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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.63 by root, Tue May 1 22:25:13 2012 UTC vs.
Revision 1.83 by root, Mon Apr 7 22:36:54 2014 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
6
7#define ECB_NO_THREADS 1
8#define ECB_NO_LIBM 1
9#include "ecb.h"
4 10
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 11#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 12#include "xthread.h"
7#include "schmorp.h" 13#include "schmorp.h"
8 14
26#if PREFER_1_1 32#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 33 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 34#endif
29 35
30#ifdef __APPLE__ 36#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 37 #define CLHDR(name) <OpenCL/name>
32#else 38#else
39 #define CLHDR(name) <CL/name>
40#endif
41
33 #include <CL/opencl.h> 42#include CLHDR(opencl.h)
43
44#if 0
45#ifndef CL_VERSION_1_2
46 #include CLHDR(cl_d3d9.h)
47#endif
48#endif
49
50#if _WIN32
51 #include CLHDR(cl_d3d10.h)
52 #if CL_VERSION_1_2
53 #include CLHDR(cl_d3d11.h)
54 #endif
55 #include CLHDR(cl_dx9_media_sharing.h)
34#endif 56#endif
35 57
36#ifndef CL_VERSION_1_2 58#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 59 #undef PREFER_1_1
38 #define PREFER_1_1 1 60 #define PREFER_1_1 1
39#endif 61#endif
40 62
63// make sure all constants we might use are actually defined
64#include "default.h"
65
41typedef cl_platform_id OpenCL__Platform; 66typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 67typedef cl_device_id OpenCL__Device;
68typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 69typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 70typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 71typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 72typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 73typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 74typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 75typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 76typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 77typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 78typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 79typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 80typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 81typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 82typedef cl_event OpenCL__UserEvent;
61 83
62typedef struct mapped * OpenCL__Mapped; 84typedef struct mapped * OpenCL__Mapped;
63 85
64typedef SV *FUTURE;
65
66static HV 86static HV
67 *stash_platform, 87 *stash_platform,
68 *stash_device, 88 *stash_device,
89 *stash_subdevice,
69 *stash_context, 90 *stash_context,
70 *stash_queue, 91 *stash_queue,
71 *stash_program, 92 *stash_program,
72 *stash_kernel, 93 *stash_kernel,
73 *stash_sampler, 94 *stash_sampler,
133} 154}
134 155
135/*****************************************************************************/ 156/*****************************************************************************/
136 157
137/* up to two temporary buffers */ 158/* up to two temporary buffers */
138static void * 159static void * ecb_noinline
139tmpbuf (size_t size) 160tmpbuf (size_t size)
140{ 161{
141 enum { buffers = 3 }; 162 enum { buffers = 4 };
142 static int idx; 163 static int idx;
143 static void *buf [buffers]; 164 static void *buf [buffers];
144 static size_t len [buffers]; 165 static size_t len [buffers];
145 166
146 idx = (idx + 1) % buffers; 167 idx = (idx + 1) % buffers;
153 } 174 }
154 175
155 return buf [idx]; 176 return buf [idx];
156} 177}
157 178
179static const char * ecb_noinline
180cv_get_name (CV *cv)
181{
182 static char fullname [256];
183
184 GV *gv = CvGV (cv); // gv better be nonzero
185
186 HV *stash = GvSTASH (gv);
187 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
188 const char *gvname = GvNAME (gv);
189
190 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
191 return fullname;
192}
193
158/*****************************************************************************/ 194/*****************************************************************************/
159 195
160typedef struct 196typedef struct
161{ 197{
162 IV iv; 198 IV iv;
163 const char *name; 199 const char *name;
164 #define const_iv(name) { (IV)CL_ ## name, # name }, 200 #define const_iv(name) { (IV)CL_ ## name, # name },
165} ivstr; 201} ivstr;
202
203typedef struct
204{
205 NV nv;
206 const char *name;
207 #define const_nv(name) { (NV)CL_ ## name, # name },
208} nvstr;
166 209
167static const char * 210static const char *
168iv2str (IV value, const ivstr *base, int count, const char *fallback) 211iv2str (IV value, const ivstr *base, int count, const char *fallback)
169{ 212{
170 int i; 213 int i;
219 if (res) \ 262 if (res) \
220 FAIL (name); 263 FAIL (name);
221 264
222/*****************************************************************************/ 265/*****************************************************************************/
223 266
267static SV *
268new_clobj (HV *stash, IV id)
269{
270 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
271}
272
273#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
274#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
275
276/* cl objects are either \$iv, or [$iv, ...] */
277/* they can be upgraded at runtime to the array form */
278static void * ecb_noinline
279SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
280{
281 // sv_derived_from is quite slow :(
282 if (SvROK (sv) && sv_derived_from (sv, pkg))
283 return (void *)SvIV (SvRV (sv));
284
285 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
286}
287
288// the "no-inherit" version of the above
289static void * ecb_noinline
290SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
291{
292 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
293 return (void *)SvIV (SvRV (sv));
294
295 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
296}
297
298/*****************************************************************************/
299
224static cl_context_properties * 300static cl_context_properties * ecb_noinline
225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 301SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
226{ 302{
227 if (!sv || !SvOK (sv)) 303 if (!sv || !SvOK (sv))
228 if (extra) 304 if (extra)
229 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 305 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
230 else 306 else
236 int i, len = av_len (av) + 1; 312 int i, len = av_len (av) + 1;
237 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 313 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
238 cl_context_properties *l = p; 314 cl_context_properties *l = p;
239 315
240 if (len & 1) 316 if (len & 1)
241 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 317 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
242 318
243 while (extracount--) 319 while (extracount--)
244 *l++ = *extra++; 320 *l++ = *extra++;
245 321
246 for (i = 0; i < len; i += 2) 322 for (i = 0; i < len; i += 2)
249 SV *p_sv = *av_fetch (av, i + 1, 0); 325 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override 326 cl_context_properties v = SvIV (p_sv); // code below can override
251 327
252 switch (t) 328 switch (t)
253 { 329 {
330 case CL_CONTEXT_PLATFORM:
331 if (SvROK (p_sv))
332 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
333 break;
334
254 case CL_GLX_DISPLAY_KHR: 335 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv)) 336 if (!SvOK (p_sv))
256 { 337 {
257 void *func = glsym ("_glXGetCurrentDisplay"); 338 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func) 339 if (func)
281 *l = 0; 362 *l = 0;
282 363
283 return p; 364 return p;
284 } 365 }
285 366
286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 367 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
287} 368}
288 369
289/*****************************************************************************/ 370// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
290 371// is a pointer (and only on well-behaved systems).
291static SV * 372static void * ecb_noinline
292new_clobj (HV *stash, IV id) 373object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
293{ 374{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash)); 375 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
295} 376 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
377 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
296 378
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id))) 379 AV *av = (AV *)SvRV (arg);
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id))) 380 void **list = 0;
381 cl_uint count = av_len (av) + 1;
299 382
300/* cl objects are either \$iv, or [$iv, ...] */ 383 if (count)
301/* they can be upgraded at runtime to the array form */ 384 {
302static void * 385 list = tmpbuf (sizeof (*list) * count);
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 386 int i;
304{ 387 for (i = 0; i < count; ++i)
305 if (SvROK (sv) && sv_derived_from (sv, pkg)) 388 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
306 return (void *)SvIV (SvRV (sv)); 389 }
307 390
308 croak ("%s: %s is not of type %s", func, svname, pkg); 391 if (!count && !or_undef)
392 croak ("%s: '%s' must contain at least one %s object",
393 cv_get_name (cv), argname, klass);
394
395 *rcount = count;
396 return (void *)list;
309} 397}
310 398
311/*****************************************************************************/ 399/*****************************************************************************/
312/* callback stuff */ 400/* callback stuff */
313 401
335static void (*eq_signal_func)(void *signal_arg, int value); 423static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg; 424static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT; 425static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail; 426static eq_item *eq_head, *eq_tail;
339 427
340static void 428static void ecb_noinline
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 429eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{ 430{
343 eq_item *item = malloc (sizeof (eq_item)); 431 eq_item *item = malloc (sizeof (eq_item));
344 432
345 item->next = 0; 433 item->next = 0;
412} 500}
413 501
414/*****************************************************************************/ 502/*****************************************************************************/
415/* context notify */ 503/* context notify */
416 504
417static void 505static void ecb_noinline
418eq_context_push (void *data1, void *data2, void *data3) 506eq_context_push (void *data1, void *data2, void *data3)
419{ 507{
420 dSP; 508 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0))); 509 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 510 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
444 { \ 532 { \
445 pfn_notify = eq_context_notify; \ 533 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \ 534 user_data = s_get_cv (notify); \
447 } 535 }
448 536
449static SV * 537static SV * ecb_noinline
450new_clobj_context (cl_context ctx, void *user_data) 538new_clobj_context (cl_context ctx, void *user_data)
451{ 539{
452 SV *sv = new_clobj (stash_context, (IV)ctx); 540 SV *sv = new_clobj (stash_context, (IV)ctx);
453 541
454 if (user_data) 542 if (user_data)
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 561static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474 562
475static void CL_CALLBACK 563static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data) 564eq_program_notify (cl_program program, void *user_data)
477{ 565{
566 clRetainProgram (program);
567
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 568 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
569}
570
571typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
572
573static program_callback ecb_noinline
574make_program_callback (SV *notify, void **ruser_data)
575{
576 if (SvOK (notify))
577 {
578 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
579 return eq_program_notify;
580 }
581 else
582 {
583 *ruser_data = 0;
584 return 0;
585 }
479} 586}
480 587
481struct build_args 588struct build_args
482{ 589{
483 cl_program program; 590 cl_program program;
497 else 604 else
498 clReleaseProgram (arg->program); 605 clReleaseProgram (arg->program);
499 606
500 free (arg->options); 607 free (arg->options);
501 free (arg); 608 free (arg);
609
610 return 0;
502} 611}
503 612
504static void 613static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 614build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{ 615{
511 arg->user_data = user_data; 620 arg->user_data = user_data;
512 arg->num_devices = num_devices; 621 arg->num_devices = num_devices;
513 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices); 622 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
514 623
515 xthread_t id; 624 xthread_t id;
516 thread_create (&id, build_program_thread, arg); 625 xthread_create (&id, build_program_thread, arg);
626}
627
628/*****************************************************************************/
629/* mem object destructor notify */
630
631static void ecb_noinline
632eq_destructor_push (void *data1, void *data2, void *data3)
633{
634}
635
636static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
637
638static void CL_CALLBACK
639eq_destructor_notify (cl_mem memobj, void *user_data)
640{
641 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
517} 642}
518 643
519/*****************************************************************************/ 644/*****************************************************************************/
520/* event objects */ 645/* event objects */
521 646
546 size_t res; 671 size_t res;
547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 672 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
548 return res; 673 return res;
549} 674}
550 675
551static cl_event * 676static cl_event * ecb_noinline
552event_list (SV **items, cl_uint *rcount, cl_event extra) 677event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
553{ 678{
554 cl_uint count = *rcount; 679 cl_uint count = *rcount;
555 680
556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed 681 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
557 count = 0; 682 *rcount = count = 0;
558 683
559 if (!count && !extra) 684 if (!count && !extra)
560 return 0; 685 return 0;
561 686
562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 687 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
563 int i = 0; 688 int i = 0;
564 689
565 while (count--) 690 while (count--)
566 if (SvOK (items [count])) 691 if (SvOK (items [count]))
567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 692 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
568 693
569 if (extra) 694 if (extra)
570 list [i++] = extra; 695 list [i++] = extra;
571 696
572 *rcount = i; 697 *rcount = i;
574 return i ? list : 0; 699 return i ? list : 0;
575} 700}
576 701
577#define EVENT_LIST(skip) \ 702#define EVENT_LIST(skip) \
578 cl_uint event_list_count = items - (skip); \ 703 cl_uint event_list_count = items - (skip); \
579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 704 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
580 705
581#define INFO(class) \ 706#define INFO(class) \
582{ \ 707{ \
583 size_t size; \ 708 size_t size; \
584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 709 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
607 void *ptr; 732 void *ptr;
608 size_t cb; 733 size_t cb;
609 cl_event event; 734 cl_event event;
610 size_t row_pitch; 735 size_t row_pitch;
611 size_t slice_pitch; 736 size_t slice_pitch;
737
738 size_t element_size;
739 size_t width, height, depth;
612}; 740};
613 741
614static SV * 742static SV *
615mapped_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) 743mapped_new (
744 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
745 void *ptr, size_t cb, cl_event ev,
746 size_t row_pitch, size_t slice_pitch, size_t element_size,
747 size_t width, size_t height, size_t depth
748)
616{ 749{
617 SV *data = newSV (0); 750 SV *data = newSV (0);
618 SvUPGRADE (data, SVt_PVMG); 751 SvUPGRADE (data, SVt_PVMG);
619 752
620 OpenCL__Mapped mapped; 753 OpenCL__Mapped mapped;
621 New (0, mapped, 1, struct mapped); 754 New (0, mapped, 1, struct mapped);
622 755
623 clRetainCommandQueue (queue); 756 clRetainCommandQueue (queue);
624 757
625 mapped->queue = queue; 758 mapped->queue = queue;
626 mapped->memobj = memobj; 759 mapped->memobj = memobj;
627 mapped->ptr = ptr; 760 mapped->ptr = ptr;
628 mapped->cb = cb; 761 mapped->cb = cb;
629 mapped->event = ev; 762 mapped->event = ev;
630 mapped->row_pitch = row_pitch; 763 mapped->row_pitch = row_pitch;
631 mapped->slice_pitch = slice_pitch; 764 mapped->slice_pitch = slice_pitch;
765
766 mapped->element_size = element_size;
767 mapped->width = width;
768 mapped->height = height;
769 mapped->depth = depth;
632 770
633 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0); 771 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
634 772
635 if (SvLEN (data)) 773 if (SvLEN (data))
636 Safefree (data); 774 Safefree (data);
666 804
667 mapped->ptr = 0; 805 mapped->ptr = 0;
668} 806}
669 807
670static void 808static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count) 809mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{ 810{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event); 811 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
674 cl_event ev; 812 cl_event ev;
675 813
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev)); 814 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677 815
678 clReleaseEvent (mapped->event); 816 clReleaseEvent (mapped->event);
679 mapped->event = ev; 817 mapped->event = ev;
680 818
681 mapped_detach (self, mapped); 819 mapped_detach (self, mapped);
682} 820}
683 821
822static size_t
823mapped_element_size (OpenCL__Mapped self)
824{
825 if (!self->element_size)
826 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
827
828 return self->element_size;
829}
830
684/*****************************************************************************/ 831/*****************************************************************************/
685 832
686MODULE = OpenCL PACKAGE = OpenCL 833MODULE = OpenCL PACKAGE = OpenCL
687 834
688PROTOTYPES: ENABLE 835PROTOTYPES: ENABLE
699 eq_signal_arg = (void*)arg; 846 eq_signal_arg = (void*)arg;
700 847
701BOOT: 848BOOT:
702{ 849{
703 HV *stash = gv_stashpv ("OpenCL", 1); 850 HV *stash = gv_stashpv ("OpenCL", 1);
851
704 static const ivstr *civ, const_iv[] = { 852 static const ivstr *civ, const_iv[] = {
705 { sizeof (cl_char ), "SIZEOF_CHAR" }, 853 { sizeof (cl_char ), "SIZEOF_CHAR" },
706 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 854 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
707 { sizeof (cl_short ), "SIZEOF_SHORT" }, 855 { sizeof (cl_short ), "SIZEOF_SHORT" },
708 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 856 { sizeof (cl_ushort), "SIZEOF_USHORT" },
711 { sizeof (cl_long ), "SIZEOF_LONG" }, 859 { sizeof (cl_long ), "SIZEOF_LONG" },
712 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 860 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
713 { sizeof (cl_half ), "SIZEOF_HALF" }, 861 { sizeof (cl_half ), "SIZEOF_HALF" },
714 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 862 { sizeof (cl_float ), "SIZEOF_FLOAT" },
715 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 863 { sizeof (cl_double), "SIZEOF_DOUBLE" },
864 { PREFER_1_1 , "PREFER_1_1" },
716#include "constiv.h" 865#include "constiv.h"
717 }; 866 };
718 867
719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 868 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 869 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721 870
871 static const nvstr *cnv, const_nv[] = {
872#include "constnv.h"
873 };
874
875 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
876 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
877
878 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
879
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 880 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 881 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
882 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 883 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 884 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 885 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 886 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 887 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
769 list = tmpbuf (sizeof (*list) * count); 928 list = tmpbuf (sizeof (*list) * count);
770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 929 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
771 930
772 EXTEND (SP, count); 931 EXTEND (SP, count);
773 for (i = 0; i < count; ++i) 932 for (i = 0; i < count; ++i)
774 PUSH_CLOBJ (stash_platform, (IV)list [i]); 933 PUSH_CLOBJ (stash_platform, list [i]);
775 934
776void 935void
777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 936context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
778 PPCODE: 937 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK; 938 CONTEXT_NOTIFY_CALLBACK;
780 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 939 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
781 XPUSH_CLOBJ_CONTEXT; 940 XPUSH_CLOBJ_CONTEXT;
782 941
783void 942void
784context (FUTURE properties, FUTURE devices, FUTURE notify) 943context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
785 PPCODE: 944 PPCODE:
786 /* der Gipfel der Kunst */ 945 cl_uint device_count;
946 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
947
948 CONTEXT_NOTIFY_CALLBACK;
949 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
950 XPUSH_CLOBJ_CONTEXT;
787 951
788void 952void
789wait_for_events (...) 953wait_for_events (...)
790 CODE: 954 CODE:
791 EVENT_LIST (0); 955 EVENT_LIST (0);
842 EXTEND (SP, count); 1006 EXTEND (SP, count);
843 for (i = 0; i < count; ++i) 1007 for (i = 0; i < count; ++i)
844 PUSH_CLOBJ (stash_device, list [i]); 1008 PUSH_CLOBJ (stash_device, list [i]);
845 1009
846void 1010void
847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 1011context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
848 PPCODE: 1012 PPCODE:
1013 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1014 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1015
1016 cl_uint device_count;
1017 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1018
1019 CONTEXT_NOTIFY_CALLBACK;
1020 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1021 XPUSH_CLOBJ_CONTEXT;
1022
1023void
1024context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1025 PPCODE:
1026 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1027 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1028
1029 CONTEXT_NOTIFY_CALLBACK;
1030 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1031 XPUSH_CLOBJ_CONTEXT;
1032
1033MODULE = OpenCL PACKAGE = OpenCL::Device
1034
1035void
1036info (OpenCL::Device self, cl_device_info name)
1037 PPCODE:
1038 INFO (Device)
1039
1040#if CL_VERSION_1_2
1041
1042void
1043sub_devices (OpenCL::Device self, SV *properties)
1044 PPCODE:
849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1045 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
850 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1046 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
851 1047
852 AV *av = (AV *)SvRV (devices); 1048 properties = SvRV (properties);
853 cl_uint num_devices = av_len (av) + 1; 1049
854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1050 cl_uint count = av_len ((AV *)properties) + 1;
1051 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
855 1052
856 int i; 1053 int i;
857 for (i = num_devices; i--; ) 1054 for (i = 0; i < count; ++i)
858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1055 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
859 1056
860 CONTEXT_NOTIFY_CALLBACK; 1057 props [count] = 0;
861 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
862 XPUSH_CLOBJ_CONTEXT;
863 1058
864void 1059 cl_uint num_devices;
865context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 1060 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
866 PPCODE: 1061 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1062 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
868 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
869 1063
870 CONTEXT_NOTIFY_CALLBACK; 1064 EXTEND (SP, num_devices);
871 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1065 for (i = 0; i < count; ++i)
872 XPUSH_CLOBJ_CONTEXT; 1066 PUSH_CLOBJ (stash_subdevice, list [i]);
873 1067
874MODULE = OpenCL PACKAGE = OpenCL::Device 1068#endif
875
876void
877info (OpenCL::Device self, cl_device_info name)
878 PPCODE:
879 INFO (Device)
880 1069
881#BEGIN:device 1070#BEGIN:device
882 1071
883void 1072void
884type (OpenCL::Device self) 1073type (OpenCL::Device self)
1074 ALIAS:
1075 type = CL_DEVICE_TYPE
1076 address_bits = CL_DEVICE_ADDRESS_BITS
1077 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1078 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1079 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1080 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1081 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1082 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1083 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1084 properties = CL_DEVICE_QUEUE_PROPERTIES
1085 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1086 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
885 PPCODE: 1087 PPCODE:
886 cl_device_type value [1]; 1088 cl_ulong value [1];
887 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1089 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
888 EXTEND (SP, 1); 1090 EXTEND (SP, 1);
889 const int i = 0; 1091 const int i = 0;
890 PUSHs (sv_2mortal (newSVuv (value [i]))); 1092 PUSHs (sv_2mortal (newSVuv (value [i])));
891 1093
892void 1094void
906 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1108 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
907 image_support = CL_DEVICE_IMAGE_SUPPORT 1109 image_support = CL_DEVICE_IMAGE_SUPPORT
908 max_samplers = CL_DEVICE_MAX_SAMPLERS 1110 max_samplers = CL_DEVICE_MAX_SAMPLERS
909 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1111 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
910 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1112 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1113 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
911 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1114 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
912 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1115 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1116 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
913 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1117 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
914 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1118 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
915 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1119 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
916 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1120 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
917 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1121 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
955 EXTEND (SP, n); 1159 EXTEND (SP, n);
956 for (i = 0; i < n; ++i) 1160 for (i = 0; i < n; ++i)
957 PUSHs (sv_2mortal (newSVuv (value [i]))); 1161 PUSHs (sv_2mortal (newSVuv (value [i])));
958 1162
959void 1163void
960address_bits (OpenCL::Device self)
961 PPCODE:
962 cl_bitfield value [1];
963 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
964 EXTEND (SP, 1);
965 const int i = 0;
966 PUSHs (sv_2mortal (newSVuv (value [i])));
967
968void
969max_mem_alloc_size (OpenCL::Device self)
970 ALIAS:
971 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
972 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
973 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
974 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
975 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
976 PPCODE:
977 cl_ulong value [1];
978 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
979 EXTEND (SP, 1);
980 const int i = 0;
981 PUSHs (sv_2mortal (newSVuv (value [i])));
982
983void
984single_fp_config (OpenCL::Device self)
985 ALIAS:
986 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
987 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
988 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
989 PPCODE:
990 cl_device_fp_config value [1];
991 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
992 EXTEND (SP, 1);
993 const int i = 0;
994 PUSHs (sv_2mortal (newSVuv (value [i])));
995
996void
997global_mem_cache_type (OpenCL::Device self)
998 PPCODE:
999 cl_device_mem_cache_type value [1];
1000 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
1001 EXTEND (SP, 1);
1002 const int i = 0;
1003 PUSHs (sv_2mortal (newSVuv (value [i])));
1004
1005void
1006local_mem_type (OpenCL::Device self)
1007 PPCODE:
1008 cl_device_local_mem_type value [1];
1009 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
1010 EXTEND (SP, 1);
1011 const int i = 0;
1012 PUSHs (sv_2mortal (newSVuv (value [i])));
1013
1014void
1015error_correction_support (OpenCL::Device self) 1164error_correction_support (OpenCL::Device self)
1016 ALIAS: 1165 ALIAS:
1017 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1166 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1018 endian_little = CL_DEVICE_ENDIAN_LITTLE 1167 endian_little = CL_DEVICE_ENDIAN_LITTLE
1019 available = CL_DEVICE_AVAILABLE 1168 available = CL_DEVICE_AVAILABLE
1020 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1169 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1021 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1170 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1022 PPCODE: 1171 PPCODE:
1023 cl_bool value [1]; 1172 cl_uint value [1];
1024 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1173 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1025 EXTEND (SP, 1); 1174 EXTEND (SP, 1);
1026 const int i = 0; 1175 const int i = 0;
1027 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1176 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1028
1029void
1030execution_capabilities (OpenCL::Device self)
1031 PPCODE:
1032 cl_device_exec_capabilities value [1];
1033 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
1034 EXTEND (SP, 1);
1035 const int i = 0;
1036 PUSHs (sv_2mortal (newSVuv (value [i])));
1037
1038void
1039properties (OpenCL::Device self)
1040 PPCODE:
1041 cl_command_queue_properties value [1];
1042 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
1043 EXTEND (SP, 1);
1044 const int i = 0;
1045 PUSHs (sv_2mortal (newSVuv (value [i])));
1046 1177
1047void 1178void
1048platform (OpenCL::Device self) 1179platform (OpenCL::Device self)
1049 PPCODE: 1180 PPCODE:
1050 cl_platform_id value [1]; 1181 cl_platform_id value [1];
1051 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1182 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1052 EXTEND (SP, 1); 1183 EXTEND (SP, 1);
1053 const int i = 0; 1184 const int i = 0;
1054 {
1055 PUSH_CLOBJ (stash_platform, value [i]); 1185 PUSH_CLOBJ (stash_platform, value [i]);
1056 }
1057 1186
1058void 1187void
1059name (OpenCL::Device self) 1188name (OpenCL::Device self)
1060 ALIAS: 1189 ALIAS:
1061 name = CL_DEVICE_NAME 1190 name = CL_DEVICE_NAME
1078 PPCODE: 1207 PPCODE:
1079 cl_device_id value [1]; 1208 cl_device_id value [1];
1080 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1209 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1081 EXTEND (SP, 1); 1210 EXTEND (SP, 1);
1082 const int i = 0; 1211 const int i = 0;
1083 {
1084 PUSH_CLOBJ (stash_device, value [i]); 1212 PUSH_CLOBJ (stash_device, value [i]);
1085 }
1086 1213
1087void 1214void
1088partition_types_ext (OpenCL::Device self) 1215partition_types_ext (OpenCL::Device self)
1089 ALIAS: 1216 ALIAS:
1090 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1217 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1100 for (i = 0; i < n; ++i) 1227 for (i = 0; i < n; ++i)
1101 PUSHs (sv_2mortal (newSVuv (value [i]))); 1228 PUSHs (sv_2mortal (newSVuv (value [i])));
1102 1229
1103#END:device 1230#END:device
1104 1231
1232MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1233
1234#if CL_VERSION_1_2
1235
1236void
1237DESTROY (OpenCL::SubDevice self)
1238 CODE:
1239 clReleaseDevice (self);
1240
1241#endif
1242
1105MODULE = OpenCL PACKAGE = OpenCL::Context 1243MODULE = OpenCL PACKAGE = OpenCL::Context
1106 1244
1107void 1245void
1108DESTROY (OpenCL::Context context) 1246DESTROY (OpenCL::Context self)
1109 CODE: 1247 CODE:
1110 clReleaseContext (context); 1248 clReleaseContext (self);
1111 1249
1112void 1250void
1113info (OpenCL::Context self, cl_context_info name) 1251info (OpenCL::Context self, cl_context_info name)
1114 PPCODE: 1252 PPCODE:
1115 INFO (Context) 1253 INFO (Context)
1156 const cl_image_desc desc = { 1294 const cl_image_desc desc = {
1157 type, 1295 type,
1158 width, height, depth, 1296 width, height, depth,
1159 array_size, row_pitch, slice_pitch, 1297 array_size, row_pitch, slice_pitch,
1160 num_mip_level, num_samples, 1298 num_mip_level, num_samples,
1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1299 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1162 }; 1300 };
1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1301 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1164 HV *stash = stash_image; 1302 HV *stash = stash_image;
1165 switch (type) 1303 switch (type)
1166 { 1304 {
1299 1437
1300 len2 = len; 1438 len2 = len;
1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1439 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1302 XPUSH_CLOBJ (stash_program, prog); 1440 XPUSH_CLOBJ (stash_program, prog);
1303 1441
1442void
1443program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1444 PPCODE:
1445 cl_uint device_count;
1446 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1447
1448 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1449 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1450
1451 binaries = SvRV (binaries);
1452
1453 if (device_count != av_len ((AV *)binaries) + 1)
1454 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1455
1456 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1457 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1458 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1459
1460 int i;
1461 for (i = 0; i < device_count; ++i)
1462 {
1463 STRLEN len;
1464 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1465 length_list [i] = len;
1466 }
1467
1468 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1469 length_list, binary_list,
1470 GIMME_V == G_ARRAY ? status_list : 0, &res));
1471
1472 EXTEND (SP, 2);
1473 PUSH_CLOBJ (stash_program, prog);
1474
1475 if (GIMME_V == G_ARRAY)
1476 {
1477 AV *av = newAV ();
1478 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1479
1480 for (i = device_count; i--; )
1481 av_store (av, i, newSViv (status_list [i]));
1482 }
1483
1484#if CL_VERSION_1_2
1485
1486void
1487program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1488 PPCODE:
1489 cl_uint device_count;
1490 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1491
1492 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1493
1494 XPUSH_CLOBJ (stash_program, prog);
1495
1496void
1497link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1498 CODE:
1499 cl_uint device_count = 0;
1500 cl_device_id *device_list = 0;
1501
1502 if (SvOK (devices))
1503 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1504
1505 cl_uint program_count;
1506 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1507
1508 void *user_data;
1509 program_callback pfn_notify = make_program_callback (notify, &user_data);
1510
1511 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1512 program_count, program_list, pfn_notify, user_data, &res));
1513
1514 XPUSH_CLOBJ (stash_program, prog);
1515
1516#endif
1517
1304#BEGIN:context 1518#BEGIN:context
1305 1519
1306void 1520void
1307reference_count (OpenCL::Context self) 1521reference_count (OpenCL::Context self)
1308 ALIAS: 1522 ALIAS:
1323 cl_device_id *value = tmpbuf (size); 1537 cl_device_id *value = tmpbuf (size);
1324 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1538 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1325 int i, n = size / sizeof (*value); 1539 int i, n = size / sizeof (*value);
1326 EXTEND (SP, n); 1540 EXTEND (SP, n);
1327 for (i = 0; i < n; ++i) 1541 for (i = 0; i < n; ++i)
1328 {
1329 PUSH_CLOBJ (stash_device, value [i]); 1542 PUSH_CLOBJ (stash_device, value [i]);
1330 }
1331 1543
1332void 1544void
1333properties (OpenCL::Context self) 1545properties (OpenCL::Context self)
1334 PPCODE: 1546 PPCODE:
1335 size_t size; 1547 size_t size;
1353void 1565void
1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1566read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS: 1567 ALIAS:
1356 enqueue_read_buffer = 0 1568 enqueue_read_buffer = 0
1357 PPCODE: 1569 PPCODE:
1358 cl_event ev = 0;
1359 EVENT_LIST (6); 1570 EVENT_LIST (6);
1360 1571
1361 SvUPGRADE (data, SVt_PV); 1572 SvUPGRADE (data, SVt_PV);
1362 SvGROW (data, len); 1573 SvGROW (data, len);
1363 SvPOK_only (data); 1574 SvPOK_only (data);
1364 SvCUR_set (data, len); 1575 SvCUR_set (data, len);
1576
1577 cl_event ev = 0;
1365 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1578 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1366 1579
1367 if (ev) 1580 if (ev)
1368 XPUSH_CLOBJ (stash_event, ev); 1581 XPUSH_CLOBJ (stash_event, ev);
1369 1582
1370void 1583void
1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1584write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS: 1585 ALIAS:
1373 enqueue_write_buffer = 0 1586 enqueue_write_buffer = 0
1374 PPCODE: 1587 PPCODE:
1375 cl_event ev = 0; 1588 EVENT_LIST (5);
1589
1376 STRLEN len; 1590 STRLEN len;
1377 char *ptr = SvPVbyte (data, len); 1591 char *ptr = SvPVbyte (data, len);
1378 EVENT_LIST (5);
1379 1592
1593 cl_event ev = 0;
1380 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1594 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1381 1595
1382 if (ev) 1596 if (ev)
1383 XPUSH_CLOBJ (stash_event, ev); 1597 XPUSH_CLOBJ (stash_event, ev);
1384 1598
1387void 1601void
1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1602fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS: 1603 ALIAS:
1390 enqueue_fill_buffer = 0 1604 enqueue_fill_buffer = 0
1391 PPCODE: 1605 PPCODE:
1392 cl_event ev = 0; 1606 EVENT_LIST (5);
1607
1393 STRLEN len; 1608 STRLEN len;
1394 char *ptr = SvPVbyte (data, len); 1609 char *ptr = SvPVbyte (data, len);
1395 EVENT_LIST (5);
1396 1610
1611 cl_event ev = 0;
1397 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1612 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1398 1613
1399 if (ev) 1614 if (ev)
1400 XPUSH_CLOBJ (stash_event, ev); 1615 XPUSH_CLOBJ (stash_event, ev);
1401 1616
1402void 1617void
1403fill_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, ...) 1618fill_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, ...)
1404 ALIAS: 1619 ALIAS:
1405 enqueue_fill_image = 0 1620 enqueue_fill_image = 0
1406 PPCODE: 1621 PPCODE:
1407 cl_event ev = 0; 1622 EVENT_LIST (12);
1408 STRLEN len; 1623
1409 const size_t origin [3] = { x, y, z }; 1624 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth }; 1625 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412 1626
1413 const cl_float c_f [4] = { r, g, b, a }; 1627 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a }; 1628 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a }; 1629 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1630 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1633 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1634 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1421 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1635 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1422 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1636 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423 1637
1638 cl_event ev = 0;
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1639 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1425 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1640 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426 1641
1427 if (ev) 1642 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev); 1643 XPUSH_CLOBJ (stash_event, ev);
1432void 1647void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1648copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS: 1649 ALIAS:
1435 enqueue_copy_buffer = 0 1650 enqueue_copy_buffer = 0
1436 PPCODE: 1651 PPCODE:
1652 EVENT_LIST (6);
1653
1437 cl_event ev = 0; 1654 cl_event ev = 0;
1438 EVENT_LIST (6);
1439
1440 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1655 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1656
1442 if (ev) 1657 if (ev)
1443 XPUSH_CLOBJ (stash_event, ev); 1658 XPUSH_CLOBJ (stash_event, ev);
1444 1659
1445void 1660void
1446read_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, ...) 1661read_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, ...)
1447 ALIAS: 1662 ALIAS:
1448 enqueue_read_buffer_rect = 0 1663 enqueue_read_buffer_rect = 0
1449 PPCODE: 1664 PPCODE:
1450 cl_event ev = 0; 1665 EVENT_LIST (17);
1666
1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1667 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1452 const size_t host_origin[3] = { host_x, host_y, host_z }; 1668 const size_t host_origin[3] = { host_x, host_y, host_z };
1453 const size_t region[3] = { width, height, depth }; 1669 const size_t region[3] = { width, height, depth };
1454 EVENT_LIST (17);
1455 1670
1456 if (!buf_row_pitch) 1671 if (!buf_row_pitch)
1457 buf_row_pitch = region [0]; 1672 buf_row_pitch = region [0];
1458 1673
1459 if (!buf_slice_pitch) 1674 if (!buf_slice_pitch)
1469 1684
1470 SvUPGRADE (data, SVt_PV); 1685 SvUPGRADE (data, SVt_PV);
1471 SvGROW (data, len); 1686 SvGROW (data, len);
1472 SvPOK_only (data); 1687 SvPOK_only (data);
1473 SvCUR_set (data, len); 1688 SvCUR_set (data, len);
1689
1690 cl_event ev = 0;
1474 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)); 1691 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));
1475 1692
1476 if (ev) 1693 if (ev)
1477 XPUSH_CLOBJ (stash_event, ev); 1694 XPUSH_CLOBJ (stash_event, ev);
1478 1695
1479void 1696void
1480write_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, ...) 1697write_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, ...)
1481 ALIAS: 1698 ALIAS:
1482 enqueue_write_buffer_rect = 0 1699 enqueue_write_buffer_rect = 0
1483 PPCODE: 1700 PPCODE:
1484 cl_event ev = 0; 1701 EVENT_LIST (17);
1702
1485 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 };
1486 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 };
1487 const size_t region[3] = { width, height, depth }; 1705 const size_t region[3] = { width, height, depth };
1488 STRLEN len; 1706 STRLEN len;
1489 char *ptr = SvPVbyte (data, len); 1707 char *ptr = SvPVbyte (data, len);
1490 EVENT_LIST (17);
1491 1708
1492 if (!buf_row_pitch) 1709 if (!buf_row_pitch)
1493 buf_row_pitch = region [0]; 1710 buf_row_pitch = region [0];
1494 1711
1495 if (!buf_slice_pitch) 1712 if (!buf_slice_pitch)
1504 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1721 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1505 1722
1506 if (len < min_len) 1723 if (len < min_len)
1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1724 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1508 1725
1726 cl_event ev = 0;
1509 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)); 1727 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));
1510 1728
1511 if (ev) 1729 if (ev)
1512 XPUSH_CLOBJ (stash_event, ev); 1730 XPUSH_CLOBJ (stash_event, ev);
1513 1731
1514void 1732void
1515copy_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, ...) 1733copy_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, ...)
1516 ALIAS: 1734 ALIAS:
1517 enqueue_copy_buffer_rect = 0 1735 enqueue_copy_buffer_rect = 0
1518 PPCODE: 1736 PPCODE:
1519 cl_event ev = 0; 1737 EVENT_LIST (16);
1738
1520 const size_t src_origin[3] = { src_x, src_y, src_z }; 1739 const size_t src_origin[3] = { src_x, src_y, src_z };
1521 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1740 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1522 const size_t region[3] = { width, height, depth }; 1741 const size_t region[3] = { width, height, depth };
1523 EVENT_LIST (16);
1524 1742
1743 cl_event ev = 0;
1525 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)); 1744 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));
1526 1745
1527 if (ev) 1746 if (ev)
1528 XPUSH_CLOBJ (stash_event, ev); 1747 XPUSH_CLOBJ (stash_event, ev);
1529 1748
1530void 1749void
1531read_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, ...) 1750read_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, ...)
1532 ALIAS: 1751 ALIAS:
1533 enqueue_read_image = 0 1752 enqueue_read_image = 0
1534 PPCODE: 1753 PPCODE:
1535 cl_event ev = 0; 1754 EVENT_LIST (12);
1755
1536 const size_t src_origin[3] = { src_x, src_y, src_z }; 1756 const size_t src_origin[3] = { src_x, src_y, src_z };
1537 const size_t region[3] = { width, height, depth }; 1757 const size_t region[3] = { width, height, depth };
1538 EVENT_LIST (12);
1539 1758
1540 if (!row_pitch) 1759 if (!row_pitch)
1541 row_pitch = img_row_pitch (src); 1760 row_pitch = img_row_pitch (src);
1542 1761
1543 if (depth > 1 && !slice_pitch) 1762 if (depth > 1 && !slice_pitch)
1547 1766
1548 SvUPGRADE (data, SVt_PV); 1767 SvUPGRADE (data, SVt_PV);
1549 SvGROW (data, len); 1768 SvGROW (data, len);
1550 SvPOK_only (data); 1769 SvPOK_only (data);
1551 SvCUR_set (data, len); 1770 SvCUR_set (data, len);
1771
1772 cl_event ev = 0;
1552 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)); 1773 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));
1553 1774
1554 if (ev) 1775 if (ev)
1555 XPUSH_CLOBJ (stash_event, ev); 1776 XPUSH_CLOBJ (stash_event, ev);
1556 1777
1557void 1778void
1558write_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, ...) 1779write_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, ...)
1559 ALIAS: 1780 ALIAS:
1560 enqueue_write_image = 0 1781 enqueue_write_image = 0
1561 PPCODE: 1782 PPCODE:
1562 cl_event ev = 0; 1783 EVENT_LIST (12);
1784
1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1785 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1564 const size_t region[3] = { width, height, depth }; 1786 const size_t region[3] = { width, height, depth };
1565 STRLEN len; 1787 STRLEN len;
1566 char *ptr = SvPVbyte (data, len); 1788 char *ptr = SvPVbyte (data, len);
1567 EVENT_LIST (12);
1568 1789
1569 if (!row_pitch) 1790 if (!row_pitch)
1570 row_pitch = img_row_pitch (dst); 1791 row_pitch = img_row_pitch (dst);
1571 1792
1572 if (depth > 1 && !slice_pitch) 1793 if (depth > 1 && !slice_pitch)
1575 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1796 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1576 1797
1577 if (len < min_len) 1798 if (len < min_len)
1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1799 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1579 1800
1801 cl_event ev = 0;
1580 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)); 1802 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));
1581 1803
1582 if (ev) 1804 if (ev)
1583 XPUSH_CLOBJ (stash_event, ev); 1805 XPUSH_CLOBJ (stash_event, ev);
1584 1806
1585void 1807void
1586copy_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, ...) 1808copy_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, ...)
1587 ALIAS: 1809 ALIAS:
1588 enqueue_copy_image = 0 1810 enqueue_copy_image = 0
1589 PPCODE: 1811 PPCODE:
1590 cl_event ev = 0; 1812 EVENT_LIST (12);
1813
1591 const size_t src_origin[3] = { src_x, src_y, src_z }; 1814 const size_t src_origin[3] = { src_x, src_y, src_z };
1592 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1815 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1593 const size_t region[3] = { width, height, depth }; 1816 const size_t region[3] = { width, height, depth };
1594 EVENT_LIST (12);
1595 1817
1818 cl_event ev = 0;
1596 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1819 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1597 1820
1598 if (ev) 1821 if (ev)
1599 XPUSH_CLOBJ (stash_event, ev); 1822 XPUSH_CLOBJ (stash_event, ev);
1600 1823
1601void 1824void
1602copy_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, ...) 1825copy_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, ...)
1603 ALIAS: 1826 ALIAS:
1604 enqueue_copy_image_to_buffer = 0 1827 enqueue_copy_image_to_buffer = 0
1605 PPCODE: 1828 PPCODE:
1606 cl_event ev = 0; 1829 EVENT_LIST (10);
1830
1607 const size_t src_origin[3] = { src_x, src_y, src_z }; 1831 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth }; 1832 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610 1833
1834 cl_event ev = 0;
1611 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1835 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612 1836
1613 if (ev) 1837 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev); 1838 XPUSH_CLOBJ (stash_event, ev);
1615 1839
1616void 1840void
1617copy_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, ...) 1841copy_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, ...)
1618 ALIAS: 1842 ALIAS:
1619 enqueue_copy_buffer_to_image = 0 1843 enqueue_copy_buffer_to_image = 0
1620 PPCODE: 1844 PPCODE:
1621 cl_event ev = 0; 1845 EVENT_LIST (10);
1846
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1847 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth }; 1848 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625 1849
1850 cl_event ev = 0;
1626 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1851 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627 1852
1628 if (ev) 1853 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev); 1854 XPUSH_CLOBJ (stash_event, ev);
1630 1855
1631void 1856void
1632map_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, size_t cb = 0, ...) 1857map_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, ...)
1633 ALIAS: 1858 ALIAS:
1634 enqueue_map_buffer = 0 1859 enqueue_map_buffer = 0
1635 PPCODE: 1860 PPCODE:
1861 EVENT_LIST (6);
1862
1863 size_t cb = SvIV (cb_);
1864
1865 if (!SvOK (cb_))
1866 {
1867 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1868 cb -= offset;
1869 }
1870
1636 cl_event ev; 1871 cl_event ev;
1637 EVENT_LIST (6);
1638
1639 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res)); 1872 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1640 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0)); 1873 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1641 1874
1642void 1875void
1643map_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, size_t width = 0, size_t height = 0, size_t depth = 0, ...) 1876map_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, ...)
1644 ALIAS: 1877 ALIAS:
1645 enqueue_map_image = 0 1878 enqueue_map_image = 0
1646 PPCODE: 1879 PPCODE:
1647 cl_event ev; 1880 size_t width = SvIV (width_);
1881 if (!SvOK (width_))
1882 {
1883 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1884 width -= x;
1885 }
1886
1887 size_t height = SvIV (width_);
1888 if (!SvOK (height_))
1889 {
1890 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1891 height -= y;
1892
1893 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1894 if (!height)
1895 height = 1;
1896 }
1897
1898 size_t depth = SvIV (width_);
1899 if (!SvOK (depth_))
1900 {
1901 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1902 depth -= z;
1903
1904 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1905 if (!depth)
1906 depth = 1;
1907 }
1908
1648 const size_t origin[3] = { x, y, z }; 1909 const size_t origin[3] = { x, y, z };
1649 const size_t region[3] = { width, height, depth }; 1910 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch; 1911 size_t row_pitch, slice_pitch;
1651 EVENT_LIST (10); 1912 EVENT_LIST (10);
1652 1913
1914 cl_event ev;
1653 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res)); 1915 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));
1654 1916
1655 size_t cb = slice_pitch ? slice_pitch * region [2] 1917 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1] 1918 : row_pitch ? row_pitch * region [1]
1657 : region [0]; 1919 : region [0];
1658 1920
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch)); 1921 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1660 1922
1661void 1923void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1924unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE: 1925 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2); 1926 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID) 1927 if (GIMME_V != G_VOID)
1666 { 1928 {
1667 clRetainEvent (mapped->event); 1929 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event); 1930 XPUSH_CLOBJ (stash_event, mapped->event);
1669 } 1931 }
1671void 1933void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1934task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS: 1935 ALIAS:
1674 enqueue_task = 0 1936 enqueue_task = 0
1675 PPCODE: 1937 PPCODE:
1938 EVENT_LIST (2);
1939
1676 cl_event ev = 0; 1940 cl_event ev = 0;
1677 EVENT_LIST (2);
1678
1679 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1941 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1680 1942
1681 if (ev) 1943 if (ev)
1682 XPUSH_CLOBJ (stash_event, ev); 1944 XPUSH_CLOBJ (stash_event, ev);
1683 1945
1684void 1946void
1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1947nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1686 ALIAS: 1948 ALIAS:
1687 enqueue_nd_range_kernel = 0 1949 enqueue_nd_range_kernel = 0
1688 PPCODE: 1950 PPCODE:
1689 cl_event ev = 0; 1951 EVENT_LIST (5);
1952
1690 size_t *gwo = 0, *gws, *lws = 0; 1953 size_t *gwo = 0, *gws, *lws = 0;
1691 int gws_len; 1954 int gws_len;
1692 size_t *lists; 1955 size_t *lists;
1693 int i; 1956 int i;
1694 EVENT_LIST (5);
1695 1957
1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1958 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1959 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1698 1960
1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1961 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1738 if (!lws [i]) 2000 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 2001 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 } 2002 }
1741 } 2003 }
1742 2004
2005 cl_event ev = 0;
1743 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2006 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1744 2007
1745 if (ev) 2008 if (ev)
1746 XPUSH_CLOBJ (stash_event, ev); 2009 XPUSH_CLOBJ (stash_event, ev);
2010
2011#if CL_VERSION_1_2
2012
2013void
2014migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2015 ALIAS:
2016 enqueue_migrate_mem_objects = 0
2017 PPCODE:
2018 EVENT_LIST (3);
2019
2020 cl_uint obj_count;
2021 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2022
2023 cl_event ev = 0;
2024 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2025
2026 if (ev)
2027 XPUSH_CLOBJ (stash_event, ev);
2028
2029#endif
1747 2030
1748#if cl_apple_gl_sharing || cl_khr_gl_sharing 2031#if cl_apple_gl_sharing || cl_khr_gl_sharing
1749 2032
1750void 2033void
1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2034acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1752 ALIAS: 2035 ALIAS:
1753 release_gl_objects = 1 2036 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0 2037 enqueue_acquire_gl_objects = 0
1755 enqueue_release_gl_objects = 1 2038 enqueue_release_gl_objects = 1
1756 PPCODE: 2039 PPCODE:
1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2040 EVENT_LIST (2);
1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2041
2042 cl_uint obj_count;
2043 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1759 2044
1760 cl_event ev = 0; 2045 cl_event ev = 0;
1761 EVENT_LIST (2);
1762 AV *av = (AV *)SvRV (objects);
1763 cl_uint num_objects = av_len (av) + 1;
1764 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1765 int i;
1766
1767 for (i = num_objects; i--; )
1768 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1769 2046
1770 if (ix) 2047 if (ix)
1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2048 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1772 else 2049 else
1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2050 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1774 2051
1775 if (ev) 2052 if (ev)
1776 XPUSH_CLOBJ (stash_event, ev); 2053 XPUSH_CLOBJ (stash_event, ev);
1777 2054
1778#endif 2055#endif
1792void 2069void
1793marker (OpenCL::Queue self, ...) 2070marker (OpenCL::Queue self, ...)
1794 ALIAS: 2071 ALIAS:
1795 enqueue_marker = 0 2072 enqueue_marker = 0
1796 PPCODE: 2073 PPCODE:
2074 EVENT_LIST (1);
1797 cl_event ev = 0; 2075 cl_event ev = 0;
1798 EVENT_LIST (1);
1799#if PREFER_1_1 2076#if PREFER_1_1
1800 if (!event_list_count) 2077 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2078 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else 2079 else
1803#if CL_VERSION_1_2 2080#if CL_VERSION_1_2
1817void 2094void
1818barrier (OpenCL::Queue self, ...) 2095barrier (OpenCL::Queue self, ...)
1819 ALIAS: 2096 ALIAS:
1820 enqueue_barrier = 0 2097 enqueue_barrier = 0
1821 PPCODE: 2098 PPCODE:
2099 EVENT_LIST (1);
1822 cl_event ev = 0; 2100 cl_event ev = 0;
1823 EVENT_LIST (1);
1824#if PREFER_1_1 2101#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID) 2102 if (!event_list_count && GIMME_V == G_VOID)
1826 NEED_SUCCESS (EnqueueBarrier, (self)); 2103 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else 2104 else
1828#if CL_VERSION_1_2 2105#if CL_VERSION_1_2
1864 PPCODE: 2141 PPCODE:
1865 cl_context value [1]; 2142 cl_context value [1];
1866 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2143 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1867 EXTEND (SP, 1); 2144 EXTEND (SP, 1);
1868 const int i = 0; 2145 const int i = 0;
1869 {
1870 NEED_SUCCESS (RetainContext, (value [i])); 2146 NEED_SUCCESS (RetainContext, (value [i]));
1871 PUSH_CLOBJ (stash_context, value [i]); 2147 PUSH_CLOBJ (stash_context, value [i]);
1872 }
1873 2148
1874void 2149void
1875device (OpenCL::Queue self) 2150device (OpenCL::Queue self)
1876 PPCODE: 2151 PPCODE:
1877 cl_device_id value [1]; 2152 cl_device_id value [1];
1878 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2153 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1879 EXTEND (SP, 1); 2154 EXTEND (SP, 1);
1880 const int i = 0; 2155 const int i = 0;
1881 {
1882 PUSH_CLOBJ (stash_device, value [i]); 2156 PUSH_CLOBJ (stash_device, value [i]);
1883 }
1884 2157
1885void 2158void
1886reference_count (OpenCL::Queue self) 2159reference_count (OpenCL::Queue self)
1887 PPCODE: 2160 PPCODE:
1888 cl_uint value [1]; 2161 cl_uint value [1];
1892 PUSHs (sv_2mortal (newSVuv (value [i]))); 2165 PUSHs (sv_2mortal (newSVuv (value [i])));
1893 2166
1894void 2167void
1895properties (OpenCL::Queue self) 2168properties (OpenCL::Queue self)
1896 PPCODE: 2169 PPCODE:
1897 cl_command_queue_properties value [1]; 2170 cl_ulong value [1];
1898 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2171 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1899 EXTEND (SP, 1); 2172 EXTEND (SP, 1);
1900 const int i = 0; 2173 const int i = 0;
1901 PUSHs (sv_2mortal (newSVuv (value [i]))); 2174 PUSHs (sv_2mortal (newSVuv (value [i])));
1902 2175
1912void 2185void
1913info (OpenCL::Memory self, cl_mem_info name) 2186info (OpenCL::Memory self, cl_mem_info name)
1914 PPCODE: 2187 PPCODE:
1915 INFO (MemObject) 2188 INFO (MemObject)
1916 2189
2190void
2191destructor_callback (OpenCL::Memory self, SV *notify)
2192 PPCODE:
2193 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2194
1917#BEGIN:mem 2195#BEGIN:mem
1918 2196
1919void 2197void
1920type (OpenCL::Memory self) 2198type (OpenCL::Memory self)
2199 ALIAS:
2200 type = CL_MEM_TYPE
2201 map_count = CL_MEM_MAP_COUNT
2202 reference_count = CL_MEM_REFERENCE_COUNT
1921 PPCODE: 2203 PPCODE:
1922 cl_mem_object_type value [1]; 2204 cl_uint value [1];
1923 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2205 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1924 EXTEND (SP, 1); 2206 EXTEND (SP, 1);
1925 const int i = 0; 2207 const int i = 0;
1926 PUSHs (sv_2mortal (newSVuv (value [i]))); 2208 PUSHs (sv_2mortal (newSVuv (value [i])));
1927 2209
1928void 2210void
1929flags (OpenCL::Memory self) 2211flags (OpenCL::Memory self)
1930 PPCODE: 2212 PPCODE:
1931 cl_mem_flags value [1]; 2213 cl_ulong value [1];
1932 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2214 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1933 EXTEND (SP, 1); 2215 EXTEND (SP, 1);
1934 const int i = 0; 2216 const int i = 0;
1935 PUSHs (sv_2mortal (newSVuv (value [i]))); 2217 PUSHs (sv_2mortal (newSVuv (value [i])));
1936 2218
1954 EXTEND (SP, 1); 2236 EXTEND (SP, 1);
1955 const int i = 0; 2237 const int i = 0;
1956 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2238 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1957 2239
1958void 2240void
1959map_count (OpenCL::Memory self)
1960 ALIAS:
1961 map_count = CL_MEM_MAP_COUNT
1962 reference_count = CL_MEM_REFERENCE_COUNT
1963 PPCODE:
1964 cl_uint value [1];
1965 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1966 EXTEND (SP, 1);
1967 const int i = 0;
1968 PUSHs (sv_2mortal (newSVuv (value [i])));
1969
1970void
1971context (OpenCL::Memory self) 2241context (OpenCL::Memory self)
1972 PPCODE: 2242 PPCODE:
1973 cl_context value [1]; 2243 cl_context value [1];
1974 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2244 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1975 EXTEND (SP, 1); 2245 EXTEND (SP, 1);
1976 const int i = 0; 2246 const int i = 0;
1977 {
1978 NEED_SUCCESS (RetainContext, (value [i])); 2247 NEED_SUCCESS (RetainContext, (value [i]));
1979 PUSH_CLOBJ (stash_context, value [i]); 2248 PUSH_CLOBJ (stash_context, value [i]);
1980 }
1981 2249
1982void 2250void
1983associated_memobject (OpenCL::Memory self) 2251associated_memobject (OpenCL::Memory self)
1984 PPCODE: 2252 PPCODE:
1985 cl_mem value [1]; 2253 cl_mem value [1];
1986 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2254 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1987 EXTEND (SP, 1); 2255 EXTEND (SP, 1);
1988 const int i = 0; 2256 const int i = 0;
1989 {
1990 NEED_SUCCESS (RetainMemObject, (value [i])); 2257 NEED_SUCCESS (RetainMemObject, (value [i]));
1991 PUSH_CLOBJ (stash_memory, value [i]); 2258 PUSH_CLOBJ (stash_memory, value [i]);
1992 }
1993 2259
1994#END:mem 2260#END:mem
1995 2261
1996#if cl_apple_gl_sharing || cl_khr_gl_sharing 2262#if cl_apple_gl_sharing || cl_khr_gl_sharing
1997 2263
2096 2362
2097#BEGIN:sampler 2363#BEGIN:sampler
2098 2364
2099void 2365void
2100reference_count (OpenCL::Sampler self) 2366reference_count (OpenCL::Sampler self)
2367 ALIAS:
2368 reference_count = CL_SAMPLER_REFERENCE_COUNT
2369 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2370 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
2101 PPCODE: 2371 PPCODE:
2102 cl_uint value [1]; 2372 cl_uint value [1];
2103 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2373 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
2104 EXTEND (SP, 1); 2374 EXTEND (SP, 1);
2105 const int i = 0; 2375 const int i = 0;
2106 PUSHs (sv_2mortal (newSVuv (value [i]))); 2376 PUSHs (sv_2mortal (newSVuv (value [i])));
2107 2377
2108void 2378void
2110 PPCODE: 2380 PPCODE:
2111 cl_context value [1]; 2381 cl_context value [1];
2112 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2382 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2113 EXTEND (SP, 1); 2383 EXTEND (SP, 1);
2114 const int i = 0; 2384 const int i = 0;
2115 {
2116 NEED_SUCCESS (RetainContext, (value [i])); 2385 NEED_SUCCESS (RetainContext, (value [i]));
2117 PUSH_CLOBJ (stash_context, value [i]); 2386 PUSH_CLOBJ (stash_context, value [i]);
2118 }
2119
2120void
2121normalized_coords (OpenCL::Sampler self)
2122 PPCODE:
2123 cl_addressing_mode value [1];
2124 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
2125 EXTEND (SP, 1);
2126 const int i = 0;
2127 PUSHs (sv_2mortal (newSVuv (value [i])));
2128
2129void
2130addressing_mode (OpenCL::Sampler self)
2131 PPCODE:
2132 cl_filter_mode value [1];
2133 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
2134 EXTEND (SP, 1);
2135 const int i = 0;
2136 PUSHs (sv_2mortal (newSVuv (value [i])));
2137 2387
2138void 2388void
2139filter_mode (OpenCL::Sampler self) 2389filter_mode (OpenCL::Sampler self)
2140 PPCODE: 2390 PPCODE:
2141 cl_bool value [1]; 2391 cl_uint value [1];
2142 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2392 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2143 EXTEND (SP, 1); 2393 EXTEND (SP, 1);
2144 const int i = 0; 2394 const int i = 0;
2145 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2395 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2146 2396
2156void 2406void
2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2407build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS: 2408 ALIAS:
2159 build_async = 1 2409 build_async = 1
2160 CODE: 2410 CODE:
2161 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2162 void *user_data = 0;
2163 cl_uint num_devices = 0; 2411 cl_uint device_count = 0;
2164 cl_device_id *device_list = 0; 2412 cl_device_id *device_list = 0;
2165 2413
2166 if (SvOK (devices)) 2414 if (SvOK (devices))
2167 { 2415 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2416
2417 void *user_data;
2418 program_callback pfn_notify = make_program_callback (notify, &user_data);
2419
2420 if (ix)
2421 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2422 else
2423 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2424
2425#if CL_VERSION_1_2
2426
2427void
2428compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2429 CODE:
2430 cl_uint device_count = 0;
2431 cl_device_id *device_list = 0;
2432
2433 if (SvOK (devices))
2434 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2435
2436 cl_uint header_count = 0;
2437 cl_program *header_list = 0;
2438 const char **header_name = 0;
2439
2440 if (SvOK (headers))
2441 {
2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2442 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2443 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2170 2444
2171 AV *av = (AV *)SvRV (devices); 2445 HV *hv = (HV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173 2446
2174 if (num_devices) 2447 header_count = hv_iterinit (hv);
2448 header_list = tmpbuf (sizeof (*header_list) * header_count);
2449 header_name = tmpbuf (sizeof (*header_name) * header_count);
2450
2451 HE *he;
2452 int i = 0;
2453 while (he = hv_iternext (hv))
2175 { 2454 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2455 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2456 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2177 int count; 2457 ++i;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 } 2458 }
2181 } 2459 }
2182 2460
2183 if (SvOK (notify)) 2461 void *user_data;
2184 { 2462 program_callback pfn_notify = make_program_callback (notify, &user_data);
2185 NEED_SUCCESS (RetainProgram, (self));
2186 pfn_notify = eq_program_notify;
2187 user_data = SvREFCNT_inc (s_get_cv (notify));
2188 }
2189 2463
2190 if (ix)
2191 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2192 else
2193 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2464 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2465 header_count, header_list, header_name, pfn_notify, user_data));
2466
2467#endif
2194 2468
2195void 2469void
2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2470build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2197 PPCODE: 2471 PPCODE:
2198 size_t size; 2472 size_t size;
2207#BEGIN:program_build 2481#BEGIN:program_build
2208 2482
2209void 2483void
2210build_status (OpenCL::Program self, OpenCL::Device device) 2484build_status (OpenCL::Program self, OpenCL::Device device)
2211 PPCODE: 2485 PPCODE:
2212 cl_build_status value [1]; 2486 cl_int value [1];
2213 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2487 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2214 EXTEND (SP, 1); 2488 EXTEND (SP, 1);
2215 const int i = 0; 2489 const int i = 0;
2216 PUSHs (sv_2mortal (newSViv (value [i]))); 2490 PUSHs (sv_2mortal (newSViv (value [i])));
2217 2491
2226 char *value = tmpbuf (size); 2500 char *value = tmpbuf (size);
2227 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2501 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2228 EXTEND (SP, 1); 2502 EXTEND (SP, 1);
2229 const int i = 0; 2503 const int i = 0;
2230 PUSHs (sv_2mortal (newSVpv (value, 0))); 2504 PUSHs (sv_2mortal (newSVpv (value, 0)));
2505
2506void
2507binary_type (OpenCL::Program self, OpenCL::Device device)
2508 PPCODE:
2509 cl_uint value [1];
2510 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2511 EXTEND (SP, 1);
2512 const int i = 0;
2513 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2231 2514
2232#END:program_build 2515#END:program_build
2233 2516
2234void 2517void
2235kernel (OpenCL::Program program, SV *function) 2518kernel (OpenCL::Program program, SV *function)
2302 PPCODE: 2585 PPCODE:
2303 cl_context value [1]; 2586 cl_context value [1];
2304 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2587 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2305 EXTEND (SP, 1); 2588 EXTEND (SP, 1);
2306 const int i = 0; 2589 const int i = 0;
2307 {
2308 NEED_SUCCESS (RetainContext, (value [i])); 2590 NEED_SUCCESS (RetainContext, (value [i]));
2309 PUSH_CLOBJ (stash_context, value [i]); 2591 PUSH_CLOBJ (stash_context, value [i]);
2310 }
2311 2592
2312void 2593void
2313devices (OpenCL::Program self) 2594devices (OpenCL::Program self)
2314 PPCODE: 2595 PPCODE:
2315 size_t size; 2596 size_t size;
2317 cl_device_id *value = tmpbuf (size); 2598 cl_device_id *value = tmpbuf (size);
2318 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2599 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2319 int i, n = size / sizeof (*value); 2600 int i, n = size / sizeof (*value);
2320 EXTEND (SP, n); 2601 EXTEND (SP, n);
2321 for (i = 0; i < n; ++i) 2602 for (i = 0; i < n; ++i)
2322 {
2323 PUSH_CLOBJ (stash_device, value [i]); 2603 PUSH_CLOBJ (stash_device, value [i]);
2324 }
2325 2604
2326void 2605void
2327source (OpenCL::Program self) 2606source (OpenCL::Program self)
2328 PPCODE: 2607 PPCODE:
2329 size_t size; 2608 size_t size;
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2681 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2682 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404 2683
2405 case 'z': nullarg = 1; size = SvIV (sv); break; 2684 case 'z': nullarg = 1; size = SvIV (sv); break;
2406 2685
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2686 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2408 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2687 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2409 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2688 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410 2689
2411 default: 2690 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2691 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 } 2692 }
2414 2693
2485 CODE: 2764 CODE:
2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2765 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2487 2766
2488void 2767void
2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2768set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
2493 CODE: 2769 CODE:
2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2770 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2495 2771
2496void 2772void
2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2773set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2543 PPCODE: 2819 PPCODE:
2544 cl_context value [1]; 2820 cl_context value [1];
2545 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2821 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2546 EXTEND (SP, 1); 2822 EXTEND (SP, 1);
2547 const int i = 0; 2823 const int i = 0;
2548 {
2549 NEED_SUCCESS (RetainContext, (value [i])); 2824 NEED_SUCCESS (RetainContext, (value [i]));
2550 PUSH_CLOBJ (stash_context, value [i]); 2825 PUSH_CLOBJ (stash_context, value [i]);
2551 }
2552 2826
2553void 2827void
2554program (OpenCL::Kernel self) 2828program (OpenCL::Kernel self)
2555 PPCODE: 2829 PPCODE:
2556 cl_program value [1]; 2830 cl_program value [1];
2557 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2831 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2558 EXTEND (SP, 1); 2832 EXTEND (SP, 1);
2559 const int i = 0; 2833 const int i = 0;
2560 {
2561 NEED_SUCCESS (RetainProgram, (value [i])); 2834 NEED_SUCCESS (RetainProgram, (value [i]));
2562 PUSH_CLOBJ (stash_program, value [i]); 2835 PUSH_CLOBJ (stash_program, value [i]);
2563 }
2564 2836
2565#END:kernel 2837#END:kernel
2566 2838
2567void 2839void
2568work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2840work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2614 const int i = 0; 2886 const int i = 0;
2615 PUSHs (sv_2mortal (newSVuv (value [i]))); 2887 PUSHs (sv_2mortal (newSVuv (value [i])));
2616 2888
2617#END:kernel_work_group 2889#END:kernel_work_group
2618 2890
2891#if CL_VERSION_1_2
2892
2893void
2894arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2895 PPCODE:
2896 size_t size;
2897 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2898 SV *sv = sv_2mortal (newSV (size));
2899 SvUPGRADE (sv, SVt_PV);
2900 SvPOK_only (sv);
2901 SvCUR_set (sv, size);
2902 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2903 XPUSHs (sv);
2904
2905#BEGIN:kernel_arg
2906
2907void
2908arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2909 ALIAS:
2910 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2911 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2912 PPCODE:
2913 cl_uint value [1];
2914 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2915 EXTEND (SP, 1);
2916 const int i = 0;
2917 PUSHs (sv_2mortal (newSVuv (value [i])));
2918
2919void
2920arg_type_name (OpenCL::Kernel self, cl_uint idx)
2921 ALIAS:
2922 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2923 arg_name = CL_KERNEL_ARG_NAME
2924 PPCODE:
2925 size_t size;
2926 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2927 char *value = tmpbuf (size);
2928 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2929 EXTEND (SP, 1);
2930 const int i = 0;
2931 PUSHs (sv_2mortal (newSVpv (value, 0)));
2932
2933void
2934arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2935 PPCODE:
2936 cl_ulong value [1];
2937 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2938 EXTEND (SP, 1);
2939 const int i = 0;
2940 PUSHs (sv_2mortal (newSVuv (value [i])));
2941
2942#END:kernel_arg
2943
2944#endif
2945
2619MODULE = OpenCL PACKAGE = OpenCL::Event 2946MODULE = OpenCL PACKAGE = OpenCL::Event
2620 2947
2621void 2948void
2622DESTROY (OpenCL::Event self) 2949DESTROY (OpenCL::Event self)
2623 CODE: 2950 CODE:
2645 PPCODE: 2972 PPCODE:
2646 cl_command_queue value [1]; 2973 cl_command_queue value [1];
2647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2974 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2648 EXTEND (SP, 1); 2975 EXTEND (SP, 1);
2649 const int i = 0; 2976 const int i = 0;
2650 {
2651 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2977 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2652 PUSH_CLOBJ (stash_queue, value [i]); 2978 PUSH_CLOBJ (stash_queue, value [i]);
2653 }
2654 2979
2655void 2980void
2656command_type (OpenCL::Event self) 2981command_type (OpenCL::Event self)
2657 PPCODE:
2658 cl_command_type value [1];
2659 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
2660 EXTEND (SP, 1);
2661 const int i = 0;
2662 PUSHs (sv_2mortal (newSVuv (value [i])));
2663
2664void
2665reference_count (OpenCL::Event self)
2666 ALIAS: 2982 ALIAS:
2983 command_type = CL_EVENT_COMMAND_TYPE
2667 reference_count = CL_EVENT_REFERENCE_COUNT 2984 reference_count = CL_EVENT_REFERENCE_COUNT
2668 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2985 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2669 PPCODE: 2986 PPCODE:
2670 cl_uint value [1]; 2987 cl_uint value [1];
2671 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2988 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2678 PPCODE: 2995 PPCODE:
2679 cl_context value [1]; 2996 cl_context value [1];
2680 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2997 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2681 EXTEND (SP, 1); 2998 EXTEND (SP, 1);
2682 const int i = 0; 2999 const int i = 0;
2683 {
2684 NEED_SUCCESS (RetainContext, (value [i])); 3000 NEED_SUCCESS (RetainContext, (value [i]));
2685 PUSH_CLOBJ (stash_context, value [i]); 3001 PUSH_CLOBJ (stash_context, value [i]);
2686 }
2687 3002
2688#END:event 3003#END:event
2689 3004
2690void 3005void
2691profiling_info (OpenCL::Event self, cl_profiling_info name) 3006profiling_info (OpenCL::Event self, cl_profiling_info name)
2732 Safefree (mapped); 3047 Safefree (mapped);
2733 3048
2734void 3049void
2735unmap (OpenCL::Mapped self, ...) 3050unmap (OpenCL::Mapped self, ...)
2736 CODE: 3051 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1); 3052 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2738 3053
2739bool 3054bool
2740mapped (OpenCL::Mapped self) 3055mapped (OpenCL::Mapped self)
2741 CODE: 3056 CODE:
2742 RETVAL = !!self->ptr; 3057 RETVAL = !!self->ptr;
2756 XSRETURN_UNDEF; 3071 XSRETURN_UNDEF;
2757 3072
2758 clRetainEvent (self->event); 3073 clRetainEvent (self->event);
2759 XPUSH_CLOBJ (stash_event, self->event); 3074 XPUSH_CLOBJ (stash_event, self->event);
2760 3075
2761size_t 3076#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3077#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3078#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3079#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3080#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3081#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3082
3083IV
2762size (OpenCL::Mapped self) 3084size (OpenCL::Mapped self)
3085 ALIAS:
3086 size = MAPPED_OFFSET_CB
3087 row_pitch = MAPPED_OFFSET_ROW_PITCH
3088 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3089 width = MAPPED_OFFSET_WIDTH
3090 height = MAPPED_OFFSET_HEIGHT
3091 depth = MAPPED_OFFSET_DEPTH
2763 CODE: 3092 CODE:
2764 RETVAL = self->cb; 3093 RETVAL = *(size_t *)((char *)self + ix);
2765 OUTPUT: 3094 OUTPUT:
2766 RETVAL 3095 RETVAL
2767 3096
2768IV 3097IV
2769ptr (OpenCL::Mapped self) 3098ptr (OpenCL::Mapped self)
2776set (OpenCL::Mapped self, size_t offset, SV *data) 3105set (OpenCL::Mapped self, size_t offset, SV *data)
2777 CODE: 3106 CODE:
2778 STRLEN len; 3107 STRLEN len;
2779 const char *ptr = SvPVbyte (data, len); 3108 const char *ptr = SvPVbyte (data, len);
2780 3109
2781 if (offset < 0 || offset + len > self->cb) 3110 if (offset + len > self->cb)
2782 croak ("OpenCL::Mapped::set out of bound condition detected"); 3111 croak ("OpenCL::Mapped::set out of bound condition detected");
2783 3112
2784 memcpy (offset + (char *)self->ptr, ptr, len); 3113 memcpy (offset + (char *)self->ptr, ptr, len);
2785 3114
3115void
3116get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3117 PPCODE:
3118 if (!SvOK (ST (1)))
3119 count = self->width - x;
3120
3121 if (x + count > self->width)
3122 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3123
3124 if (y >= self->height)
3125 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3126
3127 if (z >= self->depth)
3128 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3129
3130 size_t element = mapped_element_size (self);
3131
3132 count *= element;
3133 x *= element;
3134
3135 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3136 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3137
3138void
3139set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3140 PPCODE:
3141 STRLEN count;
3142 char *dataptr = SvPVbyte (data, count);
3143 size_t element = mapped_element_size (self);
3144
3145 x *= element;
3146
3147 if (x + count > self->width * element)
3148 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3149
3150 if (y >= self->height)
3151 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3152
3153 if (z >= self->depth)
3154 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3155
3156 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3157 memcpy (ptr, dataptr, count);
3158
2786MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer 3159MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2787 3160
2788MODULE = OpenCL PACKAGE = OpenCL::MappedImage 3161MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2789 3162
2790IV 3163IV
2791row_pitch (OpenCL::Mapped self) 3164element_size (OpenCL::Mapped self)
2792 ALIAS:
2793 slice_pitch = 1
2794 CODE: 3165 CODE:
2795 RETVAL = ix ? self->slice_pitch : self->row_pitch; 3166 RETVAL = mapped_element_size (self);
2796 OUTPUT: 3167 OUTPUT:
2797 RETVAL 3168 RETVAL
2798 3169
2799
2800

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines