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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.51 by root, Tue Apr 24 22:45:38 2012 UTC vs.
Revision 1.74 by root, Sat May 5 13:56:00 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
5#include "ecb.h"//D
4 6
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 8#include "xthread.h"
7#include "schmorp.h" 9#include "schmorp.h"
8 10
26#if PREFER_1_1 28#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 30#endif
29 31
30#ifdef __APPLE__ 32#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
32#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
33 #include <CL/opencl.h> 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>
34#endif 50#endif
35 51
36#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 53 #undef PREFER_1_1
38 #define PREFER_1_1 1 54 #define PREFER_1_1 1
39#endif 55#endif
40 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
41typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
61 77
62typedef 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;
63 104
64/*****************************************************************************/ 105/*****************************************************************************/
65 106
66// name must include a leading underscore 107// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
107} 148}
108 149
109/*****************************************************************************/ 150/*****************************************************************************/
110 151
111/* up to two temporary buffers */ 152/* up to two temporary buffers */
112static void * 153static void * ecb_noinline
113tmpbuf (size_t size) 154tmpbuf (size_t size)
114{ 155{
115 enum { buffers = 3 }; 156 enum { buffers = 4 };
116 static int idx; 157 static int idx;
117 static void *buf [buffers]; 158 static void *buf [buffers];
118 static size_t len [buffers]; 159 static size_t len [buffers];
119 160
120 idx = (idx + 1) % buffers; 161 idx = (idx + 1) % buffers;
125 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
126 buf [idx] = malloc (len [idx]); 167 buf [idx] = malloc (len [idx]);
127 } 168 }
128 169
129 return buf [idx]; 170 return buf [idx];
171}
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;
130} 186}
131 187
132/*****************************************************************************/ 188/*****************************************************************************/
133 189
134typedef struct 190typedef struct
193 if (res) \ 249 if (res) \
194 FAIL (name); 250 FAIL (name);
195 251
196/*****************************************************************************/ 252/*****************************************************************************/
197 253
254static SV *
255new_clobj (HV *stash, IV id)
256{
257 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
258}
259
260#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
261#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
262
263/* cl objects are either \$iv, or [$iv, ...] */
264/* they can be upgraded at runtime to the array form */
265static void * ecb_noinline
266SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
267{
268 // sv_derived_from is quite slow :(
269 if (SvROK (sv) && sv_derived_from (sv, pkg))
270 return (void *)SvIV (SvRV (sv));
271
272 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
273}
274
275// the "no-inherit" version of the above
276static void * ecb_noinline
277SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
278{
279 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
280 return (void *)SvIV (SvRV (sv));
281
282 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
283}
284
285/*****************************************************************************/
286
198static cl_context_properties * 287static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 288SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 289{
201 if (!sv || !SvOK (sv)) 290 if (!sv || !SvOK (sv))
202 if (extra) 291 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 292 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 293 else
210 int i, len = av_len (av) + 1; 299 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 300 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 301 cl_context_properties *l = p;
213 302
214 if (len & 1) 303 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 304 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 305
217 while (extracount--) 306 while (extracount--)
218 *l++ = *extra++; 307 *l++ = *extra++;
219 308
220 for (i = 0; i < len; i += 2) 309 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 312 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 313 cl_context_properties v = SvIV (p_sv); // code below can override
225 314
226 switch (t) 315 switch (t)
227 { 316 {
317 case CL_CONTEXT_PLATFORM:
318 if (SvROK (p_sv))
319 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
320 break;
321
228 case CL_GLX_DISPLAY_KHR: 322 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 323 if (!SvOK (p_sv))
230 { 324 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 325 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 326 if (func)
255 *l = 0; 349 *l = 0;
256 350
257 return p; 351 return p;
258 } 352 }
259 353
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 354 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 355}
262 356
263/*****************************************************************************/ 357// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 358// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 359static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 360object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr))
268
269/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */
271static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
273{ 361{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 362 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
363 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
364 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
365
366 AV *av = (AV *)SvRV (arg);
367 void **list = 0;
368 cl_uint count = av_len (av) + 1;
369
370 if (count)
275 { 371 {
276 SV *rv = SvRV (sv); 372 list = tmpbuf (sizeof (*list) * count);
277 373 int i;
278 if (SvTYPE (rv) == SVt_PVAV) 374 for (i = 0; i < count; ++i)
279 rv = AvARRAY (rv)[0]; 375 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
280
281 return (void *)SvIV (SvRV (sv));
282 } 376 }
283 377
284 croak ("%s: %s is not of type %s", func, svname, pkg); 378 if (!count && !or_undef)
285} 379 croak ("%s: '%s' must contain at least one %s object",
380 cv_get_name (cv), argname, klass);
286 381
287static void 382 *rcount = count;
288CLOBJ_push (SV *self, SV *data) 383 return (void *)list;
289{
290 SV *rv = SvRV (self);
291
292 if (SvTYPE (rv) != SVt_PVAV)
293 {
294 AV *av = newAV ();
295 av_push (av, rv);
296 rv = (SV *)av;
297 SvRV_set (self, rv);
298 }
299
300 av_push ((AV *)rv, data);
301}
302
303static SV *
304sv_struct (STRLEN size)
305{
306 SV *sv = newSV (size);
307 SvPOK_only (sv);
308 return sv;
309}
310
311static void *
312CLOBJ_push_struct (SV *self, STRLEN size)
313{
314 SV *sv = sv_struct (size);
315 CLOBJ_push (self, sv);
316 return SvPVX (sv);
317} 384}
318 385
319/*****************************************************************************/ 386/*****************************************************************************/
320/* callback stuff */ 387/* callback stuff */
321 388
343static void (*eq_signal_func)(void *signal_arg, int value); 410static void (*eq_signal_func)(void *signal_arg, int value);
344static void *eq_signal_arg; 411static void *eq_signal_arg;
345static xmutex_t eq_lock = X_MUTEX_INIT; 412static xmutex_t eq_lock = X_MUTEX_INIT;
346static eq_item *eq_head, *eq_tail; 413static eq_item *eq_head, *eq_tail;
347 414
348static void 415static void ecb_noinline
349eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 416eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
350{ 417{
351 eq_item *item = malloc (sizeof (eq_item)); 418 eq_item *item = malloc (sizeof (eq_item));
352
353 printf ("enq(%p,%p,%p,%p,%p)\n", vtbl, cb, data1, data2, data3);//D
354 419
355 item->next = 0; 420 item->next = 0;
356 item->vtbl = vtbl; 421 item->vtbl = vtbl;
357 item->cb = cb; 422 item->cb = cb;
358 item->data1 = data1; 423 item->data1 = data1;
383 X_UNLOCK (eq_lock); 448 X_UNLOCK (eq_lock);
384 449
385 return res; 450 return res;
386} 451}
387 452
388#if 0
389static void
390mem_free (pTHX_ void *p)
391{
392 free (p);
393}
394//SAVEDESTRUCTOR_X (mem_free, item);
395#endif
396
397static void 453static void
398eq_poll (void) 454eq_poll (void)
399{ 455{
400 eq_item *item; 456 eq_item *item;
401 457
428eq_poll_interrupt (pTHX_ void *c_arg, int value) 484eq_poll_interrupt (pTHX_ void *c_arg, int value)
429{ 485{
430 eq_poll (); 486 eq_poll ();
431} 487}
432 488
489/*****************************************************************************/
433/* context notify */ 490/* context notify */
434 491
435static void 492static void ecb_noinline
436eq_context_push (void *data1, void *data2, void *data3) 493eq_context_push (void *data1, void *data2, void *data3)
437{ 494{
438 dSP; 495 dSP;
439 PUSHs (sv_2mortal (newSVpv (data1, 0))); 496 PUSHs (sv_2mortal (newSVpv (data1, 0)));
440 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 497 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
441 PUTBACK; 498 PUTBACK;
499
500 free (data1);
501 free (data2);
442} 502}
443 503
444static eq_vtbl eq_context_vtbl = { 0, eq_context_push }; 504static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
445 505
506static void CL_CALLBACK
507eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
508{
509 void *pvt_copy = malloc (cb);
510 memcpy (pvt_copy, pvt, cb);
511 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
512}
513
514#define CONTEXT_NOTIFY_CALLBACK \
515 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
516 void *user_data = 0; \
517 \
518 if (SvOK (notify)) \
519 { \
520 pfn_notify = eq_context_notify; \
521 user_data = s_get_cv (notify); \
522 }
523
524static SV * ecb_noinline
525new_clobj_context (cl_context ctx, void *user_data)
526{
527 SV *sv = new_clobj (stash_context, (IV)ctx);
528
529 if (user_data)
530 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
531
532 return sv;
533}
534
535#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
536
537/*****************************************************************************/
446/* build/compile/link notify */ 538/* build/compile/link notify */
447 539
448static void 540static void
449eq_program_push (void *data1, void *data2, void *data3) 541eq_program_push (void *data1, void *data2, void *data3)
450{ 542{
451 dSP; 543 dSP;
452 PUSH_CLOBJ ("OpenCL::Program", data1); 544 PUSH_CLOBJ (stash_program, data1);
453 PUTBACK; 545 PUTBACK;
454} 546}
455 547
456static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 548static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
457 549
458static void CL_CALLBACK 550static void CL_CALLBACK
459eq_program_notify (cl_program program, void *user_data) 551eq_program_notify (cl_program program, void *user_data)
460{ 552{
553 clRetainProgram (program);
554
461 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 555 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
556}
557
558typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
559
560static program_callback ecb_noinline
561make_program_callback (SV *notify, void **ruser_data)
562{
563 if (SvOK (notify))
564 {
565 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
566 return eq_program_notify;
567 }
568 else
569 {
570 *ruser_data = 0;
571 return 0;
572 }
462} 573}
463 574
464struct build_args 575struct build_args
465{ 576{
466 cl_program program; 577 cl_program program;
480 else 591 else
481 clReleaseProgram (arg->program); 592 clReleaseProgram (arg->program);
482 593
483 free (arg->options); 594 free (arg->options);
484 free (arg); 595 free (arg);
596
597 return 0;
485} 598}
486 599
487static void 600static void
488build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 601build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
489{ 602{
497 610
498 xthread_t id; 611 xthread_t id;
499 thread_create (&id, build_program_thread, arg); 612 thread_create (&id, build_program_thread, arg);
500} 613}
501 614
615/*****************************************************************************/
502/* event objects */ 616/* event objects */
503 617
504static void 618static void
505eq_event_push (void *data1, void *data2, void *data3) 619eq_event_push (void *data1, void *data2, void *data3)
506{ 620{
507 dSP; 621 dSP;
508 PUSH_CLOBJ ("OpenCL::Event", data1); 622 PUSH_CLOBJ (stash_event, data1);
509 PUSHs (sv_2mortal (newSViv ((IV)data2))); 623 PUSHs (sv_2mortal (newSViv ((IV)data2)));
510 PUTBACK; 624 PUTBACK;
511} 625}
512 626
513static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 627static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
514 628
515static void CL_CALLBACK 629static void CL_CALLBACK
516eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data) 630eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
517{ 631{
518 clRetainEvent (event); 632 clRetainEvent (event);
519 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)event_command_exec_status, 0); 633 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
520} 634}
521 635
522/*****************************************************************************/ 636/*****************************************************************************/
637/* utilities for XS code */
523 638
524static size_t 639static size_t
525img_row_pitch (cl_mem img) 640img_row_pitch (cl_mem img)
526{ 641{
527 size_t res; 642 size_t res;
528 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 643 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
529 return res; 644 return res;
530} 645}
531 646
532static cl_event * 647static cl_event * ecb_noinline
533event_list (SV **items, cl_uint *rcount) 648event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
534{ 649{
535 cl_uint count = *rcount; 650 cl_uint count = *rcount;
536 651
537 if (!count) 652 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
653 *rcount = count = 0;
654
655 if (!count && !extra)
538 return 0; 656 return 0;
539 657
540 cl_event *list = tmpbuf (sizeof (cl_event) * count); 658 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
541 int i = 0; 659 int i = 0;
542 660
543 do 661 while (count--)
544 {
545 --count;
546 if (SvOK (items [count])) 662 if (SvOK (items [count]))
547 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 663 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
548 } 664
549 while (count); 665 if (extra)
666 list [i++] = extra;
550 667
551 *rcount = i; 668 *rcount = i;
552 669
553 return i ? list : 0; 670 return i ? list : 0;
554} 671}
555 672
556#define EVENT_LIST(items,count) \ 673#define EVENT_LIST(skip) \
557 cl_uint event_list_count = (count); \ 674 cl_uint event_list_count = items - (skip); \
558 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 675 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
559 676
560#define INFO(class) \ 677#define INFO(class) \
561{ \ 678{ \
562 size_t size; \ 679 size_t size; \
563 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 680 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
566 SvPOK_only (sv); \ 683 SvPOK_only (sv); \
567 SvCUR_set (sv, size); \ 684 SvCUR_set (sv, size); \
568 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 685 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
569 XPUSHs (sv); \ 686 XPUSHs (sv); \
570} 687}
688
689/*****************************************************************************/
690/* mapped_xxx */
691
692static OpenCL__Mapped
693SvMAPPED (SV *sv)
694{
695 // no typechecking atm., keep your fingers crossed
696 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
697}
698
699struct mapped
700{
701 cl_command_queue queue;
702 cl_mem memobj;
703 void *ptr;
704 size_t cb;
705 cl_event event;
706 size_t row_pitch;
707 size_t slice_pitch;
708};
709
710static SV *
711mapped_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)
712{
713 SV *data = newSV (0);
714 SvUPGRADE (data, SVt_PVMG);
715
716 OpenCL__Mapped mapped;
717 New (0, mapped, 1, struct mapped);
718
719 clRetainCommandQueue (queue);
720
721 mapped->queue = queue;
722 mapped->memobj = memobj;
723 mapped->ptr = ptr;
724 mapped->cb = cb;
725 mapped->event = ev;
726 mapped->row_pitch = row_pitch;
727 mapped->slice_pitch = slice_pitch;
728
729 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
730
731 if (SvLEN (data))
732 Safefree (data);
733
734 SvPVX (data) = (char *)ptr;
735 SvCUR_set (data, cb);
736 SvLEN_set (data, 0);
737 SvPOK_only (data);
738
739 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
740
741 if (!(flags & CL_MAP_WRITE))
742 SvREADONLY_on (data);
743
744 return obj;
745}
746
747static void
748mapped_detach (SV *sv, OpenCL__Mapped mapped)
749{
750 SV *data = SvRV (sv);
751
752 // the next check checks both before AND after detach, where SvPVX should STILL be 0
753 if (SvPVX (data) != (char *)mapped->ptr)
754 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
755 else
756 {
757 SvREADONLY_off (data);
758 SvCUR_set (data, 0);
759 SvPVX (data) = 0;
760 SvOK_off (data);
761 }
762
763 mapped->ptr = 0;
764}
765
766static void
767mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
768{
769 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
770 cl_event ev;
771
772 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
773
774 clReleaseEvent (mapped->event);
775 mapped->event = ev;
776
777 mapped_detach (self, mapped);
778}
779
780/*****************************************************************************/
571 781
572MODULE = OpenCL PACKAGE = OpenCL 782MODULE = OpenCL PACKAGE = OpenCL
573 783
574PROTOTYPES: ENABLE 784PROTOTYPES: ENABLE
575 785
597 { sizeof (cl_long ), "SIZEOF_LONG" }, 807 { sizeof (cl_long ), "SIZEOF_LONG" },
598 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 808 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
599 { sizeof (cl_half ), "SIZEOF_HALF" }, 809 { sizeof (cl_half ), "SIZEOF_HALF" },
600 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 810 { sizeof (cl_float ), "SIZEOF_FLOAT" },
601 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 811 { sizeof (cl_double), "SIZEOF_DOUBLE" },
812 { PREFER_1_1 , "PREFER_1_1" },
602#include "constiv.h" 813#include "constiv.h"
603 }; 814 };
604 815
605 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 816 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
606 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 817 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
818
819 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
820 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
821 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
822 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
823 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
824 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
825 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
826 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
827 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
828 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
829 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
830 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
831 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
832 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
833 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
834 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
835 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
836 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
837 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
838 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
839 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
840 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
841 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
607 842
608 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 843 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
609} 844}
610 845
611cl_int 846cl_int
614 RETVAL = res; 849 RETVAL = res;
615 OUTPUT: 850 OUTPUT:
616 RETVAL 851 RETVAL
617 852
618const char * 853const char *
619err2str (cl_int err) 854err2str (cl_int err = res)
620 855
621const char * 856const char *
622enum2str (cl_uint value) 857enum2str (cl_uint value)
623 858
624void 859void
632 list = tmpbuf (sizeof (*list) * count); 867 list = tmpbuf (sizeof (*list) * count);
633 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 868 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
634 869
635 EXTEND (SP, count); 870 EXTEND (SP, count);
636 for (i = 0; i < count; ++i) 871 for (i = 0; i < count; ++i)
637 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 872 PUSH_CLOBJ (stash_platform, list [i]);
638 873
639void 874void
640context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 875context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
641 PPCODE: 876 PPCODE:
877 CONTEXT_NOTIFY_CALLBACK;
642 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 878 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
643 XPUSH_CLOBJ ("OpenCL::Context", ctx); 879 XPUSH_CLOBJ_CONTEXT;
644 880
645#if 0
646
647void 881void
648context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 882context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
649 PPCODE: 883 PPCODE:
650 /* der Gipfel der Kunst */ 884 cl_uint device_count;
885 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
651 886
652#endif 887 CONTEXT_NOTIFY_CALLBACK;
888 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
889 XPUSH_CLOBJ_CONTEXT;
653 890
654void 891void
655wait_for_events (...) 892wait_for_events (...)
656 CODE: 893 CODE:
657 EVENT_LIST (0, items); 894 EVENT_LIST (0);
658 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 895 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
659 896
660PROTOTYPES: DISABLE 897PROTOTYPES: DISABLE
661 898
662MODULE = OpenCL PACKAGE = OpenCL::Platform 899MODULE = OpenCL PACKAGE = OpenCL::Platform
705 list = tmpbuf (sizeof (*list) * count); 942 list = tmpbuf (sizeof (*list) * count);
706 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 943 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
707 944
708 EXTEND (SP, count); 945 EXTEND (SP, count);
709 for (i = 0; i < count; ++i) 946 for (i = 0; i < count; ++i)
710 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 947 PUSH_CLOBJ (stash_device, list [i]);
711 948
712void 949void
713context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = 0) 950context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
714 PPCODE: 951 PPCODE:
952 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
953 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
954
955 cl_uint device_count;
956 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
957
958 CONTEXT_NOTIFY_CALLBACK;
959 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
960 XPUSH_CLOBJ_CONTEXT;
961
962void
963context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
964 PPCODE:
965 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
966 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
967
968 CONTEXT_NOTIFY_CALLBACK;
969 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
970 XPUSH_CLOBJ_CONTEXT;
971
972MODULE = OpenCL PACKAGE = OpenCL::Device
973
974void
975info (OpenCL::Device self, cl_device_info name)
976 PPCODE:
977 INFO (Device)
978
979#if CL_VERSION_1_2
980
981void
982sub_devices (OpenCL::Device self, SV *properties)
983 PPCODE:
715 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 984 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
716 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 985 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
717 986
718 AV *av = (AV *)SvRV (devices); 987 properties = SvRV (properties);
719 cl_uint num_devices = av_len (av) + 1; 988
720 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 989 cl_uint count = av_len ((AV *)properties) + 1;
990 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
721 991
722 int i; 992 int i;
723 for (i = num_devices; i--; ) 993 for (i = 0; i < count; ++i)
724 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 994 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
725 995
726 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; 996 props [count] = 0;
727 void *user_data = 0;
728
729 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
730 XPUSH_CLOBJ ("OpenCL::Context", ctx);
731 997
732void 998 cl_uint num_devices;
733context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 999 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
734 PPCODE: 1000 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
735 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1001 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
736 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
737 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
738 XPUSH_CLOBJ ("OpenCL::Context", ctx);
739 1002
740MODULE = OpenCL PACKAGE = OpenCL::Device 1003 EXTEND (SP, num_devices);
1004 for (i = 0; i < count; ++i)
1005 PUSH_CLOBJ (stash_subdevice, list [i]);
741 1006
742void 1007#endif
743info (OpenCL::Device self, cl_device_info name)
744 PPCODE:
745 INFO (Device)
746 1008
747#BEGIN:device 1009#BEGIN:device
748 1010
749void 1011void
750type (OpenCL::Device self) 1012type (OpenCL::Device self)
751 PPCODE: 1013 PPCODE:
752 cl_device_type value [1]; 1014 cl_device_type value [1];
753 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1015 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
754 EXTEND (SP, 1); 1016 EXTEND (SP, 1);
755 const int i = 0; 1017 const int i = 0;
756 PUSHs (sv_2mortal (newSViv (value [i]))); 1018 PUSHs (sv_2mortal (newSVuv (value [i])));
757 1019
758void 1020void
759vendor_id (OpenCL::Device self) 1021vendor_id (OpenCL::Device self)
760 ALIAS: 1022 ALIAS:
761 vendor_id = CL_DEVICE_VENDOR_ID 1023 vendor_id = CL_DEVICE_VENDOR_ID
906 PPCODE: 1168 PPCODE:
907 cl_command_queue_properties value [1]; 1169 cl_command_queue_properties value [1];
908 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1170 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
909 EXTEND (SP, 1); 1171 EXTEND (SP, 1);
910 const int i = 0; 1172 const int i = 0;
911 PUSHs (sv_2mortal (newSViv (value [i]))); 1173 PUSHs (sv_2mortal (newSVuv (value [i])));
912 1174
913void 1175void
914platform (OpenCL::Device self) 1176platform (OpenCL::Device self)
915 PPCODE: 1177 PPCODE:
916 cl_platform_id value [1]; 1178 cl_platform_id value [1];
917 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1179 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
918 EXTEND (SP, 1); 1180 EXTEND (SP, 1);
919 const int i = 0; 1181 const int i = 0;
920 { 1182 {
921 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1183 PUSH_CLOBJ (stash_platform, value [i]);
922 } 1184 }
923 1185
924void 1186void
925name (OpenCL::Device self) 1187name (OpenCL::Device self)
926 ALIAS: 1188 ALIAS:
945 cl_device_id value [1]; 1207 cl_device_id value [1];
946 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1208 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
947 EXTEND (SP, 1); 1209 EXTEND (SP, 1);
948 const int i = 0; 1210 const int i = 0;
949 { 1211 {
950 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1212 PUSH_CLOBJ (stash_device, value [i]);
951 } 1213 }
952 1214
953void 1215void
954partition_types_ext (OpenCL::Device self) 1216partition_types_ext (OpenCL::Device self)
955 ALIAS: 1217 ALIAS:
966 for (i = 0; i < n; ++i) 1228 for (i = 0; i < n; ++i)
967 PUSHs (sv_2mortal (newSVuv (value [i]))); 1229 PUSHs (sv_2mortal (newSVuv (value [i])));
968 1230
969#END:device 1231#END:device
970 1232
1233MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1234
1235#if CL_VERSION_1_2
1236
1237void
1238DESTROY (OpenCL::SubDevice self)
1239 CODE:
1240 clReleaseDevice (self);
1241
1242#endif
1243
971MODULE = OpenCL PACKAGE = OpenCL::Context 1244MODULE = OpenCL PACKAGE = OpenCL::Context
972 1245
973void 1246void
974DESTROY (OpenCL::Context context) 1247DESTROY (OpenCL::Context self)
975 CODE: 1248 CODE:
976 clReleaseContext (context); 1249 clReleaseContext (self);
977 1250
978void 1251void
979info (OpenCL::Context self, cl_context_info name) 1252info (OpenCL::Context self, cl_context_info name)
980 PPCODE: 1253 PPCODE:
981 INFO (Context) 1254 INFO (Context)
982 1255
983void 1256void
984queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1257queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
985 PPCODE: 1258 PPCODE:
986 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1259 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
987 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1260 XPUSH_CLOBJ (stash_queue, queue);
988 1261
989void 1262void
990user_event (OpenCL::Context self) 1263user_event (OpenCL::Context self)
991 PPCODE: 1264 PPCODE:
992 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1265 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
993 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1266 XPUSH_CLOBJ (stash_userevent, ev);
994 1267
995void 1268void
996buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1269buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
997 PPCODE: 1270 PPCODE:
998 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1271 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
999 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1272 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
1000 1273
1001 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1274 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1002 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1275 XPUSH_CLOBJ (stash_bufferobj, mem);
1003 1276
1004void 1277void
1005buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1278buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
1006 PPCODE: 1279 PPCODE:
1007 STRLEN len; 1280 STRLEN len;
1008 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1281 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1009 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1282 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
1010 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1283 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
1011 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1284 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
1012 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1285 XPUSH_CLOBJ (stash_bufferobj, mem);
1013 1286
1014#if CL_VERSION_1_2 1287#if CL_VERSION_1_2
1015 1288
1016void 1289void
1017image (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, 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) 1290image (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)
1018 PPCODE: 1291 PPCODE:
1019 STRLEN len; 1292 STRLEN len;
1020 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1293 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1021 const cl_image_format format = { channel_order, channel_type }; 1294 const cl_image_format format = { channel_order, channel_type };
1022 const cl_image_desc desc = { 1295 const cl_image_desc desc = {
1023 type, 1296 type,
1024 width, height, depth, 1297 width, height, depth,
1025 array_size, row_pitch, slice_pitch, 1298 array_size, row_pitch, slice_pitch,
1026 num_mip_level, num_samples, 1299 num_mip_level, num_samples,
1027 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1300 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1028 }; 1301 };
1029 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1302 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1030 char *klass = "OpenCL::Image"; 1303 HV *stash = stash_image;
1031 switch (type) 1304 switch (type)
1032 { 1305 {
1033 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1306 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1034 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1307 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1035 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1308 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1036 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1309 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1037 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1310 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1038 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1311 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1039 } 1312 }
1040 XPUSH_CLOBJ (klass, mem); 1313 XPUSH_CLOBJ (stash, mem);
1041 1314
1042#endif 1315#endif
1043 1316
1044void 1317void
1045image2d (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) 1318image2d (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)
1051 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1052#else 1325#else
1053 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1326 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1054 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1327 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1055#endif 1328#endif
1056 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1329 XPUSH_CLOBJ (stash_image2d, mem);
1057 1330
1058void 1331void
1059image3d (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) 1332image3d (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)
1060 PPCODE: 1333 PPCODE:
1061 STRLEN len; 1334 STRLEN len;
1065 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1066#else 1339#else
1067 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1340 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1068 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1069#endif 1342#endif
1070 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1343 XPUSH_CLOBJ (stash_image3d, mem);
1071 1344
1072#if cl_apple_gl_sharing || cl_khr_gl_sharing 1345#if cl_apple_gl_sharing || cl_khr_gl_sharing
1073 1346
1074void 1347void
1075gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1348gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1076 PPCODE: 1349 PPCODE:
1077 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1350 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1078 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1351 XPUSH_CLOBJ (stash_bufferobj, mem);
1079 1352
1080void 1353void
1081gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1354gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1082 PPCODE: 1355 PPCODE:
1083 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1356 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1084 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1357 XPUSH_CLOBJ (stash_image2d, mem);
1085 1358
1086#if CL_VERSION_1_2 1359#if CL_VERSION_1_2
1087 1360
1088void 1361void
1089gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1362gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1090 ALIAS: 1363 ALIAS:
1091 PPCODE: 1364 PPCODE:
1092 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1365 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1093 cl_gl_object_type type; 1366 cl_gl_object_type type;
1094 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1367 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1095 char *klass = "OpenCL::Memory"; 1368 HV *stash = stash_memory;
1096 switch (type) 1369 switch (type)
1097 { 1370 {
1098 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1371 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1099 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1372 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1100 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1373 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1101 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1374 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1102 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1375 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1103 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1376 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1104 } 1377 }
1105 XPUSH_CLOBJ (klass, mem); 1378 XPUSH_CLOBJ (stash, mem);
1106 1379
1107#endif 1380#endif
1108 1381
1109void 1382void
1110gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1383gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1112#if PREFER_1_1 1385#if PREFER_1_1
1113 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1386 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1114#else 1387#else
1115 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1388 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1116#endif 1389#endif
1117 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1390 XPUSH_CLOBJ (stash_image2d, mem);
1118 1391
1119void 1392void
1120gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1393gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1121 PPCODE: 1394 PPCODE:
1122#if PREFER_1_1 1395#if PREFER_1_1
1123 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1396 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1124#else 1397#else
1125 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1398 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1126#endif 1399#endif
1127 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1400 XPUSH_CLOBJ (stash_image3d, mem);
1128 1401
1129#endif 1402#endif
1130 1403
1131void 1404void
1132supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1405supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1152 1425
1153void 1426void
1154sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1427sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1155 PPCODE: 1428 PPCODE:
1156 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1429 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1157 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1430 XPUSH_CLOBJ (stash_sampler, sampler);
1158 1431
1159void 1432void
1160program_with_source (OpenCL::Context self, SV *program) 1433program_with_source (OpenCL::Context self, SV *program)
1161 PPCODE: 1434 PPCODE:
1162 STRLEN len; 1435 STRLEN len;
1163 size_t len2; 1436 size_t len2;
1164 const char *ptr = SvPVbyte (program, len); 1437 const char *ptr = SvPVbyte (program, len);
1165 1438
1166 len2 = len; 1439 len2 = len;
1167 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1440 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1168 XPUSH_CLOBJ ("OpenCL::Program", prog); 1441 XPUSH_CLOBJ (stash_program, prog);
1442
1443void
1444program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1445 PPCODE:
1446 cl_uint device_count;
1447 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1448
1449 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1450 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1451
1452 binaries = SvRV (binaries);
1453
1454 if (device_count != av_len ((AV *)binaries) + 1)
1455 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1456
1457 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1458 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1459 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1460
1461 int i;
1462 for (i = 0; i < device_count; ++i)
1463 {
1464 STRLEN len;
1465 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1466 length_list [i] = len;
1467 }
1468
1469 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1470 length_list, binary_list,
1471 GIMME_V == G_ARRAY ? status_list : 0, &res));
1472
1473 EXTEND (SP, 2);
1474 PUSH_CLOBJ (stash_program, prog);
1475
1476 if (GIMME_V == G_ARRAY)
1477 {
1478 AV *av = newAV ();
1479 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1480
1481 for (i = device_count; i--; )
1482 av_store (av, i, newSViv (status_list [i]));
1483 }
1484
1485#if CL_VERSION_1_2
1486
1487void
1488program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1489 PPCODE:
1490 cl_uint device_count;
1491 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1492
1493 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1494
1495 XPUSH_CLOBJ (stash_program, prog);
1496
1497void
1498link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1499 CODE:
1500 cl_uint device_count = 0;
1501 cl_device_id *device_list = 0;
1502
1503 if (SvOK (devices))
1504 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1505
1506 cl_uint program_count;
1507 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1508
1509 void *user_data;
1510 program_callback pfn_notify = make_program_callback (notify, &user_data);
1511
1512 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1513 program_count, program_list, pfn_notify, user_data, &res));
1514
1515 XPUSH_CLOBJ (stash_program, prog);
1516
1517#endif
1169 1518
1170#BEGIN:context 1519#BEGIN:context
1171 1520
1172void 1521void
1173reference_count (OpenCL::Context self) 1522reference_count (OpenCL::Context self)
1190 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1539 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1191 int i, n = size / sizeof (*value); 1540 int i, n = size / sizeof (*value);
1192 EXTEND (SP, n); 1541 EXTEND (SP, n);
1193 for (i = 0; i < n; ++i) 1542 for (i = 0; i < n; ++i)
1194 { 1543 {
1195 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1544 PUSH_CLOBJ (stash_device, value [i]);
1196 } 1545 }
1197 1546
1198void 1547void
1199properties (OpenCL::Context self) 1548properties (OpenCL::Context self)
1200 PPCODE: 1549 PPCODE:
1215DESTROY (OpenCL::Queue self) 1564DESTROY (OpenCL::Queue self)
1216 CODE: 1565 CODE:
1217 clReleaseCommandQueue (self); 1566 clReleaseCommandQueue (self);
1218 1567
1219void 1568void
1220enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1569read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1570 ALIAS:
1571 enqueue_read_buffer = 0
1221 PPCODE: 1572 PPCODE:
1222 cl_event ev = 0;
1223 EVENT_LIST (6, items - 6); 1573 EVENT_LIST (6);
1224 1574
1225 SvUPGRADE (data, SVt_PV); 1575 SvUPGRADE (data, SVt_PV);
1226 SvGROW (data, len); 1576 SvGROW (data, len);
1227 SvPOK_only (data); 1577 SvPOK_only (data);
1228 SvCUR_set (data, len); 1578 SvCUR_set (data, len);
1579
1580 cl_event ev = 0;
1229 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1581 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1230 1582
1231 if (ev) 1583 if (ev)
1232 XPUSH_CLOBJ ("OpenCL::Event", ev); 1584 XPUSH_CLOBJ (stash_event, ev);
1233 1585
1234void 1586void
1235enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1587write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1588 ALIAS:
1589 enqueue_write_buffer = 0
1236 PPCODE: 1590 PPCODE:
1237 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1238 STRLEN len; 1593 STRLEN len;
1239 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1240 EVENT_LIST (5, items - 5);
1241 1595
1596 cl_event ev = 0;
1242 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1597 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1243 1598
1244 if (ev) 1599 if (ev)
1245 XPUSH_CLOBJ ("OpenCL::Event", ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1246 1601
1247#if CL_VERSION_1_2 1602#if CL_VERSION_1_2
1248 1603
1249void 1604void
1250enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1605fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1606 ALIAS:
1607 enqueue_fill_buffer = 0
1251 PPCODE: 1608 PPCODE:
1252 cl_event ev = 0; 1609 EVENT_LIST (5);
1610
1253 STRLEN len; 1611 STRLEN len;
1254 char *ptr = SvPVbyte (data, len); 1612 char *ptr = SvPVbyte (data, len);
1255 EVENT_LIST (5, items - 5);
1256 1613
1614 cl_event ev = 0;
1257 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1615 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1258 1616
1259 if (ev) 1617 if (ev)
1260 XPUSH_CLOBJ ("OpenCL::Event", ev); 1618 XPUSH_CLOBJ (stash_event, ev);
1261 1619
1262void 1620void
1263enqueue_fill_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, ...) 1621fill_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, ...)
1622 ALIAS:
1623 enqueue_fill_image = 0
1264 PPCODE: 1624 PPCODE:
1265 cl_event ev = 0; 1625 EVENT_LIST (12);
1266 STRLEN len; 1626
1267 const size_t origin [3] = { x, y, z }; 1627 const size_t origin [3] = { x, y, z };
1268 const size_t region [3] = { width, height, depth }; 1628 const size_t region [3] = { width, height, depth };
1269 EVENT_LIST (12, items - 12);
1270 1629
1271 const cl_float c_f [4] = { r, g, b, a }; 1630 const cl_float c_f [4] = { r, g, b, a };
1272 const cl_uint c_u [4] = { r, g, b, a }; 1631 const cl_uint c_u [4] = { r, g, b, a };
1273 const cl_int c_s [4] = { r, g, b, a }; 1632 const cl_int c_s [4] = { r, g, b, a };
1274 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1633 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1277 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1636 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1278 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1637 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1279 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1638 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1280 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1639 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1281 1640
1641 cl_event ev = 0;
1282 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1642 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1283 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1643 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1284 1644
1285 if (ev) 1645 if (ev)
1286 XPUSH_CLOBJ ("OpenCL::Event", ev); 1646 XPUSH_CLOBJ (stash_event, ev);
1287 1647
1288#endif 1648#endif
1289 1649
1290void 1650void
1291enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1651copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1652 ALIAS:
1653 enqueue_copy_buffer = 0
1292 PPCODE: 1654 PPCODE:
1655 EVENT_LIST (6);
1656
1293 cl_event ev = 0; 1657 cl_event ev = 0;
1294 EVENT_LIST (6, items - 6);
1295
1296 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1658 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1297 1659
1298 if (ev) 1660 if (ev)
1299 XPUSH_CLOBJ ("OpenCL::Event", ev); 1661 XPUSH_CLOBJ (stash_event, ev);
1300 1662
1301void 1663void
1302enqueue_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, ...) 1664read_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, ...)
1665 ALIAS:
1666 enqueue_read_buffer_rect = 0
1303 PPCODE: 1667 PPCODE:
1304 cl_event ev = 0; 1668 EVENT_LIST (17);
1669
1305 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1670 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1306 const size_t host_origin[3] = { host_x, host_y, host_z }; 1671 const size_t host_origin[3] = { host_x, host_y, host_z };
1307 const size_t region[3] = { width, height, depth }; 1672 const size_t region[3] = { width, height, depth };
1308 EVENT_LIST (17, items - 17);
1309 1673
1310 if (!buf_row_pitch) 1674 if (!buf_row_pitch)
1311 buf_row_pitch = region [0]; 1675 buf_row_pitch = region [0];
1312 1676
1313 if (!buf_slice_pitch) 1677 if (!buf_slice_pitch)
1323 1687
1324 SvUPGRADE (data, SVt_PV); 1688 SvUPGRADE (data, SVt_PV);
1325 SvGROW (data, len); 1689 SvGROW (data, len);
1326 SvPOK_only (data); 1690 SvPOK_only (data);
1327 SvCUR_set (data, len); 1691 SvCUR_set (data, len);
1692
1693 cl_event ev = 0;
1328 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)); 1694 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));
1329 1695
1330 if (ev) 1696 if (ev)
1331 XPUSH_CLOBJ ("OpenCL::Event", ev); 1697 XPUSH_CLOBJ (stash_event, ev);
1332 1698
1333void 1699void
1334enqueue_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, ...) 1700write_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, ...)
1701 ALIAS:
1702 enqueue_write_buffer_rect = 0
1335 PPCODE: 1703 PPCODE:
1336 cl_event ev = 0; 1704 EVENT_LIST (17);
1705
1337 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1706 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1338 const size_t host_origin[3] = { host_x, host_y, host_z }; 1707 const size_t host_origin[3] = { host_x, host_y, host_z };
1339 const size_t region[3] = { width, height, depth }; 1708 const size_t region[3] = { width, height, depth };
1340 STRLEN len; 1709 STRLEN len;
1341 char *ptr = SvPVbyte (data, len); 1710 char *ptr = SvPVbyte (data, len);
1342 EVENT_LIST (17, items - 17);
1343 1711
1344 if (!buf_row_pitch) 1712 if (!buf_row_pitch)
1345 buf_row_pitch = region [0]; 1713 buf_row_pitch = region [0];
1346 1714
1347 if (!buf_slice_pitch) 1715 if (!buf_slice_pitch)
1356 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1724 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1357 1725
1358 if (len < min_len) 1726 if (len < min_len)
1359 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1727 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1360 1728
1729 cl_event ev = 0;
1361 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)); 1730 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));
1362 1731
1363 if (ev) 1732 if (ev)
1364 XPUSH_CLOBJ ("OpenCL::Event", ev); 1733 XPUSH_CLOBJ (stash_event, ev);
1365 1734
1366void 1735void
1367enqueue_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, ...) 1736copy_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, ...)
1737 ALIAS:
1738 enqueue_copy_buffer_rect = 0
1368 PPCODE: 1739 PPCODE:
1369 cl_event ev = 0; 1740 EVENT_LIST (16);
1741
1370 const size_t src_origin[3] = { src_x, src_y, src_z }; 1742 const size_t src_origin[3] = { src_x, src_y, src_z };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1743 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth }; 1744 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16);
1374 1745
1746 cl_event ev = 0;
1375 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)); 1747 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));
1376 1748
1377 if (ev) 1749 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1379 1751
1380void 1752void
1381enqueue_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, ...) 1753read_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, ...)
1754 ALIAS:
1755 enqueue_read_image = 0
1382 PPCODE: 1756 PPCODE:
1383 cl_event ev = 0; 1757 EVENT_LIST (12);
1758
1384 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1385 const size_t region[3] = { width, height, depth }; 1760 const size_t region[3] = { width, height, depth };
1386 EVENT_LIST (12, items - 12);
1387 1761
1388 if (!row_pitch) 1762 if (!row_pitch)
1389 row_pitch = img_row_pitch (src); 1763 row_pitch = img_row_pitch (src);
1390 1764
1391 if (depth > 1 && !slice_pitch) 1765 if (depth > 1 && !slice_pitch)
1395 1769
1396 SvUPGRADE (data, SVt_PV); 1770 SvUPGRADE (data, SVt_PV);
1397 SvGROW (data, len); 1771 SvGROW (data, len);
1398 SvPOK_only (data); 1772 SvPOK_only (data);
1399 SvCUR_set (data, len); 1773 SvCUR_set (data, len);
1774
1775 cl_event ev = 0;
1400 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)); 1776 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));
1401 1777
1402 if (ev) 1778 if (ev)
1403 XPUSH_CLOBJ ("OpenCL::Event", ev); 1779 XPUSH_CLOBJ (stash_event, ev);
1404 1780
1405void 1781void
1406enqueue_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, ...) 1782write_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, ...)
1783 ALIAS:
1784 enqueue_write_image = 0
1407 PPCODE: 1785 PPCODE:
1408 cl_event ev = 0; 1786 EVENT_LIST (12);
1787
1409 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1788 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1410 const size_t region[3] = { width, height, depth }; 1789 const size_t region[3] = { width, height, depth };
1411 STRLEN len; 1790 STRLEN len;
1412 char *ptr = SvPVbyte (data, len); 1791 char *ptr = SvPVbyte (data, len);
1413 EVENT_LIST (12, items - 12);
1414 1792
1415 if (!row_pitch) 1793 if (!row_pitch)
1416 row_pitch = img_row_pitch (dst); 1794 row_pitch = img_row_pitch (dst);
1417 1795
1418 if (depth > 1 && !slice_pitch) 1796 if (depth > 1 && !slice_pitch)
1421 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1799 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1422 1800
1423 if (len < min_len) 1801 if (len < min_len)
1424 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1802 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1425 1803
1804 cl_event ev = 0;
1426 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)); 1805 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));
1427 1806
1428 if (ev) 1807 if (ev)
1429 XPUSH_CLOBJ ("OpenCL::Event", ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1430 1809
1431void 1810void
1432enqueue_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, ...) 1811copy_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, ...)
1812 ALIAS:
1813 enqueue_copy_image = 0
1433 PPCODE: 1814 PPCODE:
1434 cl_event ev = 0; 1815 EVENT_LIST (12);
1816
1435 const size_t src_origin[3] = { src_x, src_y, src_z }; 1817 const size_t src_origin[3] = { src_x, src_y, src_z };
1436 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1818 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1437 const size_t region[3] = { width, height, depth }; 1819 const size_t region[3] = { width, height, depth };
1438 EVENT_LIST (12, items - 12);
1439 1820
1821 cl_event ev = 0;
1440 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1822 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1823
1442 if (ev) 1824 if (ev)
1443 XPUSH_CLOBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1444 1826
1445void 1827void
1446enqueue_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, ...) 1828copy_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, ...)
1829 ALIAS:
1830 enqueue_copy_image_to_buffer = 0
1447 PPCODE: 1831 PPCODE:
1832 EVENT_LIST (10);
1833
1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1835 const size_t region [3] = { width, height, depth };
1836
1448 cl_event ev = 0; 1837 cl_event ev = 0;
1838 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1839
1840 if (ev)
1841 XPUSH_CLOBJ (stash_event, ev);
1842
1843void
1844copy_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, ...)
1845 ALIAS:
1846 enqueue_copy_buffer_to_image = 0
1847 PPCODE:
1848 EVENT_LIST (10);
1849
1449 const size_t src_origin[3] = { src_x, src_y, src_z }; 1850 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1851 const size_t region [3] = { width, height, depth };
1852
1853 cl_event ev = 0;
1854 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1855
1856 if (ev)
1857 XPUSH_CLOBJ (stash_event, ev);
1858
1859void
1860map_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, ...)
1861 ALIAS:
1862 enqueue_map_buffer = 0
1863 PPCODE:
1864 EVENT_LIST (6);
1865
1866 size_t cb = SvIV (cb_);
1867
1868 if (!SvOK (cb_))
1869 {
1870 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1871 cb -= offset;
1872 }
1873
1874 cl_event ev;
1875 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1876 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1877
1878void
1879map_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, ...)
1880 ALIAS:
1881 enqueue_map_image = 0
1882 PPCODE:
1883 size_t width = SvIV (width_);
1884 if (!SvOK (width_))
1885 {
1886 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1887 width -= x;
1888 }
1889
1890 size_t height = SvIV (width_);
1891 if (!SvOK (height_))
1892 {
1893 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1894 height -= y;
1895 }
1896
1897 size_t depth = SvIV (width_);
1898 if (!SvOK (depth_))
1899 {
1900 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1901 depth -= z;
1902
1903 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1904 if (!depth)
1905 depth = 1;
1906 }
1907
1908 const size_t origin[3] = { x, y, z };
1450 const size_t region[3] = { width, height, depth }; 1909 const size_t region[3] = { width, height, depth };
1910 size_t row_pitch, slice_pitch;
1451 EVENT_LIST (10, items - 10); 1911 EVENT_LIST (10);
1452 1912
1913 cl_event ev;
1914 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));
1915
1916 size_t cb = slice_pitch ? slice_pitch * region [2]
1917 : row_pitch ? row_pitch * region [1]
1918 : region [0];
1919
1920 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1921
1922void
1923unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1924 PPCODE:
1925 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1926 if (GIMME_V != G_VOID)
1927 {
1928 clRetainEvent (mapped->event);
1929 XPUSH_CLOBJ (stash_event, mapped->event);
1930 }
1931
1932void
1933task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1934 ALIAS:
1935 enqueue_task = 0
1936 PPCODE:
1937 EVENT_LIST (2);
1938
1939 cl_event ev = 0;
1453 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1940 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1454 1941
1455 if (ev) 1942 if (ev)
1456 XPUSH_CLOBJ ("OpenCL::Event", ev); 1943 XPUSH_CLOBJ (stash_event, ev);
1457 1944
1458void 1945void
1459enqueue_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, ...)
1460 PPCODE:
1461 cl_event ev = 0;
1462 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1463 const size_t region[3] = { width, height, depth };
1464 EVENT_LIST (10, items - 10);
1465
1466 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1467
1468 if (ev)
1469 XPUSH_CLOBJ ("OpenCL::Event", ev);
1470
1471void
1472enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1473 PPCODE:
1474 cl_event ev = 0;
1475 EVENT_LIST (2, items - 2);
1476
1477 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1478
1479 if (ev)
1480 XPUSH_CLOBJ ("OpenCL::Event", ev);
1481
1482void
1483enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1946nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1947 ALIAS:
1948 enqueue_nd_range_kernel = 0
1484 PPCODE: 1949 PPCODE:
1485 cl_event ev = 0; 1950 EVENT_LIST (5);
1951
1486 size_t *gwo = 0, *gws, *lws = 0; 1952 size_t *gwo = 0, *gws, *lws = 0;
1487 int gws_len; 1953 int gws_len;
1488 size_t *lists; 1954 size_t *lists;
1489 int i; 1955 int i;
1490 EVENT_LIST (5, items - 5);
1491 1956
1492 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1957 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1493 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1958 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1494 1959
1495 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1960 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1496 1961
1497 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1962 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1498 1963
1499 gws = lists + gws_len * 0; 1964 gws = lists + gws_len * 0;
1500 for (i = 0; i < gws_len; ++i) 1965 for (i = 0; i < gws_len; ++i)
1966 {
1501 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1967 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1968 // at least nvidia crashes for 0-sized work group sizes, work around
1969 if (!gws [i])
1970 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1971 }
1502 1972
1503 if (SvOK (global_work_offset)) 1973 if (SvOK (global_work_offset))
1504 { 1974 {
1505 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1975 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1506 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1976 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1514 } 1984 }
1515 1985
1516 if (SvOK (local_work_size)) 1986 if (SvOK (local_work_size))
1517 { 1987 {
1518 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1988 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1519 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1989 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1520 1990
1521 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1991 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1522 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1992 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1523 1993
1524 lws = lists + gws_len * 2; 1994 lws = lists + gws_len * 2;
1525 for (i = 0; i < gws_len; ++i) 1995 for (i = 0; i < gws_len; ++i)
1996 {
1526 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1997 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1998 // at least nvidia crashes for 0-sized work group sizes, work around
1999 if (!lws [i])
2000 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2001 }
1527 } 2002 }
1528 2003
2004 cl_event ev = 0;
1529 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2005 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1530 2006
1531 if (ev) 2007 if (ev)
1532 XPUSH_CLOBJ ("OpenCL::Event", ev); 2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#if CL_VERSION_1_2
2011
2012void
2013migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2014 ALIAS:
2015 enqueue_migrate_mem_objects = 0
2016 PPCODE:
2017 EVENT_LIST (3);
2018
2019 cl_uint obj_count;
2020 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2021
2022 cl_event ev = 0;
2023 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2024
2025 if (ev)
2026 XPUSH_CLOBJ (stash_event, ev);
2027
2028#endif
1533 2029
1534#if cl_apple_gl_sharing || cl_khr_gl_sharing 2030#if cl_apple_gl_sharing || cl_khr_gl_sharing
1535 2031
1536void 2032void
1537enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2033acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1538 ALIAS: 2034 ALIAS:
2035 release_gl_objects = 1
2036 enqueue_acquire_gl_objects = 0
1539 enqueue_release_gl_objects = 1 2037 enqueue_release_gl_objects = 1
1540 PPCODE: 2038 PPCODE:
1541 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2039 EVENT_LIST (2);
1542 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2040
2041 cl_uint obj_count;
2042 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1543 2043
1544 cl_event ev = 0; 2044 cl_event ev = 0;
1545 EVENT_LIST (2, items - 2);
1546 AV *av = (AV *)SvRV (objects);
1547 cl_uint num_objects = av_len (av) + 1;
1548 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1549 int i;
1550
1551 for (i = num_objects; i--; )
1552 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1553 2045
1554 if (ix) 2046 if (ix)
1555 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2047 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1556 else 2048 else
1557 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2049 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1558 2050
1559 if (ev) 2051 if (ev)
1560 XPUSH_CLOBJ ("OpenCL::Event", ev); 2052 XPUSH_CLOBJ (stash_event, ev);
1561 2053
1562#endif 2054#endif
1563 2055
1564void 2056void
1565enqueue_wait_for_events (OpenCL::Queue self, ...) 2057wait_for_events (OpenCL::Queue self, ...)
2058 ALIAS:
2059 enqueue_wait_for_events = 0
1566 CODE: 2060 CODE:
1567 EVENT_LIST (1, items - 1); 2061 EVENT_LIST (1);
1568#if PREFER_1_1 2062#if PREFER_1_1
1569 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2063 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1570#else 2064#else
1571 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2065 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1572#endif 2066#endif
1573 2067
1574void 2068void
1575enqueue_marker (OpenCL::Queue self, ...) 2069marker (OpenCL::Queue self, ...)
2070 ALIAS:
2071 enqueue_marker = 0
1576 PPCODE: 2072 PPCODE:
2073 EVENT_LIST (1);
1577 cl_event ev = 0; 2074 cl_event ev = 0;
1578 EVENT_LIST (1, items - 1);
1579#if PREFER_1_1 2075#if PREFER_1_1
1580 if (!event_list_count) 2076 if (!event_list_count)
1581 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2077 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1582 else 2078 else
1583#if CL_VERSION_1_2 2079#if CL_VERSION_1_2
1590#endif 2086#endif
1591#else 2087#else
1592 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2088 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1593#endif 2089#endif
1594 if (ev) 2090 if (ev)
1595 XPUSH_CLOBJ ("OpenCL::Event", ev); 2091 XPUSH_CLOBJ (stash_event, ev);
1596 2092
1597void 2093void
1598enqueue_barrier (OpenCL::Queue self, ...) 2094barrier (OpenCL::Queue self, ...)
2095 ALIAS:
2096 enqueue_barrier = 0
1599 PPCODE: 2097 PPCODE:
2098 EVENT_LIST (1);
1600 cl_event ev = 0; 2099 cl_event ev = 0;
1601 EVENT_LIST (1, items - 1);
1602#if PREFER_1_1 2100#if PREFER_1_1
1603 if (!event_list_count && GIMME_V == G_VOID) 2101 if (!event_list_count && GIMME_V == G_VOID)
1604 NEED_SUCCESS (EnqueueBarrier, (self)); 2102 NEED_SUCCESS (EnqueueBarrier, (self));
1605 else 2103 else
1606#if CL_VERSION_1_2 2104#if CL_VERSION_1_2
1616#endif 2114#endif
1617#else 2115#else
1618 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2116 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1619#endif 2117#endif
1620 if (ev) 2118 if (ev)
1621 XPUSH_CLOBJ ("OpenCL::Event", ev); 2119 XPUSH_CLOBJ (stash_event, ev);
1622 2120
1623void 2121void
1624flush (OpenCL::Queue self) 2122flush (OpenCL::Queue self)
1625 CODE: 2123 CODE:
1626 NEED_SUCCESS (Flush, (self)); 2124 NEED_SUCCESS (Flush, (self));
1644 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2142 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1645 EXTEND (SP, 1); 2143 EXTEND (SP, 1);
1646 const int i = 0; 2144 const int i = 0;
1647 { 2145 {
1648 NEED_SUCCESS (RetainContext, (value [i])); 2146 NEED_SUCCESS (RetainContext, (value [i]));
1649 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2147 PUSH_CLOBJ (stash_context, value [i]);
1650 } 2148 }
1651 2149
1652void 2150void
1653device (OpenCL::Queue self) 2151device (OpenCL::Queue self)
1654 PPCODE: 2152 PPCODE:
1655 cl_device_id value [1]; 2153 cl_device_id value [1];
1656 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2154 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1657 EXTEND (SP, 1); 2155 EXTEND (SP, 1);
1658 const int i = 0; 2156 const int i = 0;
1659 { 2157 {
1660 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2158 PUSH_CLOBJ (stash_device, value [i]);
1661 } 2159 }
1662 2160
1663void 2161void
1664reference_count (OpenCL::Queue self) 2162reference_count (OpenCL::Queue self)
1665 PPCODE: 2163 PPCODE:
1674 PPCODE: 2172 PPCODE:
1675 cl_command_queue_properties value [1]; 2173 cl_command_queue_properties value [1];
1676 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2174 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1677 EXTEND (SP, 1); 2175 EXTEND (SP, 1);
1678 const int i = 0; 2176 const int i = 0;
1679 PUSHs (sv_2mortal (newSViv (value [i]))); 2177 PUSHs (sv_2mortal (newSVuv (value [i])));
1680 2178
1681#END:command_queue 2179#END:command_queue
1682 2180
1683MODULE = OpenCL PACKAGE = OpenCL::Memory 2181MODULE = OpenCL PACKAGE = OpenCL::Memory
1684 2182
1699 PPCODE: 2197 PPCODE:
1700 cl_mem_object_type value [1]; 2198 cl_mem_object_type value [1];
1701 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2199 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1702 EXTEND (SP, 1); 2200 EXTEND (SP, 1);
1703 const int i = 0; 2201 const int i = 0;
1704 PUSHs (sv_2mortal (newSViv (value [i]))); 2202 PUSHs (sv_2mortal (newSVuv (value [i])));
1705 2203
1706void 2204void
1707flags (OpenCL::Memory self) 2205flags (OpenCL::Memory self)
1708 PPCODE: 2206 PPCODE:
1709 cl_mem_flags value [1]; 2207 cl_mem_flags value [1];
1710 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2208 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1711 EXTEND (SP, 1); 2209 EXTEND (SP, 1);
1712 const int i = 0; 2210 const int i = 0;
1713 PUSHs (sv_2mortal (newSViv (value [i]))); 2211 PUSHs (sv_2mortal (newSVuv (value [i])));
1714 2212
1715void 2213void
1716size (OpenCL::Memory self) 2214size (OpenCL::Memory self)
1717 ALIAS: 2215 ALIAS:
1718 size = CL_MEM_SIZE 2216 size = CL_MEM_SIZE
1752 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2250 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1753 EXTEND (SP, 1); 2251 EXTEND (SP, 1);
1754 const int i = 0; 2252 const int i = 0;
1755 { 2253 {
1756 NEED_SUCCESS (RetainContext, (value [i])); 2254 NEED_SUCCESS (RetainContext, (value [i]));
1757 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2255 PUSH_CLOBJ (stash_context, value [i]);
1758 } 2256 }
1759 2257
1760void 2258void
1761associated_memobject (OpenCL::Memory self) 2259associated_memobject (OpenCL::Memory self)
1762 PPCODE: 2260 PPCODE:
1764 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2262 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1765 EXTEND (SP, 1); 2263 EXTEND (SP, 1);
1766 const int i = 0; 2264 const int i = 0;
1767 { 2265 {
1768 NEED_SUCCESS (RetainMemObject, (value [i])); 2266 NEED_SUCCESS (RetainMemObject, (value [i]));
1769 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2267 PUSH_CLOBJ (stash_memory, value [i]);
1770 } 2268 }
1771 2269
1772#END:mem 2270#END:mem
1773 2271
1774#if cl_apple_gl_sharing || cl_khr_gl_sharing 2272#if cl_apple_gl_sharing || cl_khr_gl_sharing
1794 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2292 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1795 2293
1796 cl_buffer_region crdata = { origin, size }; 2294 cl_buffer_region crdata = { origin, size };
1797 2295
1798 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2296 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1799 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2297 XPUSH_CLOBJ (stash_buffer, mem);
1800 2298
1801MODULE = OpenCL PACKAGE = OpenCL::Image 2299MODULE = OpenCL PACKAGE = OpenCL::Image
1802 2300
1803void 2301void
1804image_info (OpenCL::Image self, cl_image_info name) 2302image_info (OpenCL::Image self, cl_image_info name)
1890 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1891 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1892 const int i = 0; 2390 const int i = 0;
1893 { 2391 {
1894 NEED_SUCCESS (RetainContext, (value [i])); 2392 NEED_SUCCESS (RetainContext, (value [i]));
1895 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2393 PUSH_CLOBJ (stash_context, value [i]);
1896 } 2394 }
1897 2395
1898void 2396void
1899normalized_coords (OpenCL::Sampler self) 2397normalized_coords (OpenCL::Sampler self)
1900 PPCODE: 2398 PPCODE:
1901 cl_addressing_mode value [1]; 2399 cl_addressing_mode value [1];
1902 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2400 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1903 EXTEND (SP, 1); 2401 EXTEND (SP, 1);
1904 const int i = 0; 2402 const int i = 0;
1905 PUSHs (sv_2mortal (newSViv (value [i]))); 2403 PUSHs (sv_2mortal (newSVuv (value [i])));
1906 2404
1907void 2405void
1908addressing_mode (OpenCL::Sampler self) 2406addressing_mode (OpenCL::Sampler self)
1909 PPCODE: 2407 PPCODE:
1910 cl_filter_mode value [1]; 2408 cl_filter_mode value [1];
1911 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2409 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1912 EXTEND (SP, 1); 2410 EXTEND (SP, 1);
1913 const int i = 0; 2411 const int i = 0;
1914 PUSHs (sv_2mortal (newSViv (value [i]))); 2412 PUSHs (sv_2mortal (newSVuv (value [i])));
1915 2413
1916void 2414void
1917filter_mode (OpenCL::Sampler self) 2415filter_mode (OpenCL::Sampler self)
1918 PPCODE: 2416 PPCODE:
1919 cl_bool value [1]; 2417 cl_bool value [1];
1934void 2432void
1935build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2433build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1936 ALIAS: 2434 ALIAS:
1937 build_async = 1 2435 build_async = 1
1938 CODE: 2436 CODE:
1939 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1940 void *user_data = 0;
1941 cl_uint num_devices = 0; 2437 cl_uint device_count = 0;
1942 cl_device_id *device_list = 0; 2438 cl_device_id *device_list = 0;
1943 2439
1944 if (SvOK (devices)) 2440 if (SvOK (devices))
1945 { 2441 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2442
2443 void *user_data;
2444 program_callback pfn_notify = make_program_callback (notify, &user_data);
2445
2446 if (ix)
2447 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2448 else
2449 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2450
2451#if CL_VERSION_1_2
2452
2453void
2454compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2455 CODE:
2456 cl_uint device_count = 0;
2457 cl_device_id *device_list = 0;
2458
2459 if (SvOK (devices))
2460 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2461
2462 cl_uint header_count = 0;
2463 cl_program *header_list = 0;
2464 const char **header_name = 0;
2465
2466 if (SvOK (headers))
2467 {
1946 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2468 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1947 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2469 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1948 2470
1949 AV *av = (AV *)SvRV (devices); 2471 HV *hv = (HV *)SvRV (devices);
1950 num_devices = av_len (av) + 1;
1951 2472
1952 if (num_devices) 2473 header_count = hv_iterinit (hv);
2474 header_list = tmpbuf (sizeof (*header_list) * header_count);
2475 header_name = tmpbuf (sizeof (*header_name) * header_count);
2476
2477 HE *he;
2478 int i = 0;
2479 while (he = hv_iternext (hv))
1953 { 2480 {
1954 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2481 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2482 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1955 int count; 2483 ++i;
1956 for (count = 0; count < num_devices; ++count)
1957 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1958 } 2484 }
1959 } 2485 }
1960 2486
1961 if (SvOK (notify)) 2487 void *user_data;
1962 { 2488 program_callback pfn_notify = make_program_callback (notify, &user_data);
1963 NEED_SUCCESS (RetainProgram, (self));
1964 pfn_notify = eq_program_notify;
1965 user_data = SvREFCNT_inc (s_get_cv (notify));
1966 }
1967 2489
1968 if (ix)
1969 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1970 else
1971 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2490 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2491 header_count, header_list, header_name, pfn_notify, user_data));
2492
2493#endif
1972 2494
1973void 2495void
1974build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2496build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1975 PPCODE: 2497 PPCODE:
1976 size_t size; 2498 size_t size;
2005 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2527 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2006 EXTEND (SP, 1); 2528 EXTEND (SP, 1);
2007 const int i = 0; 2529 const int i = 0;
2008 PUSHs (sv_2mortal (newSVpv (value, 0))); 2530 PUSHs (sv_2mortal (newSVpv (value, 0)));
2009 2531
2532void
2533binary_type (OpenCL::Program self, OpenCL::Device device)
2534 PPCODE:
2535 cl_program_binary_type value [1];
2536 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2537 EXTEND (SP, 1);
2538 const int i = 0;
2539 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2540
2010#END:program_build 2541#END:program_build
2011 2542
2012void 2543void
2013kernel (OpenCL::Program program, SV *function) 2544kernel (OpenCL::Program program, SV *function)
2014 PPCODE: 2545 PPCODE:
2015 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2546 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2016 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2547 XPUSH_CLOBJ (stash_kernel, kernel);
2017 2548
2018void 2549void
2019kernels_in_program (OpenCL::Program program) 2550kernels_in_program (OpenCL::Program program)
2020 PPCODE: 2551 PPCODE:
2021 cl_uint num_kernels; 2552 cl_uint num_kernels;
2024 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2555 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2025 2556
2026 int i; 2557 int i;
2027 EXTEND (SP, num_kernels); 2558 EXTEND (SP, num_kernels);
2028 for (i = 0; i < num_kernels; ++i) 2559 for (i = 0; i < num_kernels; ++i)
2029 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2560 PUSH_CLOBJ (stash_kernel, kernels [i]);
2030 2561
2031void 2562void
2032info (OpenCL::Program self, cl_program_info name) 2563info (OpenCL::Program self, cl_program_info name)
2033 PPCODE: 2564 PPCODE:
2034 INFO (Program) 2565 INFO (Program)
2082 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2613 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2083 EXTEND (SP, 1); 2614 EXTEND (SP, 1);
2084 const int i = 0; 2615 const int i = 0;
2085 { 2616 {
2086 NEED_SUCCESS (RetainContext, (value [i])); 2617 NEED_SUCCESS (RetainContext, (value [i]));
2087 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2618 PUSH_CLOBJ (stash_context, value [i]);
2088 } 2619 }
2089 2620
2090void 2621void
2091devices (OpenCL::Program self) 2622devices (OpenCL::Program self)
2092 PPCODE: 2623 PPCODE:
2096 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2627 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2097 int i, n = size / sizeof (*value); 2628 int i, n = size / sizeof (*value);
2098 EXTEND (SP, n); 2629 EXTEND (SP, n);
2099 for (i = 0; i < n; ++i) 2630 for (i = 0; i < n; ++i)
2100 { 2631 {
2101 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2632 PUSH_CLOBJ (stash_device, value [i]);
2102 } 2633 }
2103 2634
2104void 2635void
2105source (OpenCL::Program self) 2636source (OpenCL::Program self)
2106 PPCODE: 2637 PPCODE:
2132DESTROY (OpenCL::Kernel self) 2663DESTROY (OpenCL::Kernel self)
2133 CODE: 2664 CODE:
2134 clReleaseKernel (self); 2665 clReleaseKernel (self);
2135 2666
2136void 2667void
2668setf (OpenCL::Kernel self, const char *format, ...)
2669 CODE:
2670 int i;
2671 for (i = 2; ; ++i)
2672 {
2673 while (*format == ' ')
2674 ++format;
2675
2676 char type = *format++;
2677
2678 if (!type)
2679 break;
2680
2681 if (i >= items)
2682 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2683
2684 SV *sv = ST (i);
2685
2686 union
2687 {
2688 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2689 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2690 cl_half ch; cl_float cf; cl_double cd;
2691 cl_mem cm;
2692 cl_sampler ca;
2693 size_t cz;
2694 cl_event ce;
2695 } arg;
2696 size_t size;
2697 int nullarg = 0;
2698
2699 switch (type)
2700 {
2701 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2702 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2703 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2704 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2705 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2706 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2707 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2708 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2709
2710 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2711 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2712 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2713
2714 case 'z': nullarg = 1; size = SvIV (sv); break;
2715
2716 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2717 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2718 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2719
2720 default:
2721 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2722 }
2723
2724 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2725 if (res)
2726 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2727 }
2728
2729 if (i != items)
2730 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2731
2732void
2137set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2733set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2138 CODE: 2734 CODE:
2139 clSetKernelArg (self, idx, sizeof (value), &value); 2735 clSetKernelArg (self, idx, sizeof (value), &value);
2140 2736
2141void 2737void
2189 clSetKernelArg (self, idx, sizeof (value), &value); 2785 clSetKernelArg (self, idx, sizeof (value), &value);
2190 2786
2191void 2787void
2192set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2788set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2193 CODE: 2789 CODE:
2194 clSetKernelArg (self, idx, sizeof (value), &value); 2790 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2195 2791
2196void 2792void
2197set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2793set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2198 CODE: 2794 CODE:
2199 clSetKernelArg (self, idx, sizeof (value), &value); 2795 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2200 2796
2201void 2797void
2202set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2798set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2799 ALIAS:
2800 set_image2d = 0
2801 set_image3d = 0
2203 CODE: 2802 CODE:
2204 clSetKernelArg (self, idx, sizeof (value), &value); 2803 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2205
2206void
2207set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
2208 CODE:
2209 clSetKernelArg (self, idx, sizeof (value), &value);
2210 2804
2211void 2805void
2212set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2806set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2213 CODE: 2807 CODE:
2214 clSetKernelArg (self, idx, sizeof (value), &value); 2808 clSetKernelArg (self, idx, sizeof (value), &value);
2260 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2854 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2261 EXTEND (SP, 1); 2855 EXTEND (SP, 1);
2262 const int i = 0; 2856 const int i = 0;
2263 { 2857 {
2264 NEED_SUCCESS (RetainContext, (value [i])); 2858 NEED_SUCCESS (RetainContext, (value [i]));
2265 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2859 PUSH_CLOBJ (stash_context, value [i]);
2266 } 2860 }
2267 2861
2268void 2862void
2269program (OpenCL::Kernel self) 2863program (OpenCL::Kernel self)
2270 PPCODE: 2864 PPCODE:
2272 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2866 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2273 EXTEND (SP, 1); 2867 EXTEND (SP, 1);
2274 const int i = 0; 2868 const int i = 0;
2275 { 2869 {
2276 NEED_SUCCESS (RetainProgram, (value [i])); 2870 NEED_SUCCESS (RetainProgram, (value [i]));
2277 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2871 PUSH_CLOBJ (stash_program, value [i]);
2278 } 2872 }
2279 2873
2280#END:kernel 2874#END:kernel
2281 2875
2282void 2876void
2329 const int i = 0; 2923 const int i = 0;
2330 PUSHs (sv_2mortal (newSVuv (value [i]))); 2924 PUSHs (sv_2mortal (newSVuv (value [i])));
2331 2925
2332#END:kernel_work_group 2926#END:kernel_work_group
2333 2927
2928#if CL_VERSION_1_2
2929
2930void
2931arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2932 PPCODE:
2933 size_t size;
2934 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2935 SV *sv = sv_2mortal (newSV (size));
2936 SvUPGRADE (sv, SVt_PV);
2937 SvPOK_only (sv);
2938 SvCUR_set (sv, size);
2939 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2940 XPUSHs (sv);
2941
2942#BEGIN:kernel_arg
2943
2944void
2945arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2946 PPCODE:
2947 cl_kernel_arg_address_qualifier value [1];
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2949 EXTEND (SP, 1);
2950 const int i = 0;
2951 PUSHs (sv_2mortal (newSVuv (value [i])));
2952
2953void
2954arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2955 PPCODE:
2956 cl_kernel_arg_access_qualifier value [1];
2957 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2958 EXTEND (SP, 1);
2959 const int i = 0;
2960 PUSHs (sv_2mortal (newSVuv (value [i])));
2961
2962void
2963arg_type_name (OpenCL::Kernel self, cl_uint idx)
2964 ALIAS:
2965 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2966 arg_name = CL_KERNEL_ARG_NAME
2967 PPCODE:
2968 size_t size;
2969 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2970 char *value = tmpbuf (size);
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVpv (value, 0)));
2975
2976void
2977arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2978 PPCODE:
2979 cl_kernel_arg_type_qualifier value [1];
2980 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2981 EXTEND (SP, 1);
2982 const int i = 0;
2983 PUSHs (sv_2mortal (newSVuv (value [i])));
2984
2985#END:kernel_arg
2986
2987#endif
2988
2334MODULE = OpenCL PACKAGE = OpenCL::Event 2989MODULE = OpenCL PACKAGE = OpenCL::Event
2335 2990
2336void 2991void
2337DESTROY (OpenCL::Event self) 2992DESTROY (OpenCL::Event self)
2338 CODE: 2993 CODE:
2362 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3017 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2363 EXTEND (SP, 1); 3018 EXTEND (SP, 1);
2364 const int i = 0; 3019 const int i = 0;
2365 { 3020 {
2366 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3021 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2367 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 3022 PUSH_CLOBJ (stash_queue, value [i]);
2368 } 3023 }
2369 3024
2370void 3025void
2371command_type (OpenCL::Event self) 3026command_type (OpenCL::Event self)
2372 PPCODE: 3027 PPCODE:
2395 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3050 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2396 EXTEND (SP, 1); 3051 EXTEND (SP, 1);
2397 const int i = 0; 3052 const int i = 0;
2398 { 3053 {
2399 NEED_SUCCESS (RetainContext, (value [i])); 3054 NEED_SUCCESS (RetainContext, (value [i]));
2400 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3055 PUSH_CLOBJ (stash_context, value [i]);
2401 } 3056 }
2402 3057
2403#END:event 3058#END:event
2404 3059
2405void 3060void
2430void 3085void
2431set_status (OpenCL::UserEvent self, cl_int execution_status) 3086set_status (OpenCL::UserEvent self, cl_int execution_status)
2432 CODE: 3087 CODE:
2433 clSetUserEventStatus (self, execution_status); 3088 clSetUserEventStatus (self, execution_status);
2434 3089
3090MODULE = OpenCL PACKAGE = OpenCL::Mapped
3091
3092void
3093DESTROY (SV *self)
3094 CODE:
3095 OpenCL__Mapped mapped = SvMAPPED (self);
3096
3097 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3098 mapped_detach (self, mapped);
3099
3100 clReleaseCommandQueue (mapped->queue);
3101 clReleaseEvent (mapped->event);
3102 Safefree (mapped);
3103
3104void
3105unmap (OpenCL::Mapped self, ...)
3106 CODE:
3107 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3108
3109bool
3110mapped (OpenCL::Mapped self)
3111 CODE:
3112 RETVAL = !!self->ptr;
3113 OUTPUT:
3114 RETVAL
3115
3116void
3117wait (OpenCL::Mapped self)
3118 PPCODE:
3119 if (self->event)
3120 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3121
3122void
3123event (OpenCL::Mapped self)
3124 PPCODE:
3125 if (!self->event)
3126 XSRETURN_UNDEF;
3127
3128 clRetainEvent (self->event);
3129 XPUSH_CLOBJ (stash_event, self->event);
3130
3131size_t
3132size (OpenCL::Mapped self)
3133 CODE:
3134 RETVAL = self->cb;
3135 OUTPUT:
3136 RETVAL
3137
3138IV
3139ptr (OpenCL::Mapped self)
3140 CODE:
3141 RETVAL = PTR2IV (self->ptr);
3142 OUTPUT:
3143 RETVAL
3144
3145void
3146set (OpenCL::Mapped self, size_t offset, SV *data)
3147 CODE:
3148 STRLEN len;
3149 const char *ptr = SvPVbyte (data, len);
3150
3151 if (offset + len > self->cb)
3152 croak ("OpenCL::Mapped::set out of bound condition detected");
3153
3154 memcpy (offset + (char *)self->ptr, ptr, len);
3155
3156MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3157
3158MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3159
3160IV
3161row_pitch (OpenCL::Mapped self)
3162 ALIAS:
3163 slice_pitch = 1
3164 CODE:
3165 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3166 OUTPUT:
3167 RETVAL
3168
3169
3170

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines