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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.56 by root, Wed Apr 25 21:48:45 2012 UTC vs.
Revision 1.78 by root, Sat May 5 15:11:19 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)
275 return (void *)SvIV (SvRV (sv)); 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);
276 372
277 croak ("%s: %s is not of type %s", func, svname, pkg); 373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
378 {
379 list = tmpbuf (sizeof (*list) * count);
380 int i;
381 for (i = 0; i < count; ++i)
382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
383 }
384
385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
278} 391}
279 392
280/*****************************************************************************/ 393/*****************************************************************************/
281/* callback stuff */ 394/* callback stuff */
282 395
304static void (*eq_signal_func)(void *signal_arg, int value); 417static void (*eq_signal_func)(void *signal_arg, int value);
305static void *eq_signal_arg; 418static void *eq_signal_arg;
306static xmutex_t eq_lock = X_MUTEX_INIT; 419static xmutex_t eq_lock = X_MUTEX_INIT;
307static eq_item *eq_head, *eq_tail; 420static eq_item *eq_head, *eq_tail;
308 421
309static void 422static void ecb_noinline
310eq_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)
311{ 424{
312 eq_item *item = malloc (sizeof (eq_item)); 425 eq_item *item = malloc (sizeof (eq_item));
313 426
314 item->next = 0; 427 item->next = 0;
381} 494}
382 495
383/*****************************************************************************/ 496/*****************************************************************************/
384/* context notify */ 497/* context notify */
385 498
386static void 499static void ecb_noinline
387eq_context_push (void *data1, void *data2, void *data3) 500eq_context_push (void *data1, void *data2, void *data3)
388{ 501{
389 dSP; 502 dSP;
390 PUSHs (sv_2mortal (newSVpv (data1, 0))); 503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
391 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
413 { \ 526 { \
414 pfn_notify = eq_context_notify; \ 527 pfn_notify = eq_context_notify; \
415 user_data = s_get_cv (notify); \ 528 user_data = s_get_cv (notify); \
416 } 529 }
417 530
418static SV * 531static SV * ecb_noinline
419new_clobj_context (cl_context ctx, void *user_data) 532new_clobj_context (cl_context ctx, void *user_data)
420{ 533{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 534 SV *sv = new_clobj (stash_context, (IV)ctx);
422 535
423 if (user_data) 536 if (user_data)
424 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);
425 538
426 return sv; 539 return sv;
433 546
434static void 547static void
435eq_program_push (void *data1, void *data2, void *data3) 548eq_program_push (void *data1, void *data2, void *data3)
436{ 549{
437 dSP; 550 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 551 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 552 PUTBACK;
440} 553}
441 554
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 556
444static void CL_CALLBACK 557static void CL_CALLBACK
445eq_program_notify (cl_program program, void *user_data) 558eq_program_notify (cl_program program, void *user_data)
446{ 559{
560 clRetainProgram (program);
561
447 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 }
448} 580}
449 581
450struct build_args 582struct build_args
451{ 583{
452 cl_program program; 584 cl_program program;
466 else 598 else
467 clReleaseProgram (arg->program); 599 clReleaseProgram (arg->program);
468 600
469 free (arg->options); 601 free (arg->options);
470 free (arg); 602 free (arg);
603
604 return 0;
471} 605}
472 606
473static void 607static void
474build_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)
475{ 609{
484 xthread_t id; 618 xthread_t id;
485 thread_create (&id, build_program_thread, arg); 619 thread_create (&id, build_program_thread, arg);
486} 620}
487 621
488/*****************************************************************************/ 622/*****************************************************************************/
623/* mem object destructor notify */
624
625static void ecb_noinline
626eq_destructor_push (void *data1, void *data2, void *data3)
627{
628}
629
630static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
631
632static void CL_CALLBACK
633eq_destructor_notify (cl_mem memobj, void *user_data)
634{
635 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
636}
637
638/*****************************************************************************/
489/* event objects */ 639/* event objects */
490 640
491static void 641static void
492eq_event_push (void *data1, void *data2, void *data3) 642eq_event_push (void *data1, void *data2, void *data3)
493{ 643{
494 dSP; 644 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 645 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 646 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 647 PUTBACK;
498} 648}
499 649
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 650static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 655 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 656 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 657}
508 658
509/*****************************************************************************/ 659/*****************************************************************************/
660/* utilities for XS code */
510 661
511static size_t 662static size_t
512img_row_pitch (cl_mem img) 663img_row_pitch (cl_mem img)
513{ 664{
514 size_t res; 665 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 666 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 667 return res;
517} 668}
518 669
519static cl_event * 670static cl_event * ecb_noinline
520event_list (SV **items, cl_uint *rcount) 671event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
521{ 672{
522 cl_uint count = *rcount; 673 cl_uint count = *rcount;
523 674
524 if (!count) 675 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
676 *rcount = count = 0;
677
678 if (!count && !extra)
525 return 0; 679 return 0;
526 680
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 681 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 682 int i = 0;
529 683
530 do 684 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 685 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 686 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
535 } 687
536 while (count); 688 if (extra)
689 list [i++] = extra;
537 690
538 *rcount = i; 691 *rcount = i;
539 692
540 return i ? list : 0; 693 return i ? list : 0;
541} 694}
542 695
543#define EVENT_LIST(items,count) \ 696#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 697 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 698 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
546 699
547#define INFO(class) \ 700#define INFO(class) \
548{ \ 701{ \
549 size_t size; \ 702 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 703 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
554 SvCUR_set (sv, size); \ 707 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 708 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 709 XPUSHs (sv); \
557} 710}
558 711
712/*****************************************************************************/
713/* mapped_xxx */
714
715static OpenCL__Mapped
716SvMAPPED (SV *sv)
717{
718 // no typechecking atm., keep your fingers crossed
719 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
720}
721
722struct mapped
723{
724 cl_command_queue queue;
725 cl_mem memobj;
726 void *ptr;
727 size_t cb;
728 cl_event event;
729 size_t row_pitch;
730 size_t slice_pitch;
731};
732
733static SV *
734mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
735{
736 SV *data = newSV (0);
737 SvUPGRADE (data, SVt_PVMG);
738
739 OpenCL__Mapped mapped;
740 New (0, mapped, 1, struct mapped);
741
742 clRetainCommandQueue (queue);
743
744 mapped->queue = queue;
745 mapped->memobj = memobj;
746 mapped->ptr = ptr;
747 mapped->cb = cb;
748 mapped->event = ev;
749 mapped->row_pitch = row_pitch;
750 mapped->slice_pitch = slice_pitch;
751
752 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
753
754 if (SvLEN (data))
755 Safefree (data);
756
757 SvPVX (data) = (char *)ptr;
758 SvCUR_set (data, cb);
759 SvLEN_set (data, 0);
760 SvPOK_only (data);
761
762 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
763
764 if (!(flags & CL_MAP_WRITE))
765 SvREADONLY_on (data);
766
767 return obj;
768}
769
770static void
771mapped_detach (SV *sv, OpenCL__Mapped mapped)
772{
773 SV *data = SvRV (sv);
774
775 // the next check checks both before AND after detach, where SvPVX should STILL be 0
776 if (SvPVX (data) != (char *)mapped->ptr)
777 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
778 else
779 {
780 SvREADONLY_off (data);
781 SvCUR_set (data, 0);
782 SvPVX (data) = 0;
783 SvOK_off (data);
784 }
785
786 mapped->ptr = 0;
787}
788
789static void
790mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
791{
792 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
793 cl_event ev;
794
795 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
796
797 clReleaseEvent (mapped->event);
798 mapped->event = ev;
799
800 mapped_detach (self, mapped);
801}
802
803/*****************************************************************************/
804
559MODULE = OpenCL PACKAGE = OpenCL 805MODULE = OpenCL PACKAGE = OpenCL
560 806
561PROTOTYPES: ENABLE 807PROTOTYPES: ENABLE
562 808
563void 809void
572 eq_signal_arg = (void*)arg; 818 eq_signal_arg = (void*)arg;
573 819
574BOOT: 820BOOT:
575{ 821{
576 HV *stash = gv_stashpv ("OpenCL", 1); 822 HV *stash = gv_stashpv ("OpenCL", 1);
823
577 static const ivstr *civ, const_iv[] = { 824 static const ivstr *civ, const_iv[] = {
578 { sizeof (cl_char ), "SIZEOF_CHAR" }, 825 { sizeof (cl_char ), "SIZEOF_CHAR" },
579 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 826 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
580 { sizeof (cl_short ), "SIZEOF_SHORT" }, 827 { sizeof (cl_short ), "SIZEOF_SHORT" },
581 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 828 { sizeof (cl_ushort), "SIZEOF_USHORT" },
584 { sizeof (cl_long ), "SIZEOF_LONG" }, 831 { sizeof (cl_long ), "SIZEOF_LONG" },
585 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 832 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
586 { sizeof (cl_half ), "SIZEOF_HALF" }, 833 { sizeof (cl_half ), "SIZEOF_HALF" },
587 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 834 { sizeof (cl_float ), "SIZEOF_FLOAT" },
588 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 835 { sizeof (cl_double), "SIZEOF_DOUBLE" },
836 { PREFER_1_1 , "PREFER_1_1" },
589#include "constiv.h" 837#include "constiv.h"
590 }; 838 };
591 839
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 840 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
593 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 841 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
842
843 static const nvstr *cnv, const_nv[] = {
844#include "constnv.h"
845 };
846
847 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
848 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
849
850 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
851
852 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
853 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
854 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
855 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
856 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
857 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
858 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
859 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
860 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
861 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
862 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
863 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
864 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
865 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
866 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
867 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
868 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
869 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
870 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
871 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
872 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
873 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
874 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
594 875
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 876 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 877}
597 878
598cl_int 879cl_int
601 RETVAL = res; 882 RETVAL = res;
602 OUTPUT: 883 OUTPUT:
603 RETVAL 884 RETVAL
604 885
605const char * 886const char *
606err2str (cl_int err) 887err2str (cl_int err = res)
607 888
608const char * 889const char *
609enum2str (cl_uint value) 890enum2str (cl_uint value)
610 891
611void 892void
619 list = tmpbuf (sizeof (*list) * count); 900 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 901 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 902
622 EXTEND (SP, count); 903 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 904 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 905 PUSH_CLOBJ (stash_platform, list [i]);
625 906
626void 907void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 908context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 909 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 910 CONTEXT_NOTIFY_CALLBACK;
630 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 911 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
631 XPUSH_CLOBJ_CONTEXT; 912 XPUSH_CLOBJ_CONTEXT;
632 913
633void 914void
634context (FUTURE properties, FUTURE devices, FUTURE notify) 915context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
635 PPCODE: 916 PPCODE:
636 /* der Gipfel der Kunst */ 917 cl_uint device_count;
918 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
919
920 CONTEXT_NOTIFY_CALLBACK;
921 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
922 XPUSH_CLOBJ_CONTEXT;
637 923
638void 924void
639wait_for_events (...) 925wait_for_events (...)
640 CODE: 926 CODE:
641 EVENT_LIST (0, items); 927 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 928 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 929
644PROTOTYPES: DISABLE 930PROTOTYPES: DISABLE
645 931
646MODULE = OpenCL PACKAGE = OpenCL::Platform 932MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 975 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 976 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 977
692 EXTEND (SP, count); 978 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 979 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 980 PUSH_CLOBJ (stash_device, list [i]);
695 981
696void 982void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 983context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 984 PPCODE:
985 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
986 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
987
988 cl_uint device_count;
989 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
990
991 CONTEXT_NOTIFY_CALLBACK;
992 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
993 XPUSH_CLOBJ_CONTEXT;
994
995void
996context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
997 PPCODE:
998 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
999 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1000
1001 CONTEXT_NOTIFY_CALLBACK;
1002 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1003 XPUSH_CLOBJ_CONTEXT;
1004
1005MODULE = OpenCL PACKAGE = OpenCL::Device
1006
1007void
1008info (OpenCL::Device self, cl_device_info name)
1009 PPCODE:
1010 INFO (Device)
1011
1012#if CL_VERSION_1_2
1013
1014void
1015sub_devices (OpenCL::Device self, SV *properties)
1016 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1017 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
700 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1018 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
701 1019
702 AV *av = (AV *)SvRV (devices); 1020 properties = SvRV (properties);
703 cl_uint num_devices = av_len (av) + 1; 1021
704 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1022 cl_uint count = av_len ((AV *)properties) + 1;
1023 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
705 1024
706 int i; 1025 int i;
707 for (i = num_devices; i--; ) 1026 for (i = 0; i < count; ++i)
708 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1027 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
709 1028
710 CONTEXT_NOTIFY_CALLBACK; 1029 props [count] = 0;
711 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
712 XPUSH_CLOBJ_CONTEXT;
713 1030
714void 1031 cl_uint num_devices;
715context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 1032 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
716 PPCODE: 1033 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
717 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1034 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
718 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
719 1035
720 CONTEXT_NOTIFY_CALLBACK; 1036 EXTEND (SP, num_devices);
721 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1037 for (i = 0; i < count; ++i)
722 XPUSH_CLOBJ_CONTEXT; 1038 PUSH_CLOBJ (stash_subdevice, list [i]);
723 1039
724MODULE = OpenCL PACKAGE = OpenCL::Device 1040#endif
725
726void
727info (OpenCL::Device self, cl_device_info name)
728 PPCODE:
729 INFO (Device)
730 1041
731#BEGIN:device 1042#BEGIN:device
732 1043
733void 1044void
734type (OpenCL::Device self) 1045type (OpenCL::Device self)
735 PPCODE: 1046 PPCODE:
736 cl_device_type value [1]; 1047 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1048 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 1049 EXTEND (SP, 1);
739 const int i = 0; 1050 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 1051 PUSHs (sv_2mortal (newSVuv (value [i])));
741 1052
742void 1053void
743vendor_id (OpenCL::Device self) 1054vendor_id (OpenCL::Device self)
744 ALIAS: 1055 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 1056 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1201 PPCODE:
891 cl_command_queue_properties value [1]; 1202 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1203 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1204 EXTEND (SP, 1);
894 const int i = 0; 1205 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1206 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1207
897void 1208void
898platform (OpenCL::Device self) 1209platform (OpenCL::Device self)
899 PPCODE: 1210 PPCODE:
900 cl_platform_id value [1]; 1211 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1212 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1213 EXTEND (SP, 1);
903 const int i = 0; 1214 const int i = 0;
904 { 1215 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1216 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1217 }
907 1218
908void 1219void
909name (OpenCL::Device self) 1220name (OpenCL::Device self)
910 ALIAS: 1221 ALIAS:
929 cl_device_id value [1]; 1240 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1241 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1242 EXTEND (SP, 1);
932 const int i = 0; 1243 const int i = 0;
933 { 1244 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1245 PUSH_CLOBJ (stash_device, value [i]);
935 } 1246 }
936 1247
937void 1248void
938partition_types_ext (OpenCL::Device self) 1249partition_types_ext (OpenCL::Device self)
939 ALIAS: 1250 ALIAS:
950 for (i = 0; i < n; ++i) 1261 for (i = 0; i < n; ++i)
951 PUSHs (sv_2mortal (newSVuv (value [i]))); 1262 PUSHs (sv_2mortal (newSVuv (value [i])));
952 1263
953#END:device 1264#END:device
954 1265
1266MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1267
1268#if CL_VERSION_1_2
1269
1270void
1271DESTROY (OpenCL::SubDevice self)
1272 CODE:
1273 clReleaseDevice (self);
1274
1275#endif
1276
955MODULE = OpenCL PACKAGE = OpenCL::Context 1277MODULE = OpenCL PACKAGE = OpenCL::Context
956 1278
957void 1279void
958DESTROY (OpenCL::Context context) 1280DESTROY (OpenCL::Context self)
959 CODE: 1281 CODE:
960 clReleaseContext (context); 1282 clReleaseContext (self);
961 1283
962void 1284void
963info (OpenCL::Context self, cl_context_info name) 1285info (OpenCL::Context self, cl_context_info name)
964 PPCODE: 1286 PPCODE:
965 INFO (Context) 1287 INFO (Context)
966 1288
967void 1289void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1290queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1291 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1292 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1293 XPUSH_CLOBJ (stash_queue, queue);
972 1294
973void 1295void
974user_event (OpenCL::Context self) 1296user_event (OpenCL::Context self)
975 PPCODE: 1297 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1298 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1299 XPUSH_CLOBJ (stash_userevent, ev);
978 1300
979void 1301void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1302buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1303 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1304 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
983 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1305 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1306
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1307 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1308 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1309
988void 1310void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1311buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1312 PPCODE:
991 STRLEN len; 1313 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1314 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1315 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
994 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1316 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
995 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1317 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1318 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1319
998#if CL_VERSION_1_2 1320#if CL_VERSION_1_2
999 1321
1000void 1322void
1001image (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) 1323image (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)
1006 const cl_image_desc desc = { 1328 const cl_image_desc desc = {
1007 type, 1329 type,
1008 width, height, depth, 1330 width, height, depth,
1009 array_size, row_pitch, slice_pitch, 1331 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1332 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1333 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1012 }; 1334 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1336 HV *stash = stash_image;
1015 switch (type) 1337 switch (type)
1016 { 1338 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1339 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1340 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1341 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1342 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1343 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1344 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1345 }
1024 XPUSH_CLOBJ (klass, mem); 1346 XPUSH_CLOBJ (stash, mem);
1025 1347
1026#endif 1348#endif
1027 1349
1028void 1350void
1029image2d (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) 1351image2d (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)
1035 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1357 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1358#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1359 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1038 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1360 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1361#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1362 XPUSH_CLOBJ (stash_image2d, mem);
1041 1363
1042void 1364void
1043image3d (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) 1365image3d (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)
1044 PPCODE: 1366 PPCODE:
1045 STRLEN len; 1367 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1371 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1372#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1373 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1052 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1374 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1375#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1376 XPUSH_CLOBJ (stash_image3d, mem);
1055 1377
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1378#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1379
1058void 1380void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1381gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1382 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1384 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1385
1064void 1386void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1387gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1388 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1389 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1390 XPUSH_CLOBJ (stash_image2d, mem);
1069 1391
1070#if CL_VERSION_1_2 1392#if CL_VERSION_1_2
1071 1393
1072void 1394void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1395gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1396 ALIAS:
1075 PPCODE: 1397 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1398 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1399 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1400 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1401 HV *stash = stash_memory;
1080 switch (type) 1402 switch (type)
1081 { 1403 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1404 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1405 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1406 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1407 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1408 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1409 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1410 }
1089 XPUSH_CLOBJ (klass, mem); 1411 XPUSH_CLOBJ (stash, mem);
1090 1412
1091#endif 1413#endif
1092 1414
1093void 1415void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1416gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1418#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1419 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1420#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1421 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1422#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1423 XPUSH_CLOBJ (stash_image2d, mem);
1102 1424
1103void 1425void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1426gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1427 PPCODE:
1106#if PREFER_1_1 1428#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1429 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1430#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1431 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1432#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1433 XPUSH_CLOBJ (stash_image3d, mem);
1112 1434
1113#endif 1435#endif
1114 1436
1115void 1437void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1438supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1458
1137void 1459void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1460sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1461 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1462 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1463 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1464
1143void 1465void
1144program_with_source (OpenCL::Context self, SV *program) 1466program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1467 PPCODE:
1146 STRLEN len; 1468 STRLEN len;
1147 size_t len2; 1469 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1470 const char *ptr = SvPVbyte (program, len);
1149 1471
1150 len2 = len; 1472 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1473 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1474 XPUSH_CLOBJ (stash_program, prog);
1475
1476void
1477program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1478 PPCODE:
1479 cl_uint device_count;
1480 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1481
1482 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1483 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1484
1485 binaries = SvRV (binaries);
1486
1487 if (device_count != av_len ((AV *)binaries) + 1)
1488 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1489
1490 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1491 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1492 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1493
1494 int i;
1495 for (i = 0; i < device_count; ++i)
1496 {
1497 STRLEN len;
1498 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1499 length_list [i] = len;
1500 }
1501
1502 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1503 length_list, binary_list,
1504 GIMME_V == G_ARRAY ? status_list : 0, &res));
1505
1506 EXTEND (SP, 2);
1507 PUSH_CLOBJ (stash_program, prog);
1508
1509 if (GIMME_V == G_ARRAY)
1510 {
1511 AV *av = newAV ();
1512 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1513
1514 for (i = device_count; i--; )
1515 av_store (av, i, newSViv (status_list [i]));
1516 }
1517
1518#if CL_VERSION_1_2
1519
1520void
1521program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1522 PPCODE:
1523 cl_uint device_count;
1524 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1525
1526 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1527
1528 XPUSH_CLOBJ (stash_program, prog);
1529
1530void
1531link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1532 CODE:
1533 cl_uint device_count = 0;
1534 cl_device_id *device_list = 0;
1535
1536 if (SvOK (devices))
1537 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1538
1539 cl_uint program_count;
1540 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1541
1542 void *user_data;
1543 program_callback pfn_notify = make_program_callback (notify, &user_data);
1544
1545 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1546 program_count, program_list, pfn_notify, user_data, &res));
1547
1548 XPUSH_CLOBJ (stash_program, prog);
1549
1550#endif
1153 1551
1154#BEGIN:context 1552#BEGIN:context
1155 1553
1156void 1554void
1157reference_count (OpenCL::Context self) 1555reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1572 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1573 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1574 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1575 for (i = 0; i < n; ++i)
1178 { 1576 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1577 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1578 }
1181 1579
1182void 1580void
1183properties (OpenCL::Context self) 1581properties (OpenCL::Context self)
1184 PPCODE: 1582 PPCODE:
1203void 1601void
1204read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1602read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1205 ALIAS: 1603 ALIAS:
1206 enqueue_read_buffer = 0 1604 enqueue_read_buffer = 0
1207 PPCODE: 1605 PPCODE:
1208 cl_event ev = 0;
1209 EVENT_LIST (6, items - 6); 1606 EVENT_LIST (6);
1210 1607
1211 SvUPGRADE (data, SVt_PV); 1608 SvUPGRADE (data, SVt_PV);
1212 SvGROW (data, len); 1609 SvGROW (data, len);
1213 SvPOK_only (data); 1610 SvPOK_only (data);
1214 SvCUR_set (data, len); 1611 SvCUR_set (data, len);
1612
1613 cl_event ev = 0;
1215 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1216 1615
1217 if (ev) 1616 if (ev)
1218 XPUSH_CLOBJ ("OpenCL::Event", ev); 1617 XPUSH_CLOBJ (stash_event, ev);
1219 1618
1220void 1619void
1221write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1620write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1222 ALIAS: 1621 ALIAS:
1223 enqueue_write_buffer = 0 1622 enqueue_write_buffer = 0
1224 PPCODE: 1623 PPCODE:
1225 cl_event ev = 0; 1624 EVENT_LIST (5);
1625
1226 STRLEN len; 1626 STRLEN len;
1227 char *ptr = SvPVbyte (data, len); 1627 char *ptr = SvPVbyte (data, len);
1228 EVENT_LIST (5, items - 5);
1229 1628
1629 cl_event ev = 0;
1230 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1630 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1231 1631
1232 if (ev) 1632 if (ev)
1233 XPUSH_CLOBJ ("OpenCL::Event", ev); 1633 XPUSH_CLOBJ (stash_event, ev);
1234 1634
1235#if CL_VERSION_1_2 1635#if CL_VERSION_1_2
1236 1636
1237void 1637void
1238fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1638fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1239 ALIAS: 1639 ALIAS:
1240 enqueue_fill_buffer = 0 1640 enqueue_fill_buffer = 0
1241 PPCODE: 1641 PPCODE:
1242 cl_event ev = 0; 1642 EVENT_LIST (5);
1643
1243 STRLEN len; 1644 STRLEN len;
1244 char *ptr = SvPVbyte (data, len); 1645 char *ptr = SvPVbyte (data, len);
1245 EVENT_LIST (5, items - 5);
1246 1646
1647 cl_event ev = 0;
1247 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1648 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1248 1649
1249 if (ev) 1650 if (ev)
1250 XPUSH_CLOBJ ("OpenCL::Event", ev); 1651 XPUSH_CLOBJ (stash_event, ev);
1251 1652
1252void 1653void
1253fill_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, ...) 1654fill_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, ...)
1254 ALIAS: 1655 ALIAS:
1255 enqueue_fill_image = 0 1656 enqueue_fill_image = 0
1256 PPCODE: 1657 PPCODE:
1257 cl_event ev = 0; 1658 EVENT_LIST (12);
1258 STRLEN len; 1659
1259 const size_t origin [3] = { x, y, z }; 1660 const size_t origin [3] = { x, y, z };
1260 const size_t region [3] = { width, height, depth }; 1661 const size_t region [3] = { width, height, depth };
1261 EVENT_LIST (12, items - 12);
1262 1662
1263 const cl_float c_f [4] = { r, g, b, a }; 1663 const cl_float c_f [4] = { r, g, b, a };
1264 const cl_uint c_u [4] = { r, g, b, a }; 1664 const cl_uint c_u [4] = { r, g, b, a };
1265 const cl_int c_s [4] = { r, g, b, a }; 1665 const cl_int c_s [4] = { r, g, b, a };
1266 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1666 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1269 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1669 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1270 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1670 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1271 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1671 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1272 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1672 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1273 1673
1674 cl_event ev = 0;
1274 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1675 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1275 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1676 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1276 1677
1277 if (ev) 1678 if (ev)
1278 XPUSH_CLOBJ ("OpenCL::Event", ev); 1679 XPUSH_CLOBJ (stash_event, ev);
1279 1680
1280#endif 1681#endif
1281 1682
1282void 1683void
1283copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1684copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1284 ALIAS: 1685 ALIAS:
1285 enqueue_copy_buffer = 0 1686 enqueue_copy_buffer = 0
1286 PPCODE: 1687 PPCODE:
1688 EVENT_LIST (6);
1689
1287 cl_event ev = 0; 1690 cl_event ev = 0;
1288 EVENT_LIST (6, items - 6);
1289
1290 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1691 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1291 1692
1292 if (ev) 1693 if (ev)
1293 XPUSH_CLOBJ ("OpenCL::Event", ev); 1694 XPUSH_CLOBJ (stash_event, ev);
1294 1695
1295void 1696void
1296read_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, ...) 1697read_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, ...)
1297 ALIAS: 1698 ALIAS:
1298 enqueue_read_buffer_rect = 0 1699 enqueue_read_buffer_rect = 0
1299 PPCODE: 1700 PPCODE:
1300 cl_event ev = 0; 1701 EVENT_LIST (17);
1702
1301 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1703 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1302 const size_t host_origin[3] = { host_x, host_y, host_z }; 1704 const size_t host_origin[3] = { host_x, host_y, host_z };
1303 const size_t region[3] = { width, height, depth }; 1705 const size_t region[3] = { width, height, depth };
1304 EVENT_LIST (17, items - 17);
1305 1706
1306 if (!buf_row_pitch) 1707 if (!buf_row_pitch)
1307 buf_row_pitch = region [0]; 1708 buf_row_pitch = region [0];
1308 1709
1309 if (!buf_slice_pitch) 1710 if (!buf_slice_pitch)
1319 1720
1320 SvUPGRADE (data, SVt_PV); 1721 SvUPGRADE (data, SVt_PV);
1321 SvGROW (data, len); 1722 SvGROW (data, len);
1322 SvPOK_only (data); 1723 SvPOK_only (data);
1323 SvCUR_set (data, len); 1724 SvCUR_set (data, len);
1725
1726 cl_event ev = 0;
1324 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)); 1727 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));
1325 1728
1326 if (ev) 1729 if (ev)
1327 XPUSH_CLOBJ ("OpenCL::Event", ev); 1730 XPUSH_CLOBJ (stash_event, ev);
1328 1731
1329void 1732void
1330write_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, ...) 1733write_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, ...)
1331 ALIAS: 1734 ALIAS:
1332 enqueue_write_buffer_rect = 0 1735 enqueue_write_buffer_rect = 0
1333 PPCODE: 1736 PPCODE:
1334 cl_event ev = 0; 1737 EVENT_LIST (17);
1738
1335 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1739 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1336 const size_t host_origin[3] = { host_x, host_y, host_z }; 1740 const size_t host_origin[3] = { host_x, host_y, host_z };
1337 const size_t region[3] = { width, height, depth }; 1741 const size_t region[3] = { width, height, depth };
1338 STRLEN len; 1742 STRLEN len;
1339 char *ptr = SvPVbyte (data, len); 1743 char *ptr = SvPVbyte (data, len);
1340 EVENT_LIST (17, items - 17);
1341 1744
1342 if (!buf_row_pitch) 1745 if (!buf_row_pitch)
1343 buf_row_pitch = region [0]; 1746 buf_row_pitch = region [0];
1344 1747
1345 if (!buf_slice_pitch) 1748 if (!buf_slice_pitch)
1354 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1757 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1355 1758
1356 if (len < min_len) 1759 if (len < min_len)
1357 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1760 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1358 1761
1762 cl_event ev = 0;
1359 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)); 1763 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));
1360 1764
1361 if (ev) 1765 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1766 XPUSH_CLOBJ (stash_event, ev);
1363 1767
1364void 1768void
1365copy_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, ...) 1769copy_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, ...)
1366 ALIAS: 1770 ALIAS:
1367 enqueue_copy_buffer_rect = 0 1771 enqueue_copy_buffer_rect = 0
1368 PPCODE: 1772 PPCODE:
1369 cl_event ev = 0; 1773 EVENT_LIST (16);
1774
1370 const size_t src_origin[3] = { src_x, src_y, src_z }; 1775 const size_t src_origin[3] = { src_x, src_y, src_z };
1371 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1776 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1372 const size_t region[3] = { width, height, depth }; 1777 const size_t region[3] = { width, height, depth };
1373 EVENT_LIST (16, items - 16);
1374 1778
1779 cl_event ev = 0;
1375 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1780 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1376 1781
1377 if (ev) 1782 if (ev)
1378 XPUSH_CLOBJ ("OpenCL::Event", ev); 1783 XPUSH_CLOBJ (stash_event, ev);
1379 1784
1380void 1785void
1381read_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, ...) 1786read_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, ...)
1382 ALIAS: 1787 ALIAS:
1383 enqueue_read_image = 0 1788 enqueue_read_image = 0
1384 PPCODE: 1789 PPCODE:
1385 cl_event ev = 0; 1790 EVENT_LIST (12);
1791
1386 const size_t src_origin[3] = { src_x, src_y, src_z }; 1792 const size_t src_origin[3] = { src_x, src_y, src_z };
1387 const size_t region[3] = { width, height, depth }; 1793 const size_t region[3] = { width, height, depth };
1388 EVENT_LIST (12, items - 12);
1389 1794
1390 if (!row_pitch) 1795 if (!row_pitch)
1391 row_pitch = img_row_pitch (src); 1796 row_pitch = img_row_pitch (src);
1392 1797
1393 if (depth > 1 && !slice_pitch) 1798 if (depth > 1 && !slice_pitch)
1397 1802
1398 SvUPGRADE (data, SVt_PV); 1803 SvUPGRADE (data, SVt_PV);
1399 SvGROW (data, len); 1804 SvGROW (data, len);
1400 SvPOK_only (data); 1805 SvPOK_only (data);
1401 SvCUR_set (data, len); 1806 SvCUR_set (data, len);
1807
1808 cl_event ev = 0;
1402 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)); 1809 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));
1403 1810
1404 if (ev) 1811 if (ev)
1405 XPUSH_CLOBJ ("OpenCL::Event", ev); 1812 XPUSH_CLOBJ (stash_event, ev);
1406 1813
1407void 1814void
1408write_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, ...) 1815write_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, ...)
1409 ALIAS: 1816 ALIAS:
1410 enqueue_write_image = 0 1817 enqueue_write_image = 0
1411 PPCODE: 1818 PPCODE:
1412 cl_event ev = 0; 1819 EVENT_LIST (12);
1820
1413 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1821 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1414 const size_t region[3] = { width, height, depth }; 1822 const size_t region[3] = { width, height, depth };
1415 STRLEN len; 1823 STRLEN len;
1416 char *ptr = SvPVbyte (data, len); 1824 char *ptr = SvPVbyte (data, len);
1417 EVENT_LIST (12, items - 12);
1418 1825
1419 if (!row_pitch) 1826 if (!row_pitch)
1420 row_pitch = img_row_pitch (dst); 1827 row_pitch = img_row_pitch (dst);
1421 1828
1422 if (depth > 1 && !slice_pitch) 1829 if (depth > 1 && !slice_pitch)
1425 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1832 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1426 1833
1427 if (len < min_len) 1834 if (len < min_len)
1428 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1835 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1429 1836
1837 cl_event ev = 0;
1430 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)); 1838 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));
1431 1839
1432 if (ev) 1840 if (ev)
1433 XPUSH_CLOBJ ("OpenCL::Event", ev); 1841 XPUSH_CLOBJ (stash_event, ev);
1434 1842
1435void 1843void
1436copy_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, ...) 1844copy_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, ...)
1437 ALIAS: 1845 ALIAS:
1438 enqueue_copy_image = 0 1846 enqueue_copy_image = 0
1439 PPCODE: 1847 PPCODE:
1440 cl_event ev = 0; 1848 EVENT_LIST (12);
1849
1441 const size_t src_origin[3] = { src_x, src_y, src_z }; 1850 const size_t src_origin[3] = { src_x, src_y, src_z };
1442 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1851 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1443 const size_t region[3] = { width, height, depth }; 1852 const size_t region[3] = { width, height, depth };
1444 EVENT_LIST (12, items - 12);
1445 1853
1854 cl_event ev = 0;
1446 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1855 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1447 1856
1448 if (ev) 1857 if (ev)
1449 XPUSH_CLOBJ ("OpenCL::Event", ev); 1858 XPUSH_CLOBJ (stash_event, ev);
1450 1859
1451void 1860void
1452copy_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, ...) 1861copy_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, ...)
1453 ALIAS: 1862 ALIAS:
1454 enqueue_copy_image_to_buffer = 0 1863 enqueue_copy_image_to_buffer = 0
1455 PPCODE: 1864 PPCODE:
1865 EVENT_LIST (10);
1866
1867 const size_t src_origin[3] = { src_x, src_y, src_z };
1868 const size_t region [3] = { width, height, depth };
1869
1456 cl_event ev = 0; 1870 cl_event ev = 0;
1457 const size_t src_origin[3] = { src_x, src_y, src_z };
1458 const size_t region[3] = { width, height, depth };
1459 EVENT_LIST (10, items - 10);
1460
1461 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1871 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1462 1872
1463 if (ev) 1873 if (ev)
1464 XPUSH_CLOBJ ("OpenCL::Event", ev); 1874 XPUSH_CLOBJ (stash_event, ev);
1465 1875
1466void 1876void
1467copy_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, ...) 1877copy_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, ...)
1468 ALIAS: 1878 ALIAS:
1469 enqueue_copy_buffer_to_image = 0 1879 enqueue_copy_buffer_to_image = 0
1470 PPCODE: 1880 PPCODE:
1881 EVENT_LIST (10);
1882
1883 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1884 const size_t region [3] = { width, height, depth };
1885
1471 cl_event ev = 0; 1886 cl_event ev = 0;
1887 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1888
1889 if (ev)
1890 XPUSH_CLOBJ (stash_event, ev);
1891
1892void
1893map_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, ...)
1894 ALIAS:
1895 enqueue_map_buffer = 0
1896 PPCODE:
1897 EVENT_LIST (6);
1898
1899 size_t cb = SvIV (cb_);
1900
1901 if (!SvOK (cb_))
1902 {
1903 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1904 cb -= offset;
1905 }
1906
1907 cl_event ev;
1908 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1909 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1910
1911void
1912map_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, ...)
1913 ALIAS:
1914 enqueue_map_image = 0
1915 PPCODE:
1916 size_t width = SvIV (width_);
1917 if (!SvOK (width_))
1918 {
1919 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1920 width -= x;
1921 }
1922
1923 size_t height = SvIV (width_);
1924 if (!SvOK (height_))
1925 {
1926 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1927 height -= y;
1928 }
1929
1930 size_t depth = SvIV (width_);
1931 if (!SvOK (depth_))
1932 {
1933 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1934 depth -= z;
1935
1936 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1937 if (!depth)
1938 depth = 1;
1939 }
1940
1472 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1941 const size_t origin[3] = { x, y, z };
1473 const size_t region[3] = { width, height, depth }; 1942 const size_t region[3] = { width, height, depth };
1943 size_t row_pitch, slice_pitch;
1474 EVENT_LIST (10, items - 10); 1944 EVENT_LIST (10);
1475 1945
1476 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1946 cl_event ev;
1947 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));
1477 1948
1478 if (ev) 1949 size_t cb = slice_pitch ? slice_pitch * region [2]
1479 XPUSH_CLOBJ ("OpenCL::Event", ev); 1950 : row_pitch ? row_pitch * region [1]
1951 : region [0];
1952
1953 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1954
1955void
1956unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1957 PPCODE:
1958 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1959 if (GIMME_V != G_VOID)
1960 {
1961 clRetainEvent (mapped->event);
1962 XPUSH_CLOBJ (stash_event, mapped->event);
1963 }
1480 1964
1481void 1965void
1482task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1966task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1483 ALIAS: 1967 ALIAS:
1484 enqueue_task = 0 1968 enqueue_task = 0
1485 PPCODE: 1969 PPCODE:
1970 EVENT_LIST (2);
1971
1486 cl_event ev = 0; 1972 cl_event ev = 0;
1487 EVENT_LIST (2, items - 2);
1488
1489 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1973 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1490 1974
1491 if (ev) 1975 if (ev)
1492 XPUSH_CLOBJ ("OpenCL::Event", ev); 1976 XPUSH_CLOBJ (stash_event, ev);
1493 1977
1494void 1978void
1495nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1979nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1496 ALIAS: 1980 ALIAS:
1497 enqueue_nd_range_kernel = 0 1981 enqueue_nd_range_kernel = 0
1498 PPCODE: 1982 PPCODE:
1499 cl_event ev = 0; 1983 EVENT_LIST (5);
1984
1500 size_t *gwo = 0, *gws, *lws = 0; 1985 size_t *gwo = 0, *gws, *lws = 0;
1501 int gws_len; 1986 int gws_len;
1502 size_t *lists; 1987 size_t *lists;
1503 int i; 1988 int i;
1504 EVENT_LIST (5, items - 5);
1505 1989
1506 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1990 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1507 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1991 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1508 1992
1509 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1993 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1510 1994
1511 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1995 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1512 1996
1513 gws = lists + gws_len * 0; 1997 gws = lists + gws_len * 0;
1514 for (i = 0; i < gws_len; ++i) 1998 for (i = 0; i < gws_len; ++i)
1999 {
1515 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 2000 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
2001 // at least nvidia crashes for 0-sized work group sizes, work around
2002 if (!gws [i])
2003 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
2004 }
1516 2005
1517 if (SvOK (global_work_offset)) 2006 if (SvOK (global_work_offset))
1518 { 2007 {
1519 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 2008 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1520 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 2009 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1528 } 2017 }
1529 2018
1530 if (SvOK (local_work_size)) 2019 if (SvOK (local_work_size))
1531 { 2020 {
1532 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 2021 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1533 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 2022 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1534 2023
1535 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 2024 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1536 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 2025 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1537 2026
1538 lws = lists + gws_len * 2; 2027 lws = lists + gws_len * 2;
1539 for (i = 0; i < gws_len; ++i) 2028 for (i = 0; i < gws_len; ++i)
2029 {
1540 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 2030 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
2031 // at least nvidia crashes for 0-sized work group sizes, work around
2032 if (!lws [i])
2033 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2034 }
1541 } 2035 }
1542 2036
2037 cl_event ev = 0;
1543 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2038 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1544 2039
1545 if (ev) 2040 if (ev)
1546 XPUSH_CLOBJ ("OpenCL::Event", ev); 2041 XPUSH_CLOBJ (stash_event, ev);
2042
2043#if CL_VERSION_1_2
2044
2045void
2046migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2047 ALIAS:
2048 enqueue_migrate_mem_objects = 0
2049 PPCODE:
2050 EVENT_LIST (3);
2051
2052 cl_uint obj_count;
2053 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2054
2055 cl_event ev = 0;
2056 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2057
2058 if (ev)
2059 XPUSH_CLOBJ (stash_event, ev);
2060
2061#endif
1547 2062
1548#if cl_apple_gl_sharing || cl_khr_gl_sharing 2063#if cl_apple_gl_sharing || cl_khr_gl_sharing
1549 2064
1550void 2065void
1551acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2066acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1552 ALIAS: 2067 ALIAS:
2068 release_gl_objects = 1
1553 enqueue_acquire_gl_objects = 0 2069 enqueue_acquire_gl_objects = 0
1554 ALIAS:
1555 enqueue_release_gl_objects = 1 2070 enqueue_release_gl_objects = 1
1556 PPCODE: 2071 PPCODE:
1557 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2072 EVENT_LIST (2);
1558 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2073
2074 cl_uint obj_count;
2075 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1559 2076
1560 cl_event ev = 0; 2077 cl_event ev = 0;
1561 EVENT_LIST (2, items - 2);
1562 AV *av = (AV *)SvRV (objects);
1563 cl_uint num_objects = av_len (av) + 1;
1564 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1565 int i;
1566
1567 for (i = num_objects; i--; )
1568 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1569 2078
1570 if (ix) 2079 if (ix)
1571 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2080 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1572 else 2081 else
1573 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2082 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1574 2083
1575 if (ev) 2084 if (ev)
1576 XPUSH_CLOBJ ("OpenCL::Event", ev); 2085 XPUSH_CLOBJ (stash_event, ev);
1577 2086
1578#endif 2087#endif
1579 2088
1580void 2089void
1581wait_for_events (OpenCL::Queue self, ...) 2090wait_for_events (OpenCL::Queue self, ...)
1582 ALIAS: 2091 ALIAS:
1583 enqueue_wait_for_events = 0 2092 enqueue_wait_for_events = 0
1584 CODE: 2093 CODE:
1585 EVENT_LIST (1, items - 1); 2094 EVENT_LIST (1);
1586#if PREFER_1_1 2095#if PREFER_1_1
1587 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2096 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1588#else 2097#else
1589 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 2098 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1590#endif 2099#endif
1592void 2101void
1593marker (OpenCL::Queue self, ...) 2102marker (OpenCL::Queue self, ...)
1594 ALIAS: 2103 ALIAS:
1595 enqueue_marker = 0 2104 enqueue_marker = 0
1596 PPCODE: 2105 PPCODE:
2106 EVENT_LIST (1);
1597 cl_event ev = 0; 2107 cl_event ev = 0;
1598 EVENT_LIST (1, items - 1);
1599#if PREFER_1_1 2108#if PREFER_1_1
1600 if (!event_list_count) 2109 if (!event_list_count)
1601 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2110 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1602 else 2111 else
1603#if CL_VERSION_1_2 2112#if CL_VERSION_1_2
1610#endif 2119#endif
1611#else 2120#else
1612 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2121 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1613#endif 2122#endif
1614 if (ev) 2123 if (ev)
1615 XPUSH_CLOBJ ("OpenCL::Event", ev); 2124 XPUSH_CLOBJ (stash_event, ev);
1616 2125
1617void 2126void
1618barrier (OpenCL::Queue self, ...) 2127barrier (OpenCL::Queue self, ...)
1619 ALIAS: 2128 ALIAS:
1620 enqueue_barrier = 0 2129 enqueue_barrier = 0
1621 PPCODE: 2130 PPCODE:
2131 EVENT_LIST (1);
1622 cl_event ev = 0; 2132 cl_event ev = 0;
1623 EVENT_LIST (1, items - 1);
1624#if PREFER_1_1 2133#if PREFER_1_1
1625 if (!event_list_count && GIMME_V == G_VOID) 2134 if (!event_list_count && GIMME_V == G_VOID)
1626 NEED_SUCCESS (EnqueueBarrier, (self)); 2135 NEED_SUCCESS (EnqueueBarrier, (self));
1627 else 2136 else
1628#if CL_VERSION_1_2 2137#if CL_VERSION_1_2
1638#endif 2147#endif
1639#else 2148#else
1640 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2149 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1641#endif 2150#endif
1642 if (ev) 2151 if (ev)
1643 XPUSH_CLOBJ ("OpenCL::Event", ev); 2152 XPUSH_CLOBJ (stash_event, ev);
1644 2153
1645void 2154void
1646flush (OpenCL::Queue self) 2155flush (OpenCL::Queue self)
1647 CODE: 2156 CODE:
1648 NEED_SUCCESS (Flush, (self)); 2157 NEED_SUCCESS (Flush, (self));
1666 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2175 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1667 EXTEND (SP, 1); 2176 EXTEND (SP, 1);
1668 const int i = 0; 2177 const int i = 0;
1669 { 2178 {
1670 NEED_SUCCESS (RetainContext, (value [i])); 2179 NEED_SUCCESS (RetainContext, (value [i]));
1671 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2180 PUSH_CLOBJ (stash_context, value [i]);
1672 } 2181 }
1673 2182
1674void 2183void
1675device (OpenCL::Queue self) 2184device (OpenCL::Queue self)
1676 PPCODE: 2185 PPCODE:
1677 cl_device_id value [1]; 2186 cl_device_id value [1];
1678 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2187 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1679 EXTEND (SP, 1); 2188 EXTEND (SP, 1);
1680 const int i = 0; 2189 const int i = 0;
1681 { 2190 {
1682 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2191 PUSH_CLOBJ (stash_device, value [i]);
1683 } 2192 }
1684 2193
1685void 2194void
1686reference_count (OpenCL::Queue self) 2195reference_count (OpenCL::Queue self)
1687 PPCODE: 2196 PPCODE:
1696 PPCODE: 2205 PPCODE:
1697 cl_command_queue_properties value [1]; 2206 cl_command_queue_properties value [1];
1698 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2207 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1699 EXTEND (SP, 1); 2208 EXTEND (SP, 1);
1700 const int i = 0; 2209 const int i = 0;
1701 PUSHs (sv_2mortal (newSViv (value [i]))); 2210 PUSHs (sv_2mortal (newSVuv (value [i])));
1702 2211
1703#END:command_queue 2212#END:command_queue
1704 2213
1705MODULE = OpenCL PACKAGE = OpenCL::Memory 2214MODULE = OpenCL PACKAGE = OpenCL::Memory
1706 2215
1711 2220
1712void 2221void
1713info (OpenCL::Memory self, cl_mem_info name) 2222info (OpenCL::Memory self, cl_mem_info name)
1714 PPCODE: 2223 PPCODE:
1715 INFO (MemObject) 2224 INFO (MemObject)
2225
2226void
2227destructor_callback (OpenCL::Memory self, SV *notify)
2228 PPCODE:
2229 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
1716 2230
1717#BEGIN:mem 2231#BEGIN:mem
1718 2232
1719void 2233void
1720type (OpenCL::Memory self) 2234type (OpenCL::Memory self)
1721 PPCODE: 2235 PPCODE:
1722 cl_mem_object_type value [1]; 2236 cl_mem_object_type value [1];
1723 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2237 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1724 EXTEND (SP, 1); 2238 EXTEND (SP, 1);
1725 const int i = 0; 2239 const int i = 0;
1726 PUSHs (sv_2mortal (newSViv (value [i]))); 2240 PUSHs (sv_2mortal (newSVuv (value [i])));
1727 2241
1728void 2242void
1729flags (OpenCL::Memory self) 2243flags (OpenCL::Memory self)
1730 PPCODE: 2244 PPCODE:
1731 cl_mem_flags value [1]; 2245 cl_mem_flags value [1];
1732 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2246 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1733 EXTEND (SP, 1); 2247 EXTEND (SP, 1);
1734 const int i = 0; 2248 const int i = 0;
1735 PUSHs (sv_2mortal (newSViv (value [i]))); 2249 PUSHs (sv_2mortal (newSVuv (value [i])));
1736 2250
1737void 2251void
1738size (OpenCL::Memory self) 2252size (OpenCL::Memory self)
1739 ALIAS: 2253 ALIAS:
1740 size = CL_MEM_SIZE 2254 size = CL_MEM_SIZE
1774 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2288 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1775 EXTEND (SP, 1); 2289 EXTEND (SP, 1);
1776 const int i = 0; 2290 const int i = 0;
1777 { 2291 {
1778 NEED_SUCCESS (RetainContext, (value [i])); 2292 NEED_SUCCESS (RetainContext, (value [i]));
1779 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2293 PUSH_CLOBJ (stash_context, value [i]);
1780 } 2294 }
1781 2295
1782void 2296void
1783associated_memobject (OpenCL::Memory self) 2297associated_memobject (OpenCL::Memory self)
1784 PPCODE: 2298 PPCODE:
1786 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2300 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1787 EXTEND (SP, 1); 2301 EXTEND (SP, 1);
1788 const int i = 0; 2302 const int i = 0;
1789 { 2303 {
1790 NEED_SUCCESS (RetainMemObject, (value [i])); 2304 NEED_SUCCESS (RetainMemObject, (value [i]));
1791 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 2305 PUSH_CLOBJ (stash_memory, value [i]);
1792 } 2306 }
1793 2307
1794#END:mem 2308#END:mem
1795 2309
1796#if cl_apple_gl_sharing || cl_khr_gl_sharing 2310#if cl_apple_gl_sharing || cl_khr_gl_sharing
1816 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2330 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1817 2331
1818 cl_buffer_region crdata = { origin, size }; 2332 cl_buffer_region crdata = { origin, size };
1819 2333
1820 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2334 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1821 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2335 XPUSH_CLOBJ (stash_buffer, mem);
1822 2336
1823MODULE = OpenCL PACKAGE = OpenCL::Image 2337MODULE = OpenCL PACKAGE = OpenCL::Image
1824 2338
1825void 2339void
1826image_info (OpenCL::Image self, cl_image_info name) 2340image_info (OpenCL::Image self, cl_image_info name)
1912 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2426 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1913 EXTEND (SP, 1); 2427 EXTEND (SP, 1);
1914 const int i = 0; 2428 const int i = 0;
1915 { 2429 {
1916 NEED_SUCCESS (RetainContext, (value [i])); 2430 NEED_SUCCESS (RetainContext, (value [i]));
1917 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2431 PUSH_CLOBJ (stash_context, value [i]);
1918 } 2432 }
1919 2433
1920void 2434void
1921normalized_coords (OpenCL::Sampler self) 2435normalized_coords (OpenCL::Sampler self)
1922 PPCODE: 2436 PPCODE:
1923 cl_addressing_mode value [1]; 2437 cl_addressing_mode value [1];
1924 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2438 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1925 EXTEND (SP, 1); 2439 EXTEND (SP, 1);
1926 const int i = 0; 2440 const int i = 0;
1927 PUSHs (sv_2mortal (newSViv (value [i]))); 2441 PUSHs (sv_2mortal (newSVuv (value [i])));
1928 2442
1929void 2443void
1930addressing_mode (OpenCL::Sampler self) 2444addressing_mode (OpenCL::Sampler self)
1931 PPCODE: 2445 PPCODE:
1932 cl_filter_mode value [1]; 2446 cl_filter_mode value [1];
1933 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2447 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1934 EXTEND (SP, 1); 2448 EXTEND (SP, 1);
1935 const int i = 0; 2449 const int i = 0;
1936 PUSHs (sv_2mortal (newSViv (value [i]))); 2450 PUSHs (sv_2mortal (newSVuv (value [i])));
1937 2451
1938void 2452void
1939filter_mode (OpenCL::Sampler self) 2453filter_mode (OpenCL::Sampler self)
1940 PPCODE: 2454 PPCODE:
1941 cl_bool value [1]; 2455 cl_bool value [1];
1956void 2470void
1957build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2471build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
1958 ALIAS: 2472 ALIAS:
1959 build_async = 1 2473 build_async = 1
1960 CODE: 2474 CODE:
1961 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
1962 void *user_data = 0;
1963 cl_uint num_devices = 0; 2475 cl_uint device_count = 0;
1964 cl_device_id *device_list = 0; 2476 cl_device_id *device_list = 0;
1965 2477
1966 if (SvOK (devices)) 2478 if (SvOK (devices))
1967 { 2479 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2480
2481 void *user_data;
2482 program_callback pfn_notify = make_program_callback (notify, &user_data);
2483
2484 if (ix)
2485 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2486 else
2487 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2488
2489#if CL_VERSION_1_2
2490
2491void
2492compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2493 CODE:
2494 cl_uint device_count = 0;
2495 cl_device_id *device_list = 0;
2496
2497 if (SvOK (devices))
2498 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2499
2500 cl_uint header_count = 0;
2501 cl_program *header_list = 0;
2502 const char **header_name = 0;
2503
2504 if (SvOK (headers))
2505 {
1968 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2506 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
1969 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2507 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
1970 2508
1971 AV *av = (AV *)SvRV (devices); 2509 HV *hv = (HV *)SvRV (devices);
1972 num_devices = av_len (av) + 1;
1973 2510
1974 if (num_devices) 2511 header_count = hv_iterinit (hv);
2512 header_list = tmpbuf (sizeof (*header_list) * header_count);
2513 header_name = tmpbuf (sizeof (*header_name) * header_count);
2514
2515 HE *he;
2516 int i = 0;
2517 while (he = hv_iternext (hv))
1975 { 2518 {
1976 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2519 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2520 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
1977 int count; 2521 ++i;
1978 for (count = 0; count < num_devices; ++count)
1979 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
1980 } 2522 }
1981 } 2523 }
1982 2524
1983 if (SvOK (notify)) 2525 void *user_data;
1984 { 2526 program_callback pfn_notify = make_program_callback (notify, &user_data);
1985 NEED_SUCCESS (RetainProgram, (self));
1986 pfn_notify = eq_program_notify;
1987 user_data = SvREFCNT_inc (s_get_cv (notify));
1988 }
1989 2527
1990 if (ix)
1991 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
1992 else
1993 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2528 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2529 header_count, header_list, header_name, pfn_notify, user_data));
2530
2531#endif
1994 2532
1995void 2533void
1996build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2534build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1997 PPCODE: 2535 PPCODE:
1998 size_t size; 2536 size_t size;
2027 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2565 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2028 EXTEND (SP, 1); 2566 EXTEND (SP, 1);
2029 const int i = 0; 2567 const int i = 0;
2030 PUSHs (sv_2mortal (newSVpv (value, 0))); 2568 PUSHs (sv_2mortal (newSVpv (value, 0)));
2031 2569
2570void
2571binary_type (OpenCL::Program self, OpenCL::Device device)
2572 PPCODE:
2573 cl_program_binary_type value [1];
2574 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2575 EXTEND (SP, 1);
2576 const int i = 0;
2577 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2578
2032#END:program_build 2579#END:program_build
2033 2580
2034void 2581void
2035kernel (OpenCL::Program program, SV *function) 2582kernel (OpenCL::Program program, SV *function)
2036 PPCODE: 2583 PPCODE:
2037 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2584 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2038 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2585 XPUSH_CLOBJ (stash_kernel, kernel);
2039 2586
2040void 2587void
2041kernels_in_program (OpenCL::Program program) 2588kernels_in_program (OpenCL::Program program)
2042 PPCODE: 2589 PPCODE:
2043 cl_uint num_kernels; 2590 cl_uint num_kernels;
2046 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2593 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2047 2594
2048 int i; 2595 int i;
2049 EXTEND (SP, num_kernels); 2596 EXTEND (SP, num_kernels);
2050 for (i = 0; i < num_kernels; ++i) 2597 for (i = 0; i < num_kernels; ++i)
2051 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2598 PUSH_CLOBJ (stash_kernel, kernels [i]);
2052 2599
2053void 2600void
2054info (OpenCL::Program self, cl_program_info name) 2601info (OpenCL::Program self, cl_program_info name)
2055 PPCODE: 2602 PPCODE:
2056 INFO (Program) 2603 INFO (Program)
2104 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2651 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2105 EXTEND (SP, 1); 2652 EXTEND (SP, 1);
2106 const int i = 0; 2653 const int i = 0;
2107 { 2654 {
2108 NEED_SUCCESS (RetainContext, (value [i])); 2655 NEED_SUCCESS (RetainContext, (value [i]));
2109 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2656 PUSH_CLOBJ (stash_context, value [i]);
2110 } 2657 }
2111 2658
2112void 2659void
2113devices (OpenCL::Program self) 2660devices (OpenCL::Program self)
2114 PPCODE: 2661 PPCODE:
2118 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2665 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2119 int i, n = size / sizeof (*value); 2666 int i, n = size / sizeof (*value);
2120 EXTEND (SP, n); 2667 EXTEND (SP, n);
2121 for (i = 0; i < n; ++i) 2668 for (i = 0; i < n; ++i)
2122 { 2669 {
2123 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2670 PUSH_CLOBJ (stash_device, value [i]);
2124 } 2671 }
2125 2672
2126void 2673void
2127source (OpenCL::Program self) 2674source (OpenCL::Program self)
2128 PPCODE: 2675 PPCODE:
2183 cl_sampler ca; 2730 cl_sampler ca;
2184 size_t cz; 2731 size_t cz;
2185 cl_event ce; 2732 cl_event ce;
2186 } arg; 2733 } arg;
2187 size_t size; 2734 size_t size;
2735 int nullarg = 0;
2188 2736
2189 switch (type) 2737 switch (type)
2190 { 2738 {
2191 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break; 2739 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2192 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break; 2740 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2198 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break; 2746 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2199 2747
2200 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break; 2748 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2201 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2749 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2202 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2750 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2203 case 'z': arg.cz = SvUV (sv); size = sizeof (arg.cz); break;
2204 2751
2752 case 'z': nullarg = 1; size = SvIV (sv); break;
2753
2205 case 'm': arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2754 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2206 case 'a': arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2755 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2207 case 'e': arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2756 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2208 2757
2209 default: 2758 default:
2210 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2759 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2211 } 2760 }
2212 2761
2213 clSetKernelArg (self, i - 2, size, &arg); 2762 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2763 if (res)
2764 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2214 } 2765 }
2215 2766
2216 if (i != items) 2767 if (i != items)
2217 croak ("OpenCL::Kernel::setf format string too short (too many arguments)"); 2768 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2218 2769
2272 clSetKernelArg (self, idx, sizeof (value), &value); 2823 clSetKernelArg (self, idx, sizeof (value), &value);
2273 2824
2274void 2825void
2275set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2826set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2276 CODE: 2827 CODE:
2277 clSetKernelArg (self, idx, sizeof (value), &value); 2828 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2278 2829
2279void 2830void
2280set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2831set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2281 CODE: 2832 CODE:
2282 clSetKernelArg (self, idx, sizeof (value), &value); 2833 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2283 2834
2284void 2835void
2285set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2836set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2286 ALIAS:
2287 set_image2d = 0
2288 set_image3d = 0
2289 CODE: 2837 CODE:
2290 clSetKernelArg (self, idx, sizeof (value), &value); 2838 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2291 2839
2292void 2840void
2293set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2841set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2294 CODE: 2842 CODE:
2295 clSetKernelArg (self, idx, sizeof (value), &value); 2843 clSetKernelArg (self, idx, sizeof (value), &value);
2341 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2889 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2342 EXTEND (SP, 1); 2890 EXTEND (SP, 1);
2343 const int i = 0; 2891 const int i = 0;
2344 { 2892 {
2345 NEED_SUCCESS (RetainContext, (value [i])); 2893 NEED_SUCCESS (RetainContext, (value [i]));
2346 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2894 PUSH_CLOBJ (stash_context, value [i]);
2347 } 2895 }
2348 2896
2349void 2897void
2350program (OpenCL::Kernel self) 2898program (OpenCL::Kernel self)
2351 PPCODE: 2899 PPCODE:
2353 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2901 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2354 EXTEND (SP, 1); 2902 EXTEND (SP, 1);
2355 const int i = 0; 2903 const int i = 0;
2356 { 2904 {
2357 NEED_SUCCESS (RetainProgram, (value [i])); 2905 NEED_SUCCESS (RetainProgram, (value [i]));
2358 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2906 PUSH_CLOBJ (stash_program, value [i]);
2359 } 2907 }
2360 2908
2361#END:kernel 2909#END:kernel
2362 2910
2363void 2911void
2410 const int i = 0; 2958 const int i = 0;
2411 PUSHs (sv_2mortal (newSVuv (value [i]))); 2959 PUSHs (sv_2mortal (newSVuv (value [i])));
2412 2960
2413#END:kernel_work_group 2961#END:kernel_work_group
2414 2962
2963#if CL_VERSION_1_2
2964
2965void
2966arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2967 PPCODE:
2968 size_t size;
2969 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2970 SV *sv = sv_2mortal (newSV (size));
2971 SvUPGRADE (sv, SVt_PV);
2972 SvPOK_only (sv);
2973 SvCUR_set (sv, size);
2974 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2975 XPUSHs (sv);
2976
2977#BEGIN:kernel_arg
2978
2979void
2980arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2981 PPCODE:
2982 cl_kernel_arg_address_qualifier value [1];
2983 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2984 EXTEND (SP, 1);
2985 const int i = 0;
2986 PUSHs (sv_2mortal (newSVuv (value [i])));
2987
2988void
2989arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2990 PPCODE:
2991 cl_kernel_arg_access_qualifier value [1];
2992 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2993 EXTEND (SP, 1);
2994 const int i = 0;
2995 PUSHs (sv_2mortal (newSVuv (value [i])));
2996
2997void
2998arg_type_name (OpenCL::Kernel self, cl_uint idx)
2999 ALIAS:
3000 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
3001 arg_name = CL_KERNEL_ARG_NAME
3002 PPCODE:
3003 size_t size;
3004 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
3005 char *value = tmpbuf (size);
3006 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
3007 EXTEND (SP, 1);
3008 const int i = 0;
3009 PUSHs (sv_2mortal (newSVpv (value, 0)));
3010
3011void
3012arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
3013 PPCODE:
3014 cl_kernel_arg_type_qualifier value [1];
3015 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
3016 EXTEND (SP, 1);
3017 const int i = 0;
3018 PUSHs (sv_2mortal (newSVuv (value [i])));
3019
3020#END:kernel_arg
3021
3022#endif
3023
2415MODULE = OpenCL PACKAGE = OpenCL::Event 3024MODULE = OpenCL PACKAGE = OpenCL::Event
2416 3025
2417void 3026void
2418DESTROY (OpenCL::Event self) 3027DESTROY (OpenCL::Event self)
2419 CODE: 3028 CODE:
2443 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3052 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2444 EXTEND (SP, 1); 3053 EXTEND (SP, 1);
2445 const int i = 0; 3054 const int i = 0;
2446 { 3055 {
2447 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3056 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2448 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 3057 PUSH_CLOBJ (stash_queue, value [i]);
2449 } 3058 }
2450 3059
2451void 3060void
2452command_type (OpenCL::Event self) 3061command_type (OpenCL::Event self)
2453 PPCODE: 3062 PPCODE:
2476 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3085 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2477 EXTEND (SP, 1); 3086 EXTEND (SP, 1);
2478 const int i = 0; 3087 const int i = 0;
2479 { 3088 {
2480 NEED_SUCCESS (RetainContext, (value [i])); 3089 NEED_SUCCESS (RetainContext, (value [i]));
2481 PUSH_CLOBJ ("OpenCL::Context", value [i]); 3090 PUSH_CLOBJ (stash_context, value [i]);
2482 } 3091 }
2483 3092
2484#END:event 3093#END:event
2485 3094
2486void 3095void
2511void 3120void
2512set_status (OpenCL::UserEvent self, cl_int execution_status) 3121set_status (OpenCL::UserEvent self, cl_int execution_status)
2513 CODE: 3122 CODE:
2514 clSetUserEventStatus (self, execution_status); 3123 clSetUserEventStatus (self, execution_status);
2515 3124
3125MODULE = OpenCL PACKAGE = OpenCL::Mapped
3126
3127void
3128DESTROY (SV *self)
3129 CODE:
3130 OpenCL__Mapped mapped = SvMAPPED (self);
3131
3132 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3133 mapped_detach (self, mapped);
3134
3135 clReleaseCommandQueue (mapped->queue);
3136 clReleaseEvent (mapped->event);
3137 Safefree (mapped);
3138
3139void
3140unmap (OpenCL::Mapped self, ...)
3141 CODE:
3142 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3143
3144bool
3145mapped (OpenCL::Mapped self)
3146 CODE:
3147 RETVAL = !!self->ptr;
3148 OUTPUT:
3149 RETVAL
3150
3151void
3152wait (OpenCL::Mapped self)
3153 PPCODE:
3154 if (self->event)
3155 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3156
3157void
3158event (OpenCL::Mapped self)
3159 PPCODE:
3160 if (!self->event)
3161 XSRETURN_UNDEF;
3162
3163 clRetainEvent (self->event);
3164 XPUSH_CLOBJ (stash_event, self->event);
3165
3166size_t
3167size (OpenCL::Mapped self)
3168 CODE:
3169 RETVAL = self->cb;
3170 OUTPUT:
3171 RETVAL
3172
3173IV
3174ptr (OpenCL::Mapped self)
3175 CODE:
3176 RETVAL = PTR2IV (self->ptr);
3177 OUTPUT:
3178 RETVAL
3179
3180void
3181set (OpenCL::Mapped self, size_t offset, SV *data)
3182 CODE:
3183 STRLEN len;
3184 const char *ptr = SvPVbyte (data, len);
3185
3186 if (offset + len > self->cb)
3187 croak ("OpenCL::Mapped::set out of bound condition detected");
3188
3189 memcpy (offset + (char *)self->ptr, ptr, len);
3190
3191MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3192
3193MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3194
3195IV
3196row_pitch (OpenCL::Mapped self)
3197 ALIAS:
3198 slice_pitch = 1
3199 CODE:
3200 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3201 OUTPUT:
3202 RETVAL
3203
3204
3205

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines