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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.26 by root, Thu Apr 19 13:06:55 2012 UTC vs.
Revision 1.79 by root, Sat May 5 15:43:02 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
10
11#ifdef I_DLFCN
12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
30#endif
31
5#ifdef __APPLE__ 32#ifdef __APPLE__
6 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
7#else 34#else
8 #include <CL/opencl.h> 35 #define CLHDR(name) <CL/name>
9#endif 36#endif
37
38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
50#endif
51
52#ifndef CL_VERSION_1_2
53 #undef PREFER_1_1
54 #define PREFER_1_1 1
55#endif
56
57// make sure all constants we might use are actually defined
58#include "default.h"
10 59
11typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
12typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
13typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
14typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
15typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
16typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
17typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
18typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
19typedef cl_mem OpenCL__Image2D;
20typedef cl_mem OpenCL__Image3D;
21typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
22typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
23typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
24typedef cl_mem OpenCL__Image2D_ornull;
25typedef cl_mem OpenCL__Image3D_ornull;
26typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
27typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
28typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
29typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
30typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
31 77
32typedef SV *FUTURE; 78typedef struct mapped * OpenCL__Mapped;
79
80static HV
81 *stash_platform,
82 *stash_device,
83 *stash_subdevice,
84 *stash_context,
85 *stash_queue,
86 *stash_program,
87 *stash_kernel,
88 *stash_sampler,
89 *stash_event,
90 *stash_userevent,
91 *stash_memory,
92 *stash_buffer,
93 *stash_bufferobj,
94 *stash_image,
95 *stash_image1d,
96 *stash_image1darray,
97 *stash_image1dbuffer,
98 *stash_image2d,
99 *stash_image2darray,
100 *stash_image3d,
101 *stash_mapped,
102 *stash_mappedbuffer,
103 *stash_mappedimage;
33 104
34/*****************************************************************************/ 105/*****************************************************************************/
35 106
107// name must include a leading underscore
108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
109// for perl. doh.
110static void *
111glsym (const char *name)
112{
113 void *fun = 0;
114
115 #if defined I_DLFCN && defined RTLD_DEFAULT
116 fun = dlsym (RTLD_DEFAULT, name + 1);
117 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
118
119 if (!fun)
120 {
121 static void *libgl;
122 static const char *glso[] = {
123 "libGL.so.1",
124 "libGL.so.3",
125 "libGL.so.4.0",
126 "libGL.so",
127 "/usr/lib/libGL.so",
128 "/usr/X11R6/lib/libGL.1.dylib"
129 };
130 int i;
131
132 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
133 {
134 libgl = dlopen (glso [i], RTLD_LAZY);
135 if (libgl)
136 break;
137 }
138
139 if (libgl)
140 {
141 fun = dlsym (libgl, name + 1);
142 if (!fun) fun = dlsym (libgl, name);
143 }
144 }
145 #endif
146
147 return fun;
148}
149
150/*****************************************************************************/
151
36/* up to two temporary buffers */ 152/* up to two temporary buffers */
37static void * 153static void * ecb_noinline
38tmpbuf (size_t size) 154tmpbuf (size_t size)
39{ 155{
40 enum { buffers = 3 }; 156 enum { buffers = 4 };
41 static int idx; 157 static int idx;
42 static void *buf [buffers]; 158 static void *buf [buffers];
43 static size_t len [buffers]; 159 static size_t len [buffers];
44 160
45 idx = ++idx % buffers; 161 idx = (idx + 1) % buffers;
46 162
47 if (len [idx] < size) 163 if (len [idx] < size)
48 { 164 {
49 free (buf [idx]); 165 free (buf [idx]);
50 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
52 } 168 }
53 169
54 return buf [idx]; 170 return buf [idx];
55} 171}
56 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
57/*****************************************************************************/ 188/*****************************************************************************/
58 189
59typedef struct 190typedef struct
60{ 191{
61 IV iv; 192 IV iv;
62 const char *name; 193 const char *name;
63 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
64} ivstr; 195} ivstr;
196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
65 203
66static const char * 204static const char *
67iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
68{ 206{
69 int i; 207 int i;
118 if (res) \ 256 if (res) \
119 FAIL (name); 257 FAIL (name);
120 258
121/*****************************************************************************/ 259/*****************************************************************************/
122 260
123#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
124#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
125
126static void * 261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
127SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
128{ 274{
275 // sv_derived_from is quite slow :(
129 if (SvROK (sv) && sv_derived_from (sv, pkg)) 276 if (SvROK (sv) && sv_derived_from (sv, pkg))
130 return (void *)SvIV (SvRV (sv)); 277 return (void *)SvIV (SvRV (sv));
131 278
132 croak ("%s: %s is not of type %s", func, svname, pkg); 279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
133} 290}
134 291
135/*****************************************************************************/ 292/*****************************************************************************/
136 293
137static cl_context_properties * 294static cl_context_properties * ecb_noinline
138SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
139{ 296{
140 if (!sv || !SvOK (sv)) 297 if (!sv || !SvOK (sv))
141 if (extra) 298 if (extra)
142 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
143 else 300 else
144 return 0; 301 return 0;
145 302
146 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV) 303 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
147 { 304 {
148 AV *av = (AV *)SvRV (sv); 305 AV *av = (AV *)SvRV (sv);
149 int i, len = av_len (av); 306 int i, len = av_len (av) + 1;
150 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
151 cl_context_properties *l = p; 308 cl_context_properties *l = p;
152 309
153 if (len & 1) 310 if (len & 1)
154 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
155 312
156 while (extracount--) 313 while (extracount--)
157 *l++ = *extra++; 314 *l++ = *extra++;
158 315
159 for (i = 0; i < len; i += 2) 316 for (i = 0; i < len; i += 2)
160 { 317 {
161 cl_context_properties t = SvIV (*av_fetch (av, i, 0)); 318 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
162 cl_context_properties v; 319 SV *p_sv = *av_fetch (av, i + 1, 0);
320 cl_context_properties v = SvIV (p_sv); // code below can override
163 321
164 switch (t) 322 switch (t)
165 { 323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
329 case CL_GLX_DISPLAY_KHR:
330 if (!SvOK (p_sv))
331 {
332 void *func = glsym ("_glXGetCurrentDisplay");
333 if (func)
334 v = (cl_context_properties)((void *(*)(void))func)();
335 }
336 break;
337
338 case CL_GL_CONTEXT_KHR:
339 if (!SvOK (p_sv))
340 {
341 void *func = glsym ("_glXGetCurrentContext");
342 if (func)
343 v = (cl_context_properties)((void *(*)(void))func)();
344 }
345 break;
346
166 default: 347 default:
167 /* unknown property, treat as int */ 348 /* unknown property, treat as int */
168 v = SvIV (*av_fetch (av, i, 0));
169 break; 349 break;
170 } 350 }
171 351
172 *l++ = t; 352 *l++ = t;
173 *l++ = v; 353 *l++ = v;
176 *l = 0; 356 *l = 0;
177 357
178 return p; 358 return p;
179 } 359 }
180 360
181 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
362}
363
364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
365// is a pointer (and only on well-behaved systems).
366static void * ecb_noinline
367object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
368{
369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
372
373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
378 {
379 list = tmpbuf (sizeof (*list) * count);
380 int i;
381 for (i = 0; i < count; ++i)
382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
383 }
384
385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
182} 391}
183 392
184/*****************************************************************************/ 393/*****************************************************************************/
394/* callback stuff */
395
396/* default context callback, log to stderr */
397static void CL_CALLBACK
398context_default_notify (const char *msg, const void *info, size_t cb, void *data)
399{
400 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
401}
402
403typedef struct
404{
405 int free_cb;
406 void (*push)(void *data1, void *data2, void *data3);
407} eq_vtbl;
408
409typedef struct eq_item
410{
411 struct eq_item *next;
412 eq_vtbl *vtbl;
413 SV *cb;
414 void *data1, *data2, *data3;
415} eq_item;
416
417static void (*eq_signal_func)(void *signal_arg, int value);
418static void *eq_signal_arg;
419static xmutex_t eq_lock = X_MUTEX_INIT;
420static eq_item *eq_head, *eq_tail;
421
422static void ecb_noinline
423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
424{
425 eq_item *item = malloc (sizeof (eq_item));
426
427 item->next = 0;
428 item->vtbl = vtbl;
429 item->cb = cb;
430 item->data1 = data1;
431 item->data2 = data2;
432 item->data3 = data3;
433
434 X_LOCK (eq_lock);
435
436 *(eq_head ? &eq_tail->next : &eq_head) = item;
437 eq_tail = item;
438
439 X_UNLOCK (eq_lock);
440
441 eq_signal_func (eq_signal_arg, 0);
442}
443
444static eq_item *
445eq_dec (void)
446{
447 eq_item *res;
448
449 X_LOCK (eq_lock);
450
451 res = eq_head;
452 if (res)
453 eq_head = res->next;
454
455 X_UNLOCK (eq_lock);
456
457 return res;
458}
459
460static void
461eq_poll (void)
462{
463 eq_item *item;
464
465 while ((item = eq_dec ()))
466 {
467 ENTER;
468 SAVETMPS;
469
470 dSP;
471 PUSHMARK (SP);
472 EXTEND (SP, 2);
473
474 if (item->vtbl->free_cb)
475 sv_2mortal (item->cb);
476
477 PUTBACK;
478 item->vtbl->push (item->data1, item->data2, item->data3);
479
480 SV *cb = item->cb;
481 free (item);
482
483 call_sv (cb, G_DISCARD | G_VOID);
484
485 FREETMPS;
486 LEAVE;
487 }
488}
489
490static void
491eq_poll_interrupt (pTHX_ void *c_arg, int value)
492{
493 eq_poll ();
494}
495
496/*****************************************************************************/
497/* context notify */
498
499static void ecb_noinline
500eq_context_push (void *data1, void *data2, void *data3)
501{
502 dSP;
503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
505 PUTBACK;
506
507 free (data1);
508 free (data2);
509}
510
511static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
512
513static void CL_CALLBACK
514eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
515{
516 void *pvt_copy = malloc (cb);
517 memcpy (pvt_copy, pvt, cb);
518 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
519}
520
521#define CONTEXT_NOTIFY_CALLBACK \
522 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
523 void *user_data = 0; \
524 \
525 if (SvOK (notify)) \
526 { \
527 pfn_notify = eq_context_notify; \
528 user_data = s_get_cv (notify); \
529 }
530
531static SV * ecb_noinline
532new_clobj_context (cl_context ctx, void *user_data)
533{
534 SV *sv = new_clobj (stash_context, (IV)ctx);
535
536 if (user_data)
537 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
538
539 return sv;
540}
541
542#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
543
544/*****************************************************************************/
545/* build/compile/link notify */
546
547static void
548eq_program_push (void *data1, void *data2, void *data3)
549{
550 dSP;
551 PUSH_CLOBJ (stash_program, data1);
552 PUTBACK;
553}
554
555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
556
557static void CL_CALLBACK
558eq_program_notify (cl_program program, void *user_data)
559{
560 clRetainProgram (program);
561
562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
580}
581
582struct build_args
583{
584 cl_program program;
585 char *options;
586 void *user_data;
587 cl_uint num_devices;
588};
589
590X_THREAD_PROC (build_program_thread)
591{
592 struct build_args *arg = thr_arg;
593
594 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
595
596 if (arg->user_data)
597 eq_program_notify (arg->program, arg->user_data);
598 else
599 clReleaseProgram (arg->program);
600
601 free (arg->options);
602 free (arg);
603
604 return 0;
605}
606
607static void
608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
609{
610 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
611
612 arg->program = program;
613 arg->options = strdup (options);
614 arg->user_data = user_data;
615 arg->num_devices = num_devices;
616 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
617
618 xthread_t id;
619 thread_create (&id, build_program_thread, arg);
620}
621
622/*****************************************************************************/
623/* mem object destructor notify */
624
625static void ecb_noinline
626eq_destructor_push (void *data1, void *data2, void *data3)
627{
628}
629
630static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
631
632static void CL_CALLBACK
633eq_destructor_notify (cl_mem memobj, void *user_data)
634{
635 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
636}
637
638/*****************************************************************************/
639/* event objects */
640
641static void
642eq_event_push (void *data1, void *data2, void *data3)
643{
644 dSP;
645 PUSH_CLOBJ (stash_event, data1);
646 PUSHs (sv_2mortal (newSViv ((IV)data2)));
647 PUTBACK;
648}
649
650static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
651
652static void CL_CALLBACK
653eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
654{
655 clRetainEvent (event);
656 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
657}
658
659/*****************************************************************************/
660/* utilities for XS code */
185 661
186static size_t 662static size_t
187img_row_pitch (cl_mem img) 663img_row_pitch (cl_mem img)
188{ 664{
189 size_t res; 665 size_t res;
190 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 666 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
191 return res; 667 return res;
192} 668}
193 669
194static cl_event * 670static cl_event * ecb_noinline
195event_list (SV **items, int count) 671event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
196{ 672{
197 if (!count) 673 cl_uint count = *rcount;
674
675 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
676 *rcount = count = 0;
677
678 if (!count && !extra)
198 return 0; 679 return 0;
199 680
200 cl_event *list = tmpbuf (sizeof (cl_event) * count); 681 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
682 int i = 0;
201 683
202 while (count--) 684 while (count--)
685 if (SvOK (items [count]))
203 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 686 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
204 687
688 if (extra)
689 list [i++] = extra;
690
691 *rcount = i;
692
205 return list; 693 return i ? list : 0;
206} 694}
207 695
208#define EVENT_LIST(items,count) \ 696#define EVENT_LIST(skip) \
209 cl_uint event_list_count = (count); \ 697 cl_uint event_list_count = items - (skip); \
210 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 698 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
211 699
212#define INFO(class) \ 700#define INFO(class) \
213{ \ 701{ \
214 size_t size; \ 702 size_t size; \
215 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 703 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
219 SvCUR_set (sv, size); \ 707 SvCUR_set (sv, size); \
220 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 708 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
221 XPUSHs (sv); \ 709 XPUSHs (sv); \
222} 710}
223 711
712/*****************************************************************************/
713/* mapped_xxx */
714
715static OpenCL__Mapped
716SvMAPPED (SV *sv)
717{
718 // no typechecking atm., keep your fingers crossed
719 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
720}
721
722struct mapped
723{
724 cl_command_queue queue;
725 cl_mem memobj;
726 void *ptr;
727 size_t cb;
728 cl_event event;
729 size_t row_pitch;
730 size_t slice_pitch;
731};
732
733static SV *
734mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
735{
736 SV *data = newSV (0);
737 SvUPGRADE (data, SVt_PVMG);
738
739 OpenCL__Mapped mapped;
740 New (0, mapped, 1, struct mapped);
741
742 clRetainCommandQueue (queue);
743
744 mapped->queue = queue;
745 mapped->memobj = memobj;
746 mapped->ptr = ptr;
747 mapped->cb = cb;
748 mapped->event = ev;
749 mapped->row_pitch = row_pitch;
750 mapped->slice_pitch = slice_pitch;
751
752 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
753
754 if (SvLEN (data))
755 Safefree (data);
756
757 SvPVX (data) = (char *)ptr;
758 SvCUR_set (data, cb);
759 SvLEN_set (data, 0);
760 SvPOK_only (data);
761
762 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
763
764 if (!(flags & CL_MAP_WRITE))
765 SvREADONLY_on (data);
766
767 return obj;
768}
769
770static void
771mapped_detach (SV *sv, OpenCL__Mapped mapped)
772{
773 SV *data = SvRV (sv);
774
775 // the next check checks both before AND after detach, where SvPVX should STILL be 0
776 if (SvPVX (data) != (char *)mapped->ptr)
777 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
778 else
779 {
780 SvREADONLY_off (data);
781 SvCUR_set (data, 0);
782 SvPVX (data) = 0;
783 SvOK_off (data);
784 }
785
786 mapped->ptr = 0;
787}
788
789static void
790mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
791{
792 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
793 cl_event ev;
794
795 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
796
797 clReleaseEvent (mapped->event);
798 mapped->event = ev;
799
800 mapped_detach (self, mapped);
801}
802
803/*****************************************************************************/
804
224MODULE = OpenCL PACKAGE = OpenCL 805MODULE = OpenCL PACKAGE = OpenCL
225 806
226PROTOTYPES: ENABLE 807PROTOTYPES: ENABLE
227 808
809void
810poll ()
811 CODE:
812 eq_poll ();
813
814void
815_eq_initialise (IV func, IV arg)
816 CODE:
817 eq_signal_func = (void (*)(void *, int))func;
818 eq_signal_arg = (void*)arg;
819
228BOOT: 820BOOT:
229{ 821{
230 HV *stash = gv_stashpv ("OpenCL", 1); 822 HV *stash = gv_stashpv ("OpenCL", 1);
823
231 static const ivstr *civ, const_iv[] = { 824 static const ivstr *civ, const_iv[] = {
232 { sizeof (cl_char ), "SIZEOF_CHAR" }, 825 { sizeof (cl_char ), "SIZEOF_CHAR" },
233 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 826 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
234 { sizeof (cl_short ), "SIZEOF_SHORT" }, 827 { sizeof (cl_short ), "SIZEOF_SHORT" },
235 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 828 { sizeof (cl_ushort), "SIZEOF_USHORT" },
238 { sizeof (cl_long ), "SIZEOF_LONG" }, 831 { sizeof (cl_long ), "SIZEOF_LONG" },
239 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 832 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
240 { sizeof (cl_half ), "SIZEOF_HALF" }, 833 { sizeof (cl_half ), "SIZEOF_HALF" },
241 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 834 { sizeof (cl_float ), "SIZEOF_FLOAT" },
242 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 835 { sizeof (cl_double), "SIZEOF_DOUBLE" },
836 { PREFER_1_1 , "PREFER_1_1" },
243#include "constiv.h" 837#include "constiv.h"
244 }; 838 };
839
245 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 840 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
246 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 841 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
842
843 static const nvstr *cnv, const_nv[] = {
844#include "constnv.h"
845 };
846
847 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
848 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
849
850 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
851
852 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
853 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
854 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
855 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
856 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
857 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
858 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
859 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
860 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
861 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
862 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
863 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
864 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
865 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
866 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
867 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
868 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
869 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
870 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
871 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
872 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
873 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
874 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
875
876 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
247} 877}
248 878
249cl_int 879cl_int
250errno () 880errno ()
251 CODE: 881 CODE:
252 errno = res; 882 RETVAL = res;
883 OUTPUT:
884 RETVAL
253 885
254const char * 886const char *
255err2str (cl_int err) 887err2str (cl_int err = res)
256 888
257const char * 889const char *
258enum2str (cl_uint value) 890enum2str (cl_uint value)
259 891
260void 892void
268 list = tmpbuf (sizeof (*list) * count); 900 list = tmpbuf (sizeof (*list) * count);
269 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 901 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
270 902
271 EXTEND (SP, count); 903 EXTEND (SP, count);
272 for (i = 0; i < count; ++i) 904 for (i = 0; i < count; ++i)
273 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 905 PUSH_CLOBJ (stash_platform, list [i]);
274 906
275void 907void
276context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 908context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
277 PPCODE: 909 PPCODE:
910 CONTEXT_NOTIFY_CALLBACK;
278 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 911 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
279 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 912 XPUSH_CLOBJ_CONTEXT;
280 913
281void 914void
282context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 915context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
283 PPCODE: 916 PPCODE:
284 /* der Gipfel der Kunst */ 917 cl_uint device_count;
918 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
919
920 CONTEXT_NOTIFY_CALLBACK;
921 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
922 XPUSH_CLOBJ_CONTEXT;
285 923
286void 924void
287wait_for_events (...) 925wait_for_events (...)
288 CODE: 926 CODE:
289 EVENT_LIST (0, items); 927 EVENT_LIST (0);
290 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 928 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
291 929
292PROTOTYPES: DISABLE 930PROTOTYPES: DISABLE
293 931
294MODULE = OpenCL PACKAGE = OpenCL::Platform 932MODULE = OpenCL PACKAGE = OpenCL::Platform
295 933
296void 934void
297info (OpenCL::Platform self, cl_platform_info name) 935info (OpenCL::Platform self, cl_platform_info name)
298 PPCODE: 936 PPCODE:
299 INFO (Platform) 937 INFO (Platform)
938
939void
940unload_compiler (OpenCL::Platform self)
941 CODE:
942#if CL_VERSION_1_2
943 clUnloadPlatformCompiler (self);
944#endif
300 945
301#BEGIN:platform 946#BEGIN:platform
302 947
303void 948void
304profile (OpenCL::Platform self) 949profile (OpenCL::Platform self)
330 list = tmpbuf (sizeof (*list) * count); 975 list = tmpbuf (sizeof (*list) * count);
331 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 976 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
332 977
333 EXTEND (SP, count); 978 EXTEND (SP, count);
334 for (i = 0; i < count; ++i) 979 for (i = 0; i < count; ++i)
335 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 980 PUSH_CLOBJ (stash_device, list [i]);
336 981
337void 982void
338context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 983context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
339 PPCODE: 984 PPCODE:
985 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
986 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
987
988 cl_uint device_count;
989 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
990
991 CONTEXT_NOTIFY_CALLBACK;
992 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
993 XPUSH_CLOBJ_CONTEXT;
994
995void
996context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
997 PPCODE:
998 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
999 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1000
1001 CONTEXT_NOTIFY_CALLBACK;
1002 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1003 XPUSH_CLOBJ_CONTEXT;
1004
1005MODULE = OpenCL PACKAGE = OpenCL::Device
1006
1007void
1008info (OpenCL::Device self, cl_device_info name)
1009 PPCODE:
1010 INFO (Device)
1011
1012#if CL_VERSION_1_2
1013
1014void
1015sub_devices (OpenCL::Device self, SV *properties)
1016 PPCODE:
340 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1017 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
341 croak ("OpenCL::Platform argument 'device' must be an arrayref with device objects, in call"); 1018 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
342 1019
343 AV *av = (AV *)SvRV (devices); 1020 properties = SvRV (properties);
344 cl_uint num_devices = av_len (av) + 1; 1021
345 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1022 cl_uint count = av_len ((AV *)properties) + 1;
1023 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1024
346 int i; 1025 int i;
1026 for (i = 0; i < count; ++i)
1027 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
347 1028
348 for (i = num_devices; i--; ) 1029 props [count] = 0;
349 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
350 1030
351 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 1031 cl_uint num_devices;
352 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1032 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1033 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1034 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
353 1035
354void 1036 EXTEND (SP, num_devices);
355context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1037 for (i = 0; i < count; ++i)
356 PPCODE: 1038 PUSH_CLOBJ (stash_subdevice, list [i]);
357 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
358 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
359 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
360 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
361 1039
362MODULE = OpenCL PACKAGE = OpenCL::Device 1040#endif
363
364void
365info (OpenCL::Device self, cl_device_info name)
366 PPCODE:
367 INFO (Device)
368 1041
369#BEGIN:device 1042#BEGIN:device
370 1043
371void 1044void
372type (OpenCL::Device self) 1045type (OpenCL::Device self)
1046 ALIAS:
1047 type = CL_DEVICE_TYPE
1048 address_bits = CL_DEVICE_ADDRESS_BITS
1049 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1050 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1051 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1052 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1053 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1054 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1055 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1056 properties = CL_DEVICE_QUEUE_PROPERTIES
1057 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1058 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
373 PPCODE: 1059 PPCODE:
374 cl_device_type value [1]; 1060 cl_ulong value [1];
375 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1061 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
376 EXTEND (SP, 1); 1062 EXTEND (SP, 1);
377 const int i = 0; 1063 const int i = 0;
378 PUSHs (sv_2mortal (newSViv (value [i]))); 1064 PUSHs (sv_2mortal (newSVuv (value [i])));
379 1065
380void 1066void
381vendor_id (OpenCL::Device self) 1067vendor_id (OpenCL::Device self)
382 ALIAS: 1068 ALIAS:
383 vendor_id = CL_DEVICE_VENDOR_ID 1069 vendor_id = CL_DEVICE_VENDOR_ID
394 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1080 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
395 image_support = CL_DEVICE_IMAGE_SUPPORT 1081 image_support = CL_DEVICE_IMAGE_SUPPORT
396 max_samplers = CL_DEVICE_MAX_SAMPLERS 1082 max_samplers = CL_DEVICE_MAX_SAMPLERS
397 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1083 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
398 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1084 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1085 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
399 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1086 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
400 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1087 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1088 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
401 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1089 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
402 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1090 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
403 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1091 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
404 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1092 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
405 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1093 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
406 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1094 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
407 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1095 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
408 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1096 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
409 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1097 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
410 PPCODE: 1098 PPCODE:
411 cl_uint value [1]; 1099 cl_uint value [1];
412 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1100 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
413 EXTEND (SP, 1); 1101 EXTEND (SP, 1);
414 const int i = 0; 1102 const int i = 0;
443 EXTEND (SP, n); 1131 EXTEND (SP, n);
444 for (i = 0; i < n; ++i) 1132 for (i = 0; i < n; ++i)
445 PUSHs (sv_2mortal (newSVuv (value [i]))); 1133 PUSHs (sv_2mortal (newSVuv (value [i])));
446 1134
447void 1135void
448address_bits (OpenCL::Device self)
449 PPCODE:
450 cl_bitfield value [1];
451 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
452 EXTEND (SP, 1);
453 const int i = 0;
454 PUSHs (sv_2mortal (newSVuv (value [i])));
455
456void
457max_mem_alloc_size (OpenCL::Device self)
458 ALIAS:
459 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
460 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
461 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
462 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
463 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
464 PPCODE:
465 cl_ulong value [1];
466 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
467 EXTEND (SP, 1);
468 const int i = 0;
469 PUSHs (sv_2mortal (newSVuv (value [i])));
470
471void
472single_fp_config (OpenCL::Device self)
473 ALIAS:
474 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
475 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
476 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
477 PPCODE:
478 cl_device_fp_config value [1];
479 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
480 EXTEND (SP, 1);
481 const int i = 0;
482 PUSHs (sv_2mortal (newSVuv (value [i])));
483
484void
485global_mem_cache_type (OpenCL::Device self)
486 PPCODE:
487 cl_device_mem_cache_type value [1];
488 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
489 EXTEND (SP, 1);
490 const int i = 0;
491 PUSHs (sv_2mortal (newSVuv (value [i])));
492
493void
494local_mem_type (OpenCL::Device self)
495 PPCODE:
496 cl_device_local_mem_type value [1];
497 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
498 EXTEND (SP, 1);
499 const int i = 0;
500 PUSHs (sv_2mortal (newSVuv (value [i])));
501
502void
503error_correction_support (OpenCL::Device self) 1136error_correction_support (OpenCL::Device self)
504 ALIAS: 1137 ALIAS:
505 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1138 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
506 endian_little = CL_DEVICE_ENDIAN_LITTLE 1139 endian_little = CL_DEVICE_ENDIAN_LITTLE
507 available = CL_DEVICE_AVAILABLE 1140 available = CL_DEVICE_AVAILABLE
508 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1141 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
509 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1142 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
510 PPCODE: 1143 PPCODE:
511 cl_bool value [1]; 1144 cl_uint value [1];
512 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1145 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
513 EXTEND (SP, 1); 1146 EXTEND (SP, 1);
514 const int i = 0; 1147 const int i = 0;
515 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1148 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
516
517void
518execution_capabilities (OpenCL::Device self)
519 PPCODE:
520 cl_device_exec_capabilities value [1];
521 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
522 EXTEND (SP, 1);
523 const int i = 0;
524 PUSHs (sv_2mortal (newSVuv (value [i])));
525
526void
527properties (OpenCL::Device self)
528 PPCODE:
529 cl_command_queue_properties value [1];
530 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
531 EXTEND (SP, 1);
532 const int i = 0;
533 PUSHs (sv_2mortal (newSViv (value [i])));
534 1149
535void 1150void
536platform (OpenCL::Device self) 1151platform (OpenCL::Device self)
537 PPCODE: 1152 PPCODE:
538 cl_platform_id value [1]; 1153 cl_platform_id value [1];
539 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1154 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
540 EXTEND (SP, 1); 1155 EXTEND (SP, 1);
541 const int i = 0; 1156 const int i = 0;
542 { 1157 PUSH_CLOBJ (stash_platform, value [i]);
543 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i]));
544 }
545 1158
546void 1159void
547name (OpenCL::Device self) 1160name (OpenCL::Device self)
548 ALIAS: 1161 ALIAS:
549 name = CL_DEVICE_NAME 1162 name = CL_DEVICE_NAME
566 PPCODE: 1179 PPCODE:
567 cl_device_id value [1]; 1180 cl_device_id value [1];
568 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1181 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
569 EXTEND (SP, 1); 1182 EXTEND (SP, 1);
570 const int i = 0; 1183 const int i = 0;
571 { 1184 PUSH_CLOBJ (stash_device, value [i]);
572 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
573 }
574 1185
575void 1186void
576partition_types_ext (OpenCL::Device self) 1187partition_types_ext (OpenCL::Device self)
577 ALIAS: 1188 ALIAS:
578 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1189 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
588 for (i = 0; i < n; ++i) 1199 for (i = 0; i < n; ++i)
589 PUSHs (sv_2mortal (newSVuv (value [i]))); 1200 PUSHs (sv_2mortal (newSVuv (value [i])));
590 1201
591#END:device 1202#END:device
592 1203
1204MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1205
1206#if CL_VERSION_1_2
1207
1208void
1209DESTROY (OpenCL::SubDevice self)
1210 CODE:
1211 clReleaseDevice (self);
1212
1213#endif
1214
593MODULE = OpenCL PACKAGE = OpenCL::Context 1215MODULE = OpenCL PACKAGE = OpenCL::Context
594 1216
595void 1217void
596DESTROY (OpenCL::Context context) 1218DESTROY (OpenCL::Context self)
597 CODE: 1219 CODE:
598 clReleaseContext (context); 1220 clReleaseContext (self);
599 1221
600void 1222void
601info (OpenCL::Context self, cl_context_info name) 1223info (OpenCL::Context self, cl_context_info name)
602 PPCODE: 1224 PPCODE:
603 INFO (Context) 1225 INFO (Context)
604 1226
605void 1227void
606queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1228queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
607 PPCODE: 1229 PPCODE:
608 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1230 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
609 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1231 XPUSH_CLOBJ (stash_queue, queue);
610 1232
611void 1233void
612user_event (OpenCL::Context self) 1234user_event (OpenCL::Context self)
613 PPCODE: 1235 PPCODE:
614 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1236 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
615 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1237 XPUSH_CLOBJ (stash_userevent, ev);
616 1238
617void 1239void
618buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1240buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
619 PPCODE: 1241 PPCODE:
620 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1242 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
621 croak ("clCreateBuffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1243 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
622 1244
623 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1245 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
624 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1246 XPUSH_CLOBJ (stash_bufferobj, mem);
625 1247
626void 1248void
627buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1249buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
628 PPCODE: 1250 PPCODE:
629 STRLEN len; 1251 STRLEN len;
630 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1252 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
631 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1253 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
632 croak ("clCreateBuffer: have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1254 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
633 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1255 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
634 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1256 XPUSH_CLOBJ (stash_bufferobj, mem);
635 1257
1258#if CL_VERSION_1_2
1259
636void 1260void
637image2d (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) 1261image (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)
638 PPCODE: 1262 PPCODE:
639 STRLEN len; 1263 STRLEN len;
640 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1264 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
641 const cl_image_format format = { channel_order, channel_type }; 1265 const cl_image_format format = { channel_order, channel_type };
1266 const cl_image_desc desc = {
1267 type,
1268 width, height, depth,
1269 array_size, row_pitch, slice_pitch,
1270 num_mip_level, num_samples,
1271 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1272 };
642 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1273 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
643 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1274 HV *stash = stash_image;
1275 switch (type)
1276 {
1277 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1278 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1279 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1280 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1281 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1282 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1283 }
1284 XPUSH_CLOBJ (stash, mem);
644 1285
1286#endif
1287
645void 1288void
646image3d (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) 1289image2d (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)
647 PPCODE: 1290 PPCODE:
648 STRLEN len; 1291 STRLEN len;
649 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1292 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
650 const cl_image_format format = { channel_order, channel_type }; 1293 const cl_image_format format = { channel_order, channel_type };
1294#if PREFER_1_1
1295 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1296#else
1297 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1298 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1299#endif
1300 XPUSH_CLOBJ (stash_image2d, mem);
1301
1302void
1303image3d (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)
1304 PPCODE:
1305 STRLEN len;
1306 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1307 const cl_image_format format = { channel_order, channel_type };
1308#if PREFER_1_1
651 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1309 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
652 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1310#else
1311 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1312 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1313#endif
1314 XPUSH_CLOBJ (stash_image3d, mem);
653 1315
654#if cl_apple_gl_sharing || cl_khr_gl_sharing 1316#if cl_apple_gl_sharing || cl_khr_gl_sharing
655 1317
656void 1318void
657gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1319gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
658 PPCODE: 1320 PPCODE:
659 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1321 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
660 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1322 XPUSH_CLOBJ (stash_bufferobj, mem);
1323
1324void
1325gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1326 PPCODE:
1327 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1328 XPUSH_CLOBJ (stash_image2d, mem);
1329
1330#if CL_VERSION_1_2
1331
1332void
1333gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1334 ALIAS:
1335 PPCODE:
1336 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1337 cl_gl_object_type type;
1338 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1339 HV *stash = stash_memory;
1340 switch (type)
1341 {
1342 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1343 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1344 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1345 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1346 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1347 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1348 }
1349 XPUSH_CLOBJ (stash, mem);
1350
1351#endif
661 1352
662void 1353void
663gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1354gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
664 PPCODE: 1355 PPCODE:
1356#if PREFER_1_1
665 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1357 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
666 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1358#else
1359 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1360#endif
1361 XPUSH_CLOBJ (stash_image2d, mem);
667 1362
668void 1363void
669gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1364gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
670 PPCODE: 1365 PPCODE:
1366#if PREFER_1_1
671 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
672 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1368#else
673 1369 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
674void 1370#endif
675gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1371 XPUSH_CLOBJ (stash_image3d, mem);
676 PPCODE:
677 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
678 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
679 1372
680#endif 1373#endif
681 1374
682void 1375void
683supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1376supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
703 1396
704void 1397void
705sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1398sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
706 PPCODE: 1399 PPCODE:
707 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1400 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
708 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1401 XPUSH_CLOBJ (stash_sampler, sampler);
709 1402
710void 1403void
711program_with_source (OpenCL::Context self, SV *program) 1404program_with_source (OpenCL::Context self, SV *program)
712 PPCODE: 1405 PPCODE:
713 STRLEN len; 1406 STRLEN len;
714 size_t len2; 1407 size_t len2;
715 const char *ptr = SvPVbyte (program, len); 1408 const char *ptr = SvPVbyte (program, len);
716 1409
717 len2 = len; 1410 len2 = len;
718 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1411 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
719 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1412 XPUSH_CLOBJ (stash_program, prog);
1413
1414void
1415program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1416 PPCODE:
1417 cl_uint device_count;
1418 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1419
1420 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1421 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1422
1423 binaries = SvRV (binaries);
1424
1425 if (device_count != av_len ((AV *)binaries) + 1)
1426 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1427
1428 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1429 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1430 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1431
1432 int i;
1433 for (i = 0; i < device_count; ++i)
1434 {
1435 STRLEN len;
1436 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1437 length_list [i] = len;
1438 }
1439
1440 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1441 length_list, binary_list,
1442 GIMME_V == G_ARRAY ? status_list : 0, &res));
1443
1444 EXTEND (SP, 2);
1445 PUSH_CLOBJ (stash_program, prog);
1446
1447 if (GIMME_V == G_ARRAY)
1448 {
1449 AV *av = newAV ();
1450 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1451
1452 for (i = device_count; i--; )
1453 av_store (av, i, newSViv (status_list [i]));
1454 }
1455
1456#if CL_VERSION_1_2
1457
1458void
1459program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1460 PPCODE:
1461 cl_uint device_count;
1462 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1463
1464 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1465
1466 XPUSH_CLOBJ (stash_program, prog);
1467
1468void
1469link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1470 CODE:
1471 cl_uint device_count = 0;
1472 cl_device_id *device_list = 0;
1473
1474 if (SvOK (devices))
1475 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1476
1477 cl_uint program_count;
1478 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1479
1480 void *user_data;
1481 program_callback pfn_notify = make_program_callback (notify, &user_data);
1482
1483 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1484 program_count, program_list, pfn_notify, user_data, &res));
1485
1486 XPUSH_CLOBJ (stash_program, prog);
1487
1488#endif
720 1489
721#BEGIN:context 1490#BEGIN:context
722 1491
723void 1492void
724reference_count (OpenCL::Context self) 1493reference_count (OpenCL::Context self)
740 cl_device_id *value = tmpbuf (size); 1509 cl_device_id *value = tmpbuf (size);
741 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1510 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
742 int i, n = size / sizeof (*value); 1511 int i, n = size / sizeof (*value);
743 EXTEND (SP, n); 1512 EXTEND (SP, n);
744 for (i = 0; i < n; ++i) 1513 for (i = 0; i < n; ++i)
745 { 1514 PUSH_CLOBJ (stash_device, value [i]);
746 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
747 }
748 1515
749void 1516void
750properties (OpenCL::Context self) 1517properties (OpenCL::Context self)
751 PPCODE: 1518 PPCODE:
752 size_t size; 1519 size_t size;
766DESTROY (OpenCL::Queue self) 1533DESTROY (OpenCL::Queue self)
767 CODE: 1534 CODE:
768 clReleaseCommandQueue (self); 1535 clReleaseCommandQueue (self);
769 1536
770void 1537void
771enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1538read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1539 ALIAS:
1540 enqueue_read_buffer = 0
772 PPCODE: 1541 PPCODE:
773 cl_event ev = 0;
774 EVENT_LIST (6, items - 6); 1542 EVENT_LIST (6);
775 1543
776 SvUPGRADE (data, SVt_PV); 1544 SvUPGRADE (data, SVt_PV);
777 SvGROW (data, len); 1545 SvGROW (data, len);
778 SvPOK_only (data); 1546 SvPOK_only (data);
779 SvCUR_set (data, len); 1547 SvCUR_set (data, len);
1548
1549 cl_event ev = 0;
780 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1550 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
781 1551
782 if (ev) 1552 if (ev)
783 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1553 XPUSH_CLOBJ (stash_event, ev);
784 1554
785void 1555void
786enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1556write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1557 ALIAS:
1558 enqueue_write_buffer = 0
787 PPCODE: 1559 PPCODE:
788 cl_event ev = 0; 1560 EVENT_LIST (5);
1561
789 STRLEN len; 1562 STRLEN len;
790 char *ptr = SvPVbyte (data, len); 1563 char *ptr = SvPVbyte (data, len);
791 EVENT_LIST (5, items - 5);
792 1564
1565 cl_event ev = 0;
793 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1566 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
794 1567
795 if (ev) 1568 if (ev)
796 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1569 XPUSH_CLOBJ (stash_event, ev);
797 1570
1571#if CL_VERSION_1_2
1572
798void 1573void
799enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1574fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1575 ALIAS:
1576 enqueue_fill_buffer = 0
800 PPCODE: 1577 PPCODE:
1578 EVENT_LIST (5);
1579
1580 STRLEN len;
1581 char *ptr = SvPVbyte (data, len);
1582
801 cl_event ev = 0; 1583 cl_event ev = 0;
1584 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1585
1586 if (ev)
1587 XPUSH_CLOBJ (stash_event, ev);
1588
1589void
1590fill_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, ...)
1591 ALIAS:
1592 enqueue_fill_image = 0
1593 PPCODE:
1594 EVENT_LIST (12);
1595
1596 const size_t origin [3] = { x, y, z };
1597 const size_t region [3] = { width, height, depth };
1598
1599 const cl_float c_f [4] = { r, g, b, a };
1600 const cl_uint c_u [4] = { r, g, b, a };
1601 const cl_int c_s [4] = { r, g, b, a };
1602 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1603 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1604 cl_image_format format;
1605 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1606 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1607 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1608 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1609
1610 cl_event ev = 0;
1611 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1612 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1613
1614 if (ev)
1615 XPUSH_CLOBJ (stash_event, ev);
1616
1617#endif
1618
1619void
1620copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1621 ALIAS:
1622 enqueue_copy_buffer = 0
1623 PPCODE:
802 EVENT_LIST (6, items - 6); 1624 EVENT_LIST (6);
803 1625
1626 cl_event ev = 0;
804 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1627 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
805 1628
806 if (ev) 1629 if (ev)
807 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1630 XPUSH_CLOBJ (stash_event, ev);
808 1631
809void 1632void
810enqueue_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, ...) 1633read_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, ...)
1634 ALIAS:
1635 enqueue_read_buffer_rect = 0
811 PPCODE: 1636 PPCODE:
812 cl_event ev = 0; 1637 EVENT_LIST (17);
1638
813 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1639 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
814 const size_t host_origin[3] = { host_x, host_y, host_z }; 1640 const size_t host_origin[3] = { host_x, host_y, host_z };
815 const size_t region[3] = { width, height, depth }; 1641 const size_t region[3] = { width, height, depth };
816 EVENT_LIST (17, items - 17);
817 1642
818 if (!buf_row_pitch) 1643 if (!buf_row_pitch)
819 buf_row_pitch = region [0]; 1644 buf_row_pitch = region [0];
820 1645
821 if (!buf_slice_pitch) 1646 if (!buf_slice_pitch)
831 1656
832 SvUPGRADE (data, SVt_PV); 1657 SvUPGRADE (data, SVt_PV);
833 SvGROW (data, len); 1658 SvGROW (data, len);
834 SvPOK_only (data); 1659 SvPOK_only (data);
835 SvCUR_set (data, len); 1660 SvCUR_set (data, len);
1661
1662 cl_event ev = 0;
836 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)); 1663 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));
837 1664
838 if (ev) 1665 if (ev)
839 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1666 XPUSH_CLOBJ (stash_event, ev);
840 1667
841void 1668void
842enqueue_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, ...) 1669write_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, ...)
1670 ALIAS:
1671 enqueue_write_buffer_rect = 0
843 PPCODE: 1672 PPCODE:
844 cl_event ev = 0; 1673 EVENT_LIST (17);
1674
845 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1675 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
846 const size_t host_origin[3] = { host_x, host_y, host_z }; 1676 const size_t host_origin[3] = { host_x, host_y, host_z };
847 const size_t region[3] = { width, height, depth }; 1677 const size_t region[3] = { width, height, depth };
848 STRLEN len; 1678 STRLEN len;
849 char *ptr = SvPVbyte (data, len); 1679 char *ptr = SvPVbyte (data, len);
850 EVENT_LIST (17, items - 17);
851 1680
852 if (!buf_row_pitch) 1681 if (!buf_row_pitch)
853 buf_row_pitch = region [0]; 1682 buf_row_pitch = region [0];
854 1683
855 if (!buf_slice_pitch) 1684 if (!buf_slice_pitch)
864 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1693 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
865 1694
866 if (len < min_len) 1695 if (len < min_len)
867 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1696 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
868 1697
1698 cl_event ev = 0;
869 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)); 1699 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));
870 1700
871 if (ev) 1701 if (ev)
872 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1702 XPUSH_CLOBJ (stash_event, ev);
873 1703
874void 1704void
875enqueue_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, ...) 1705copy_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, ...)
1706 ALIAS:
1707 enqueue_copy_buffer_rect = 0
876 PPCODE: 1708 PPCODE:
877 cl_event ev = 0; 1709 EVENT_LIST (16);
1710
878 const size_t src_origin[3] = { src_x, src_y, src_z }; 1711 const size_t src_origin[3] = { src_x, src_y, src_z };
879 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1712 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
880 const size_t region[3] = { width, height, depth }; 1713 const size_t region[3] = { width, height, depth };
881 EVENT_LIST (16, items - 16);
882 1714
1715 cl_event ev = 0;
883 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)); 1716 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));
884 1717
885 if (ev) 1718 if (ev)
886 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1719 XPUSH_CLOBJ (stash_event, ev);
887 1720
888void 1721void
889enqueue_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, ...) 1722read_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, ...)
1723 ALIAS:
1724 enqueue_read_image = 0
890 PPCODE: 1725 PPCODE:
891 cl_event ev = 0; 1726 EVENT_LIST (12);
1727
892 const size_t src_origin[3] = { src_x, src_y, src_z }; 1728 const size_t src_origin[3] = { src_x, src_y, src_z };
893 const size_t region[3] = { width, height, depth }; 1729 const size_t region[3] = { width, height, depth };
894 EVENT_LIST (12, items - 12);
895 1730
896 if (!row_pitch) 1731 if (!row_pitch)
897 row_pitch = img_row_pitch (src); 1732 row_pitch = img_row_pitch (src);
898 1733
899 if (depth > 1 && !slice_pitch) 1734 if (depth > 1 && !slice_pitch)
903 1738
904 SvUPGRADE (data, SVt_PV); 1739 SvUPGRADE (data, SVt_PV);
905 SvGROW (data, len); 1740 SvGROW (data, len);
906 SvPOK_only (data); 1741 SvPOK_only (data);
907 SvCUR_set (data, len); 1742 SvCUR_set (data, len);
1743
1744 cl_event ev = 0;
908 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)); 1745 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));
909 1746
910 if (ev) 1747 if (ev)
911 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1748 XPUSH_CLOBJ (stash_event, ev);
912 1749
913void 1750void
914enqueue_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, ...) 1751write_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, ...)
1752 ALIAS:
1753 enqueue_write_image = 0
915 PPCODE: 1754 PPCODE:
916 cl_event ev = 0; 1755 EVENT_LIST (12);
1756
917 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1757 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
918 const size_t region[3] = { width, height, depth }; 1758 const size_t region[3] = { width, height, depth };
919 STRLEN len; 1759 STRLEN len;
920 char *ptr = SvPVbyte (data, len); 1760 char *ptr = SvPVbyte (data, len);
921 EVENT_LIST (12, items - 12);
922 1761
923 if (!row_pitch) 1762 if (!row_pitch)
924 row_pitch = img_row_pitch (dst); 1763 row_pitch = img_row_pitch (dst);
925 1764
926 if (depth > 1 && !slice_pitch) 1765 if (depth > 1 && !slice_pitch)
929 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1768 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
930 1769
931 if (len < min_len) 1770 if (len < min_len)
932 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1771 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
933 1772
1773 cl_event ev = 0;
934 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)); 1774 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));
935 1775
936 if (ev) 1776 if (ev)
937 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1777 XPUSH_CLOBJ (stash_event, ev);
938 1778
939void 1779void
940enqueue_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, ...) 1780copy_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, ...)
1781 ALIAS:
1782 enqueue_copy_image = 0
941 PPCODE: 1783 PPCODE:
942 cl_event ev = 0; 1784 EVENT_LIST (12);
1785
943 const size_t src_origin[3] = { src_x, src_y, src_z }; 1786 const size_t src_origin[3] = { src_x, src_y, src_z };
944 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1787 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
945 const size_t region[3] = { width, height, depth }; 1788 const size_t region[3] = { width, height, depth };
946 EVENT_LIST (12, items - 12);
947 1789
1790 cl_event ev = 0;
948 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1791 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
949 1792
950 if (ev) 1793 if (ev)
951 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1794 XPUSH_CLOBJ (stash_event, ev);
952 1795
953void 1796void
954enqueue_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, ...) 1797copy_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, ...)
1798 ALIAS:
1799 enqueue_copy_image_to_buffer = 0
955 PPCODE: 1800 PPCODE:
1801 EVENT_LIST (10);
1802
1803 const size_t src_origin[3] = { src_x, src_y, src_z };
1804 const size_t region [3] = { width, height, depth };
1805
956 cl_event ev = 0; 1806 cl_event ev = 0;
1807 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1808
1809 if (ev)
1810 XPUSH_CLOBJ (stash_event, ev);
1811
1812void
1813copy_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, ...)
1814 ALIAS:
1815 enqueue_copy_buffer_to_image = 0
1816 PPCODE:
1817 EVENT_LIST (10);
1818
957 const size_t src_origin[3] = { src_x, src_y, src_z }; 1819 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1820 const size_t region [3] = { width, height, depth };
1821
1822 cl_event ev = 0;
1823 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1824
1825 if (ev)
1826 XPUSH_CLOBJ (stash_event, ev);
1827
1828void
1829map_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, ...)
1830 ALIAS:
1831 enqueue_map_buffer = 0
1832 PPCODE:
1833 EVENT_LIST (6);
1834
1835 size_t cb = SvIV (cb_);
1836
1837 if (!SvOK (cb_))
1838 {
1839 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1840 cb -= offset;
1841 }
1842
1843 cl_event ev;
1844 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1845 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1846
1847void
1848map_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, ...)
1849 ALIAS:
1850 enqueue_map_image = 0
1851 PPCODE:
1852 size_t width = SvIV (width_);
1853 if (!SvOK (width_))
1854 {
1855 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1856 width -= x;
1857 }
1858
1859 size_t height = SvIV (width_);
1860 if (!SvOK (height_))
1861 {
1862 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1863 height -= y;
1864 }
1865
1866 size_t depth = SvIV (width_);
1867 if (!SvOK (depth_))
1868 {
1869 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1870 depth -= z;
1871
1872 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1873 if (!depth)
1874 depth = 1;
1875 }
1876
1877 const size_t origin[3] = { x, y, z };
958 const size_t region[3] = { width, height, depth }; 1878 const size_t region[3] = { width, height, depth };
1879 size_t row_pitch, slice_pitch;
959 EVENT_LIST (10, items - 10); 1880 EVENT_LIST (10);
960 1881
1882 cl_event ev;
1883 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));
1884
1885 size_t cb = slice_pitch ? slice_pitch * region [2]
1886 : row_pitch ? row_pitch * region [1]
1887 : region [0];
1888
1889 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1890
1891void
1892unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1893 PPCODE:
1894 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1895 if (GIMME_V != G_VOID)
1896 {
1897 clRetainEvent (mapped->event);
1898 XPUSH_CLOBJ (stash_event, mapped->event);
1899 }
1900
1901void
1902task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1903 ALIAS:
1904 enqueue_task = 0
1905 PPCODE:
1906 EVENT_LIST (2);
1907
1908 cl_event ev = 0;
961 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1909 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
962 1910
963 if (ev) 1911 if (ev)
964 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1912 XPUSH_CLOBJ (stash_event, ev);
965 1913
966void 1914void
967enqueue_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, ...)
968 PPCODE:
969 cl_event ev = 0;
970 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
971 const size_t region[3] = { width, height, depth };
972 EVENT_LIST (10, items - 10);
973
974 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
975
976 if (ev)
977 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
978
979void
980enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
981 PPCODE:
982 cl_event ev = 0;
983 EVENT_LIST (2, items - 2);
984
985 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
986
987 if (ev)
988 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
989
990void
991enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1915nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1916 ALIAS:
1917 enqueue_nd_range_kernel = 0
992 PPCODE: 1918 PPCODE:
993 cl_event ev = 0; 1919 EVENT_LIST (5);
1920
994 size_t *gwo = 0, *gws, *lws = 0; 1921 size_t *gwo = 0, *gws, *lws = 0;
995 int gws_len; 1922 int gws_len;
996 size_t *lists; 1923 size_t *lists;
997 int i; 1924 int i;
998 EVENT_LIST (5, items - 5);
999 1925
1000 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1926 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1001 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1927 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1002 1928
1003 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1929 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1004 1930
1005 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1931 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1006 1932
1007 gws = lists + gws_len * 0; 1933 gws = lists + gws_len * 0;
1008 for (i = 0; i < gws_len; ++i) 1934 for (i = 0; i < gws_len; ++i)
1935 {
1009 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1936 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1937 // at least nvidia crashes for 0-sized work group sizes, work around
1938 if (!gws [i])
1939 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1940 }
1010 1941
1011 if (SvOK (global_work_offset)) 1942 if (SvOK (global_work_offset))
1012 { 1943 {
1013 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1944 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1014 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1945 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1021 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1952 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1022 } 1953 }
1023 1954
1024 if (SvOK (local_work_size)) 1955 if (SvOK (local_work_size))
1025 { 1956 {
1026 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1957 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1027 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1958 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1028 1959
1029 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1960 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1030 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1961 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1031 1962
1032 lws = lists + gws_len * 2; 1963 lws = lists + gws_len * 2;
1033 for (i = 0; i < gws_len; ++i) 1964 for (i = 0; i < gws_len; ++i)
1965 {
1034 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1966 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1967 // at least nvidia crashes for 0-sized work group sizes, work around
1968 if (!lws [i])
1969 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1970 }
1035 } 1971 }
1036 1972
1973 cl_event ev = 0;
1037 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1974 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1038 1975
1039 if (ev) 1976 if (ev)
1040 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1977 XPUSH_CLOBJ (stash_event, ev);
1041 1978
1042#if 0 1979#if CL_VERSION_1_2
1980
1981void
1982migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1983 ALIAS:
1984 enqueue_migrate_mem_objects = 0
1985 PPCODE:
1986 EVENT_LIST (3);
1987
1988 cl_uint obj_count;
1989 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1990
1991 cl_event ev = 0;
1992 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1993
1994 if (ev)
1995 XPUSH_CLOBJ (stash_event, ev);
1996
1997#endif
1998
1043#if cl_apple_gl_sharing || cl_khr_gl_sharing 1999#if cl_apple_gl_sharing || cl_khr_gl_sharing
1044 2000
1045void 2001void
1046enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2002acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2003 ALIAS:
2004 release_gl_objects = 1
2005 enqueue_acquire_gl_objects = 0
2006 enqueue_release_gl_objects = 1
1047 CODE: 2007 PPCODE:
2008 EVENT_LIST (2);
2009
2010 cl_uint obj_count;
2011 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2012
1048 cl_event ev = 0; 2013 cl_event ev = 0;
1049 EVENT_LIST (2, items - 2);
1050 2014
2015 if (ix)
2016 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2017 else
1051 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, ..., event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2018 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1052 2019
1053 if (ev) 2020 if (ev)
1054 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2021 XPUSH_CLOBJ (stash_event, ev);
1055 2022
2023#endif
2024
1056void 2025void
1057enqueue_release_gl_objects (OpenCL::Queue self, SV *objects, ...) 2026wait_for_events (OpenCL::Queue self, ...)
2027 ALIAS:
2028 enqueue_wait_for_events = 0
2029 CODE:
2030 EVENT_LIST (1);
2031#if PREFER_1_1
2032 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2033#else
2034 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2035#endif
2036
2037void
2038marker (OpenCL::Queue self, ...)
2039 ALIAS:
2040 enqueue_marker = 0
1058 CODE: 2041 PPCODE:
2042 EVENT_LIST (1);
1059 cl_event ev = 0; 2043 cl_event ev = 0;
1060 EVENT_LIST (2, items - 2); 2044#if PREFER_1_1
1061 2045 if (!event_list_count)
2046 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2047 else
2048#if CL_VERSION_1_2
2049 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2050#else
2051 {
2052 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2053 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2054 }
2055#endif
2056#else
1062 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, ..., event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2057 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1063 2058#endif
1064 if (ev) 2059 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2060 XPUSH_CLOBJ (stash_event, ev);
1066 2061
2062void
2063barrier (OpenCL::Queue self, ...)
2064 ALIAS:
2065 enqueue_barrier = 0
2066 PPCODE:
2067 EVENT_LIST (1);
2068 cl_event ev = 0;
2069#if PREFER_1_1
2070 if (!event_list_count && GIMME_V == G_VOID)
2071 NEED_SUCCESS (EnqueueBarrier, (self));
2072 else
2073#if CL_VERSION_1_2
2074 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2075#else
2076 {
2077 if (event_list_count)
2078 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2079
2080 if (GIMME_V != G_VOID)
2081 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2082 }
1067#endif 2083#endif
2084#else
2085 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1068#endif 2086#endif
1069 2087 if (ev)
1070void 2088 XPUSH_CLOBJ (stash_event, ev);
1071enqueue_marker (OpenCL::Queue self)
1072 PPCODE:
1073 cl_event ev;
1074 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1075 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1076
1077void
1078enqueue_wait_for_events (OpenCL::Queue self, ...)
1079 CODE:
1080 EVENT_LIST (1, items - 1);
1081 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1082
1083void
1084enqueue_barrier (OpenCL::Queue self)
1085 CODE:
1086 NEED_SUCCESS (EnqueueBarrier, (self));
1087 2089
1088void 2090void
1089flush (OpenCL::Queue self) 2091flush (OpenCL::Queue self)
1090 CODE: 2092 CODE:
1091 NEED_SUCCESS (Flush, (self)); 2093 NEED_SUCCESS (Flush, (self));
1107 PPCODE: 2109 PPCODE:
1108 cl_context value [1]; 2110 cl_context value [1];
1109 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2111 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1110 EXTEND (SP, 1); 2112 EXTEND (SP, 1);
1111 const int i = 0; 2113 const int i = 0;
1112 {
1113 NEED_SUCCESS (RetainContext, (value [i])); 2114 NEED_SUCCESS (RetainContext, (value [i]));
1114 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2115 PUSH_CLOBJ (stash_context, value [i]);
1115 }
1116 2116
1117void 2117void
1118device (OpenCL::Queue self) 2118device (OpenCL::Queue self)
1119 PPCODE: 2119 PPCODE:
1120 cl_device_id value [1]; 2120 cl_device_id value [1];
1121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1122 EXTEND (SP, 1); 2122 EXTEND (SP, 1);
1123 const int i = 0; 2123 const int i = 0;
1124 { 2124 PUSH_CLOBJ (stash_device, value [i]);
1125 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1126 }
1127 2125
1128void 2126void
1129reference_count (OpenCL::Queue self) 2127reference_count (OpenCL::Queue self)
1130 PPCODE: 2128 PPCODE:
1131 cl_uint value [1]; 2129 cl_uint value [1];
1135 PUSHs (sv_2mortal (newSVuv (value [i]))); 2133 PUSHs (sv_2mortal (newSVuv (value [i])));
1136 2134
1137void 2135void
1138properties (OpenCL::Queue self) 2136properties (OpenCL::Queue self)
1139 PPCODE: 2137 PPCODE:
1140 cl_command_queue_properties value [1]; 2138 cl_ulong value [1];
1141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2139 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1142 EXTEND (SP, 1); 2140 EXTEND (SP, 1);
1143 const int i = 0; 2141 const int i = 0;
1144 PUSHs (sv_2mortal (newSViv (value [i]))); 2142 PUSHs (sv_2mortal (newSVuv (value [i])));
1145 2143
1146#END:command_queue 2144#END:command_queue
1147 2145
1148MODULE = OpenCL PACKAGE = OpenCL::Memory 2146MODULE = OpenCL PACKAGE = OpenCL::Memory
1149 2147
1155void 2153void
1156info (OpenCL::Memory self, cl_mem_info name) 2154info (OpenCL::Memory self, cl_mem_info name)
1157 PPCODE: 2155 PPCODE:
1158 INFO (MemObject) 2156 INFO (MemObject)
1159 2157
2158void
2159destructor_callback (OpenCL::Memory self, SV *notify)
2160 PPCODE:
2161 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2162
1160#BEGIN:mem 2163#BEGIN:mem
1161 2164
1162void 2165void
1163type (OpenCL::Memory self) 2166type (OpenCL::Memory self)
2167 ALIAS:
2168 type = CL_MEM_TYPE
2169 map_count = CL_MEM_MAP_COUNT
2170 reference_count = CL_MEM_REFERENCE_COUNT
1164 PPCODE: 2171 PPCODE:
1165 cl_mem_object_type value [1]; 2172 cl_uint value [1];
1166 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2173 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1167 EXTEND (SP, 1); 2174 EXTEND (SP, 1);
1168 const int i = 0; 2175 const int i = 0;
1169 PUSHs (sv_2mortal (newSViv (value [i]))); 2176 PUSHs (sv_2mortal (newSVuv (value [i])));
1170 2177
1171void 2178void
1172flags (OpenCL::Memory self) 2179flags (OpenCL::Memory self)
1173 PPCODE: 2180 PPCODE:
1174 cl_mem_flags value [1]; 2181 cl_ulong value [1];
1175 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2182 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1176 EXTEND (SP, 1); 2183 EXTEND (SP, 1);
1177 const int i = 0; 2184 const int i = 0;
1178 PUSHs (sv_2mortal (newSViv (value [i]))); 2185 PUSHs (sv_2mortal (newSVuv (value [i])));
1179 2186
1180void 2187void
1181size (OpenCL::Memory self) 2188size (OpenCL::Memory self)
1182 ALIAS: 2189 ALIAS:
1183 size = CL_MEM_SIZE 2190 size = CL_MEM_SIZE
1197 EXTEND (SP, 1); 2204 EXTEND (SP, 1);
1198 const int i = 0; 2205 const int i = 0;
1199 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2206 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1200 2207
1201void 2208void
1202map_count (OpenCL::Memory self)
1203 ALIAS:
1204 map_count = CL_MEM_MAP_COUNT
1205 reference_count = CL_MEM_REFERENCE_COUNT
1206 PPCODE:
1207 cl_uint value [1];
1208 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1209 EXTEND (SP, 1);
1210 const int i = 0;
1211 PUSHs (sv_2mortal (newSVuv (value [i])));
1212
1213void
1214context (OpenCL::Memory self) 2209context (OpenCL::Memory self)
1215 PPCODE: 2210 PPCODE:
1216 cl_context value [1]; 2211 cl_context value [1];
1217 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1218 EXTEND (SP, 1); 2213 EXTEND (SP, 1);
1219 const int i = 0; 2214 const int i = 0;
1220 {
1221 NEED_SUCCESS (RetainContext, (value [i])); 2215 NEED_SUCCESS (RetainContext, (value [i]));
1222 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2216 PUSH_CLOBJ (stash_context, value [i]);
1223 }
1224 2217
1225void 2218void
1226associated_memobject (OpenCL::Memory self) 2219associated_memobject (OpenCL::Memory self)
1227 PPCODE: 2220 PPCODE:
1228 cl_mem value [1]; 2221 cl_mem value [1];
1229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2222 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1230 EXTEND (SP, 1); 2223 EXTEND (SP, 1);
1231 const int i = 0; 2224 const int i = 0;
1232 {
1233 NEED_SUCCESS (RetainMemObject, (value [i])); 2225 NEED_SUCCESS (RetainMemObject, (value [i]));
1234 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2226 PUSH_CLOBJ (stash_memory, value [i]);
1235 }
1236 2227
1237#END:mem 2228#END:mem
1238 2229
1239#if cl_apple_gl_sharing || cl_khr_gl_sharing 2230#if cl_apple_gl_sharing || cl_khr_gl_sharing
1240 2231
1241void 2232void
1242gl_object_info (OpenCL::Memory self) 2233gl_object_info (OpenCL::Memory self)
1243 PPCODE: 2234 PPCODE:
1244 cl_gl_object_type type; 2235 cl_gl_object_type type;
1245 cl_GLuint name; 2236 cl_GLuint name;
1246 NEED_SUCCESS (clGetGLObjectInfo, (self, &type, &name)); 2237 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
1247 EXTEND (SP, 2); 2238 EXTEND (SP, 2);
1248 PUSHs (sv_2mortal (newSVuv (type))); 2239 PUSHs (sv_2mortal (newSVuv (type)));
1249 PUSHs (sv_2mortal (newSVuv (name))); 2240 PUSHs (sv_2mortal (newSVuv (name)));
1250 2241
1251#endif 2242#endif
1259 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2250 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1260 2251
1261 cl_buffer_region crdata = { origin, size }; 2252 cl_buffer_region crdata = { origin, size };
1262 2253
1263 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2254 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1264 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2255 XPUSH_CLOBJ (stash_buffer, mem);
1265 2256
1266MODULE = OpenCL PACKAGE = OpenCL::Image 2257MODULE = OpenCL PACKAGE = OpenCL::Image
1267 2258
1268void 2259void
1269image_info (OpenCL::Image self, cl_image_info name) 2260image_info (OpenCL::Image self, cl_image_info name)
1270 PPCODE: 2261 PPCODE:
1271 INFO (Image) 2262 INFO (Image)
2263
2264void
2265format (OpenCL::Image self)
2266 PPCODE:
2267 cl_image_format format;
2268 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2269 EXTEND (SP, 2);
2270 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2271 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1272 2272
1273#BEGIN:image 2273#BEGIN:image
1274 2274
1275void 2275void
1276element_size (OpenCL::Image self) 2276element_size (OpenCL::Image self)
1296 2296
1297void 2297void
1298target (OpenCL::Image self) 2298target (OpenCL::Image self)
1299 PPCODE: 2299 PPCODE:
1300 cl_GLenum value [1]; 2300 cl_GLenum value [1];
1301 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0)); 2301 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
1302 EXTEND (SP, 1); 2302 EXTEND (SP, 1);
1303 const int i = 0; 2303 const int i = 0;
1304 PUSHs (sv_2mortal (newSVuv (value [i]))); 2304 PUSHs (sv_2mortal (newSVuv (value [i])));
1305 2305
1306void 2306void
1307gl_mipmap_level (OpenCL::Image self) 2307gl_mipmap_level (OpenCL::Image self)
1308 PPCODE: 2308 PPCODE:
1309 cl_GLint value [1]; 2309 cl_GLint value [1];
1310 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0)); 2310 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
1311 EXTEND (SP, 1); 2311 EXTEND (SP, 1);
1312 const int i = 0; 2312 const int i = 0;
1313 PUSHs (sv_2mortal (newSViv (value [i]))); 2313 PUSHs (sv_2mortal (newSViv (value [i])));
1314 2314
1315#END:gl_texture 2315#END:gl_texture
1330 2330
1331#BEGIN:sampler 2331#BEGIN:sampler
1332 2332
1333void 2333void
1334reference_count (OpenCL::Sampler self) 2334reference_count (OpenCL::Sampler self)
2335 ALIAS:
2336 reference_count = CL_SAMPLER_REFERENCE_COUNT
2337 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2338 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1335 PPCODE: 2339 PPCODE:
1336 cl_uint value [1]; 2340 cl_uint value [1];
1337 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2341 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1338 EXTEND (SP, 1); 2342 EXTEND (SP, 1);
1339 const int i = 0; 2343 const int i = 0;
1340 PUSHs (sv_2mortal (newSVuv (value [i]))); 2344 PUSHs (sv_2mortal (newSVuv (value [i])));
1341 2345
1342void 2346void
1344 PPCODE: 2348 PPCODE:
1345 cl_context value [1]; 2349 cl_context value [1];
1346 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2350 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1347 EXTEND (SP, 1); 2351 EXTEND (SP, 1);
1348 const int i = 0; 2352 const int i = 0;
1349 {
1350 NEED_SUCCESS (RetainContext, (value [i])); 2353 NEED_SUCCESS (RetainContext, (value [i]));
1351 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2354 PUSH_CLOBJ (stash_context, value [i]);
1352 }
1353
1354void
1355normalized_coords (OpenCL::Sampler self)
1356 PPCODE:
1357 cl_addressing_mode value [1];
1358 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1359 EXTEND (SP, 1);
1360 const int i = 0;
1361 PUSHs (sv_2mortal (newSViv (value [i])));
1362
1363void
1364addressing_mode (OpenCL::Sampler self)
1365 PPCODE:
1366 cl_filter_mode value [1];
1367 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1368 EXTEND (SP, 1);
1369 const int i = 0;
1370 PUSHs (sv_2mortal (newSViv (value [i])));
1371 2355
1372void 2356void
1373filter_mode (OpenCL::Sampler self) 2357filter_mode (OpenCL::Sampler self)
1374 PPCODE: 2358 PPCODE:
1375 cl_bool value [1]; 2359 cl_uint value [1];
1376 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2360 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1377 EXTEND (SP, 1); 2361 EXTEND (SP, 1);
1378 const int i = 0; 2362 const int i = 0;
1379 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2363 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1380 2364
1386DESTROY (OpenCL::Program self) 2370DESTROY (OpenCL::Program self)
1387 CODE: 2371 CODE:
1388 clReleaseProgram (self); 2372 clReleaseProgram (self);
1389 2373
1390void 2374void
1391build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2375build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2376 ALIAS:
2377 build_async = 1
1392 CODE: 2378 CODE:
2379 cl_uint device_count = 0;
2380 cl_device_id *device_list = 0;
2381
2382 if (SvOK (devices))
2383 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2384
2385 void *user_data;
2386 program_callback pfn_notify = make_program_callback (notify, &user_data);
2387
2388 if (ix)
2389 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2390 else
2391 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2392
2393#if CL_VERSION_1_2
2394
2395void
2396compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2397 CODE:
2398 cl_uint device_count = 0;
2399 cl_device_id *device_list = 0;
2400
2401 if (SvOK (devices))
2402 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2403
2404 cl_uint header_count = 0;
2405 cl_program *header_list = 0;
2406 const char **header_name = 0;
2407
2408 if (SvOK (headers))
2409 {
2410 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2411 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2412
2413 HV *hv = (HV *)SvRV (devices);
2414
2415 header_count = hv_iterinit (hv);
2416 header_list = tmpbuf (sizeof (*header_list) * header_count);
2417 header_name = tmpbuf (sizeof (*header_name) * header_count);
2418
2419 HE *he;
2420 int i = 0;
2421 while (he = hv_iternext (hv))
2422 {
2423 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2424 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2425 ++i;
2426 }
2427 }
2428
2429 void *user_data;
2430 program_callback pfn_notify = make_program_callback (notify, &user_data);
2431
1393 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2432 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2433 header_count, header_list, header_name, pfn_notify, user_data));
2434
2435#endif
1394 2436
1395void 2437void
1396build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2438build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1397 PPCODE: 2439 PPCODE:
1398 size_t size; 2440 size_t size;
1407#BEGIN:program_build 2449#BEGIN:program_build
1408 2450
1409void 2451void
1410build_status (OpenCL::Program self, OpenCL::Device device) 2452build_status (OpenCL::Program self, OpenCL::Device device)
1411 PPCODE: 2453 PPCODE:
1412 cl_build_status value [1]; 2454 cl_int value [1];
1413 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2455 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1414 EXTEND (SP, 1); 2456 EXTEND (SP, 1);
1415 const int i = 0; 2457 const int i = 0;
1416 PUSHs (sv_2mortal (newSViv (value [i]))); 2458 PUSHs (sv_2mortal (newSViv (value [i])));
1417 2459
1427 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2469 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1428 EXTEND (SP, 1); 2470 EXTEND (SP, 1);
1429 const int i = 0; 2471 const int i = 0;
1430 PUSHs (sv_2mortal (newSVpv (value, 0))); 2472 PUSHs (sv_2mortal (newSVpv (value, 0)));
1431 2473
2474void
2475binary_type (OpenCL::Program self, OpenCL::Device device)
2476 PPCODE:
2477 cl_uint value [1];
2478 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2479 EXTEND (SP, 1);
2480 const int i = 0;
2481 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2482
1432#END:program_build 2483#END:program_build
1433 2484
1434void 2485void
1435kernel (OpenCL::Program program, SV *function) 2486kernel (OpenCL::Program program, SV *function)
1436 PPCODE: 2487 PPCODE:
1437 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2488 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1438 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2489 XPUSH_CLOBJ (stash_kernel, kernel);
2490
2491void
2492kernels_in_program (OpenCL::Program program)
2493 PPCODE:
2494 cl_uint num_kernels;
2495 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2496 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2497 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2498
2499 int i;
2500 EXTEND (SP, num_kernels);
2501 for (i = 0; i < num_kernels; ++i)
2502 PUSH_CLOBJ (stash_kernel, kernels [i]);
1439 2503
1440void 2504void
1441info (OpenCL::Program self, cl_program_info name) 2505info (OpenCL::Program self, cl_program_info name)
1442 PPCODE: 2506 PPCODE:
1443 INFO (Program) 2507 INFO (Program)
1461 { 2525 {
1462 SV *sv = sv_2mortal (newSV (sizes [i])); 2526 SV *sv = sv_2mortal (newSV (sizes [i]));
1463 SvUPGRADE (sv, SVt_PV); 2527 SvUPGRADE (sv, SVt_PV);
1464 SvPOK_only (sv); 2528 SvPOK_only (sv);
1465 SvCUR_set (sv, sizes [i]); 2529 SvCUR_set (sv, sizes [i]);
1466 ptrs [i] = SvPVX (sv); 2530 ptrs [i] = (void *)SvPVX (sv);
1467 PUSHs (sv); 2531 PUSHs (sv);
1468 } 2532 }
1469 2533
1470 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2534 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1471 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2535 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1489 PPCODE: 2553 PPCODE:
1490 cl_context value [1]; 2554 cl_context value [1];
1491 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2555 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1492 EXTEND (SP, 1); 2556 EXTEND (SP, 1);
1493 const int i = 0; 2557 const int i = 0;
1494 {
1495 NEED_SUCCESS (RetainContext, (value [i])); 2558 NEED_SUCCESS (RetainContext, (value [i]));
1496 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2559 PUSH_CLOBJ (stash_context, value [i]);
1497 }
1498 2560
1499void 2561void
1500devices (OpenCL::Program self) 2562devices (OpenCL::Program self)
1501 PPCODE: 2563 PPCODE:
1502 size_t size; 2564 size_t size;
1504 cl_device_id *value = tmpbuf (size); 2566 cl_device_id *value = tmpbuf (size);
1505 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2567 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1506 int i, n = size / sizeof (*value); 2568 int i, n = size / sizeof (*value);
1507 EXTEND (SP, n); 2569 EXTEND (SP, n);
1508 for (i = 0; i < n; ++i) 2570 for (i = 0; i < n; ++i)
1509 { 2571 PUSH_CLOBJ (stash_device, value [i]);
1510 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1511 }
1512 2572
1513void 2573void
1514source (OpenCL::Program self) 2574source (OpenCL::Program self)
1515 PPCODE: 2575 PPCODE:
1516 size_t size; 2576 size_t size;
1541DESTROY (OpenCL::Kernel self) 2601DESTROY (OpenCL::Kernel self)
1542 CODE: 2602 CODE:
1543 clReleaseKernel (self); 2603 clReleaseKernel (self);
1544 2604
1545void 2605void
2606setf (OpenCL::Kernel self, const char *format, ...)
2607 CODE:
2608 int i;
2609 for (i = 2; ; ++i)
2610 {
2611 while (*format == ' ')
2612 ++format;
2613
2614 char type = *format++;
2615
2616 if (!type)
2617 break;
2618
2619 if (i >= items)
2620 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2621
2622 SV *sv = ST (i);
2623
2624 union
2625 {
2626 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2627 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2628 cl_half ch; cl_float cf; cl_double cd;
2629 cl_mem cm;
2630 cl_sampler ca;
2631 size_t cz;
2632 cl_event ce;
2633 } arg;
2634 size_t size;
2635 int nullarg = 0;
2636
2637 switch (type)
2638 {
2639 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2640 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2641 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2642 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2643 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2644 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2645 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2646 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2647
2648 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2649 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2650 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2651
2652 case 'z': nullarg = 1; size = SvIV (sv); break;
2653
2654 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2655 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2656 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2657
2658 default:
2659 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2660 }
2661
2662 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2663 if (res)
2664 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2665 }
2666
2667 if (i != items)
2668 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2669
2670void
1546set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2671set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1547 CODE: 2672 CODE:
1548 clSetKernelArg (self, idx, sizeof (value), &value); 2673 clSetKernelArg (self, idx, sizeof (value), &value);
1549 2674
1550void 2675void
1598 clSetKernelArg (self, idx, sizeof (value), &value); 2723 clSetKernelArg (self, idx, sizeof (value), &value);
1599 2724
1600void 2725void
1601set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2726set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1602 CODE: 2727 CODE:
2728 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2729
2730void
2731set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2732 CODE:
2733 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2734
2735void
2736set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2737 CODE:
2738 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2739
2740void
2741set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2742 CODE:
1603 clSetKernelArg (self, idx, sizeof (value), &value); 2743 clSetKernelArg (self, idx, sizeof (value), &value);
1604 2744
1605void 2745void
1606set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2746set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1607 CODE: 2747 CODE:
1608 clSetKernelArg (self, idx, sizeof (value), &value); 2748 clSetKernelArg (self, idx, size, 0);
1609
1610void
1611set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1612 CODE:
1613 clSetKernelArg (self, idx, sizeof (value), &value);
1614
1615void
1616set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1617 CODE:
1618 clSetKernelArg (self, idx, sizeof (value), &value);
1619
1620void
1621set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1622 CODE:
1623 clSetKernelArg (self, idx, sizeof (value), &value);
1624 2749
1625void 2750void
1626set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2751set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1627 CODE: 2752 CODE:
1628 clSetKernelArg (self, idx, sizeof (value), &value); 2753 clSetKernelArg (self, idx, sizeof (value), &value);
1662 PPCODE: 2787 PPCODE:
1663 cl_context value [1]; 2788 cl_context value [1];
1664 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2789 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1665 EXTEND (SP, 1); 2790 EXTEND (SP, 1);
1666 const int i = 0; 2791 const int i = 0;
1667 {
1668 NEED_SUCCESS (RetainContext, (value [i])); 2792 NEED_SUCCESS (RetainContext, (value [i]));
1669 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2793 PUSH_CLOBJ (stash_context, value [i]);
1670 }
1671 2794
1672void 2795void
1673program (OpenCL::Kernel self) 2796program (OpenCL::Kernel self)
1674 PPCODE: 2797 PPCODE:
1675 cl_program value [1]; 2798 cl_program value [1];
1676 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2799 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1677 EXTEND (SP, 1); 2800 EXTEND (SP, 1);
1678 const int i = 0; 2801 const int i = 0;
1679 {
1680 NEED_SUCCESS (RetainProgram, (value [i])); 2802 NEED_SUCCESS (RetainProgram, (value [i]));
1681 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2803 PUSH_CLOBJ (stash_program, value [i]);
1682 }
1683 2804
1684#END:kernel 2805#END:kernel
1685 2806
1686void 2807void
1687work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2808work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
1733 const int i = 0; 2854 const int i = 0;
1734 PUSHs (sv_2mortal (newSVuv (value [i]))); 2855 PUSHs (sv_2mortal (newSVuv (value [i])));
1735 2856
1736#END:kernel_work_group 2857#END:kernel_work_group
1737 2858
2859#if CL_VERSION_1_2
2860
2861void
2862arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2863 PPCODE:
2864 size_t size;
2865 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2866 SV *sv = sv_2mortal (newSV (size));
2867 SvUPGRADE (sv, SVt_PV);
2868 SvPOK_only (sv);
2869 SvCUR_set (sv, size);
2870 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2871 XPUSHs (sv);
2872
2873#BEGIN:kernel_arg
2874
2875void
2876arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2877 ALIAS:
2878 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2879 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2880 PPCODE:
2881 cl_uint value [1];
2882 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2883 EXTEND (SP, 1);
2884 const int i = 0;
2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2886
2887void
2888arg_type_name (OpenCL::Kernel self, cl_uint idx)
2889 ALIAS:
2890 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2891 arg_name = CL_KERNEL_ARG_NAME
2892 PPCODE:
2893 size_t size;
2894 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2895 char *value = tmpbuf (size);
2896 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2897 EXTEND (SP, 1);
2898 const int i = 0;
2899 PUSHs (sv_2mortal (newSVpv (value, 0)));
2900
2901void
2902arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2903 PPCODE:
2904 cl_ulong value [1];
2905 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2906 EXTEND (SP, 1);
2907 const int i = 0;
2908 PUSHs (sv_2mortal (newSVuv (value [i])));
2909
2910#END:kernel_arg
2911
2912#endif
2913
1738MODULE = OpenCL PACKAGE = OpenCL::Event 2914MODULE = OpenCL PACKAGE = OpenCL::Event
1739 2915
1740void 2916void
1741DESTROY (OpenCL::Event self) 2917DESTROY (OpenCL::Event self)
1742 CODE: 2918 CODE:
1744 2920
1745void 2921void
1746wait (OpenCL::Event self) 2922wait (OpenCL::Event self)
1747 CODE: 2923 CODE:
1748 clWaitForEvents (1, &self); 2924 clWaitForEvents (1, &self);
2925
2926void
2927cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2928 CODE:
2929 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1749 2930
1750void 2931void
1751info (OpenCL::Event self, cl_event_info name) 2932info (OpenCL::Event self, cl_event_info name)
1752 PPCODE: 2933 PPCODE:
1753 INFO (Event) 2934 INFO (Event)
1759 PPCODE: 2940 PPCODE:
1760 cl_command_queue value [1]; 2941 cl_command_queue value [1];
1761 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2942 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1762 EXTEND (SP, 1); 2943 EXTEND (SP, 1);
1763 const int i = 0; 2944 const int i = 0;
1764 {
1765 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2945 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1766 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2946 PUSH_CLOBJ (stash_queue, value [i]);
1767 }
1768 2947
1769void 2948void
1770command_type (OpenCL::Event self) 2949command_type (OpenCL::Event self)
1771 PPCODE:
1772 cl_command_type value [1];
1773 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
1774 EXTEND (SP, 1);
1775 const int i = 0;
1776 PUSHs (sv_2mortal (newSVuv (value [i])));
1777
1778void
1779reference_count (OpenCL::Event self)
1780 ALIAS: 2950 ALIAS:
2951 command_type = CL_EVENT_COMMAND_TYPE
1781 reference_count = CL_EVENT_REFERENCE_COUNT 2952 reference_count = CL_EVENT_REFERENCE_COUNT
1782 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2953 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
1783 PPCODE: 2954 PPCODE:
1784 cl_uint value [1]; 2955 cl_uint value [1];
1785 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2956 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
1792 PPCODE: 2963 PPCODE:
1793 cl_context value [1]; 2964 cl_context value [1];
1794 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2965 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1795 EXTEND (SP, 1); 2966 EXTEND (SP, 1);
1796 const int i = 0; 2967 const int i = 0;
1797 {
1798 NEED_SUCCESS (RetainContext, (value [i])); 2968 NEED_SUCCESS (RetainContext, (value [i]));
1799 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2969 PUSH_CLOBJ (stash_context, value [i]);
1800 }
1801 2970
1802#END:event 2971#END:event
1803 2972
1804void 2973void
1805profiling_info (OpenCL::Event self, cl_profiling_info name) 2974profiling_info (OpenCL::Event self, cl_profiling_info name)
1829void 2998void
1830set_status (OpenCL::UserEvent self, cl_int execution_status) 2999set_status (OpenCL::UserEvent self, cl_int execution_status)
1831 CODE: 3000 CODE:
1832 clSetUserEventStatus (self, execution_status); 3001 clSetUserEventStatus (self, execution_status);
1833 3002
3003MODULE = OpenCL PACKAGE = OpenCL::Mapped
3004
3005void
3006DESTROY (SV *self)
3007 CODE:
3008 OpenCL__Mapped mapped = SvMAPPED (self);
3009
3010 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3011 mapped_detach (self, mapped);
3012
3013 clReleaseCommandQueue (mapped->queue);
3014 clReleaseEvent (mapped->event);
3015 Safefree (mapped);
3016
3017void
3018unmap (OpenCL::Mapped self, ...)
3019 CODE:
3020 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3021
3022bool
3023mapped (OpenCL::Mapped self)
3024 CODE:
3025 RETVAL = !!self->ptr;
3026 OUTPUT:
3027 RETVAL
3028
3029void
3030wait (OpenCL::Mapped self)
3031 PPCODE:
3032 if (self->event)
3033 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3034
3035void
3036event (OpenCL::Mapped self)
3037 PPCODE:
3038 if (!self->event)
3039 XSRETURN_UNDEF;
3040
3041 clRetainEvent (self->event);
3042 XPUSH_CLOBJ (stash_event, self->event);
3043
3044size_t
3045size (OpenCL::Mapped self)
3046 CODE:
3047 RETVAL = self->cb;
3048 OUTPUT:
3049 RETVAL
3050
3051IV
3052ptr (OpenCL::Mapped self)
3053 CODE:
3054 RETVAL = PTR2IV (self->ptr);
3055 OUTPUT:
3056 RETVAL
3057
3058void
3059set (OpenCL::Mapped self, size_t offset, SV *data)
3060 CODE:
3061 STRLEN len;
3062 const char *ptr = SvPVbyte (data, len);
3063
3064 if (offset + len > self->cb)
3065 croak ("OpenCL::Mapped::set out of bound condition detected");
3066
3067 memcpy (offset + (char *)self->ptr, ptr, len);
3068
3069MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3070
3071MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3072
3073IV
3074row_pitch (OpenCL::Mapped self)
3075 ALIAS:
3076 slice_pitch = 1
3077 CODE:
3078 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3079 OUTPUT:
3080 RETVAL
3081
3082
3083

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines