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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.52 by root, Tue Apr 24 23:53:12 2012 UTC vs.
Revision 1.76 by root, Sat May 5 14:49:52 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
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;
127 } 168 }
128 169
129 return buf [idx]; 170 return buf [idx];
130} 171}
131 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
132/*****************************************************************************/ 188/*****************************************************************************/
133 189
134typedef struct 190typedef struct
135{ 191{
136 IV iv; 192 IV iv;
137 const char *name; 193 const char *name;
138 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
139} ivstr; 195} ivstr;
196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
140 203
141static const char * 204static const char *
142iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
143{ 206{
144 int i; 207 int i;
193 if (res) \ 256 if (res) \
194 FAIL (name); 257 FAIL (name);
195 258
196/*****************************************************************************/ 259/*****************************************************************************/
197 260
261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
274{
275 // sv_derived_from is quite slow :(
276 if (SvROK (sv) && sv_derived_from (sv, pkg))
277 return (void *)SvIV (SvRV (sv));
278
279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
290}
291
292/*****************************************************************************/
293
198static cl_context_properties * 294static cl_context_properties * ecb_noinline
199SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
200{ 296{
201 if (!sv || !SvOK (sv)) 297 if (!sv || !SvOK (sv))
202 if (extra) 298 if (extra)
203 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
204 else 300 else
210 int i, len = av_len (av) + 1; 306 int i, len = av_len (av) + 1;
211 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
212 cl_context_properties *l = p; 308 cl_context_properties *l = p;
213 309
214 if (len & 1) 310 if (len & 1)
215 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
216 312
217 while (extracount--) 313 while (extracount--)
218 *l++ = *extra++; 314 *l++ = *extra++;
219 315
220 for (i = 0; i < len; i += 2) 316 for (i = 0; i < len; i += 2)
223 SV *p_sv = *av_fetch (av, i + 1, 0); 319 SV *p_sv = *av_fetch (av, i + 1, 0);
224 cl_context_properties v = SvIV (p_sv); // code below can override 320 cl_context_properties v = SvIV (p_sv); // code below can override
225 321
226 switch (t) 322 switch (t)
227 { 323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
228 case CL_GLX_DISPLAY_KHR: 329 case CL_GLX_DISPLAY_KHR:
229 if (!SvOK (p_sv)) 330 if (!SvOK (p_sv))
230 { 331 {
231 void *func = glsym ("_glXGetCurrentDisplay"); 332 void *func = glsym ("_glXGetCurrentDisplay");
232 if (func) 333 if (func)
255 *l = 0; 356 *l = 0;
256 357
257 return p; 358 return p;
258 } 359 }
259 360
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
261} 362}
262 363
263/*****************************************************************************/ 364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
264 365// is a pointer (and only on well-behaved systems).
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 366static void * ecb_noinline
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 367object_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{ 368{
274 if (SvROK (sv) && sv_derived_from (sv, pkg)) 369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
372
373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
275 { 378 {
276 SV *rv = SvRV (sv); 379 list = tmpbuf (sizeof (*list) * count);
277 380 int i;
278#if CLOBJ_PUSH 381 for (i = 0; i < count; ++i)
279 if (SvTYPE (rv) == SVt_PVAV) 382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
280 rv = AvARRAY (rv)[0];
281#endif
282
283 return (void *)SvIV (rv);
284 } 383 }
285 384
286 croak ("%s: %s is not of type %s", func, svname, pkg); 385 if (!count && !or_undef)
287} 386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
288 388
289#if CLOBJ_PUSH 389 *rcount = count;
290 390 return (void *)list;
291static void
292CLOBJ_push (SV *self, SV *data)
293{
294 SV *rv = SvRV (self);
295
296 if (SvTYPE (rv) != SVt_PVAV)
297 {
298 AV *av = newAV ();
299 av_push (av, rv);
300 rv = (SV *)av;
301 SvRV_set (self, rv);
302 }
303
304 av_push ((AV *)rv, data);
305} 391}
306
307static SV *
308sv_struct (STRLEN size)
309{
310 SV *sv = newSV (size);
311 SvPOK_only (sv);
312 return sv;
313}
314
315static void *
316CLOBJ_push_struct (SV *self, STRLEN size)
317{
318 SV *sv = sv_struct (size);
319 CLOBJ_push (self, sv);
320 return SvPVX (sv);
321}
322
323#endif
324 392
325/*****************************************************************************/ 393/*****************************************************************************/
326/* callback stuff */ 394/* callback stuff */
327 395
328/* default context callback, log to stderr */ 396/* default context callback, log to stderr */
349static void (*eq_signal_func)(void *signal_arg, int value); 417static void (*eq_signal_func)(void *signal_arg, int value);
350static void *eq_signal_arg; 418static void *eq_signal_arg;
351static xmutex_t eq_lock = X_MUTEX_INIT; 419static xmutex_t eq_lock = X_MUTEX_INIT;
352static eq_item *eq_head, *eq_tail; 420static eq_item *eq_head, *eq_tail;
353 421
354static void 422static void ecb_noinline
355eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
356{ 424{
357 eq_item *item = malloc (sizeof (eq_item)); 425 eq_item *item = malloc (sizeof (eq_item));
358 426
359 item->next = 0; 427 item->next = 0;
387 X_UNLOCK (eq_lock); 455 X_UNLOCK (eq_lock);
388 456
389 return res; 457 return res;
390} 458}
391 459
392#if 0
393static void
394mem_free (pTHX_ void *p)
395{
396 free (p);
397}
398//SAVEDESTRUCTOR_X (mem_free, item);
399#endif
400
401static void 460static void
402eq_poll (void) 461eq_poll (void)
403{ 462{
404 eq_item *item; 463 eq_item *item;
405 464
435} 494}
436 495
437/*****************************************************************************/ 496/*****************************************************************************/
438/* context notify */ 497/* context notify */
439 498
440static void 499static void ecb_noinline
441eq_context_push (void *data1, void *data2, void *data3) 500eq_context_push (void *data1, void *data2, void *data3)
442{ 501{
443 dSP; 502 dSP;
444 PUSHs (sv_2mortal (newSVpv (data1, 0))); 503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
445 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
467 { \ 526 { \
468 pfn_notify = eq_context_notify; \ 527 pfn_notify = eq_context_notify; \
469 user_data = s_get_cv (notify); \ 528 user_data = s_get_cv (notify); \
470 } 529 }
471 530
472static SV * 531static SV * ecb_noinline
473new_clobj_context (cl_context ctx, void *user_data) 532new_clobj_context (cl_context ctx, void *user_data)
474{ 533{
475 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 534 SV *sv = new_clobj (stash_context, (IV)ctx);
476 535
477 if (user_data) 536 if (user_data)
478 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 537 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
479 538
480 return sv; 539 return sv;
487 546
488static void 547static void
489eq_program_push (void *data1, void *data2, void *data3) 548eq_program_push (void *data1, void *data2, void *data3)
490{ 549{
491 dSP; 550 dSP;
492 PUSH_CLOBJ ("OpenCL::Program", data1); 551 PUSH_CLOBJ (stash_program, data1);
493 PUTBACK; 552 PUTBACK;
494} 553}
495 554
496static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
497 556
498static void CL_CALLBACK 557static void CL_CALLBACK
499eq_program_notify (cl_program program, void *user_data) 558eq_program_notify (cl_program program, void *user_data)
500{ 559{
560 clRetainProgram (program);
561
501 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
502} 580}
503 581
504struct build_args 582struct build_args
505{ 583{
506 cl_program program; 584 cl_program program;
520 else 598 else
521 clReleaseProgram (arg->program); 599 clReleaseProgram (arg->program);
522 600
523 free (arg->options); 601 free (arg->options);
524 free (arg); 602 free (arg);
603
604 return 0;
525} 605}
526 606
527static void 607static void
528build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
529{ 609{
544 624
545static void 625static void
546eq_event_push (void *data1, void *data2, void *data3) 626eq_event_push (void *data1, void *data2, void *data3)
547{ 627{
548 dSP; 628 dSP;
549 PUSH_CLOBJ ("OpenCL::Event", data1); 629 PUSH_CLOBJ (stash_event, data1);
550 PUSHs (sv_2mortal (newSViv ((IV)data2))); 630 PUSHs (sv_2mortal (newSViv ((IV)data2)));
551 PUTBACK; 631 PUTBACK;
552} 632}
553 633
554static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 634static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
559 clRetainEvent (event); 639 clRetainEvent (event);
560 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 640 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
561} 641}
562 642
563/*****************************************************************************/ 643/*****************************************************************************/
644/* utilities for XS code */
564 645
565static size_t 646static size_t
566img_row_pitch (cl_mem img) 647img_row_pitch (cl_mem img)
567{ 648{
568 size_t res; 649 size_t res;
569 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 650 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
570 return res; 651 return res;
571} 652}
572 653
573static cl_event * 654static cl_event * ecb_noinline
574event_list (SV **items, cl_uint *rcount) 655event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
575{ 656{
576 cl_uint count = *rcount; 657 cl_uint count = *rcount;
577 658
578 if (!count) 659 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
660 *rcount = count = 0;
661
662 if (!count && !extra)
579 return 0; 663 return 0;
580 664
581 cl_event *list = tmpbuf (sizeof (cl_event) * count); 665 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
582 int i = 0; 666 int i = 0;
583 667
584 do 668 while (count--)
585 {
586 --count;
587 if (SvOK (items [count])) 669 if (SvOK (items [count]))
588 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 670 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
589 } 671
590 while (count); 672 if (extra)
673 list [i++] = extra;
591 674
592 *rcount = i; 675 *rcount = i;
593 676
594 return i ? list : 0; 677 return i ? list : 0;
595} 678}
596 679
597#define EVENT_LIST(items,count) \ 680#define EVENT_LIST(skip) \
598 cl_uint event_list_count = (count); \ 681 cl_uint event_list_count = items - (skip); \
599 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 682 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
600 683
601#define INFO(class) \ 684#define INFO(class) \
602{ \ 685{ \
603 size_t size; \ 686 size_t size; \
604 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 687 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
608 SvCUR_set (sv, size); \ 691 SvCUR_set (sv, size); \
609 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 692 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
610 XPUSHs (sv); \ 693 XPUSHs (sv); \
611} 694}
612 695
696/*****************************************************************************/
697/* mapped_xxx */
698
699static OpenCL__Mapped
700SvMAPPED (SV *sv)
701{
702 // no typechecking atm., keep your fingers crossed
703 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
704}
705
706struct mapped
707{
708 cl_command_queue queue;
709 cl_mem memobj;
710 void *ptr;
711 size_t cb;
712 cl_event event;
713 size_t row_pitch;
714 size_t slice_pitch;
715};
716
717static SV *
718mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
719{
720 SV *data = newSV (0);
721 SvUPGRADE (data, SVt_PVMG);
722
723 OpenCL__Mapped mapped;
724 New (0, mapped, 1, struct mapped);
725
726 clRetainCommandQueue (queue);
727
728 mapped->queue = queue;
729 mapped->memobj = memobj;
730 mapped->ptr = ptr;
731 mapped->cb = cb;
732 mapped->event = ev;
733 mapped->row_pitch = row_pitch;
734 mapped->slice_pitch = slice_pitch;
735
736 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
737
738 if (SvLEN (data))
739 Safefree (data);
740
741 SvPVX (data) = (char *)ptr;
742 SvCUR_set (data, cb);
743 SvLEN_set (data, 0);
744 SvPOK_only (data);
745
746 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
747
748 if (!(flags & CL_MAP_WRITE))
749 SvREADONLY_on (data);
750
751 return obj;
752}
753
754static void
755mapped_detach (SV *sv, OpenCL__Mapped mapped)
756{
757 SV *data = SvRV (sv);
758
759 // the next check checks both before AND after detach, where SvPVX should STILL be 0
760 if (SvPVX (data) != (char *)mapped->ptr)
761 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
762 else
763 {
764 SvREADONLY_off (data);
765 SvCUR_set (data, 0);
766 SvPVX (data) = 0;
767 SvOK_off (data);
768 }
769
770 mapped->ptr = 0;
771}
772
773static void
774mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
775{
776 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
777 cl_event ev;
778
779 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
780
781 clReleaseEvent (mapped->event);
782 mapped->event = ev;
783
784 mapped_detach (self, mapped);
785}
786
787/*****************************************************************************/
788
613MODULE = OpenCL PACKAGE = OpenCL 789MODULE = OpenCL PACKAGE = OpenCL
614 790
615PROTOTYPES: ENABLE 791PROTOTYPES: ENABLE
616 792
617void 793void
626 eq_signal_arg = (void*)arg; 802 eq_signal_arg = (void*)arg;
627 803
628BOOT: 804BOOT:
629{ 805{
630 HV *stash = gv_stashpv ("OpenCL", 1); 806 HV *stash = gv_stashpv ("OpenCL", 1);
807
631 static const ivstr *civ, const_iv[] = { 808 static const ivstr *civ, const_iv[] = {
632 { sizeof (cl_char ), "SIZEOF_CHAR" }, 809 { sizeof (cl_char ), "SIZEOF_CHAR" },
633 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 810 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
634 { sizeof (cl_short ), "SIZEOF_SHORT" }, 811 { sizeof (cl_short ), "SIZEOF_SHORT" },
635 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 812 { sizeof (cl_ushort), "SIZEOF_USHORT" },
638 { sizeof (cl_long ), "SIZEOF_LONG" }, 815 { sizeof (cl_long ), "SIZEOF_LONG" },
639 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 816 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
640 { sizeof (cl_half ), "SIZEOF_HALF" }, 817 { sizeof (cl_half ), "SIZEOF_HALF" },
641 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 818 { sizeof (cl_float ), "SIZEOF_FLOAT" },
642 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 819 { sizeof (cl_double), "SIZEOF_DOUBLE" },
820 { PREFER_1_1 , "PREFER_1_1" },
643#include "constiv.h" 821#include "constiv.h"
644 }; 822 };
645 823
646 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 824 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
647 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 825 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
826
827 static const nvstr *cnv, const_nv[] = {
828#include "constnv.h"
829 };
830
831 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
832 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
833
834 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
835
836 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
837 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
838 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
839 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
840 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
841 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
842 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
843 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
844 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
845 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
846 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
847 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
848 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
849 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
850 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
851 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
852 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
853 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
854 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
855 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
856 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
857 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
858 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
648 859
649 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 860 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
650} 861}
651 862
652cl_int 863cl_int
655 RETVAL = res; 866 RETVAL = res;
656 OUTPUT: 867 OUTPUT:
657 RETVAL 868 RETVAL
658 869
659const char * 870const char *
660err2str (cl_int err) 871err2str (cl_int err = res)
661 872
662const char * 873const char *
663enum2str (cl_uint value) 874enum2str (cl_uint value)
664 875
665void 876void
673 list = tmpbuf (sizeof (*list) * count); 884 list = tmpbuf (sizeof (*list) * count);
674 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 885 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
675 886
676 EXTEND (SP, count); 887 EXTEND (SP, count);
677 for (i = 0; i < count; ++i) 888 for (i = 0; i < count; ++i)
678 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 889 PUSH_CLOBJ (stash_platform, list [i]);
679 890
680void 891void
681context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 892context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
682 PPCODE: 893 PPCODE:
683 CONTEXT_NOTIFY_CALLBACK; 894 CONTEXT_NOTIFY_CALLBACK;
684 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 895 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
685 XPUSH_CLOBJ_CONTEXT; 896 XPUSH_CLOBJ_CONTEXT;
686 897
687void 898void
688context (FUTURE properties, FUTURE devices, FUTURE notify) 899context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
689 PPCODE: 900 PPCODE:
690 /* der Gipfel der Kunst */ 901 cl_uint device_count;
902 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
903
904 CONTEXT_NOTIFY_CALLBACK;
905 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
906 XPUSH_CLOBJ_CONTEXT;
691 907
692void 908void
693wait_for_events (...) 909wait_for_events (...)
694 CODE: 910 CODE:
695 EVENT_LIST (0, items); 911 EVENT_LIST (0);
696 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 912 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
697 913
698PROTOTYPES: DISABLE 914PROTOTYPES: DISABLE
699 915
700MODULE = OpenCL PACKAGE = OpenCL::Platform 916MODULE = OpenCL PACKAGE = OpenCL::Platform
743 list = tmpbuf (sizeof (*list) * count); 959 list = tmpbuf (sizeof (*list) * count);
744 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 960 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
745 961
746 EXTEND (SP, count); 962 EXTEND (SP, count);
747 for (i = 0; i < count; ++i) 963 for (i = 0; i < count; ++i)
748 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 964 PUSH_CLOBJ (stash_device, list [i]);
749 965
750void 966void
751context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 967context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
752 PPCODE: 968 PPCODE:
969 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
970 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
971
972 cl_uint device_count;
973 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
974
975 CONTEXT_NOTIFY_CALLBACK;
976 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
977 XPUSH_CLOBJ_CONTEXT;
978
979void
980context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
981 PPCODE:
982 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
983 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
984
985 CONTEXT_NOTIFY_CALLBACK;
986 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
987 XPUSH_CLOBJ_CONTEXT;
988
989MODULE = OpenCL PACKAGE = OpenCL::Device
990
991void
992info (OpenCL::Device self, cl_device_info name)
993 PPCODE:
994 INFO (Device)
995
996#if CL_VERSION_1_2
997
998void
999sub_devices (OpenCL::Device self, SV *properties)
1000 PPCODE:
753 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1001 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
754 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1002 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
755 1003
756 AV *av = (AV *)SvRV (devices); 1004 properties = SvRV (properties);
757 cl_uint num_devices = av_len (av) + 1; 1005
758 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1006 cl_uint count = av_len ((AV *)properties) + 1;
1007 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
759 1008
760 int i; 1009 int i;
761 for (i = num_devices; i--; ) 1010 for (i = 0; i < count; ++i)
762 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1011 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
763 1012
764 CONTEXT_NOTIFY_CALLBACK; 1013 props [count] = 0;
765 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
766 XPUSH_CLOBJ_CONTEXT;
767 1014
768void 1015 cl_uint num_devices;
769context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 1016 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
770 PPCODE: 1017 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
771 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1018 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
772 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
773 1019
774 CONTEXT_NOTIFY_CALLBACK; 1020 EXTEND (SP, num_devices);
775 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1021 for (i = 0; i < count; ++i)
776 XPUSH_CLOBJ_CONTEXT; 1022 PUSH_CLOBJ (stash_subdevice, list [i]);
777 1023
778MODULE = OpenCL PACKAGE = OpenCL::Device 1024#endif
779
780void
781info (OpenCL::Device self, cl_device_info name)
782 PPCODE:
783 INFO (Device)
784 1025
785#BEGIN:device 1026#BEGIN:device
786 1027
787void 1028void
788type (OpenCL::Device self) 1029type (OpenCL::Device self)
789 PPCODE: 1030 PPCODE:
790 cl_device_type value [1]; 1031 cl_device_type value [1];
791 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1032 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
792 EXTEND (SP, 1); 1033 EXTEND (SP, 1);
793 const int i = 0; 1034 const int i = 0;
794 PUSHs (sv_2mortal (newSViv (value [i]))); 1035 PUSHs (sv_2mortal (newSVuv (value [i])));
795 1036
796void 1037void
797vendor_id (OpenCL::Device self) 1038vendor_id (OpenCL::Device self)
798 ALIAS: 1039 ALIAS:
799 vendor_id = CL_DEVICE_VENDOR_ID 1040 vendor_id = CL_DEVICE_VENDOR_ID
944 PPCODE: 1185 PPCODE:
945 cl_command_queue_properties value [1]; 1186 cl_command_queue_properties value [1];
946 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
947 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
948 const int i = 0; 1189 const int i = 0;
949 PUSHs (sv_2mortal (newSViv (value [i]))); 1190 PUSHs (sv_2mortal (newSVuv (value [i])));
950 1191
951void 1192void
952platform (OpenCL::Device self) 1193platform (OpenCL::Device self)
953 PPCODE: 1194 PPCODE:
954 cl_platform_id value [1]; 1195 cl_platform_id value [1];
955 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1196 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
956 EXTEND (SP, 1); 1197 EXTEND (SP, 1);
957 const int i = 0; 1198 const int i = 0;
958 { 1199 {
959 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1200 PUSH_CLOBJ (stash_platform, value [i]);
960 } 1201 }
961 1202
962void 1203void
963name (OpenCL::Device self) 1204name (OpenCL::Device self)
964 ALIAS: 1205 ALIAS:
983 cl_device_id value [1]; 1224 cl_device_id value [1];
984 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1225 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
985 EXTEND (SP, 1); 1226 EXTEND (SP, 1);
986 const int i = 0; 1227 const int i = 0;
987 { 1228 {
988 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1229 PUSH_CLOBJ (stash_device, value [i]);
989 } 1230 }
990 1231
991void 1232void
992partition_types_ext (OpenCL::Device self) 1233partition_types_ext (OpenCL::Device self)
993 ALIAS: 1234 ALIAS:
1004 for (i = 0; i < n; ++i) 1245 for (i = 0; i < n; ++i)
1005 PUSHs (sv_2mortal (newSVuv (value [i]))); 1246 PUSHs (sv_2mortal (newSVuv (value [i])));
1006 1247
1007#END:device 1248#END:device
1008 1249
1250MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1251
1252#if CL_VERSION_1_2
1253
1254void
1255DESTROY (OpenCL::SubDevice self)
1256 CODE:
1257 clReleaseDevice (self);
1258
1259#endif
1260
1009MODULE = OpenCL PACKAGE = OpenCL::Context 1261MODULE = OpenCL PACKAGE = OpenCL::Context
1010 1262
1011void 1263void
1012DESTROY (OpenCL::Context context) 1264DESTROY (OpenCL::Context self)
1013 CODE: 1265 CODE:
1014 clReleaseContext (context); 1266 clReleaseContext (self);
1015 1267
1016void 1268void
1017info (OpenCL::Context self, cl_context_info name) 1269info (OpenCL::Context self, cl_context_info name)
1018 PPCODE: 1270 PPCODE:
1019 INFO (Context) 1271 INFO (Context)
1020 1272
1021void 1273void
1022queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1274queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
1023 PPCODE: 1275 PPCODE:
1024 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1276 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
1025 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1277 XPUSH_CLOBJ (stash_queue, queue);
1026 1278
1027void 1279void
1028user_event (OpenCL::Context self) 1280user_event (OpenCL::Context self)
1029 PPCODE: 1281 PPCODE:
1030 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1282 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
1031 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1283 XPUSH_CLOBJ (stash_userevent, ev);
1032 1284
1033void 1285void
1034buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1286buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
1035 PPCODE: 1287 PPCODE:
1036 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1288 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
1037 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1289 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
1038 1290
1039 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1291 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
1040 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1292 XPUSH_CLOBJ (stash_bufferobj, mem);
1041 1293
1042void 1294void
1043buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1295buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
1044 PPCODE: 1296 PPCODE:
1045 STRLEN len; 1297 STRLEN len;
1046 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1298 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1047 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1299 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
1048 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1300 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1301 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
1050 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1302 XPUSH_CLOBJ (stash_bufferobj, mem);
1051 1303
1052#if CL_VERSION_1_2 1304#if CL_VERSION_1_2
1053 1305
1054void 1306void
1055image (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) 1307image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
1056 PPCODE: 1308 PPCODE:
1057 STRLEN len; 1309 STRLEN len;
1058 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1310 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1059 const cl_image_format format = { channel_order, channel_type }; 1311 const cl_image_format format = { channel_order, channel_type };
1060 const cl_image_desc desc = { 1312 const cl_image_desc desc = {
1061 type, 1313 type,
1062 width, height, depth, 1314 width, height, depth,
1063 array_size, row_pitch, slice_pitch, 1315 array_size, row_pitch, slice_pitch,
1064 num_mip_level, num_samples, 1316 num_mip_level, num_samples,
1065 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1317 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1066 }; 1318 };
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1068 char *klass = "OpenCL::Image"; 1320 HV *stash = stash_image;
1069 switch (type) 1321 switch (type)
1070 { 1322 {
1071 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1323 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1072 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1324 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1073 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1325 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1074 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1326 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1075 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1327 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1076 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1328 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1077 } 1329 }
1078 XPUSH_CLOBJ (klass, mem); 1330 XPUSH_CLOBJ (stash, mem);
1079 1331
1080#endif 1332#endif
1081 1333
1082void 1334void
1083image2d (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) 1335image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
1089 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1090#else 1342#else
1091 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1343 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1092 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1344 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1093#endif 1345#endif
1094 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1346 XPUSH_CLOBJ (stash_image2d, mem);
1095 1347
1096void 1348void
1097image3d (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) 1349image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1098 PPCODE: 1350 PPCODE:
1099 STRLEN len; 1351 STRLEN len;
1103 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1355 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1104#else 1356#else
1105 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1357 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1106 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1358 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1107#endif 1359#endif
1108 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1360 XPUSH_CLOBJ (stash_image3d, mem);
1109 1361
1110#if cl_apple_gl_sharing || cl_khr_gl_sharing 1362#if cl_apple_gl_sharing || cl_khr_gl_sharing
1111 1363
1112void 1364void
1113gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1365gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1114 PPCODE: 1366 PPCODE:
1115 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1116 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1368 XPUSH_CLOBJ (stash_bufferobj, mem);
1117 1369
1118void 1370void
1119gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1371gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1120 PPCODE: 1372 PPCODE:
1121 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1373 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1122 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1374 XPUSH_CLOBJ (stash_image2d, mem);
1123 1375
1124#if CL_VERSION_1_2 1376#if CL_VERSION_1_2
1125 1377
1126void 1378void
1127gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1379gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1128 ALIAS: 1380 ALIAS:
1129 PPCODE: 1381 PPCODE:
1130 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1382 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1131 cl_gl_object_type type; 1383 cl_gl_object_type type;
1132 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1384 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1133 char *klass = "OpenCL::Memory"; 1385 HV *stash = stash_memory;
1134 switch (type) 1386 switch (type)
1135 { 1387 {
1136 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1388 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1137 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1389 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1138 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1390 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1139 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1391 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1140 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1392 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1141 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1393 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1142 } 1394 }
1143 XPUSH_CLOBJ (klass, mem); 1395 XPUSH_CLOBJ (stash, mem);
1144 1396
1145#endif 1397#endif
1146 1398
1147void 1399void
1148gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1400gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1150#if PREFER_1_1 1402#if PREFER_1_1
1151 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1403 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1152#else 1404#else
1153 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1405 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1154#endif 1406#endif
1155 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1407 XPUSH_CLOBJ (stash_image2d, mem);
1156 1408
1157void 1409void
1158gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1410gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1159 PPCODE: 1411 PPCODE:
1160#if PREFER_1_1 1412#if PREFER_1_1
1161 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1413 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1162#else 1414#else
1163 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1415 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1164#endif 1416#endif
1165 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1417 XPUSH_CLOBJ (stash_image3d, mem);
1166 1418
1167#endif 1419#endif
1168 1420
1169void 1421void
1170supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1422supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1190 1442
1191void 1443void
1192sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1444sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1193 PPCODE: 1445 PPCODE:
1194 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1446 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1195 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1447 XPUSH_CLOBJ (stash_sampler, sampler);
1196 1448
1197void 1449void
1198program_with_source (OpenCL::Context self, SV *program) 1450program_with_source (OpenCL::Context self, SV *program)
1199 PPCODE: 1451 PPCODE:
1200 STRLEN len; 1452 STRLEN len;
1201 size_t len2; 1453 size_t len2;
1202 const char *ptr = SvPVbyte (program, len); 1454 const char *ptr = SvPVbyte (program, len);
1203 1455
1204 len2 = len; 1456 len2 = len;
1205 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1457 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1206 XPUSH_CLOBJ ("OpenCL::Program", prog); 1458 XPUSH_CLOBJ (stash_program, prog);
1459
1460void
1461program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1462 PPCODE:
1463 cl_uint device_count;
1464 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1465
1466 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1467 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1468
1469 binaries = SvRV (binaries);
1470
1471 if (device_count != av_len ((AV *)binaries) + 1)
1472 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1473
1474 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1475 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1476 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1477
1478 int i;
1479 for (i = 0; i < device_count; ++i)
1480 {
1481 STRLEN len;
1482 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1483 length_list [i] = len;
1484 }
1485
1486 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1487 length_list, binary_list,
1488 GIMME_V == G_ARRAY ? status_list : 0, &res));
1489
1490 EXTEND (SP, 2);
1491 PUSH_CLOBJ (stash_program, prog);
1492
1493 if (GIMME_V == G_ARRAY)
1494 {
1495 AV *av = newAV ();
1496 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1497
1498 for (i = device_count; i--; )
1499 av_store (av, i, newSViv (status_list [i]));
1500 }
1501
1502#if CL_VERSION_1_2
1503
1504void
1505program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1506 PPCODE:
1507 cl_uint device_count;
1508 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1509
1510 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514void
1515link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1516 CODE:
1517 cl_uint device_count = 0;
1518 cl_device_id *device_list = 0;
1519
1520 if (SvOK (devices))
1521 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1522
1523 cl_uint program_count;
1524 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1525
1526 void *user_data;
1527 program_callback pfn_notify = make_program_callback (notify, &user_data);
1528
1529 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1530 program_count, program_list, pfn_notify, user_data, &res));
1531
1532 XPUSH_CLOBJ (stash_program, prog);
1533
1534#endif
1207 1535
1208#BEGIN:context 1536#BEGIN:context
1209 1537
1210void 1538void
1211reference_count (OpenCL::Context self) 1539reference_count (OpenCL::Context self)
1228 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1556 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1229 int i, n = size / sizeof (*value); 1557 int i, n = size / sizeof (*value);
1230 EXTEND (SP, n); 1558 EXTEND (SP, n);
1231 for (i = 0; i < n; ++i) 1559 for (i = 0; i < n; ++i)
1232 { 1560 {
1233 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1561 PUSH_CLOBJ (stash_device, value [i]);
1234 } 1562 }
1235 1563
1236void 1564void
1237properties (OpenCL::Context self) 1565properties (OpenCL::Context self)
1238 PPCODE: 1566 PPCODE:
1253DESTROY (OpenCL::Queue self) 1581DESTROY (OpenCL::Queue self)
1254 CODE: 1582 CODE:
1255 clReleaseCommandQueue (self); 1583 clReleaseCommandQueue (self);
1256 1584
1257void 1585void
1258enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1586read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1587 ALIAS:
1588 enqueue_read_buffer = 0
1259 PPCODE: 1589 PPCODE:
1260 cl_event ev = 0;
1261 EVENT_LIST (6, items - 6); 1590 EVENT_LIST (6);
1262 1591
1263 SvUPGRADE (data, SVt_PV); 1592 SvUPGRADE (data, SVt_PV);
1264 SvGROW (data, len); 1593 SvGROW (data, len);
1265 SvPOK_only (data); 1594 SvPOK_only (data);
1266 SvCUR_set (data, len); 1595 SvCUR_set (data, len);
1596
1597 cl_event ev = 0;
1267 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1598 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1599
1269 if (ev) 1600 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1601 XPUSH_CLOBJ (stash_event, ev);
1271 1602
1272void 1603void
1273enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1604write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1605 ALIAS:
1606 enqueue_write_buffer = 0
1274 PPCODE: 1607 PPCODE:
1275 cl_event ev = 0; 1608 EVENT_LIST (5);
1609
1276 STRLEN len; 1610 STRLEN len;
1277 char *ptr = SvPVbyte (data, len); 1611 char *ptr = SvPVbyte (data, len);
1278 EVENT_LIST (5, items - 5);
1279 1612
1613 cl_event ev = 0;
1280 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1615
1282 if (ev) 1616 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1617 XPUSH_CLOBJ (stash_event, ev);
1284 1618
1285#if CL_VERSION_1_2 1619#if CL_VERSION_1_2
1286 1620
1287void 1621void
1288enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1622fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1623 ALIAS:
1624 enqueue_fill_buffer = 0
1289 PPCODE: 1625 PPCODE:
1290 cl_event ev = 0; 1626 EVENT_LIST (5);
1627
1291 STRLEN len; 1628 STRLEN len;
1292 char *ptr = SvPVbyte (data, len); 1629 char *ptr = SvPVbyte (data, len);
1293 EVENT_LIST (5, items - 5);
1294 1630
1631 cl_event ev = 0;
1295 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1632 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1296 1633
1297 if (ev) 1634 if (ev)
1298 XPUSH_CLOBJ ("OpenCL::Event", ev); 1635 XPUSH_CLOBJ (stash_event, ev);
1299 1636
1300void 1637void
1301enqueue_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, ...) 1638fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1639 ALIAS:
1640 enqueue_fill_image = 0
1302 PPCODE: 1641 PPCODE:
1303 cl_event ev = 0; 1642 EVENT_LIST (12);
1304 STRLEN len; 1643
1305 const size_t origin [3] = { x, y, z }; 1644 const size_t origin [3] = { x, y, z };
1306 const size_t region [3] = { width, height, depth }; 1645 const size_t region [3] = { width, height, depth };
1307 EVENT_LIST (12, items - 12);
1308 1646
1309 const cl_float c_f [4] = { r, g, b, a }; 1647 const cl_float c_f [4] = { r, g, b, a };
1310 const cl_uint c_u [4] = { r, g, b, a }; 1648 const cl_uint c_u [4] = { r, g, b, a };
1311 const cl_int c_s [4] = { r, g, b, a }; 1649 const cl_int c_s [4] = { r, g, b, a };
1312 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1650 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1315 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1653 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1316 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1654 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1317 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1655 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1318 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1656 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1319 1657
1658 cl_event ev = 0;
1320 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1659 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1321 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1660 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1322 1661
1323 if (ev) 1662 if (ev)
1324 XPUSH_CLOBJ ("OpenCL::Event", ev); 1663 XPUSH_CLOBJ (stash_event, ev);
1325 1664
1326#endif 1665#endif
1327 1666
1328void 1667void
1329enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1668copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1669 ALIAS:
1670 enqueue_copy_buffer = 0
1330 PPCODE: 1671 PPCODE:
1672 EVENT_LIST (6);
1673
1331 cl_event ev = 0; 1674 cl_event ev = 0;
1332 EVENT_LIST (6, items - 6);
1333
1334 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1675 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1335 1676
1336 if (ev) 1677 if (ev)
1337 XPUSH_CLOBJ ("OpenCL::Event", ev); 1678 XPUSH_CLOBJ (stash_event, ev);
1338 1679
1339void 1680void
1340enqueue_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, ...) 1681read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1682 ALIAS:
1683 enqueue_read_buffer_rect = 0
1341 PPCODE: 1684 PPCODE:
1342 cl_event ev = 0; 1685 EVENT_LIST (17);
1686
1343 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1687 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1344 const size_t host_origin[3] = { host_x, host_y, host_z }; 1688 const size_t host_origin[3] = { host_x, host_y, host_z };
1345 const size_t region[3] = { width, height, depth }; 1689 const size_t region[3] = { width, height, depth };
1346 EVENT_LIST (17, items - 17);
1347 1690
1348 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
1349 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
1350 1693
1351 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
1361 1704
1362 SvUPGRADE (data, SVt_PV); 1705 SvUPGRADE (data, SVt_PV);
1363 SvGROW (data, len); 1706 SvGROW (data, len);
1364 SvPOK_only (data); 1707 SvPOK_only (data);
1365 SvCUR_set (data, len); 1708 SvCUR_set (data, len);
1709
1710 cl_event ev = 0;
1366 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)); 1711 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1367 1712
1368 if (ev) 1713 if (ev)
1369 XPUSH_CLOBJ ("OpenCL::Event", ev); 1714 XPUSH_CLOBJ (stash_event, ev);
1370 1715
1371void 1716void
1372enqueue_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, ...) 1717write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1718 ALIAS:
1719 enqueue_write_buffer_rect = 0
1373 PPCODE: 1720 PPCODE:
1374 cl_event ev = 0; 1721 EVENT_LIST (17);
1722
1375 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1723 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1376 const size_t host_origin[3] = { host_x, host_y, host_z }; 1724 const size_t host_origin[3] = { host_x, host_y, host_z };
1377 const size_t region[3] = { width, height, depth }; 1725 const size_t region[3] = { width, height, depth };
1378 STRLEN len; 1726 STRLEN len;
1379 char *ptr = SvPVbyte (data, len); 1727 char *ptr = SvPVbyte (data, len);
1380 EVENT_LIST (17, items - 17);
1381 1728
1382 if (!buf_row_pitch) 1729 if (!buf_row_pitch)
1383 buf_row_pitch = region [0]; 1730 buf_row_pitch = region [0];
1384 1731
1385 if (!buf_slice_pitch) 1732 if (!buf_slice_pitch)
1394 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1741 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1395 1742
1396 if (len < min_len) 1743 if (len < min_len)
1397 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1744 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1398 1745
1746 cl_event ev = 0;
1399 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)); 1747 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1400 1748
1401 if (ev) 1749 if (ev)
1402 XPUSH_CLOBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1403 1751
1404void 1752void
1405enqueue_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, ...) 1753copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1754 ALIAS:
1755 enqueue_copy_buffer_rect = 0
1406 PPCODE: 1756 PPCODE:
1407 cl_event ev = 0; 1757 EVENT_LIST (16);
1758
1408 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 };
1409 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1760 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1410 const size_t region[3] = { width, height, depth }; 1761 const size_t region[3] = { width, height, depth };
1411 EVENT_LIST (16, items - 16);
1412 1762
1763 cl_event ev = 0;
1413 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)); 1764 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1414 1765
1415 if (ev) 1766 if (ev)
1416 XPUSH_CLOBJ ("OpenCL::Event", ev); 1767 XPUSH_CLOBJ (stash_event, ev);
1417 1768
1418void 1769void
1419enqueue_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, ...) 1770read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1771 ALIAS:
1772 enqueue_read_image = 0
1420 PPCODE: 1773 PPCODE:
1421 cl_event ev = 0; 1774 EVENT_LIST (12);
1775
1422 const size_t src_origin[3] = { src_x, src_y, src_z }; 1776 const size_t src_origin[3] = { src_x, src_y, src_z };
1423 const size_t region[3] = { width, height, depth }; 1777 const size_t region[3] = { width, height, depth };
1424 EVENT_LIST (12, items - 12);
1425 1778
1426 if (!row_pitch) 1779 if (!row_pitch)
1427 row_pitch = img_row_pitch (src); 1780 row_pitch = img_row_pitch (src);
1428 1781
1429 if (depth > 1 && !slice_pitch) 1782 if (depth > 1 && !slice_pitch)
1433 1786
1434 SvUPGRADE (data, SVt_PV); 1787 SvUPGRADE (data, SVt_PV);
1435 SvGROW (data, len); 1788 SvGROW (data, len);
1436 SvPOK_only (data); 1789 SvPOK_only (data);
1437 SvCUR_set (data, len); 1790 SvCUR_set (data, len);
1791
1792 cl_event ev = 0;
1438 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)); 1793 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1439 1794
1440 if (ev) 1795 if (ev)
1441 XPUSH_CLOBJ ("OpenCL::Event", ev); 1796 XPUSH_CLOBJ (stash_event, ev);
1442 1797
1443void 1798void
1444enqueue_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, ...) 1799write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1800 ALIAS:
1801 enqueue_write_image = 0
1445 PPCODE: 1802 PPCODE:
1446 cl_event ev = 0; 1803 EVENT_LIST (12);
1804
1447 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1805 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1448 const size_t region[3] = { width, height, depth }; 1806 const size_t region[3] = { width, height, depth };
1449 STRLEN len; 1807 STRLEN len;
1450 char *ptr = SvPVbyte (data, len); 1808 char *ptr = SvPVbyte (data, len);
1451 EVENT_LIST (12, items - 12);
1452 1809
1453 if (!row_pitch) 1810 if (!row_pitch)
1454 row_pitch = img_row_pitch (dst); 1811 row_pitch = img_row_pitch (dst);
1455 1812
1456 if (depth > 1 && !slice_pitch) 1813 if (depth > 1 && !slice_pitch)
1459 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1816 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1460 1817
1461 if (len < min_len) 1818 if (len < min_len)
1462 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1819 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1463 1820
1821 cl_event ev = 0;
1464 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)); 1822 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1465 1823
1466 if (ev) 1824 if (ev)
1467 XPUSH_CLOBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1468 1826
1469void 1827void
1470enqueue_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, ...) 1828copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1829 ALIAS:
1830 enqueue_copy_image = 0
1471 PPCODE: 1831 PPCODE:
1472 cl_event ev = 0; 1832 EVENT_LIST (12);
1833
1473 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1474 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1835 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1475 const size_t region[3] = { width, height, depth }; 1836 const size_t region[3] = { width, height, depth };
1476 EVENT_LIST (12, items - 12);
1477 1837
1838 cl_event ev = 0;
1478 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1839 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1479 1840
1480 if (ev) 1841 if (ev)
1481 XPUSH_CLOBJ ("OpenCL::Event", ev); 1842 XPUSH_CLOBJ (stash_event, ev);
1482 1843
1483void 1844void
1484enqueue_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, ...) 1845copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1846 ALIAS:
1847 enqueue_copy_image_to_buffer = 0
1485 PPCODE: 1848 PPCODE:
1849 EVENT_LIST (10);
1850
1851 const size_t src_origin[3] = { src_x, src_y, src_z };
1852 const size_t region [3] = { width, height, depth };
1853
1486 cl_event ev = 0; 1854 cl_event ev = 0;
1855 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1856
1857 if (ev)
1858 XPUSH_CLOBJ (stash_event, ev);
1859
1860void
1861copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1862 ALIAS:
1863 enqueue_copy_buffer_to_image = 0
1864 PPCODE:
1865 EVENT_LIST (10);
1866
1487 const size_t src_origin[3] = { src_x, src_y, src_z }; 1867 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1868 const size_t region [3] = { width, height, depth };
1869
1870 cl_event ev = 0;
1871 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1872
1873 if (ev)
1874 XPUSH_CLOBJ (stash_event, ev);
1875
1876void
1877map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1878 ALIAS:
1879 enqueue_map_buffer = 0
1880 PPCODE:
1881 EVENT_LIST (6);
1882
1883 size_t cb = SvIV (cb_);
1884
1885 if (!SvOK (cb_))
1886 {
1887 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1888 cb -= offset;
1889 }
1890
1891 cl_event ev;
1892 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1893 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1894
1895void
1896map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1897 ALIAS:
1898 enqueue_map_image = 0
1899 PPCODE:
1900 size_t width = SvIV (width_);
1901 if (!SvOK (width_))
1902 {
1903 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1904 width -= x;
1905 }
1906
1907 size_t height = SvIV (width_);
1908 if (!SvOK (height_))
1909 {
1910 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1911 height -= y;
1912 }
1913
1914 size_t depth = SvIV (width_);
1915 if (!SvOK (depth_))
1916 {
1917 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1918 depth -= z;
1919
1920 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1921 if (!depth)
1922 depth = 1;
1923 }
1924
1925 const size_t origin[3] = { x, y, z };
1488 const size_t region[3] = { width, height, depth }; 1926 const size_t region[3] = { width, height, depth };
1927 size_t row_pitch, slice_pitch;
1489 EVENT_LIST (10, items - 10); 1928 EVENT_LIST (10);
1490 1929
1930 cl_event ev;
1931 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1932
1933 size_t cb = slice_pitch ? slice_pitch * region [2]
1934 : row_pitch ? row_pitch * region [1]
1935 : region [0];
1936
1937 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1938
1939void
1940unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1941 PPCODE:
1942 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1943 if (GIMME_V != G_VOID)
1944 {
1945 clRetainEvent (mapped->event);
1946 XPUSH_CLOBJ (stash_event, mapped->event);
1947 }
1948
1949void
1950task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1951 ALIAS:
1952 enqueue_task = 0
1953 PPCODE:
1954 EVENT_LIST (2);
1955
1956 cl_event ev = 0;
1491 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1957 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1492 1958
1493 if (ev) 1959 if (ev)
1494 XPUSH_CLOBJ ("OpenCL::Event", ev); 1960 XPUSH_CLOBJ (stash_event, ev);
1495 1961
1496void 1962void
1497enqueue_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, ...)
1498 PPCODE:
1499 cl_event ev = 0;
1500 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1501 const size_t region[3] = { width, height, depth };
1502 EVENT_LIST (10, items - 10);
1503
1504 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1505
1506 if (ev)
1507 XPUSH_CLOBJ ("OpenCL::Event", ev);
1508
1509void
1510enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1511 PPCODE:
1512 cl_event ev = 0;
1513 EVENT_LIST (2, items - 2);
1514
1515 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1516
1517 if (ev)
1518 XPUSH_CLOBJ ("OpenCL::Event", ev);
1519
1520void
1521enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1963nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1964 ALIAS:
1965 enqueue_nd_range_kernel = 0
1522 PPCODE: 1966 PPCODE:
1523 cl_event ev = 0; 1967 EVENT_LIST (5);
1968
1524 size_t *gwo = 0, *gws, *lws = 0; 1969 size_t *gwo = 0, *gws, *lws = 0;
1525 int gws_len; 1970 int gws_len;
1526 size_t *lists; 1971 size_t *lists;
1527 int i; 1972 int i;
1528 EVENT_LIST (5, items - 5);
1529 1973
1530 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1974 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1531 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1532 1976
1533 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1977 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1534 1978
1535 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1979 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1536 1980
1537 gws = lists + gws_len * 0; 1981 gws = lists + gws_len * 0;
1538 for (i = 0; i < gws_len; ++i) 1982 for (i = 0; i < gws_len; ++i)
1983 {
1539 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1984 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1985 // at least nvidia crashes for 0-sized work group sizes, work around
1986 if (!gws [i])
1987 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1988 }
1540 1989
1541 if (SvOK (global_work_offset)) 1990 if (SvOK (global_work_offset))
1542 { 1991 {
1543 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1992 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1544 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1993 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1552 } 2001 }
1553 2002
1554 if (SvOK (local_work_size)) 2003 if (SvOK (local_work_size))
1555 { 2004 {
1556 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 2005 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1557 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 2006 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1558 2007
1559 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 2008 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1560 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 2009 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1561 2010
1562 lws = lists + gws_len * 2; 2011 lws = lists + gws_len * 2;
1563 for (i = 0; i < gws_len; ++i) 2012 for (i = 0; i < gws_len; ++i)
2013 {
1564 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 2014 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
2015 // at least nvidia crashes for 0-sized work group sizes, work around
2016 if (!lws [i])
2017 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2018 }
1565 } 2019 }
1566 2020
2021 cl_event ev = 0;
1567 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2022 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1568 2023
1569 if (ev) 2024 if (ev)
1570 XPUSH_CLOBJ ("OpenCL::Event", ev); 2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#if CL_VERSION_1_2
2028
2029void
2030migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2031 ALIAS:
2032 enqueue_migrate_mem_objects = 0
2033 PPCODE:
2034 EVENT_LIST (3);
2035
2036 cl_uint obj_count;
2037 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2038
2039 cl_event ev = 0;
2040 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2041
2042 if (ev)
2043 XPUSH_CLOBJ (stash_event, ev);
2044
2045#endif
1571 2046
1572#if cl_apple_gl_sharing || cl_khr_gl_sharing 2047#if cl_apple_gl_sharing || cl_khr_gl_sharing
1573 2048
1574void 2049void
1575enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2050acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1576 ALIAS: 2051 ALIAS:
2052 release_gl_objects = 1
2053 enqueue_acquire_gl_objects = 0
1577 enqueue_release_gl_objects = 1 2054 enqueue_release_gl_objects = 1
1578 PPCODE: 2055 PPCODE:
1579 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2056 EVENT_LIST (2);
1580 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2057
2058 cl_uint obj_count;
2059 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1581 2060
1582 cl_event ev = 0; 2061 cl_event ev = 0;
1583 EVENT_LIST (2, items - 2);
1584 AV *av = (AV *)SvRV (objects);
1585 cl_uint num_objects = av_len (av) + 1;
1586 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1587 int i;
1588
1589 for (i = num_objects; i--; )
1590 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1591 2062
1592 if (ix) 2063 if (ix)
1593 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2064 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1594 else 2065 else
1595 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2066 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1596 2067
1597 if (ev) 2068 if (ev)
1598 XPUSH_CLOBJ ("OpenCL::Event", ev); 2069 XPUSH_CLOBJ (stash_event, ev);
1599 2070
1600#endif 2071#endif
1601 2072
1602void 2073void
1603enqueue_wait_for_events (OpenCL::Queue self, ...) 2074wait_for_events (OpenCL::Queue self, ...)
2075 ALIAS:
2076 enqueue_wait_for_events = 0
1604 CODE: 2077 CODE:
1605 EVENT_LIST (1, items - 1); 2078 EVENT_LIST (1);
1606#if PREFER_1_1 2079#if PREFER_1_1
1607 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1608#else 2081#else
1609 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2082 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1610#endif 2083#endif
1611 2084
1612void 2085void
1613enqueue_marker (OpenCL::Queue self, ...) 2086marker (OpenCL::Queue self, ...)
2087 ALIAS:
2088 enqueue_marker = 0
1614 PPCODE: 2089 PPCODE:
2090 EVENT_LIST (1);
1615 cl_event ev = 0; 2091 cl_event ev = 0;
1616 EVENT_LIST (1, items - 1);
1617#if PREFER_1_1 2092#if PREFER_1_1
1618 if (!event_list_count) 2093 if (!event_list_count)
1619 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2094 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1620 else 2095 else
1621#if CL_VERSION_1_2 2096#if CL_VERSION_1_2
1628#endif 2103#endif
1629#else 2104#else
1630 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2105 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1631#endif 2106#endif
1632 if (ev) 2107 if (ev)
1633 XPUSH_CLOBJ ("OpenCL::Event", ev); 2108 XPUSH_CLOBJ (stash_event, ev);
1634 2109
1635void 2110void
1636enqueue_barrier (OpenCL::Queue self, ...) 2111barrier (OpenCL::Queue self, ...)
2112 ALIAS:
2113 enqueue_barrier = 0
1637 PPCODE: 2114 PPCODE:
2115 EVENT_LIST (1);
1638 cl_event ev = 0; 2116 cl_event ev = 0;
1639 EVENT_LIST (1, items - 1);
1640#if PREFER_1_1 2117#if PREFER_1_1
1641 if (!event_list_count && GIMME_V == G_VOID) 2118 if (!event_list_count && GIMME_V == G_VOID)
1642 NEED_SUCCESS (EnqueueBarrier, (self)); 2119 NEED_SUCCESS (EnqueueBarrier, (self));
1643 else 2120 else
1644#if CL_VERSION_1_2 2121#if CL_VERSION_1_2
1654#endif 2131#endif
1655#else 2132#else
1656 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2133 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1657#endif 2134#endif
1658 if (ev) 2135 if (ev)
1659 XPUSH_CLOBJ ("OpenCL::Event", ev); 2136 XPUSH_CLOBJ (stash_event, ev);
1660 2137
1661void 2138void
1662flush (OpenCL::Queue self) 2139flush (OpenCL::Queue self)
1663 CODE: 2140 CODE:
1664 NEED_SUCCESS (Flush, (self)); 2141 NEED_SUCCESS (Flush, (self));
1682 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2159 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1683 EXTEND (SP, 1); 2160 EXTEND (SP, 1);
1684 const int i = 0; 2161 const int i = 0;
1685 { 2162 {
1686 NEED_SUCCESS (RetainContext, (value [i])); 2163 NEED_SUCCESS (RetainContext, (value [i]));
1687 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2164 PUSH_CLOBJ (stash_context, value [i]);
1688 } 2165 }
1689 2166
1690void 2167void
1691device (OpenCL::Queue self) 2168device (OpenCL::Queue self)
1692 PPCODE: 2169 PPCODE:
1693 cl_device_id value [1]; 2170 cl_device_id value [1];
1694 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2171 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 2172 EXTEND (SP, 1);
1696 const int i = 0; 2173 const int i = 0;
1697 { 2174 {
1698 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2175 PUSH_CLOBJ (stash_device, value [i]);
1699 } 2176 }
1700 2177
1701void 2178void
1702reference_count (OpenCL::Queue self) 2179reference_count (OpenCL::Queue self)
1703 PPCODE: 2180 PPCODE:
1712 PPCODE: 2189 PPCODE:
1713 cl_command_queue_properties value [1]; 2190 cl_command_queue_properties value [1];
1714 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2191 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1715 EXTEND (SP, 1); 2192 EXTEND (SP, 1);
1716 const int i = 0; 2193 const int i = 0;
1717 PUSHs (sv_2mortal (newSViv (value [i]))); 2194 PUSHs (sv_2mortal (newSVuv (value [i])));
1718 2195
1719#END:command_queue 2196#END:command_queue
1720 2197
1721MODULE = OpenCL PACKAGE = OpenCL::Memory 2198MODULE = OpenCL PACKAGE = OpenCL::Memory
1722 2199
1737 PPCODE: 2214 PPCODE:
1738 cl_mem_object_type value [1]; 2215 cl_mem_object_type value [1];
1739 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2216 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1740 EXTEND (SP, 1); 2217 EXTEND (SP, 1);
1741 const int i = 0; 2218 const int i = 0;
1742 PUSHs (sv_2mortal (newSViv (value [i]))); 2219 PUSHs (sv_2mortal (newSVuv (value [i])));
1743 2220
1744void 2221void
1745flags (OpenCL::Memory self) 2222flags (OpenCL::Memory self)
1746 PPCODE: 2223 PPCODE:
1747 cl_mem_flags value [1]; 2224 cl_mem_flags value [1];
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2225 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 2226 EXTEND (SP, 1);
1750 const int i = 0; 2227 const int i = 0;
1751 PUSHs (sv_2mortal (newSViv (value [i]))); 2228 PUSHs (sv_2mortal (newSVuv (value [i])));
1752 2229
1753void 2230void
1754size (OpenCL::Memory self) 2231size (OpenCL::Memory self)
1755 ALIAS: 2232 ALIAS:
1756 size = CL_MEM_SIZE 2233 size = CL_MEM_SIZE
1790 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2267 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1791 EXTEND (SP, 1); 2268 EXTEND (SP, 1);
1792 const int i = 0; 2269 const int i = 0;
1793 { 2270 {
1794 NEED_SUCCESS (RetainContext, (value [i])); 2271 NEED_SUCCESS (RetainContext, (value [i]));
1795 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2272 PUSH_CLOBJ (stash_context, value [i]);
1796 } 2273 }
1797 2274
1798void 2275void
1799associated_memobject (OpenCL::Memory self) 2276associated_memobject (OpenCL::Memory self)
1800 PPCODE: 2277 PPCODE:
1802 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2279 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1803 EXTEND (SP, 1); 2280 EXTEND (SP, 1);
1804 const int i = 0; 2281 const int i = 0;
1805 { 2282 {
1806 NEED_SUCCESS (RetainMemObject, (value [i])); 2283 NEED_SUCCESS (RetainMemObject, (value [i]));
1807 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2284 PUSH_CLOBJ (stash_memory, value [i]);
1808 } 2285 }
1809 2286
1810#END:mem 2287#END:mem
1811 2288
1812#if cl_apple_gl_sharing || cl_khr_gl_sharing 2289#if cl_apple_gl_sharing || cl_khr_gl_sharing
1832 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2309 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1833 2310
1834 cl_buffer_region crdata = { origin, size }; 2311 cl_buffer_region crdata = { origin, size };
1835 2312
1836 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2313 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1837 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2314 XPUSH_CLOBJ (stash_buffer, mem);
1838 2315
1839MODULE = OpenCL PACKAGE = OpenCL::Image 2316MODULE = OpenCL PACKAGE = OpenCL::Image
1840 2317
1841void 2318void
1842image_info (OpenCL::Image self, cl_image_info name) 2319image_info (OpenCL::Image self, cl_image_info name)
1928 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2405 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1929 EXTEND (SP, 1); 2406 EXTEND (SP, 1);
1930 const int i = 0; 2407 const int i = 0;
1931 { 2408 {
1932 NEED_SUCCESS (RetainContext, (value [i])); 2409 NEED_SUCCESS (RetainContext, (value [i]));
1933 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2410 PUSH_CLOBJ (stash_context, value [i]);
1934 } 2411 }
1935 2412
1936void 2413void
1937normalized_coords (OpenCL::Sampler self) 2414normalized_coords (OpenCL::Sampler self)
1938 PPCODE: 2415 PPCODE:
1939 cl_addressing_mode value [1]; 2416 cl_addressing_mode value [1];
1940 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2417 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1941 EXTEND (SP, 1); 2418 EXTEND (SP, 1);
1942 const int i = 0; 2419 const int i = 0;
1943 PUSHs (sv_2mortal (newSViv (value [i]))); 2420 PUSHs (sv_2mortal (newSVuv (value [i])));
1944 2421
1945void 2422void
1946addressing_mode (OpenCL::Sampler self) 2423addressing_mode (OpenCL::Sampler self)
1947 PPCODE: 2424 PPCODE:
1948 cl_filter_mode value [1]; 2425 cl_filter_mode value [1];
1949 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2426 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1950 EXTEND (SP, 1); 2427 EXTEND (SP, 1);
1951 const int i = 0; 2428 const int i = 0;
1952 PUSHs (sv_2mortal (newSViv (value [i]))); 2429 PUSHs (sv_2mortal (newSVuv (value [i])));
1953 2430
1954void 2431void
1955filter_mode (OpenCL::Sampler self) 2432filter_mode (OpenCL::Sampler self)
1956 PPCODE: 2433 PPCODE:
1957 cl_bool value [1]; 2434 cl_bool value [1];
1972void 2449void
1973build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2450build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1974 ALIAS: 2451 ALIAS:
1975 build_async = 1 2452 build_async = 1
1976 CODE: 2453 CODE:
1977 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1978 void *user_data = 0;
1979 cl_uint num_devices = 0; 2454 cl_uint device_count = 0;
1980 cl_device_id *device_list = 0; 2455 cl_device_id *device_list = 0;
1981 2456
1982 if (SvOK (devices)) 2457 if (SvOK (devices))
1983 { 2458 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2459
2460 void *user_data;
2461 program_callback pfn_notify = make_program_callback (notify, &user_data);
2462
2463 if (ix)
2464 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2465 else
2466 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2467
2468#if CL_VERSION_1_2
2469
2470void
2471compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2472 CODE:
2473 cl_uint device_count = 0;
2474 cl_device_id *device_list = 0;
2475
2476 if (SvOK (devices))
2477 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2478
2479 cl_uint header_count = 0;
2480 cl_program *header_list = 0;
2481 const char **header_name = 0;
2482
2483 if (SvOK (headers))
2484 {
1984 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2485 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1985 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2486 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1986 2487
1987 AV *av = (AV *)SvRV (devices); 2488 HV *hv = (HV *)SvRV (devices);
1988 num_devices = av_len (av) + 1;
1989 2489
1990 if (num_devices) 2490 header_count = hv_iterinit (hv);
2491 header_list = tmpbuf (sizeof (*header_list) * header_count);
2492 header_name = tmpbuf (sizeof (*header_name) * header_count);
2493
2494 HE *he;
2495 int i = 0;
2496 while (he = hv_iternext (hv))
1991 { 2497 {
1992 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2498 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2499 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1993 int count; 2500 ++i;
1994 for (count = 0; count < num_devices; ++count)
1995 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1996 } 2501 }
1997 } 2502 }
1998 2503
1999 if (SvOK (notify)) 2504 void *user_data;
2000 { 2505 program_callback pfn_notify = make_program_callback (notify, &user_data);
2001 NEED_SUCCESS (RetainProgram, (self));
2002 pfn_notify = eq_program_notify;
2003 user_data = SvREFCNT_inc (s_get_cv (notify));
2004 }
2005 2506
2006 if (ix)
2007 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2008 else
2009 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2507 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2508 header_count, header_list, header_name, pfn_notify, user_data));
2509
2510#endif
2010 2511
2011void 2512void
2012build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2513build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2013 PPCODE: 2514 PPCODE:
2014 size_t size; 2515 size_t size;
2043 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2544 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2044 EXTEND (SP, 1); 2545 EXTEND (SP, 1);
2045 const int i = 0; 2546 const int i = 0;
2046 PUSHs (sv_2mortal (newSVpv (value, 0))); 2547 PUSHs (sv_2mortal (newSVpv (value, 0)));
2047 2548
2549void
2550binary_type (OpenCL::Program self, OpenCL::Device device)
2551 PPCODE:
2552 cl_program_binary_type value [1];
2553 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2554 EXTEND (SP, 1);
2555 const int i = 0;
2556 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2557
2048#END:program_build 2558#END:program_build
2049 2559
2050void 2560void
2051kernel (OpenCL::Program program, SV *function) 2561kernel (OpenCL::Program program, SV *function)
2052 PPCODE: 2562 PPCODE:
2053 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2563 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2054 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2564 XPUSH_CLOBJ (stash_kernel, kernel);
2055 2565
2056void 2566void
2057kernels_in_program (OpenCL::Program program) 2567kernels_in_program (OpenCL::Program program)
2058 PPCODE: 2568 PPCODE:
2059 cl_uint num_kernels; 2569 cl_uint num_kernels;
2062 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2572 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2063 2573
2064 int i; 2574 int i;
2065 EXTEND (SP, num_kernels); 2575 EXTEND (SP, num_kernels);
2066 for (i = 0; i < num_kernels; ++i) 2576 for (i = 0; i < num_kernels; ++i)
2067 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2577 PUSH_CLOBJ (stash_kernel, kernels [i]);
2068 2578
2069void 2579void
2070info (OpenCL::Program self, cl_program_info name) 2580info (OpenCL::Program self, cl_program_info name)
2071 PPCODE: 2581 PPCODE:
2072 INFO (Program) 2582 INFO (Program)
2120 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2630 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2121 EXTEND (SP, 1); 2631 EXTEND (SP, 1);
2122 const int i = 0; 2632 const int i = 0;
2123 { 2633 {
2124 NEED_SUCCESS (RetainContext, (value [i])); 2634 NEED_SUCCESS (RetainContext, (value [i]));
2125 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2635 PUSH_CLOBJ (stash_context, value [i]);
2126 } 2636 }
2127 2637
2128void 2638void
2129devices (OpenCL::Program self) 2639devices (OpenCL::Program self)
2130 PPCODE: 2640 PPCODE:
2134 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2644 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2135 int i, n = size / sizeof (*value); 2645 int i, n = size / sizeof (*value);
2136 EXTEND (SP, n); 2646 EXTEND (SP, n);
2137 for (i = 0; i < n; ++i) 2647 for (i = 0; i < n; ++i)
2138 { 2648 {
2139 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2649 PUSH_CLOBJ (stash_device, value [i]);
2140 } 2650 }
2141 2651
2142void 2652void
2143source (OpenCL::Program self) 2653source (OpenCL::Program self)
2144 PPCODE: 2654 PPCODE:
2170DESTROY (OpenCL::Kernel self) 2680DESTROY (OpenCL::Kernel self)
2171 CODE: 2681 CODE:
2172 clReleaseKernel (self); 2682 clReleaseKernel (self);
2173 2683
2174void 2684void
2685setf (OpenCL::Kernel self, const char *format, ...)
2686 CODE:
2687 int i;
2688 for (i = 2; ; ++i)
2689 {
2690 while (*format == ' ')
2691 ++format;
2692
2693 char type = *format++;
2694
2695 if (!type)
2696 break;
2697
2698 if (i >= items)
2699 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2700
2701 SV *sv = ST (i);
2702
2703 union
2704 {
2705 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2706 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2707 cl_half ch; cl_float cf; cl_double cd;
2708 cl_mem cm;
2709 cl_sampler ca;
2710 size_t cz;
2711 cl_event ce;
2712 } arg;
2713 size_t size;
2714 int nullarg = 0;
2715
2716 switch (type)
2717 {
2718 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2719 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2720 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2721 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2722 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2723 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2724 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2725 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2726
2727 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2728 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2729 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2730
2731 case 'z': nullarg = 1; size = SvIV (sv); break;
2732
2733 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2734 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2735 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2736
2737 default:
2738 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2739 }
2740
2741 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2742 if (res)
2743 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2744 }
2745
2746 if (i != items)
2747 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2748
2749void
2175set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2750set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2176 CODE: 2751 CODE:
2177 clSetKernelArg (self, idx, sizeof (value), &value); 2752 clSetKernelArg (self, idx, sizeof (value), &value);
2178 2753
2179void 2754void
2227 clSetKernelArg (self, idx, sizeof (value), &value); 2802 clSetKernelArg (self, idx, sizeof (value), &value);
2228 2803
2229void 2804void
2230set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2805set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2231 CODE: 2806 CODE:
2232 clSetKernelArg (self, idx, sizeof (value), &value); 2807 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2233 2808
2234void 2809void
2235set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2810set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2236 CODE: 2811 CODE:
2237 clSetKernelArg (self, idx, sizeof (value), &value); 2812 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2238 2813
2239void 2814void
2240set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2815set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2241 CODE: 2816 CODE:
2242 clSetKernelArg (self, idx, sizeof (value), &value); 2817 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2243
2244void
2245set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
2246 CODE:
2247 clSetKernelArg (self, idx, sizeof (value), &value);
2248 2818
2249void 2819void
2250set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2820set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2251 CODE: 2821 CODE:
2252 clSetKernelArg (self, idx, sizeof (value), &value); 2822 clSetKernelArg (self, idx, sizeof (value), &value);
2298 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2868 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2299 EXTEND (SP, 1); 2869 EXTEND (SP, 1);
2300 const int i = 0; 2870 const int i = 0;
2301 { 2871 {
2302 NEED_SUCCESS (RetainContext, (value [i])); 2872 NEED_SUCCESS (RetainContext, (value [i]));
2303 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2873 PUSH_CLOBJ (stash_context, value [i]);
2304 } 2874 }
2305 2875
2306void 2876void
2307program (OpenCL::Kernel self) 2877program (OpenCL::Kernel self)
2308 PPCODE: 2878 PPCODE:
2310 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2880 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2311 EXTEND (SP, 1); 2881 EXTEND (SP, 1);
2312 const int i = 0; 2882 const int i = 0;
2313 { 2883 {
2314 NEED_SUCCESS (RetainProgram, (value [i])); 2884 NEED_SUCCESS (RetainProgram, (value [i]));
2315 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2885 PUSH_CLOBJ (stash_program, value [i]);
2316 } 2886 }
2317 2887
2318#END:kernel 2888#END:kernel
2319 2889
2320void 2890void
2367 const int i = 0; 2937 const int i = 0;
2368 PUSHs (sv_2mortal (newSVuv (value [i]))); 2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2369 2939
2370#END:kernel_work_group 2940#END:kernel_work_group
2371 2941
2942#if CL_VERSION_1_2
2943
2944void
2945arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2946 PPCODE:
2947 size_t size;
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2949 SV *sv = sv_2mortal (newSV (size));
2950 SvUPGRADE (sv, SVt_PV);
2951 SvPOK_only (sv);
2952 SvCUR_set (sv, size);
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2954 XPUSHs (sv);
2955
2956#BEGIN:kernel_arg
2957
2958void
2959arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2960 PPCODE:
2961 cl_kernel_arg_address_qualifier value [1];
2962 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2963 EXTEND (SP, 1);
2964 const int i = 0;
2965 PUSHs (sv_2mortal (newSVuv (value [i])));
2966
2967void
2968arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2969 PPCODE:
2970 cl_kernel_arg_access_qualifier value [1];
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVuv (value [i])));
2975
2976void
2977arg_type_name (OpenCL::Kernel self, cl_uint idx)
2978 ALIAS:
2979 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2980 arg_name = CL_KERNEL_ARG_NAME
2981 PPCODE:
2982 size_t size;
2983 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2984 char *value = tmpbuf (size);
2985 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2986 EXTEND (SP, 1);
2987 const int i = 0;
2988 PUSHs (sv_2mortal (newSVpv (value, 0)));
2989
2990void
2991arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2992 PPCODE:
2993 cl_kernel_arg_type_qualifier value [1];
2994 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2995 EXTEND (SP, 1);
2996 const int i = 0;
2997 PUSHs (sv_2mortal (newSVuv (value [i])));
2998
2999#END:kernel_arg
3000
3001#endif
3002
2372MODULE = OpenCL PACKAGE = OpenCL::Event 3003MODULE = OpenCL PACKAGE = OpenCL::Event
2373 3004
2374void 3005void
2375DESTROY (OpenCL::Event self) 3006DESTROY (OpenCL::Event self)
2376 CODE: 3007 CODE:
2400 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3031 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2401 EXTEND (SP, 1); 3032 EXTEND (SP, 1);
2402 const int i = 0; 3033 const int i = 0;
2403 { 3034 {
2404 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3035 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2405 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 3036 PUSH_CLOBJ (stash_queue, value [i]);
2406 } 3037 }
2407 3038
2408void 3039void
2409command_type (OpenCL::Event self) 3040command_type (OpenCL::Event self)
2410 PPCODE: 3041 PPCODE:
2433 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3064 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2434 EXTEND (SP, 1); 3065 EXTEND (SP, 1);
2435 const int i = 0; 3066 const int i = 0;
2436 { 3067 {
2437 NEED_SUCCESS (RetainContext, (value [i])); 3068 NEED_SUCCESS (RetainContext, (value [i]));
2438 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3069 PUSH_CLOBJ (stash_context, value [i]);
2439 } 3070 }
2440 3071
2441#END:event 3072#END:event
2442 3073
2443void 3074void
2468void 3099void
2469set_status (OpenCL::UserEvent self, cl_int execution_status) 3100set_status (OpenCL::UserEvent self, cl_int execution_status)
2470 CODE: 3101 CODE:
2471 clSetUserEventStatus (self, execution_status); 3102 clSetUserEventStatus (self, execution_status);
2472 3103
3104MODULE = OpenCL PACKAGE = OpenCL::Mapped
3105
3106void
3107DESTROY (SV *self)
3108 CODE:
3109 OpenCL__Mapped mapped = SvMAPPED (self);
3110
3111 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3112 mapped_detach (self, mapped);
3113
3114 clReleaseCommandQueue (mapped->queue);
3115 clReleaseEvent (mapped->event);
3116 Safefree (mapped);
3117
3118void
3119unmap (OpenCL::Mapped self, ...)
3120 CODE:
3121 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3122
3123bool
3124mapped (OpenCL::Mapped self)
3125 CODE:
3126 RETVAL = !!self->ptr;
3127 OUTPUT:
3128 RETVAL
3129
3130void
3131wait (OpenCL::Mapped self)
3132 PPCODE:
3133 if (self->event)
3134 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3135
3136void
3137event (OpenCL::Mapped self)
3138 PPCODE:
3139 if (!self->event)
3140 XSRETURN_UNDEF;
3141
3142 clRetainEvent (self->event);
3143 XPUSH_CLOBJ (stash_event, self->event);
3144
3145size_t
3146size (OpenCL::Mapped self)
3147 CODE:
3148 RETVAL = self->cb;
3149 OUTPUT:
3150 RETVAL
3151
3152IV
3153ptr (OpenCL::Mapped self)
3154 CODE:
3155 RETVAL = PTR2IV (self->ptr);
3156 OUTPUT:
3157 RETVAL
3158
3159void
3160set (OpenCL::Mapped self, size_t offset, SV *data)
3161 CODE:
3162 STRLEN len;
3163 const char *ptr = SvPVbyte (data, len);
3164
3165 if (offset + len > self->cb)
3166 croak ("OpenCL::Mapped::set out of bound condition detected");
3167
3168 memcpy (offset + (char *)self->ptr, ptr, len);
3169
3170MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3171
3172MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3173
3174IV
3175row_pitch (OpenCL::Mapped self)
3176 ALIAS:
3177 slice_pitch = 1
3178 CODE:
3179 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3180 OUTPUT:
3181 RETVAL
3182
3183
3184

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines