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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.62 by root, Tue May 1 22:04:56 2012 UTC vs.
Revision 1.82 by root, Thu Oct 11 05:02:54 2012 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#include "ecb.h"//D
4 8
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 10#include "xthread.h"
7#include "schmorp.h" 11#include "schmorp.h"
8 12
26#if PREFER_1_1 30#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 32#endif
29 33
30#ifdef __APPLE__ 34#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
32#else 36#else
37 #define CLHDR(name) <CL/name>
38#endif
39
33 #include <CL/opencl.h> 40#include CLHDR(opencl.h)
41
42#if 0
43#ifndef CL_VERSION_1_2
44 #include CLHDR(cl_d3d9.h)
45#endif
46#endif
47
48#if _WIN32
49 #include CLHDR(cl_d3d10.h)
50 #if CL_VERSION_1_2
51 #include CLHDR<cl_d3d11.h>
52 #endif
53 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 54#endif
35 55
36#ifndef CL_VERSION_1_2 56#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 57 #undef PREFER_1_1
38 #define PREFER_1_1 1 58 #define PREFER_1_1 1
39#endif 59#endif
40 60
61// make sure all constants we might use are actually defined
62#include "default.h"
63
41typedef cl_platform_id OpenCL__Platform; 64typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 65typedef cl_device_id OpenCL__Device;
66typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 67typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 68typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 69typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 70typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 71typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 72typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 73typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 74typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 75typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 76typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 77typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 78typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 79typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 80typedef cl_event OpenCL__UserEvent;
61 81
62typedef struct mapped * OpenCL__Mapped; 82typedef struct mapped * OpenCL__Mapped;
63 83
64typedef SV *FUTURE;
65
66static HV 84static HV
67 *stash_platform, 85 *stash_platform,
68 *stash_device, 86 *stash_device,
87 *stash_subdevice,
69 *stash_context, 88 *stash_context,
70 *stash_queue, 89 *stash_queue,
71 *stash_program, 90 *stash_program,
72 *stash_kernel, 91 *stash_kernel,
73 *stash_sampler, 92 *stash_sampler,
133} 152}
134 153
135/*****************************************************************************/ 154/*****************************************************************************/
136 155
137/* up to two temporary buffers */ 156/* up to two temporary buffers */
138static void * 157static void * ecb_noinline
139tmpbuf (size_t size) 158tmpbuf (size_t size)
140{ 159{
141 enum { buffers = 3 }; 160 enum { buffers = 4 };
142 static int idx; 161 static int idx;
143 static void *buf [buffers]; 162 static void *buf [buffers];
144 static size_t len [buffers]; 163 static size_t len [buffers];
145 164
146 idx = (idx + 1) % buffers; 165 idx = (idx + 1) % buffers;
153 } 172 }
154 173
155 return buf [idx]; 174 return buf [idx];
156} 175}
157 176
177static const char * ecb_noinline
178cv_get_name (CV *cv)
179{
180 static char fullname [256];
181
182 GV *gv = CvGV (cv); // gv better be nonzero
183
184 HV *stash = GvSTASH (gv);
185 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
186 const char *gvname = GvNAME (gv);
187
188 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
189 return fullname;
190}
191
158/*****************************************************************************/ 192/*****************************************************************************/
159 193
160typedef struct 194typedef struct
161{ 195{
162 IV iv; 196 IV iv;
163 const char *name; 197 const char *name;
164 #define const_iv(name) { (IV)CL_ ## name, # name }, 198 #define const_iv(name) { (IV)CL_ ## name, # name },
165} ivstr; 199} ivstr;
200
201typedef struct
202{
203 NV nv;
204 const char *name;
205 #define const_nv(name) { (NV)CL_ ## name, # name },
206} nvstr;
166 207
167static const char * 208static const char *
168iv2str (IV value, const ivstr *base, int count, const char *fallback) 209iv2str (IV value, const ivstr *base, int count, const char *fallback)
169{ 210{
170 int i; 211 int i;
219 if (res) \ 260 if (res) \
220 FAIL (name); 261 FAIL (name);
221 262
222/*****************************************************************************/ 263/*****************************************************************************/
223 264
265static SV *
266new_clobj (HV *stash, IV id)
267{
268 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
269}
270
271#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
272#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
273
274/* cl objects are either \$iv, or [$iv, ...] */
275/* they can be upgraded at runtime to the array form */
276static void * ecb_noinline
277SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
278{
279 // sv_derived_from is quite slow :(
280 if (SvROK (sv) && sv_derived_from (sv, pkg))
281 return (void *)SvIV (SvRV (sv));
282
283 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
284}
285
286// the "no-inherit" version of the above
287static void * ecb_noinline
288SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
289{
290 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
291 return (void *)SvIV (SvRV (sv));
292
293 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
294}
295
296/*****************************************************************************/
297
224static cl_context_properties * 298static cl_context_properties * ecb_noinline
225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 299SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
226{ 300{
227 if (!sv || !SvOK (sv)) 301 if (!sv || !SvOK (sv))
228 if (extra) 302 if (extra)
229 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 303 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
230 else 304 else
236 int i, len = av_len (av) + 1; 310 int i, len = av_len (av) + 1;
237 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 311 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
238 cl_context_properties *l = p; 312 cl_context_properties *l = p;
239 313
240 if (len & 1) 314 if (len & 1)
241 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 315 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
242 316
243 while (extracount--) 317 while (extracount--)
244 *l++ = *extra++; 318 *l++ = *extra++;
245 319
246 for (i = 0; i < len; i += 2) 320 for (i = 0; i < len; i += 2)
249 SV *p_sv = *av_fetch (av, i + 1, 0); 323 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override 324 cl_context_properties v = SvIV (p_sv); // code below can override
251 325
252 switch (t) 326 switch (t)
253 { 327 {
328 case CL_CONTEXT_PLATFORM:
329 if (SvROK (p_sv))
330 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
331 break;
332
254 case CL_GLX_DISPLAY_KHR: 333 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv)) 334 if (!SvOK (p_sv))
256 { 335 {
257 void *func = glsym ("_glXGetCurrentDisplay"); 336 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func) 337 if (func)
281 *l = 0; 360 *l = 0;
282 361
283 return p; 362 return p;
284 } 363 }
285 364
286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 365 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
287} 366}
288 367
289/*****************************************************************************/ 368// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
290 369// is a pointer (and only on well-behaved systems).
291static SV * 370static void * ecb_noinline
292new_clobj (HV *stash, IV id) 371object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
293{ 372{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash)); 373 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
295} 374 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
375 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
296 376
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id))) 377 AV *av = (AV *)SvRV (arg);
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id))) 378 void **list = 0;
379 cl_uint count = av_len (av) + 1;
299 380
300/* cl objects are either \$iv, or [$iv, ...] */ 381 if (count)
301/* they can be upgraded at runtime to the array form */ 382 {
302static void * 383 list = tmpbuf (sizeof (*list) * count);
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 384 int i;
304{ 385 for (i = 0; i < count; ++i)
305 if (SvROK (sv) && sv_derived_from (sv, pkg)) 386 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
306 return (void *)SvIV (SvRV (sv)); 387 }
307 388
308 croak ("%s: %s is not of type %s", func, svname, pkg); 389 if (!count && !or_undef)
390 croak ("%s: '%s' must contain at least one %s object",
391 cv_get_name (cv), argname, klass);
392
393 *rcount = count;
394 return (void *)list;
309} 395}
310 396
311/*****************************************************************************/ 397/*****************************************************************************/
312/* callback stuff */ 398/* callback stuff */
313 399
335static void (*eq_signal_func)(void *signal_arg, int value); 421static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg; 422static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT; 423static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail; 424static eq_item *eq_head, *eq_tail;
339 425
340static void 426static void ecb_noinline
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 427eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{ 428{
343 eq_item *item = malloc (sizeof (eq_item)); 429 eq_item *item = malloc (sizeof (eq_item));
344 430
345 item->next = 0; 431 item->next = 0;
412} 498}
413 499
414/*****************************************************************************/ 500/*****************************************************************************/
415/* context notify */ 501/* context notify */
416 502
417static void 503static void ecb_noinline
418eq_context_push (void *data1, void *data2, void *data3) 504eq_context_push (void *data1, void *data2, void *data3)
419{ 505{
420 dSP; 506 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0))); 507 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 508 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
444 { \ 530 { \
445 pfn_notify = eq_context_notify; \ 531 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \ 532 user_data = s_get_cv (notify); \
447 } 533 }
448 534
449static SV * 535static SV * ecb_noinline
450new_clobj_context (cl_context ctx, void *user_data) 536new_clobj_context (cl_context ctx, void *user_data)
451{ 537{
452 SV *sv = new_clobj (stash_context, (IV)ctx); 538 SV *sv = new_clobj (stash_context, (IV)ctx);
453 539
454 if (user_data) 540 if (user_data)
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 559static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474 560
475static void CL_CALLBACK 561static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data) 562eq_program_notify (cl_program program, void *user_data)
477{ 563{
564 clRetainProgram (program);
565
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 566 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
567}
568
569typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
570
571static program_callback ecb_noinline
572make_program_callback (SV *notify, void **ruser_data)
573{
574 if (SvOK (notify))
575 {
576 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
577 return eq_program_notify;
578 }
579 else
580 {
581 *ruser_data = 0;
582 return 0;
583 }
479} 584}
480 585
481struct build_args 586struct build_args
482{ 587{
483 cl_program program; 588 cl_program program;
497 else 602 else
498 clReleaseProgram (arg->program); 603 clReleaseProgram (arg->program);
499 604
500 free (arg->options); 605 free (arg->options);
501 free (arg); 606 free (arg);
607
608 return 0;
502} 609}
503 610
504static void 611static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 612build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{ 613{
511 arg->user_data = user_data; 618 arg->user_data = user_data;
512 arg->num_devices = num_devices; 619 arg->num_devices = num_devices;
513 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices); 620 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
514 621
515 xthread_t id; 622 xthread_t id;
516 thread_create (&id, build_program_thread, arg); 623 xthread_create (&id, build_program_thread, arg);
624}
625
626/*****************************************************************************/
627/* mem object destructor notify */
628
629static void ecb_noinline
630eq_destructor_push (void *data1, void *data2, void *data3)
631{
632}
633
634static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
635
636static void CL_CALLBACK
637eq_destructor_notify (cl_mem memobj, void *user_data)
638{
639 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
517} 640}
518 641
519/*****************************************************************************/ 642/*****************************************************************************/
520/* event objects */ 643/* event objects */
521 644
546 size_t res; 669 size_t res;
547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 670 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
548 return res; 671 return res;
549} 672}
550 673
551static cl_event * 674static cl_event * ecb_noinline
552event_list (SV **items, cl_uint *rcount, cl_event extra) 675event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
553{ 676{
554 cl_uint count = *rcount; 677 cl_uint count = *rcount;
555 678
556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed 679 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
557 count = 0; 680 *rcount = count = 0;
558 681
559 if (!count && !extra) 682 if (!count && !extra)
560 return 0; 683 return 0;
561 684
562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 685 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
563 int i = 0; 686 int i = 0;
564 687
565 while (count--) 688 while (count--)
566 if (SvOK (items [count])) 689 if (SvOK (items [count]))
567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 690 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
568 691
569 if (extra) 692 if (extra)
570 list [i++] = extra; 693 list [i++] = extra;
571 694
572 *rcount = i; 695 *rcount = i;
574 return i ? list : 0; 697 return i ? list : 0;
575} 698}
576 699
577#define EVENT_LIST(skip) \ 700#define EVENT_LIST(skip) \
578 cl_uint event_list_count = items - (skip); \ 701 cl_uint event_list_count = items - (skip); \
579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 702 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
580 703
581#define INFO(class) \ 704#define INFO(class) \
582{ \ 705{ \
583 size_t size; \ 706 size_t size; \
584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 707 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
607 void *ptr; 730 void *ptr;
608 size_t cb; 731 size_t cb;
609 cl_event event; 732 cl_event event;
610 size_t row_pitch; 733 size_t row_pitch;
611 size_t slice_pitch; 734 size_t slice_pitch;
735
736 size_t element_size;
737 size_t width, height, depth;
612}; 738};
613 739
614static SV * 740static 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) 741mapped_new (
742 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
743 void *ptr, size_t cb, cl_event ev,
744 size_t row_pitch, size_t slice_pitch, size_t element_size,
745 size_t width, size_t height, size_t depth
746)
616{ 747{
617 SV *data = newSV (0); 748 SV *data = newSV (0);
618 SvUPGRADE (data, SVt_PVMG); 749 SvUPGRADE (data, SVt_PVMG);
619 750
620 OpenCL__Mapped mapped; 751 OpenCL__Mapped mapped;
621 New (0, mapped, 1, struct mapped); 752 New (0, mapped, 1, struct mapped);
622 753
623 clRetainCommandQueue (queue); 754 clRetainCommandQueue (queue);
624 755
625 mapped->queue = queue; 756 mapped->queue = queue;
626 mapped->memobj = memobj; 757 mapped->memobj = memobj;
627 mapped->ptr = ptr; 758 mapped->ptr = ptr;
628 mapped->cb = cb; 759 mapped->cb = cb;
629 mapped->event = ev; 760 mapped->event = ev;
630 mapped->row_pitch = row_pitch; 761 mapped->row_pitch = row_pitch;
631 mapped->slice_pitch = slice_pitch; 762 mapped->slice_pitch = slice_pitch;
763
764 mapped->element_size = element_size;
765 mapped->width = width;
766 mapped->height = height;
767 mapped->depth = depth;
632 768
633 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0); 769 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
634 770
635 if (SvLEN (data)) 771 if (SvLEN (data))
636 Safefree (data); 772 Safefree (data);
666 802
667 mapped->ptr = 0; 803 mapped->ptr = 0;
668} 804}
669 805
670static void 806static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count) 807mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{ 808{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event); 809 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
674 cl_event ev; 810 cl_event ev;
675 811
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev)); 812 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677 813
678 clReleaseEvent (mapped->event); 814 clReleaseEvent (mapped->event);
679 mapped->event = ev; 815 mapped->event = ev;
680 816
681 mapped_detach (self, mapped); 817 mapped_detach (self, mapped);
682} 818}
683 819
820static size_t
821mapped_element_size (OpenCL__Mapped self)
822{
823 if (!self->element_size)
824 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
825
826 return self->element_size;
827}
828
684/*****************************************************************************/ 829/*****************************************************************************/
685 830
686MODULE = OpenCL PACKAGE = OpenCL 831MODULE = OpenCL PACKAGE = OpenCL
687 832
688PROTOTYPES: ENABLE 833PROTOTYPES: ENABLE
699 eq_signal_arg = (void*)arg; 844 eq_signal_arg = (void*)arg;
700 845
701BOOT: 846BOOT:
702{ 847{
703 HV *stash = gv_stashpv ("OpenCL", 1); 848 HV *stash = gv_stashpv ("OpenCL", 1);
849
704 static const ivstr *civ, const_iv[] = { 850 static const ivstr *civ, const_iv[] = {
705 { sizeof (cl_char ), "SIZEOF_CHAR" }, 851 { sizeof (cl_char ), "SIZEOF_CHAR" },
706 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 852 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
707 { sizeof (cl_short ), "SIZEOF_SHORT" }, 853 { sizeof (cl_short ), "SIZEOF_SHORT" },
708 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 854 { sizeof (cl_ushort), "SIZEOF_USHORT" },
711 { sizeof (cl_long ), "SIZEOF_LONG" }, 857 { sizeof (cl_long ), "SIZEOF_LONG" },
712 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 858 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
713 { sizeof (cl_half ), "SIZEOF_HALF" }, 859 { sizeof (cl_half ), "SIZEOF_HALF" },
714 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 860 { sizeof (cl_float ), "SIZEOF_FLOAT" },
715 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 861 { sizeof (cl_double), "SIZEOF_DOUBLE" },
862 { PREFER_1_1 , "PREFER_1_1" },
716#include "constiv.h" 863#include "constiv.h"
717 }; 864 };
718 865
719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 866 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)); 867 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721 868
869 static const nvstr *cnv, const_nv[] = {
870#include "constnv.h"
871 };
872
873 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
874 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
875
876 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
877
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 878 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 879 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
880 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 881 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 882 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 883 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 884 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 885 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
769 list = tmpbuf (sizeof (*list) * count); 926 list = tmpbuf (sizeof (*list) * count);
770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 927 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
771 928
772 EXTEND (SP, count); 929 EXTEND (SP, count);
773 for (i = 0; i < count; ++i) 930 for (i = 0; i < count; ++i)
774 PUSH_CLOBJ (stash_platform, (IV)list [i]); 931 PUSH_CLOBJ (stash_platform, list [i]);
775 932
776void 933void
777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 934context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
778 PPCODE: 935 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK; 936 CONTEXT_NOTIFY_CALLBACK;
780 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 937 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
781 XPUSH_CLOBJ_CONTEXT; 938 XPUSH_CLOBJ_CONTEXT;
782 939
783void 940void
784context (FUTURE properties, FUTURE devices, FUTURE notify) 941context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
785 PPCODE: 942 PPCODE:
786 /* der Gipfel der Kunst */ 943 cl_uint device_count;
944 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
945
946 CONTEXT_NOTIFY_CALLBACK;
947 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
948 XPUSH_CLOBJ_CONTEXT;
787 949
788void 950void
789wait_for_events (...) 951wait_for_events (...)
790 CODE: 952 CODE:
791 EVENT_LIST (0); 953 EVENT_LIST (0);
842 EXTEND (SP, count); 1004 EXTEND (SP, count);
843 for (i = 0; i < count; ++i) 1005 for (i = 0; i < count; ++i)
844 PUSH_CLOBJ (stash_device, list [i]); 1006 PUSH_CLOBJ (stash_device, list [i]);
845 1007
846void 1008void
847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 1009context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
848 PPCODE: 1010 PPCODE:
1011 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1012 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1013
1014 cl_uint device_count;
1015 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1016
1017 CONTEXT_NOTIFY_CALLBACK;
1018 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1019 XPUSH_CLOBJ_CONTEXT;
1020
1021void
1022context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1023 PPCODE:
1024 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1025 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1026
1027 CONTEXT_NOTIFY_CALLBACK;
1028 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1029 XPUSH_CLOBJ_CONTEXT;
1030
1031MODULE = OpenCL PACKAGE = OpenCL::Device
1032
1033void
1034info (OpenCL::Device self, cl_device_info name)
1035 PPCODE:
1036 INFO (Device)
1037
1038#if CL_VERSION_1_2
1039
1040void
1041sub_devices (OpenCL::Device self, SV *properties)
1042 PPCODE:
849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1043 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
850 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1044 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
851 1045
852 AV *av = (AV *)SvRV (devices); 1046 properties = SvRV (properties);
853 cl_uint num_devices = av_len (av) + 1; 1047
854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1048 cl_uint count = av_len ((AV *)properties) + 1;
1049 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
855 1050
856 int i; 1051 int i;
857 for (i = num_devices; i--; ) 1052 for (i = 0; i < count; ++i)
858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1053 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
859 1054
860 CONTEXT_NOTIFY_CALLBACK; 1055 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 1056
864void 1057 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) 1058 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
866 PPCODE: 1059 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1060 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 1061
870 CONTEXT_NOTIFY_CALLBACK; 1062 EXTEND (SP, num_devices);
871 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1063 for (i = 0; i < count; ++i)
872 XPUSH_CLOBJ_CONTEXT; 1064 PUSH_CLOBJ (stash_subdevice, list [i]);
873 1065
874MODULE = OpenCL PACKAGE = OpenCL::Device 1066#endif
875
876void
877info (OpenCL::Device self, cl_device_info name)
878 PPCODE:
879 INFO (Device)
880 1067
881#BEGIN:device 1068#BEGIN:device
882 1069
883void 1070void
884type (OpenCL::Device self) 1071type (OpenCL::Device self)
1072 ALIAS:
1073 type = CL_DEVICE_TYPE
1074 address_bits = CL_DEVICE_ADDRESS_BITS
1075 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1076 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1077 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1078 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1079 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1080 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1081 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1082 properties = CL_DEVICE_QUEUE_PROPERTIES
1083 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1084 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
885 PPCODE: 1085 PPCODE:
886 cl_device_type value [1]; 1086 cl_ulong value [1];
887 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
888 EXTEND (SP, 1); 1088 EXTEND (SP, 1);
889 const int i = 0; 1089 const int i = 0;
890 PUSHs (sv_2mortal (newSVuv (value [i]))); 1090 PUSHs (sv_2mortal (newSVuv (value [i])));
891 1091
892void 1092void
906 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1106 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
907 image_support = CL_DEVICE_IMAGE_SUPPORT 1107 image_support = CL_DEVICE_IMAGE_SUPPORT
908 max_samplers = CL_DEVICE_MAX_SAMPLERS 1108 max_samplers = CL_DEVICE_MAX_SAMPLERS
909 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1109 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
910 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1110 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1111 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
911 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1112 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
912 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1113 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1114 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
913 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1115 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
914 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1116 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
915 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1117 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
916 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1118 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
917 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1119 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
955 EXTEND (SP, n); 1157 EXTEND (SP, n);
956 for (i = 0; i < n; ++i) 1158 for (i = 0; i < n; ++i)
957 PUSHs (sv_2mortal (newSVuv (value [i]))); 1159 PUSHs (sv_2mortal (newSVuv (value [i])));
958 1160
959void 1161void
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) 1162error_correction_support (OpenCL::Device self)
1016 ALIAS: 1163 ALIAS:
1017 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1164 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
1018 endian_little = CL_DEVICE_ENDIAN_LITTLE 1165 endian_little = CL_DEVICE_ENDIAN_LITTLE
1019 available = CL_DEVICE_AVAILABLE 1166 available = CL_DEVICE_AVAILABLE
1020 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1167 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
1021 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1168 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
1022 PPCODE: 1169 PPCODE:
1023 cl_bool value [1]; 1170 cl_uint value [1];
1024 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1171 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
1025 EXTEND (SP, 1); 1172 EXTEND (SP, 1);
1026 const int i = 0; 1173 const int i = 0;
1027 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1174 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 1175
1047void 1176void
1048platform (OpenCL::Device self) 1177platform (OpenCL::Device self)
1049 PPCODE: 1178 PPCODE:
1050 cl_platform_id value [1]; 1179 cl_platform_id value [1];
1051 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1180 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
1052 EXTEND (SP, 1); 1181 EXTEND (SP, 1);
1053 const int i = 0; 1182 const int i = 0;
1054 {
1055 PUSH_CLOBJ (stash_platform, value [i]); 1183 PUSH_CLOBJ (stash_platform, value [i]);
1056 }
1057 1184
1058void 1185void
1059name (OpenCL::Device self) 1186name (OpenCL::Device self)
1060 ALIAS: 1187 ALIAS:
1061 name = CL_DEVICE_NAME 1188 name = CL_DEVICE_NAME
1078 PPCODE: 1205 PPCODE:
1079 cl_device_id value [1]; 1206 cl_device_id value [1];
1080 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1207 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
1081 EXTEND (SP, 1); 1208 EXTEND (SP, 1);
1082 const int i = 0; 1209 const int i = 0;
1083 {
1084 PUSH_CLOBJ (stash_device, value [i]); 1210 PUSH_CLOBJ (stash_device, value [i]);
1085 }
1086 1211
1087void 1212void
1088partition_types_ext (OpenCL::Device self) 1213partition_types_ext (OpenCL::Device self)
1089 ALIAS: 1214 ALIAS:
1090 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1215 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
1100 for (i = 0; i < n; ++i) 1225 for (i = 0; i < n; ++i)
1101 PUSHs (sv_2mortal (newSVuv (value [i]))); 1226 PUSHs (sv_2mortal (newSVuv (value [i])));
1102 1227
1103#END:device 1228#END:device
1104 1229
1230MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1231
1232#if CL_VERSION_1_2
1233
1234void
1235DESTROY (OpenCL::SubDevice self)
1236 CODE:
1237 clReleaseDevice (self);
1238
1239#endif
1240
1105MODULE = OpenCL PACKAGE = OpenCL::Context 1241MODULE = OpenCL PACKAGE = OpenCL::Context
1106 1242
1107void 1243void
1108DESTROY (OpenCL::Context context) 1244DESTROY (OpenCL::Context self)
1109 CODE: 1245 CODE:
1110 clReleaseContext (context); 1246 clReleaseContext (self);
1111 1247
1112void 1248void
1113info (OpenCL::Context self, cl_context_info name) 1249info (OpenCL::Context self, cl_context_info name)
1114 PPCODE: 1250 PPCODE:
1115 INFO (Context) 1251 INFO (Context)
1156 const cl_image_desc desc = { 1292 const cl_image_desc desc = {
1157 type, 1293 type,
1158 width, height, depth, 1294 width, height, depth,
1159 array_size, row_pitch, slice_pitch, 1295 array_size, row_pitch, slice_pitch,
1160 num_mip_level, num_samples, 1296 num_mip_level, num_samples,
1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1297 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1162 }; 1298 };
1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1299 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1164 HV *stash = stash_image; 1300 HV *stash = stash_image;
1165 switch (type) 1301 switch (type)
1166 { 1302 {
1299 1435
1300 len2 = len; 1436 len2 = len;
1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1437 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1302 XPUSH_CLOBJ (stash_program, prog); 1438 XPUSH_CLOBJ (stash_program, prog);
1303 1439
1440void
1441program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1442 PPCODE:
1443 cl_uint device_count;
1444 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1445
1446 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1447 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1448
1449 binaries = SvRV (binaries);
1450
1451 if (device_count != av_len ((AV *)binaries) + 1)
1452 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1453
1454 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1455 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1456 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1457
1458 int i;
1459 for (i = 0; i < device_count; ++i)
1460 {
1461 STRLEN len;
1462 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1463 length_list [i] = len;
1464 }
1465
1466 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1467 length_list, binary_list,
1468 GIMME_V == G_ARRAY ? status_list : 0, &res));
1469
1470 EXTEND (SP, 2);
1471 PUSH_CLOBJ (stash_program, prog);
1472
1473 if (GIMME_V == G_ARRAY)
1474 {
1475 AV *av = newAV ();
1476 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1477
1478 for (i = device_count; i--; )
1479 av_store (av, i, newSViv (status_list [i]));
1480 }
1481
1482#if CL_VERSION_1_2
1483
1484void
1485program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1486 PPCODE:
1487 cl_uint device_count;
1488 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1489
1490 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1491
1492 XPUSH_CLOBJ (stash_program, prog);
1493
1494void
1495link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1496 CODE:
1497 cl_uint device_count = 0;
1498 cl_device_id *device_list = 0;
1499
1500 if (SvOK (devices))
1501 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1502
1503 cl_uint program_count;
1504 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1505
1506 void *user_data;
1507 program_callback pfn_notify = make_program_callback (notify, &user_data);
1508
1509 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1510 program_count, program_list, pfn_notify, user_data, &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514#endif
1515
1304#BEGIN:context 1516#BEGIN:context
1305 1517
1306void 1518void
1307reference_count (OpenCL::Context self) 1519reference_count (OpenCL::Context self)
1308 ALIAS: 1520 ALIAS:
1323 cl_device_id *value = tmpbuf (size); 1535 cl_device_id *value = tmpbuf (size);
1324 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1536 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1325 int i, n = size / sizeof (*value); 1537 int i, n = size / sizeof (*value);
1326 EXTEND (SP, n); 1538 EXTEND (SP, n);
1327 for (i = 0; i < n; ++i) 1539 for (i = 0; i < n; ++i)
1328 {
1329 PUSH_CLOBJ (stash_device, value [i]); 1540 PUSH_CLOBJ (stash_device, value [i]);
1330 }
1331 1541
1332void 1542void
1333properties (OpenCL::Context self) 1543properties (OpenCL::Context self)
1334 PPCODE: 1544 PPCODE:
1335 size_t size; 1545 size_t size;
1353void 1563void
1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1564read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS: 1565 ALIAS:
1356 enqueue_read_buffer = 0 1566 enqueue_read_buffer = 0
1357 PPCODE: 1567 PPCODE:
1358 cl_event ev = 0;
1359 EVENT_LIST (6); 1568 EVENT_LIST (6);
1360 1569
1361 SvUPGRADE (data, SVt_PV); 1570 SvUPGRADE (data, SVt_PV);
1362 SvGROW (data, len); 1571 SvGROW (data, len);
1363 SvPOK_only (data); 1572 SvPOK_only (data);
1364 SvCUR_set (data, len); 1573 SvCUR_set (data, len);
1574
1575 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)); 1576 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1366 1577
1367 if (ev) 1578 if (ev)
1368 XPUSH_CLOBJ (stash_event, ev); 1579 XPUSH_CLOBJ (stash_event, ev);
1369 1580
1370void 1581void
1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1582write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS: 1583 ALIAS:
1373 enqueue_write_buffer = 0 1584 enqueue_write_buffer = 0
1374 PPCODE: 1585 PPCODE:
1375 cl_event ev = 0; 1586 EVENT_LIST (5);
1587
1376 STRLEN len; 1588 STRLEN len;
1377 char *ptr = SvPVbyte (data, len); 1589 char *ptr = SvPVbyte (data, len);
1378 EVENT_LIST (5);
1379 1590
1591 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)); 1592 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1381 1593
1382 if (ev) 1594 if (ev)
1383 XPUSH_CLOBJ (stash_event, ev); 1595 XPUSH_CLOBJ (stash_event, ev);
1384 1596
1387void 1599void
1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1600fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS: 1601 ALIAS:
1390 enqueue_fill_buffer = 0 1602 enqueue_fill_buffer = 0
1391 PPCODE: 1603 PPCODE:
1392 cl_event ev = 0; 1604 EVENT_LIST (5);
1605
1393 STRLEN len; 1606 STRLEN len;
1394 char *ptr = SvPVbyte (data, len); 1607 char *ptr = SvPVbyte (data, len);
1395 EVENT_LIST (5);
1396 1608
1609 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)); 1610 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1398 1611
1399 if (ev) 1612 if (ev)
1400 XPUSH_CLOBJ (stash_event, ev); 1613 XPUSH_CLOBJ (stash_event, ev);
1401 1614
1402void 1615void
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, ...) 1616fill_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: 1617 ALIAS:
1405 enqueue_fill_image = 0 1618 enqueue_fill_image = 0
1406 PPCODE: 1619 PPCODE:
1407 cl_event ev = 0; 1620 EVENT_LIST (12);
1408 STRLEN len; 1621
1409 const size_t origin [3] = { x, y, z }; 1622 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth }; 1623 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412 1624
1413 const cl_float c_f [4] = { r, g, b, a }; 1625 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a }; 1626 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a }; 1627 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1628 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1631 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1632 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) 1633 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."); 1634 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423 1635
1636 cl_event ev = 0;
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1637 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)); 1638 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426 1639
1427 if (ev) 1640 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev); 1641 XPUSH_CLOBJ (stash_event, ev);
1432void 1645void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1646copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS: 1647 ALIAS:
1435 enqueue_copy_buffer = 0 1648 enqueue_copy_buffer = 0
1436 PPCODE: 1649 PPCODE:
1650 EVENT_LIST (6);
1651
1437 cl_event ev = 0; 1652 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)); 1653 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1654
1442 if (ev) 1655 if (ev)
1443 XPUSH_CLOBJ (stash_event, ev); 1656 XPUSH_CLOBJ (stash_event, ev);
1444 1657
1445void 1658void
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, ...) 1659read_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: 1660 ALIAS:
1448 enqueue_read_buffer_rect = 0 1661 enqueue_read_buffer_rect = 0
1449 PPCODE: 1662 PPCODE:
1450 cl_event ev = 0; 1663 EVENT_LIST (17);
1664
1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1665 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 }; 1666 const size_t host_origin[3] = { host_x, host_y, host_z };
1453 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
1454 EVENT_LIST (17);
1455 1668
1456 if (!buf_row_pitch) 1669 if (!buf_row_pitch)
1457 buf_row_pitch = region [0]; 1670 buf_row_pitch = region [0];
1458 1671
1459 if (!buf_slice_pitch) 1672 if (!buf_slice_pitch)
1469 1682
1470 SvUPGRADE (data, SVt_PV); 1683 SvUPGRADE (data, SVt_PV);
1471 SvGROW (data, len); 1684 SvGROW (data, len);
1472 SvPOK_only (data); 1685 SvPOK_only (data);
1473 SvCUR_set (data, len); 1686 SvCUR_set (data, len);
1687
1688 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)); 1689 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 1690
1476 if (ev) 1691 if (ev)
1477 XPUSH_CLOBJ (stash_event, ev); 1692 XPUSH_CLOBJ (stash_event, ev);
1478 1693
1479void 1694void
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, ...) 1695write_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: 1696 ALIAS:
1482 enqueue_write_buffer_rect = 0 1697 enqueue_write_buffer_rect = 0
1483 PPCODE: 1698 PPCODE:
1484 cl_event ev = 0; 1699 EVENT_LIST (17);
1700
1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1701 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 }; 1702 const size_t host_origin[3] = { host_x, host_y, host_z };
1487 const size_t region[3] = { width, height, depth }; 1703 const size_t region[3] = { width, height, depth };
1488 STRLEN len; 1704 STRLEN len;
1489 char *ptr = SvPVbyte (data, len); 1705 char *ptr = SvPVbyte (data, len);
1490 EVENT_LIST (17);
1491 1706
1492 if (!buf_row_pitch) 1707 if (!buf_row_pitch)
1493 buf_row_pitch = region [0]; 1708 buf_row_pitch = region [0];
1494 1709
1495 if (!buf_slice_pitch) 1710 if (!buf_slice_pitch)
1504 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1719 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1505 1720
1506 if (len < min_len) 1721 if (len < min_len)
1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1722 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1508 1723
1724 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)); 1725 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 1726
1511 if (ev) 1727 if (ev)
1512 XPUSH_CLOBJ (stash_event, ev); 1728 XPUSH_CLOBJ (stash_event, ev);
1513 1729
1514void 1730void
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, ...) 1731copy_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: 1732 ALIAS:
1517 enqueue_copy_buffer_rect = 0 1733 enqueue_copy_buffer_rect = 0
1518 PPCODE: 1734 PPCODE:
1519 cl_event ev = 0; 1735 EVENT_LIST (16);
1736
1520 const size_t src_origin[3] = { src_x, src_y, src_z }; 1737 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 }; 1738 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1522 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
1523 EVENT_LIST (16);
1524 1740
1741 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)); 1742 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 1743
1527 if (ev) 1744 if (ev)
1528 XPUSH_CLOBJ (stash_event, ev); 1745 XPUSH_CLOBJ (stash_event, ev);
1529 1746
1530void 1747void
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, ...) 1748read_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: 1749 ALIAS:
1533 enqueue_read_image = 0 1750 enqueue_read_image = 0
1534 PPCODE: 1751 PPCODE:
1535 cl_event ev = 0; 1752 EVENT_LIST (12);
1753
1536 const size_t src_origin[3] = { src_x, src_y, src_z }; 1754 const size_t src_origin[3] = { src_x, src_y, src_z };
1537 const size_t region[3] = { width, height, depth }; 1755 const size_t region[3] = { width, height, depth };
1538 EVENT_LIST (12);
1539 1756
1540 if (!row_pitch) 1757 if (!row_pitch)
1541 row_pitch = img_row_pitch (src); 1758 row_pitch = img_row_pitch (src);
1542 1759
1543 if (depth > 1 && !slice_pitch) 1760 if (depth > 1 && !slice_pitch)
1547 1764
1548 SvUPGRADE (data, SVt_PV); 1765 SvUPGRADE (data, SVt_PV);
1549 SvGROW (data, len); 1766 SvGROW (data, len);
1550 SvPOK_only (data); 1767 SvPOK_only (data);
1551 SvCUR_set (data, len); 1768 SvCUR_set (data, len);
1769
1770 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)); 1771 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 1772
1554 if (ev) 1773 if (ev)
1555 XPUSH_CLOBJ (stash_event, ev); 1774 XPUSH_CLOBJ (stash_event, ev);
1556 1775
1557void 1776void
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, ...) 1777write_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: 1778 ALIAS:
1560 enqueue_write_image = 0 1779 enqueue_write_image = 0
1561 PPCODE: 1780 PPCODE:
1562 cl_event ev = 0; 1781 EVENT_LIST (12);
1782
1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1783 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1564 const size_t region[3] = { width, height, depth }; 1784 const size_t region[3] = { width, height, depth };
1565 STRLEN len; 1785 STRLEN len;
1566 char *ptr = SvPVbyte (data, len); 1786 char *ptr = SvPVbyte (data, len);
1567 EVENT_LIST (12);
1568 1787
1569 if (!row_pitch) 1788 if (!row_pitch)
1570 row_pitch = img_row_pitch (dst); 1789 row_pitch = img_row_pitch (dst);
1571 1790
1572 if (depth > 1 && !slice_pitch) 1791 if (depth > 1 && !slice_pitch)
1575 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1794 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1576 1795
1577 if (len < min_len) 1796 if (len < min_len)
1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1797 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1579 1798
1799 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)); 1800 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 1801
1582 if (ev) 1802 if (ev)
1583 XPUSH_CLOBJ (stash_event, ev); 1803 XPUSH_CLOBJ (stash_event, ev);
1584 1804
1585void 1805void
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, ...) 1806copy_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: 1807 ALIAS:
1588 enqueue_copy_image = 0 1808 enqueue_copy_image = 0
1589 PPCODE: 1809 PPCODE:
1590 cl_event ev = 0; 1810 EVENT_LIST (12);
1811
1591 const size_t src_origin[3] = { src_x, src_y, src_z }; 1812 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 }; 1813 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1593 const size_t region[3] = { width, height, depth }; 1814 const size_t region[3] = { width, height, depth };
1594 EVENT_LIST (12);
1595 1815
1816 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)); 1817 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1597 1818
1598 if (ev) 1819 if (ev)
1599 XPUSH_CLOBJ (stash_event, ev); 1820 XPUSH_CLOBJ (stash_event, ev);
1600 1821
1601void 1822void
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, ...) 1823copy_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: 1824 ALIAS:
1604 enqueue_copy_image_to_buffer = 0 1825 enqueue_copy_image_to_buffer = 0
1605 PPCODE: 1826 PPCODE:
1606 cl_event ev = 0; 1827 EVENT_LIST (10);
1828
1607 const size_t src_origin[3] = { src_x, src_y, src_z }; 1829 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth }; 1830 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610 1831
1832 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)); 1833 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612 1834
1613 if (ev) 1835 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev); 1836 XPUSH_CLOBJ (stash_event, ev);
1615 1837
1616void 1838void
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, ...) 1839copy_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: 1840 ALIAS:
1619 enqueue_copy_buffer_to_image = 0 1841 enqueue_copy_buffer_to_image = 0
1620 PPCODE: 1842 PPCODE:
1621 cl_event ev = 0; 1843 EVENT_LIST (10);
1844
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1845 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth }; 1846 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625 1847
1848 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)); 1849 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627 1850
1628 if (ev) 1851 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev); 1852 XPUSH_CLOBJ (stash_event, ev);
1630 1853
1631void 1854void
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, ...) 1855map_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: 1856 ALIAS:
1634 enqueue_map_buffer = 0 1857 enqueue_map_buffer = 0
1635 PPCODE: 1858 PPCODE:
1859 EVENT_LIST (6);
1860
1861 size_t cb = SvIV (cb_);
1862
1863 if (!SvOK (cb_))
1864 {
1865 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1866 cb -= offset;
1867 }
1868
1636 cl_event ev; 1869 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)); 1870 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)); 1871 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1641 1872
1642void 1873void
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, ...) 1874map_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: 1875 ALIAS:
1645 enqueue_map_image = 0 1876 enqueue_map_image = 0
1646 PPCODE: 1877 PPCODE:
1647 cl_event ev; 1878 size_t width = SvIV (width_);
1879 if (!SvOK (width_))
1880 {
1881 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1882 width -= x;
1883 }
1884
1885 size_t height = SvIV (width_);
1886 if (!SvOK (height_))
1887 {
1888 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1889 height -= y;
1890
1891 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1892 if (!height)
1893 height = 1;
1894 }
1895
1896 size_t depth = SvIV (width_);
1897 if (!SvOK (depth_))
1898 {
1899 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1900 depth -= z;
1901
1902 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1903 if (!depth)
1904 depth = 1;
1905 }
1906
1648 const size_t origin[3] = { x, y, z }; 1907 const size_t origin[3] = { x, y, z };
1649 const size_t region[3] = { width, height, depth }; 1908 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch; 1909 size_t row_pitch, slice_pitch;
1651 EVENT_LIST (10); 1910 EVENT_LIST (10);
1652 1911
1912 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)); 1913 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 1914
1655 size_t cb = slice_pitch ? slice_pitch * region [2] 1915 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1] 1916 : row_pitch ? row_pitch * region [1]
1657 : region [0]; 1917 : region [0];
1658 1918
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch)); 1919 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1660 1920
1661void 1921void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1922unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE: 1923 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2); 1924 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID) 1925 if (GIMME_V != G_VOID)
1666 { 1926 {
1667 clRetainEvent (mapped->event); 1927 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event); 1928 XPUSH_CLOBJ (stash_event, mapped->event);
1669 } 1929 }
1671void 1931void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1932task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS: 1933 ALIAS:
1674 enqueue_task = 0 1934 enqueue_task = 0
1675 PPCODE: 1935 PPCODE:
1936 EVENT_LIST (2);
1937
1676 cl_event ev = 0; 1938 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)); 1939 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1680 1940
1681 if (ev) 1941 if (ev)
1682 XPUSH_CLOBJ (stash_event, ev); 1942 XPUSH_CLOBJ (stash_event, ev);
1683 1943
1684void 1944void
1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1945nd_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: 1946 ALIAS:
1687 enqueue_nd_range_kernel = 0 1947 enqueue_nd_range_kernel = 0
1688 PPCODE: 1948 PPCODE:
1689 cl_event ev = 0; 1949 EVENT_LIST (5);
1950
1690 size_t *gwo = 0, *gws, *lws = 0; 1951 size_t *gwo = 0, *gws, *lws = 0;
1691 int gws_len; 1952 int gws_len;
1692 size_t *lists; 1953 size_t *lists;
1693 int i; 1954 int i;
1694 EVENT_LIST (5);
1695 1955
1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1956 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1957 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1698 1958
1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1959 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1738 if (!lws [i]) 1998 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 1999 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 } 2000 }
1741 } 2001 }
1742 2002
2003 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)); 2004 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1744 2005
1745 if (ev) 2006 if (ev)
1746 XPUSH_CLOBJ (stash_event, ev); 2007 XPUSH_CLOBJ (stash_event, ev);
2008
2009#if CL_VERSION_1_2
2010
2011void
2012migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2013 ALIAS:
2014 enqueue_migrate_mem_objects = 0
2015 PPCODE:
2016 EVENT_LIST (3);
2017
2018 cl_uint obj_count;
2019 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2020
2021 cl_event ev = 0;
2022 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2023
2024 if (ev)
2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#endif
1747 2028
1748#if cl_apple_gl_sharing || cl_khr_gl_sharing 2029#if cl_apple_gl_sharing || cl_khr_gl_sharing
1749 2030
1750void 2031void
1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2032acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1752 ALIAS: 2033 ALIAS:
1753 release_gl_objects = 1 2034 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0 2035 enqueue_acquire_gl_objects = 0
1755 enqueue_release_gl_objects = 1 2036 enqueue_release_gl_objects = 1
1756 PPCODE: 2037 PPCODE:
1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2038 EVENT_LIST (2);
1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2039
2040 cl_uint obj_count;
2041 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1759 2042
1760 cl_event ev = 0; 2043 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 2044
1770 if (ix) 2045 if (ix)
1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1772 else 2047 else
1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2048 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1774 2049
1775 if (ev) 2050 if (ev)
1776 XPUSH_CLOBJ (stash_event, ev); 2051 XPUSH_CLOBJ (stash_event, ev);
1777 2052
1778#endif 2053#endif
1792void 2067void
1793marker (OpenCL::Queue self, ...) 2068marker (OpenCL::Queue self, ...)
1794 ALIAS: 2069 ALIAS:
1795 enqueue_marker = 0 2070 enqueue_marker = 0
1796 PPCODE: 2071 PPCODE:
2072 EVENT_LIST (1);
1797 cl_event ev = 0; 2073 cl_event ev = 0;
1798 EVENT_LIST (1);
1799#if PREFER_1_1 2074#if PREFER_1_1
1800 if (!event_list_count) 2075 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2076 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else 2077 else
1803#if CL_VERSION_1_2 2078#if CL_VERSION_1_2
1817void 2092void
1818barrier (OpenCL::Queue self, ...) 2093barrier (OpenCL::Queue self, ...)
1819 ALIAS: 2094 ALIAS:
1820 enqueue_barrier = 0 2095 enqueue_barrier = 0
1821 PPCODE: 2096 PPCODE:
2097 EVENT_LIST (1);
1822 cl_event ev = 0; 2098 cl_event ev = 0;
1823 EVENT_LIST (1);
1824#if PREFER_1_1 2099#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID) 2100 if (!event_list_count && GIMME_V == G_VOID)
1826 NEED_SUCCESS (EnqueueBarrier, (self)); 2101 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else 2102 else
1828#if CL_VERSION_1_2 2103#if CL_VERSION_1_2
1864 PPCODE: 2139 PPCODE:
1865 cl_context value [1]; 2140 cl_context value [1];
1866 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1867 EXTEND (SP, 1); 2142 EXTEND (SP, 1);
1868 const int i = 0; 2143 const int i = 0;
1869 {
1870 NEED_SUCCESS (RetainContext, (value [i])); 2144 NEED_SUCCESS (RetainContext, (value [i]));
1871 PUSH_CLOBJ (stash_context, value [i]); 2145 PUSH_CLOBJ (stash_context, value [i]);
1872 }
1873 2146
1874void 2147void
1875device (OpenCL::Queue self) 2148device (OpenCL::Queue self)
1876 PPCODE: 2149 PPCODE:
1877 cl_device_id value [1]; 2150 cl_device_id value [1];
1878 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2151 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1879 EXTEND (SP, 1); 2152 EXTEND (SP, 1);
1880 const int i = 0; 2153 const int i = 0;
1881 {
1882 PUSH_CLOBJ (stash_device, value [i]); 2154 PUSH_CLOBJ (stash_device, value [i]);
1883 }
1884 2155
1885void 2156void
1886reference_count (OpenCL::Queue self) 2157reference_count (OpenCL::Queue self)
1887 PPCODE: 2158 PPCODE:
1888 cl_uint value [1]; 2159 cl_uint value [1];
1892 PUSHs (sv_2mortal (newSVuv (value [i]))); 2163 PUSHs (sv_2mortal (newSVuv (value [i])));
1893 2164
1894void 2165void
1895properties (OpenCL::Queue self) 2166properties (OpenCL::Queue self)
1896 PPCODE: 2167 PPCODE:
1897 cl_command_queue_properties value [1]; 2168 cl_ulong value [1];
1898 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2169 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1899 EXTEND (SP, 1); 2170 EXTEND (SP, 1);
1900 const int i = 0; 2171 const int i = 0;
1901 PUSHs (sv_2mortal (newSVuv (value [i]))); 2172 PUSHs (sv_2mortal (newSVuv (value [i])));
1902 2173
1912void 2183void
1913info (OpenCL::Memory self, cl_mem_info name) 2184info (OpenCL::Memory self, cl_mem_info name)
1914 PPCODE: 2185 PPCODE:
1915 INFO (MemObject) 2186 INFO (MemObject)
1916 2187
2188void
2189destructor_callback (OpenCL::Memory self, SV *notify)
2190 PPCODE:
2191 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2192
1917#BEGIN:mem 2193#BEGIN:mem
1918 2194
1919void 2195void
1920type (OpenCL::Memory self) 2196type (OpenCL::Memory self)
2197 ALIAS:
2198 type = CL_MEM_TYPE
2199 map_count = CL_MEM_MAP_COUNT
2200 reference_count = CL_MEM_REFERENCE_COUNT
1921 PPCODE: 2201 PPCODE:
1922 cl_mem_object_type value [1]; 2202 cl_uint value [1];
1923 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2203 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1924 EXTEND (SP, 1); 2204 EXTEND (SP, 1);
1925 const int i = 0; 2205 const int i = 0;
1926 PUSHs (sv_2mortal (newSVuv (value [i]))); 2206 PUSHs (sv_2mortal (newSVuv (value [i])));
1927 2207
1928void 2208void
1929flags (OpenCL::Memory self) 2209flags (OpenCL::Memory self)
1930 PPCODE: 2210 PPCODE:
1931 cl_mem_flags value [1]; 2211 cl_ulong value [1];
1932 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1933 EXTEND (SP, 1); 2213 EXTEND (SP, 1);
1934 const int i = 0; 2214 const int i = 0;
1935 PUSHs (sv_2mortal (newSVuv (value [i]))); 2215 PUSHs (sv_2mortal (newSVuv (value [i])));
1936 2216
1954 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1955 const int i = 0; 2235 const int i = 0;
1956 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2236 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1957 2237
1958void 2238void
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) 2239context (OpenCL::Memory self)
1972 PPCODE: 2240 PPCODE:
1973 cl_context value [1]; 2241 cl_context value [1];
1974 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2242 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1975 EXTEND (SP, 1); 2243 EXTEND (SP, 1);
1976 const int i = 0; 2244 const int i = 0;
1977 {
1978 NEED_SUCCESS (RetainContext, (value [i])); 2245 NEED_SUCCESS (RetainContext, (value [i]));
1979 PUSH_CLOBJ (stash_context, value [i]); 2246 PUSH_CLOBJ (stash_context, value [i]);
1980 }
1981 2247
1982void 2248void
1983associated_memobject (OpenCL::Memory self) 2249associated_memobject (OpenCL::Memory self)
1984 PPCODE: 2250 PPCODE:
1985 cl_mem value [1]; 2251 cl_mem value [1];
1986 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1987 EXTEND (SP, 1); 2253 EXTEND (SP, 1);
1988 const int i = 0; 2254 const int i = 0;
1989 {
1990 NEED_SUCCESS (RetainMemObject, (value [i])); 2255 NEED_SUCCESS (RetainMemObject, (value [i]));
1991 PUSH_CLOBJ (stash_memory, value [i]); 2256 PUSH_CLOBJ (stash_memory, value [i]);
1992 }
1993 2257
1994#END:mem 2258#END:mem
1995 2259
1996#if cl_apple_gl_sharing || cl_khr_gl_sharing 2260#if cl_apple_gl_sharing || cl_khr_gl_sharing
1997 2261
2096 2360
2097#BEGIN:sampler 2361#BEGIN:sampler
2098 2362
2099void 2363void
2100reference_count (OpenCL::Sampler self) 2364reference_count (OpenCL::Sampler self)
2365 ALIAS:
2366 reference_count = CL_SAMPLER_REFERENCE_COUNT
2367 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2368 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
2101 PPCODE: 2369 PPCODE:
2102 cl_uint value [1]; 2370 cl_uint value [1];
2103 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
2104 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
2105 const int i = 0; 2373 const int i = 0;
2106 PUSHs (sv_2mortal (newSVuv (value [i]))); 2374 PUSHs (sv_2mortal (newSVuv (value [i])));
2107 2375
2108void 2376void
2110 PPCODE: 2378 PPCODE:
2111 cl_context value [1]; 2379 cl_context value [1];
2112 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2380 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
2113 EXTEND (SP, 1); 2381 EXTEND (SP, 1);
2114 const int i = 0; 2382 const int i = 0;
2115 {
2116 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
2117 PUSH_CLOBJ (stash_context, value [i]); 2384 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 2385
2138void 2386void
2139filter_mode (OpenCL::Sampler self) 2387filter_mode (OpenCL::Sampler self)
2140 PPCODE: 2388 PPCODE:
2141 cl_bool value [1]; 2389 cl_uint value [1];
2142 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
2143 EXTEND (SP, 1); 2391 EXTEND (SP, 1);
2144 const int i = 0; 2392 const int i = 0;
2145 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2393 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
2146 2394
2156void 2404void
2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2405build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS: 2406 ALIAS:
2159 build_async = 1 2407 build_async = 1
2160 CODE: 2408 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; 2409 cl_uint device_count = 0;
2164 cl_device_id *device_list = 0; 2410 cl_device_id *device_list = 0;
2165 2411
2166 if (SvOK (devices)) 2412 if (SvOK (devices))
2167 { 2413 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2414
2415 void *user_data;
2416 program_callback pfn_notify = make_program_callback (notify, &user_data);
2417
2418 if (ix)
2419 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2420 else
2421 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2422
2423#if CL_VERSION_1_2
2424
2425void
2426compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2427 CODE:
2428 cl_uint device_count = 0;
2429 cl_device_id *device_list = 0;
2430
2431 if (SvOK (devices))
2432 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2433
2434 cl_uint header_count = 0;
2435 cl_program *header_list = 0;
2436 const char **header_name = 0;
2437
2438 if (SvOK (headers))
2439 {
2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2440 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2441 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2170 2442
2171 AV *av = (AV *)SvRV (devices); 2443 HV *hv = (HV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173 2444
2174 if (num_devices) 2445 header_count = hv_iterinit (hv);
2446 header_list = tmpbuf (sizeof (*header_list) * header_count);
2447 header_name = tmpbuf (sizeof (*header_name) * header_count);
2448
2449 HE *he;
2450 int i = 0;
2451 while (he = hv_iternext (hv))
2175 { 2452 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2453 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2454 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2177 int count; 2455 ++i;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 } 2456 }
2181 } 2457 }
2182 2458
2183 if (SvOK (notify)) 2459 void *user_data;
2184 { 2460 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 2461
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)); 2462 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2463 header_count, header_list, header_name, pfn_notify, user_data));
2464
2465#endif
2194 2466
2195void 2467void
2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2468build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2197 PPCODE: 2469 PPCODE:
2198 size_t size; 2470 size_t size;
2207#BEGIN:program_build 2479#BEGIN:program_build
2208 2480
2209void 2481void
2210build_status (OpenCL::Program self, OpenCL::Device device) 2482build_status (OpenCL::Program self, OpenCL::Device device)
2211 PPCODE: 2483 PPCODE:
2212 cl_build_status value [1]; 2484 cl_int value [1];
2213 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2485 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
2214 EXTEND (SP, 1); 2486 EXTEND (SP, 1);
2215 const int i = 0; 2487 const int i = 0;
2216 PUSHs (sv_2mortal (newSViv (value [i]))); 2488 PUSHs (sv_2mortal (newSViv (value [i])));
2217 2489
2226 char *value = tmpbuf (size); 2498 char *value = tmpbuf (size);
2227 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2499 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2228 EXTEND (SP, 1); 2500 EXTEND (SP, 1);
2229 const int i = 0; 2501 const int i = 0;
2230 PUSHs (sv_2mortal (newSVpv (value, 0))); 2502 PUSHs (sv_2mortal (newSVpv (value, 0)));
2503
2504void
2505binary_type (OpenCL::Program self, OpenCL::Device device)
2506 PPCODE:
2507 cl_uint value [1];
2508 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2509 EXTEND (SP, 1);
2510 const int i = 0;
2511 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2231 2512
2232#END:program_build 2513#END:program_build
2233 2514
2234void 2515void
2235kernel (OpenCL::Program program, SV *function) 2516kernel (OpenCL::Program program, SV *function)
2302 PPCODE: 2583 PPCODE:
2303 cl_context value [1]; 2584 cl_context value [1];
2304 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2585 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2305 EXTEND (SP, 1); 2586 EXTEND (SP, 1);
2306 const int i = 0; 2587 const int i = 0;
2307 {
2308 NEED_SUCCESS (RetainContext, (value [i])); 2588 NEED_SUCCESS (RetainContext, (value [i]));
2309 PUSH_CLOBJ (stash_context, value [i]); 2589 PUSH_CLOBJ (stash_context, value [i]);
2310 }
2311 2590
2312void 2591void
2313devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
2314 PPCODE: 2593 PPCODE:
2315 size_t size; 2594 size_t size;
2317 cl_device_id *value = tmpbuf (size); 2596 cl_device_id *value = tmpbuf (size);
2318 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2319 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
2320 EXTEND (SP, n); 2599 EXTEND (SP, n);
2321 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
2322 {
2323 PUSH_CLOBJ (stash_device, value [i]); 2601 PUSH_CLOBJ (stash_device, value [i]);
2324 }
2325 2602
2326void 2603void
2327source (OpenCL::Program self) 2604source (OpenCL::Program self)
2328 PPCODE: 2605 PPCODE:
2329 size_t size; 2606 size_t size;
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2679 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2680 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404 2681
2405 case 'z': nullarg = 1; size = SvIV (sv); break; 2682 case 'z': nullarg = 1; size = SvIV (sv); break;
2406 2683
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2684 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; 2685 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; 2686 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410 2687
2411 default: 2688 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2689 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 } 2690 }
2414 2691
2485 CODE: 2762 CODE:
2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2487 2764
2488void 2765void
2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2766set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
2493 CODE: 2767 CODE:
2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2495 2769
2496void 2770void
2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2771set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2543 PPCODE: 2817 PPCODE:
2544 cl_context value [1]; 2818 cl_context value [1];
2545 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2819 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2546 EXTEND (SP, 1); 2820 EXTEND (SP, 1);
2547 const int i = 0; 2821 const int i = 0;
2548 {
2549 NEED_SUCCESS (RetainContext, (value [i])); 2822 NEED_SUCCESS (RetainContext, (value [i]));
2550 PUSH_CLOBJ (stash_context, value [i]); 2823 PUSH_CLOBJ (stash_context, value [i]);
2551 }
2552 2824
2553void 2825void
2554program (OpenCL::Kernel self) 2826program (OpenCL::Kernel self)
2555 PPCODE: 2827 PPCODE:
2556 cl_program value [1]; 2828 cl_program value [1];
2557 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2829 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2558 EXTEND (SP, 1); 2830 EXTEND (SP, 1);
2559 const int i = 0; 2831 const int i = 0;
2560 {
2561 NEED_SUCCESS (RetainProgram, (value [i])); 2832 NEED_SUCCESS (RetainProgram, (value [i]));
2562 PUSH_CLOBJ (stash_program, value [i]); 2833 PUSH_CLOBJ (stash_program, value [i]);
2563 }
2564 2834
2565#END:kernel 2835#END:kernel
2566 2836
2567void 2837void
2568work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2838work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
2614 const int i = 0; 2884 const int i = 0;
2615 PUSHs (sv_2mortal (newSVuv (value [i]))); 2885 PUSHs (sv_2mortal (newSVuv (value [i])));
2616 2886
2617#END:kernel_work_group 2887#END:kernel_work_group
2618 2888
2889#if CL_VERSION_1_2
2890
2891void
2892arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2893 PPCODE:
2894 size_t size;
2895 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2896 SV *sv = sv_2mortal (newSV (size));
2897 SvUPGRADE (sv, SVt_PV);
2898 SvPOK_only (sv);
2899 SvCUR_set (sv, size);
2900 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2901 XPUSHs (sv);
2902
2903#BEGIN:kernel_arg
2904
2905void
2906arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2907 ALIAS:
2908 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2909 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2910 PPCODE:
2911 cl_uint value [1];
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2913 EXTEND (SP, 1);
2914 const int i = 0;
2915 PUSHs (sv_2mortal (newSVuv (value [i])));
2916
2917void
2918arg_type_name (OpenCL::Kernel self, cl_uint idx)
2919 ALIAS:
2920 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2921 arg_name = CL_KERNEL_ARG_NAME
2922 PPCODE:
2923 size_t size;
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2925 char *value = tmpbuf (size);
2926 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2927 EXTEND (SP, 1);
2928 const int i = 0;
2929 PUSHs (sv_2mortal (newSVpv (value, 0)));
2930
2931void
2932arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2933 PPCODE:
2934 cl_ulong value [1];
2935 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2936 EXTEND (SP, 1);
2937 const int i = 0;
2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2939
2940#END:kernel_arg
2941
2942#endif
2943
2619MODULE = OpenCL PACKAGE = OpenCL::Event 2944MODULE = OpenCL PACKAGE = OpenCL::Event
2620 2945
2621void 2946void
2622DESTROY (OpenCL::Event self) 2947DESTROY (OpenCL::Event self)
2623 CODE: 2948 CODE:
2645 PPCODE: 2970 PPCODE:
2646 cl_command_queue value [1]; 2971 cl_command_queue value [1];
2647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2972 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2648 EXTEND (SP, 1); 2973 EXTEND (SP, 1);
2649 const int i = 0; 2974 const int i = 0;
2650 {
2651 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2975 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2652 PUSH_CLOBJ (stash_queue, value [i]); 2976 PUSH_CLOBJ (stash_queue, value [i]);
2653 }
2654 2977
2655void 2978void
2656command_type (OpenCL::Event self) 2979command_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: 2980 ALIAS:
2981 command_type = CL_EVENT_COMMAND_TYPE
2667 reference_count = CL_EVENT_REFERENCE_COUNT 2982 reference_count = CL_EVENT_REFERENCE_COUNT
2668 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2983 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
2669 PPCODE: 2984 PPCODE:
2670 cl_uint value [1]; 2985 cl_uint value [1];
2671 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2986 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
2678 PPCODE: 2993 PPCODE:
2679 cl_context value [1]; 2994 cl_context value [1];
2680 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2995 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2681 EXTEND (SP, 1); 2996 EXTEND (SP, 1);
2682 const int i = 0; 2997 const int i = 0;
2683 {
2684 NEED_SUCCESS (RetainContext, (value [i])); 2998 NEED_SUCCESS (RetainContext, (value [i]));
2685 PUSH_CLOBJ (stash_context, value [i]); 2999 PUSH_CLOBJ (stash_context, value [i]);
2686 }
2687 3000
2688#END:event 3001#END:event
2689 3002
2690void 3003void
2691profiling_info (OpenCL::Event self, cl_profiling_info name) 3004profiling_info (OpenCL::Event self, cl_profiling_info name)
2732 Safefree (mapped); 3045 Safefree (mapped);
2733 3046
2734void 3047void
2735unmap (OpenCL::Mapped self, ...) 3048unmap (OpenCL::Mapped self, ...)
2736 CODE: 3049 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1); 3050 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2738 3051
2739bool 3052bool
2740mapped (OpenCL::Mapped self) 3053mapped (OpenCL::Mapped self)
2741 CODE: 3054 CODE:
2742 RETVAL = !!self->ptr; 3055 RETVAL = !!self->ptr;
2756 XSRETURN_UNDEF; 3069 XSRETURN_UNDEF;
2757 3070
2758 clRetainEvent (self->event); 3071 clRetainEvent (self->event);
2759 XPUSH_CLOBJ (stash_event, self->event); 3072 XPUSH_CLOBJ (stash_event, self->event);
2760 3073
2761size_t 3074#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3075#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3076#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3077#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3078#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3079#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3080
3081IV
2762size (OpenCL::Mapped self) 3082size (OpenCL::Mapped self)
3083 ALIAS:
3084 size = MAPPED_OFFSET_CB
3085 row_pitch = MAPPED_OFFSET_ROW_PITCH
3086 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3087 width = MAPPED_OFFSET_WIDTH
3088 height = MAPPED_OFFSET_HEIGHT
3089 depth = MAPPED_OFFSET_DEPTH
2763 CODE: 3090 CODE:
2764 RETVAL = self->cb; 3091 RETVAL = *(size_t *)((char *)self + ix);
2765 OUTPUT: 3092 OUTPUT:
2766 RETVAL 3093 RETVAL
2767 3094
2768IV 3095IV
2769ptr (OpenCL::Mapped self) 3096ptr (OpenCL::Mapped self)
2770 CODE: 3097 CODE:
2771 RETVAL = PTR2IV (self->ptr); 3098 RETVAL = PTR2IV (self->ptr);
2772 OUTPUT: 3099 OUTPUT:
2773 RETVAL 3100 RETVAL
2774 3101
3102void
3103set (OpenCL::Mapped self, size_t offset, SV *data)
3104 CODE:
3105 STRLEN len;
3106 const char *ptr = SvPVbyte (data, len);
3107
3108 if (offset + len > self->cb)
3109 croak ("OpenCL::Mapped::set out of bound condition detected");
3110
3111 memcpy (offset + (char *)self->ptr, ptr, len);
3112
3113void
3114get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3115 PPCODE:
3116 if (!SvOK (ST (1)))
3117 count = self->width - x;
3118
3119 if (x + count > self->width)
3120 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3121
3122 if (y >= self->height)
3123 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3124
3125 if (z >= self->depth)
3126 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3127
3128 size_t element = mapped_element_size (self);
3129
3130 count *= element;
3131 x *= element;
3132
3133 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3134 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3135
3136void
3137set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3138 PPCODE:
3139 STRLEN count;
3140 char *dataptr = SvPVbyte (data, count);
3141 size_t element = mapped_element_size (self);
3142
3143 x *= element;
3144
3145 if (x + count > self->width * element)
3146 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3147
3148 if (y >= self->height)
3149 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3150
3151 if (z >= self->depth)
3152 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3153
3154 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3155 memcpy (ptr, dataptr, count);
3156
2775MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer 3157MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2776 3158
2777MODULE = OpenCL PACKAGE = OpenCL::MappedImage 3159MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2778 3160
2779IV 3161IV
2780row_pitch (OpenCL::Mapped self) 3162element_size (OpenCL::Mapped self)
2781 ALIAS:
2782 slice_pitch = 1
2783 CODE: 3163 CODE:
2784 RETVAL = ix ? self->slice_pitch : self->row_pitch; 3164 RETVAL = mapped_element_size (self);
2785 OUTPUT: 3165 OUTPUT:
2786 RETVAL 3166 RETVAL
2787 3167
2788
2789

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines