
1634set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1635 CODE:
1636 clSetKernelArg (self, idx, sizeof (value), &value);
1637
1638void
1639set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1640 CODE:
1641 clSetKernelArg (self, idx, sizeof (value), &value);
1642
1643void
1644set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1645 CODE:
1646 clSetKernelArg (self, idx, sizeof (value), &value);
1647 2777
1648void 2778void
1649set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2779set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1650 CODE: 2780 CODE:
1651 clSetKernelArg (self, idx, sizeof (value), &value); 2781 clSetKernelArg (self, idx, sizeof (value), &value);
1685 PPCODE: 2815 PPCODE:
1686 cl_context value [1]; 2816 cl_context value [1];
1687 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2817 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1688 EXTEND (SP, 1); 2818 EXTEND (SP, 1);
1689 const int i = 0; 2819 const int i = 0;
1690 {
1691 NEED_SUCCESS (RetainContext, (value [i])); 2820 NEED_SUCCESS (RetainContext, (value [i]));
1692 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2821 PUSH_CLOBJ (stash_context, value [i]);
1693 }
1694 2822
1695void 2823void
1696program (OpenCL::Kernel self) 2824program (OpenCL::Kernel self)
1697 PPCODE: 2825 PPCODE:
1698 cl_program value [1]; 2826 cl_program value [1];
1699 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2827 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1700 EXTEND (SP, 1); 2828 EXTEND (SP, 1);
1701 const int i = 0; 2829 const int i = 0;
1702 {
1703 NEED_SUCCESS (RetainProgram, (value [i])); 2830 NEED_SUCCESS (RetainProgram, (value [i]));
1704 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2831 PUSH_CLOBJ (stash_program, value [i]);
1705 }
1706 2832
1707#END:kernel 2833#END:kernel
1708 2834
1709void 2835void
1710work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2836work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
1756 const int i = 0; 2882 const int i = 0;
1757 PUSHs (sv_2mortal (newSVuv (value [i]))); 2883 PUSHs (sv_2mortal (newSVuv (value [i])));
1758 2884
1759#END:kernel_work_group 2885#END:kernel_work_group
1760 2886
2887#if CL_VERSION_1_2
2888
2889void
2890arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2891 PPCODE:
2892 size_t size;
2893 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2894 SV *sv = sv_2mortal (newSV (size));
2895 SvUPGRADE (sv, SVt_PV);
2896 SvPOK_only (sv);
2897 SvCUR_set (sv, size);
2898 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2899 XPUSHs (sv);
2900
2901#BEGIN:kernel_arg
2902
2903void
2904arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2905 ALIAS:
2906 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2907 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2908 PPCODE:
2909 cl_uint value [1];
2910 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2911 EXTEND (SP, 1);
2912 const int i = 0;
2913 PUSHs (sv_2mortal (newSVuv (value [i])));
2914
2915void
2916arg_type_name (OpenCL::Kernel self, cl_uint idx)
2917 ALIAS:
2918 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2919 arg_name = CL_KERNEL_ARG_NAME
2920 PPCODE:
2921 size_t size;
2922 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2923 char *value = tmpbuf (size);
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2925 EXTEND (SP, 1);
2926 const int i = 0;
2927 PUSHs (sv_2mortal (newSVpv (value, 0)));
2928
2929void
2930arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2931 PPCODE:
2932 cl_ulong value [1];
2933 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2934 EXTEND (SP, 1);
2935 const int i = 0;
2936 PUSHs (sv_2mortal (newSVuv (value [i])));
2937
2938#END:kernel_arg
2939
2940#endif
2941
1761MODULE = OpenCL PACKAGE = OpenCL::Event 2942MODULE = OpenCL PACKAGE = OpenCL::Event
1762 2943
1763void 2944void
1764DESTROY (OpenCL::Event self) 2945DESTROY (OpenCL::Event self)
1765 CODE: 2946 CODE:
1767 2948
1768void 2949void
1769wait (OpenCL::Event self) 2950wait (OpenCL::Event self)
1770 CODE: 2951 CODE:
1771 clWaitForEvents (1, &self); 2952 clWaitForEvents (1, &self);
2953
2954void
2955cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2956 CODE:
2957 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1772 2958
1773void 2959void
1774info (OpenCL::Event self, cl_event_info name) 2960info (OpenCL::Event self, cl_event_info name)
1775 PPCODE: 2961 PPCODE:
1776 INFO (Event) 2962 INFO (Event)
1782 PPCODE: 2968 PPCODE:
1783 cl_command_queue value [1]; 2969 cl_command_queue value [1];
1784 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2970 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1785 EXTEND (SP, 1); 2971 EXTEND (SP, 1);
1786 const int i = 0; 2972 const int i = 0;
1787 {
1788 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2973 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1789 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2974 PUSH_CLOBJ (stash_queue, value [i]);
1790 }
1791 2975
1792void 2976void
1793command_type (OpenCL::Event self) 2977command_type (OpenCL::Event self)
1794 PPCODE:
1795 cl_command_type value [1];
1796 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
1797 EXTEND (SP, 1);
1798 const int i = 0;
1799 PUSHs (sv_2mortal (newSVuv (value [i])));
1800
1801void
1802reference_count (OpenCL::Event self)
1803 ALIAS: 2978 ALIAS:
2979 command_type = CL_EVENT_COMMAND_TYPE
1804 reference_count = CL_EVENT_REFERENCE_COUNT 2980 reference_count = CL_EVENT_REFERENCE_COUNT
1805 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2981 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
1806 PPCODE: 2982 PPCODE:
1807 cl_uint value [1]; 2983 cl_uint value [1];
1808 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2984 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
1815 PPCODE: 2991 PPCODE:
1816 cl_context value [1]; 2992 cl_context value [1];
1817 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2993 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1818 EXTEND (SP, 1); 2994 EXTEND (SP, 1);
1819 const int i = 0; 2995 const int i = 0;
1820 {
1821 NEED_SUCCESS (RetainContext, (value [i])); 2996 NEED_SUCCESS (RetainContext, (value [i]));
1822 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2997 PUSH_CLOBJ (stash_context, value [i]);
1823 }
1824 2998
1825#END:event 2999#END:event
1826 3000
1827void 3001void
1828profiling_info (OpenCL::Event self, cl_profiling_info name) 3002profiling_info (OpenCL::Event self, cl_profiling_info name)
1852void 3026void
1853set_status (OpenCL::UserEvent self, cl_int execution_status) 3027set_status (OpenCL::UserEvent self, cl_int execution_status)
1854 CODE: 3028 CODE:
1855 clSetUserEventStatus (self, execution_status); 3029 clSetUserEventStatus (self, execution_status);
1856 3030
3031MODULE = OpenCL PACKAGE = OpenCL::Mapped
3032
3033void
3034DESTROY (SV *self)
3035 CODE:
3036 OpenCL__Mapped mapped = SvMAPPED (self);
3037
3038 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3039 mapped_detach (self, mapped);
3040
3041 clReleaseCommandQueue (mapped->queue);
3042 clReleaseEvent (mapped->event);
3043 Safefree (mapped);
3044
3045void
3046unmap (OpenCL::Mapped self, ...)
3047 CODE:
3048 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3049
3050bool
3051mapped (OpenCL::Mapped self)
3052 CODE:
3053 RETVAL = !!self->ptr;
3054 OUTPUT:
3055 RETVAL
3056
3057void
3058wait (OpenCL::Mapped self)
3059 PPCODE:
3060 if (self->event)
3061 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3062
3063void
3064event (OpenCL::Mapped self)
3065 PPCODE:
3066 if (!self->event)
3067 XSRETURN_UNDEF;
3068
3069 clRetainEvent (self->event);
3070 XPUSH_CLOBJ (stash_event, self->event);
3071
3072#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3073#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3074#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3075#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3076#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3077#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3078
3079IV
3080size (OpenCL::Mapped self)
3081 ALIAS:
3082 size = MAPPED_OFFSET_CB
3083 row_pitch = MAPPED_OFFSET_ROW_PITCH
3084 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3085 width = MAPPED_OFFSET_WIDTH
3086 height = MAPPED_OFFSET_HEIGHT
3087 depth = MAPPED_OFFSET_DEPTH
3088 CODE:
3089 RETVAL = *(size_t *)((char *)self + ix);
3090 OUTPUT:
3091 RETVAL
3092
3093IV
3094ptr (OpenCL::Mapped self)
3095 CODE:
3096 RETVAL = PTR2IV (self->ptr);
3097 OUTPUT:
3098 RETVAL
3099
3100void
3101set (OpenCL::Mapped self, size_t offset, SV *data)
3102 CODE:
3103 STRLEN len;
3104 const char *ptr = SvPVbyte (data, len);
3105
3106 if (offset + len > self->cb)
3107 croak ("OpenCL::Mapped::set out of bound condition detected");
3108
3109 memcpy (offset + (char *)self->ptr, ptr, len);
3110
3111void
3112get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3113 PPCODE:
3114 if (!SvOK (ST (1)))
3115 count = self->width - x;
3116
3117 if (x + count > self->width)
3118 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3119
3120 if (y >= self->height)
3121 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3122
3123 if (z >= self->depth)
3124 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3125
3126 size_t element = mapped_element_size (self);
3127
3128 count *= element;
3129 x *= element;
3130
3131 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3132 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3133
3134void
3135set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3136 PPCODE:
3137 STRLEN count;
3138 char *dataptr = SvPVbyte (data, count);
3139 size_t element = mapped_element_size (self);
3140
3141 x *= element;
3142
3143 if (x + count > self->width * element)
3144 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3145
3146 if (y >= self->height)
3147 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3148
3149 if (z >= self->depth)
3150 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3151
3152 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3153 memcpy (ptr, dataptr, count);
3154
3155MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3158
3159IV
3160element_size (OpenCL::Mapped self)
3161 CODE:
3162 RETVAL = mapped_element_size (self);
3163 OUTPUT:
3164 RETVAL
3165

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines