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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.22 by root, Mon Apr 16 09:42:33 2012 UTC vs.
Revision 1.76 by root, Sat May 5 14:49:52 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "ecb.h"//D
6
7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
8#include "xthread.h"
9#include "schmorp.h"
10
11#ifdef I_DLFCN
12 #include <dlfcn.h>
13#endif
14
15// how stupid is that, the 1.2 header files define CL_VERSION_1_1,
16// but then fail to define the api functions unless you ALSO define
17// this. This breaks 100% of the opencl 1.1 apps, for what reason?
18// after all, the functions are deprecated, not removed.
19// in addition, you cannot test for this in any future-proof way.
20// each time a new opencl version comes out, you need to make a new
21// release.
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
23
24#ifndef PREFER_1_1
25 #define PREFER_1_1 1
26#endif
27
28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
30#endif
31
5#ifdef __APPLE__ 32#ifdef __APPLE__
6 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
7#else 34#else
8 #include <CL/opencl.h> 35 #define CLHDR(name) <CL/name>
9#endif 36#endif
37
38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
50#endif
51
52#ifndef CL_VERSION_1_2
53 #undef PREFER_1_1
54 #define PREFER_1_1 1
55#endif
56
57// make sure all constants we might use are actually defined
58#include "default.h"
10 59
11typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
12typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
13typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
14typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
15typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
16typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
17typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
18typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
19typedef cl_mem OpenCL__Image2D;
20typedef cl_mem OpenCL__Image3D;
21typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
22typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
23typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
24typedef cl_mem OpenCL__Image2D_ornull;
25typedef cl_mem OpenCL__Image3D_ornull;
26typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
27typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
28typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
29typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
30typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
31 77
32typedef SV *FUTURE; 78typedef struct mapped * OpenCL__Mapped;
79
80static HV
81 *stash_platform,
82 *stash_device,
83 *stash_subdevice,
84 *stash_context,
85 *stash_queue,
86 *stash_program,
87 *stash_kernel,
88 *stash_sampler,
89 *stash_event,
90 *stash_userevent,
91 *stash_memory,
92 *stash_buffer,
93 *stash_bufferobj,
94 *stash_image,
95 *stash_image1d,
96 *stash_image1darray,
97 *stash_image1dbuffer,
98 *stash_image2d,
99 *stash_image2darray,
100 *stash_image3d,
101 *stash_mapped,
102 *stash_mappedbuffer,
103 *stash_mappedimage;
33 104
34/*****************************************************************************/ 105/*****************************************************************************/
35 106
107// name must include a leading underscore
108// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
109// for perl. doh.
110static void *
111glsym (const char *name)
112{
113 void *fun = 0;
114
115 #if defined I_DLFCN && defined RTLD_DEFAULT
116 fun = dlsym (RTLD_DEFAULT, name + 1);
117 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
118
119 if (!fun)
120 {
121 static void *libgl;
122 static const char *glso[] = {
123 "libGL.so.1",
124 "libGL.so.3",
125 "libGL.so.4.0",
126 "libGL.so",
127 "/usr/lib/libGL.so",
128 "/usr/X11R6/lib/libGL.1.dylib"
129 };
130 int i;
131
132 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
133 {
134 libgl = dlopen (glso [i], RTLD_LAZY);
135 if (libgl)
136 break;
137 }
138
139 if (libgl)
140 {
141 fun = dlsym (libgl, name + 1);
142 if (!fun) fun = dlsym (libgl, name);
143 }
144 }
145 #endif
146
147 return fun;
148}
149
150/*****************************************************************************/
151
36/* up to two temporary buffers */ 152/* up to two temporary buffers */
37static void * 153static void * ecb_noinline
38tmpbuf (size_t size) 154tmpbuf (size_t size)
39{ 155{
156 enum { buffers = 4 };
40 static int idx; 157 static int idx;
41 static void *buf [2]; 158 static void *buf [buffers];
42 static size_t len [2]; 159 static size_t len [buffers];
43 160
44 idx ^= 1; 161 idx = (idx + 1) % buffers;
45 162
46 if (len [idx] < size) 163 if (len [idx] < size)
47 { 164 {
48 free (buf [idx]); 165 free (buf [idx]);
49 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 166 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
51 } 168 }
52 169
53 return buf [idx]; 170 return buf [idx];
54} 171}
55 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
56/*****************************************************************************/ 188/*****************************************************************************/
57 189
58typedef struct 190typedef struct
59{ 191{
60 IV iv; 192 IV iv;
61 const char *name; 193 const char *name;
62 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
63} ivstr; 195} ivstr;
196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
64 203
65static const char * 204static const char *
66iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
67{ 206{
68 int i; 207 int i;
117 if (res) \ 256 if (res) \
118 FAIL (name); 257 FAIL (name);
119 258
120/*****************************************************************************/ 259/*****************************************************************************/
121 260
122#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
123#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
124
125static void * 261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
126SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
127{ 274{
275 // sv_derived_from is quite slow :(
128 if (SvROK (sv) && sv_derived_from (sv, pkg)) 276 if (SvROK (sv) && sv_derived_from (sv, pkg))
129 return (void *)SvIV (SvRV (sv)); 277 return (void *)SvIV (SvRV (sv));
130 278
131 croak ("%s: %s is not of type %s", func, svname, pkg); 279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
132} 290}
133 291
134/*****************************************************************************/ 292/*****************************************************************************/
293
294static cl_context_properties * ecb_noinline
295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
296{
297 if (!sv || !SvOK (sv))
298 if (extra)
299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
300 else
301 return 0;
302
303 if (SvROK (sv) && SvTYPE (SvRV (sv)) == SVt_PVAV)
304 {
305 AV *av = (AV *)SvRV (sv);
306 int i, len = av_len (av) + 1;
307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
308 cl_context_properties *l = p;
309
310 if (len & 1)
311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
312
313 while (extracount--)
314 *l++ = *extra++;
315
316 for (i = 0; i < len; i += 2)
317 {
318 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
319 SV *p_sv = *av_fetch (av, i + 1, 0);
320 cl_context_properties v = SvIV (p_sv); // code below can override
321
322 switch (t)
323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
329 case CL_GLX_DISPLAY_KHR:
330 if (!SvOK (p_sv))
331 {
332 void *func = glsym ("_glXGetCurrentDisplay");
333 if (func)
334 v = (cl_context_properties)((void *(*)(void))func)();
335 }
336 break;
337
338 case CL_GL_CONTEXT_KHR:
339 if (!SvOK (p_sv))
340 {
341 void *func = glsym ("_glXGetCurrentContext");
342 if (func)
343 v = (cl_context_properties)((void *(*)(void))func)();
344 }
345 break;
346
347 default:
348 /* unknown property, treat as int */
349 break;
350 }
351
352 *l++ = t;
353 *l++ = v;
354 }
355
356 *l = 0;
357
358 return p;
359 }
360
361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
362}
363
364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
365// is a pointer (and only on well-behaved systems).
366static void * ecb_noinline
367object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
368{
369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
372
373 AV *av = (AV *)SvRV (arg);
374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
376
377 if (count)
378 {
379 list = tmpbuf (sizeof (*list) * count);
380 int i;
381 for (i = 0; i < count; ++i)
382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
383 }
384
385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
391}
392
393/*****************************************************************************/
394/* callback stuff */
395
396/* default context callback, log to stderr */
397static void CL_CALLBACK
398context_default_notify (const char *msg, const void *info, size_t cb, void *data)
399{
400 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
401}
402
403typedef struct
404{
405 int free_cb;
406 void (*push)(void *data1, void *data2, void *data3);
407} eq_vtbl;
408
409typedef struct eq_item
410{
411 struct eq_item *next;
412 eq_vtbl *vtbl;
413 SV *cb;
414 void *data1, *data2, *data3;
415} eq_item;
416
417static void (*eq_signal_func)(void *signal_arg, int value);
418static void *eq_signal_arg;
419static xmutex_t eq_lock = X_MUTEX_INIT;
420static eq_item *eq_head, *eq_tail;
421
422static void ecb_noinline
423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
424{
425 eq_item *item = malloc (sizeof (eq_item));
426
427 item->next = 0;
428 item->vtbl = vtbl;
429 item->cb = cb;
430 item->data1 = data1;
431 item->data2 = data2;
432 item->data3 = data3;
433
434 X_LOCK (eq_lock);
435
436 *(eq_head ? &eq_tail->next : &eq_head) = item;
437 eq_tail = item;
438
439 X_UNLOCK (eq_lock);
440
441 eq_signal_func (eq_signal_arg, 0);
442}
443
444static eq_item *
445eq_dec (void)
446{
447 eq_item *res;
448
449 X_LOCK (eq_lock);
450
451 res = eq_head;
452 if (res)
453 eq_head = res->next;
454
455 X_UNLOCK (eq_lock);
456
457 return res;
458}
459
460static void
461eq_poll (void)
462{
463 eq_item *item;
464
465 while ((item = eq_dec ()))
466 {
467 ENTER;
468 SAVETMPS;
469
470 dSP;
471 PUSHMARK (SP);
472 EXTEND (SP, 2);
473
474 if (item->vtbl->free_cb)
475 sv_2mortal (item->cb);
476
477 PUTBACK;
478 item->vtbl->push (item->data1, item->data2, item->data3);
479
480 SV *cb = item->cb;
481 free (item);
482
483 call_sv (cb, G_DISCARD | G_VOID);
484
485 FREETMPS;
486 LEAVE;
487 }
488}
489
490static void
491eq_poll_interrupt (pTHX_ void *c_arg, int value)
492{
493 eq_poll ();
494}
495
496/*****************************************************************************/
497/* context notify */
498
499static void ecb_noinline
500eq_context_push (void *data1, void *data2, void *data3)
501{
502 dSP;
503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
505 PUTBACK;
506
507 free (data1);
508 free (data2);
509}
510
511static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
512
513static void CL_CALLBACK
514eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
515{
516 void *pvt_copy = malloc (cb);
517 memcpy (pvt_copy, pvt, cb);
518 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
519}
520
521#define CONTEXT_NOTIFY_CALLBACK \
522 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
523 void *user_data = 0; \
524 \
525 if (SvOK (notify)) \
526 { \
527 pfn_notify = eq_context_notify; \
528 user_data = s_get_cv (notify); \
529 }
530
531static SV * ecb_noinline
532new_clobj_context (cl_context ctx, void *user_data)
533{
534 SV *sv = new_clobj (stash_context, (IV)ctx);
535
536 if (user_data)
537 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
538
539 return sv;
540}
541
542#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
543
544/*****************************************************************************/
545/* build/compile/link notify */
546
547static void
548eq_program_push (void *data1, void *data2, void *data3)
549{
550 dSP;
551 PUSH_CLOBJ (stash_program, data1);
552 PUTBACK;
553}
554
555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
556
557static void CL_CALLBACK
558eq_program_notify (cl_program program, void *user_data)
559{
560 clRetainProgram (program);
561
562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
580}
581
582struct build_args
583{
584 cl_program program;
585 char *options;
586 void *user_data;
587 cl_uint num_devices;
588};
589
590X_THREAD_PROC (build_program_thread)
591{
592 struct build_args *arg = thr_arg;
593
594 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
595
596 if (arg->user_data)
597 eq_program_notify (arg->program, arg->user_data);
598 else
599 clReleaseProgram (arg->program);
600
601 free (arg->options);
602 free (arg);
603
604 return 0;
605}
606
607static void
608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
609{
610 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
611
612 arg->program = program;
613 arg->options = strdup (options);
614 arg->user_data = user_data;
615 arg->num_devices = num_devices;
616 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
617
618 xthread_t id;
619 thread_create (&id, build_program_thread, arg);
620}
621
622/*****************************************************************************/
623/* event objects */
624
625static void
626eq_event_push (void *data1, void *data2, void *data3)
627{
628 dSP;
629 PUSH_CLOBJ (stash_event, data1);
630 PUSHs (sv_2mortal (newSViv ((IV)data2)));
631 PUTBACK;
632}
633
634static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
635
636static void CL_CALLBACK
637eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
638{
639 clRetainEvent (event);
640 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
641}
642
643/*****************************************************************************/
644/* utilities for XS code */
135 645
136static size_t 646static size_t
137img_row_pitch (cl_mem img) 647img_row_pitch (cl_mem img)
138{ 648{
139 size_t res; 649 size_t res;
140 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 650 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
141 return res; 651 return res;
142} 652}
143 653
144static cl_event * 654static cl_event * ecb_noinline
145event_list (SV **items, int count) 655event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
146{ 656{
147 if (!count) 657 cl_uint count = *rcount;
658
659 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
660 *rcount = count = 0;
661
662 if (!count && !extra)
148 return 0; 663 return 0;
149 664
150 cl_event *list = tmpbuf (sizeof (cl_event) * count); 665 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
666 int i = 0;
151 667
152 while (count--) 668 while (count--)
669 if (SvOK (items [count]))
153 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 670 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
154 671
672 if (extra)
673 list [i++] = extra;
674
675 *rcount = i;
676
155 return list; 677 return i ? list : 0;
156} 678}
157 679
158#define EVENT_LIST(items,count) \ 680#define EVENT_LIST(skip) \
159 cl_uint event_list_count = (count); \ 681 cl_uint event_list_count = items - (skip); \
160 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 682 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
161 683
162#define INFO(class) \ 684#define INFO(class) \
163{ \ 685{ \
164 size_t size; \ 686 size_t size; \
165 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 687 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
166 SV *sv = sv_2mortal (newSV (size)); \ 688 SV *sv = sv_2mortal (newSV (size)); \
167 SvUPGRADE (sv, SVt_PV); \ 689 SvUPGRADE (sv, SVt_PV); \
168 SvPOK_only (sv); \ 690 SvPOK_only (sv); \
169 SvCUR_set (sv, size); \ 691 SvCUR_set (sv, size); \
170 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 692 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
171 XPUSHs (sv); \ 693 XPUSHs (sv); \
172} 694}
173 695
696/*****************************************************************************/
697/* mapped_xxx */
698
699static OpenCL__Mapped
700SvMAPPED (SV *sv)
701{
702 // no typechecking atm., keep your fingers crossed
703 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
704}
705
706struct mapped
707{
708 cl_command_queue queue;
709 cl_mem memobj;
710 void *ptr;
711 size_t cb;
712 cl_event event;
713 size_t row_pitch;
714 size_t slice_pitch;
715};
716
717static SV *
718mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
719{
720 SV *data = newSV (0);
721 SvUPGRADE (data, SVt_PVMG);
722
723 OpenCL__Mapped mapped;
724 New (0, mapped, 1, struct mapped);
725
726 clRetainCommandQueue (queue);
727
728 mapped->queue = queue;
729 mapped->memobj = memobj;
730 mapped->ptr = ptr;
731 mapped->cb = cb;
732 mapped->event = ev;
733 mapped->row_pitch = row_pitch;
734 mapped->slice_pitch = slice_pitch;
735
736 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
737
738 if (SvLEN (data))
739 Safefree (data);
740
741 SvPVX (data) = (char *)ptr;
742 SvCUR_set (data, cb);
743 SvLEN_set (data, 0);
744 SvPOK_only (data);
745
746 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
747
748 if (!(flags & CL_MAP_WRITE))
749 SvREADONLY_on (data);
750
751 return obj;
752}
753
754static void
755mapped_detach (SV *sv, OpenCL__Mapped mapped)
756{
757 SV *data = SvRV (sv);
758
759 // the next check checks both before AND after detach, where SvPVX should STILL be 0
760 if (SvPVX (data) != (char *)mapped->ptr)
761 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
762 else
763 {
764 SvREADONLY_off (data);
765 SvCUR_set (data, 0);
766 SvPVX (data) = 0;
767 SvOK_off (data);
768 }
769
770 mapped->ptr = 0;
771}
772
773static void
774mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
775{
776 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
777 cl_event ev;
778
779 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
780
781 clReleaseEvent (mapped->event);
782 mapped->event = ev;
783
784 mapped_detach (self, mapped);
785}
786
787/*****************************************************************************/
788
174MODULE = OpenCL PACKAGE = OpenCL 789MODULE = OpenCL PACKAGE = OpenCL
175 790
176PROTOTYPES: ENABLE 791PROTOTYPES: ENABLE
177 792
793void
794poll ()
795 CODE:
796 eq_poll ();
797
798void
799_eq_initialise (IV func, IV arg)
800 CODE:
801 eq_signal_func = (void (*)(void *, int))func;
802 eq_signal_arg = (void*)arg;
803
178BOOT: 804BOOT:
179{ 805{
180 HV *stash = gv_stashpv ("OpenCL", 1); 806 HV *stash = gv_stashpv ("OpenCL", 1);
807
181 static const ivstr *civ, const_iv[] = { 808 static const ivstr *civ, const_iv[] = {
182 { sizeof (cl_char ), "SIZEOF_CHAR" }, 809 { sizeof (cl_char ), "SIZEOF_CHAR" },
183 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 810 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
184 { sizeof (cl_short ), "SIZEOF_SHORT" }, 811 { sizeof (cl_short ), "SIZEOF_SHORT" },
185 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 812 { sizeof (cl_ushort), "SIZEOF_USHORT" },
186 { sizeof (cl_int ), "SIZEOF_INT" }, 813 { sizeof (cl_int ), "SIZEOF_INT" },
187 { sizeof (cl_uint ), "SIZEOF_UINT" }, 814 { sizeof (cl_uint ), "SIZEOF_UINT" },
188 { sizeof (cl_long ), "SIZEOF_LONG" }, 815 { sizeof (cl_long ), "SIZEOF_LONG" },
189 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 816 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
190 { sizeof (cl_half ), "SIZEOF_HALF" }, 817 { sizeof (cl_half ), "SIZEOF_HALF" },
191 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 818 { sizeof (cl_float ), "SIZEOF_FLOAT" },
192 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 819 { sizeof (cl_double), "SIZEOF_DOUBLE" },
820 { PREFER_1_1 , "PREFER_1_1" },
193#include "constiv.h" 821#include "constiv.h"
194 }; 822 };
823
195 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 824 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
196 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 825 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
826
827 static const nvstr *cnv, const_nv[] = {
828#include "constnv.h"
829 };
830
831 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
832 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
833
834 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
835
836 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
837 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
838 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
839 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
840 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
841 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
842 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
843 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
844 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
845 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
846 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
847 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
848 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
849 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
850 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
851 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
852 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
853 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
854 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
855 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
856 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
857 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
858 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
859
860 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
197} 861}
198 862
199cl_int 863cl_int
200errno () 864errno ()
201 CODE: 865 CODE:
202 errno = res; 866 RETVAL = res;
867 OUTPUT:
868 RETVAL
203 869
204const char * 870const char *
205err2str (cl_int err) 871err2str (cl_int err = res)
206 872
207const char * 873const char *
208enum2str (cl_uint value) 874enum2str (cl_uint value)
209 875
210void 876void
218 list = tmpbuf (sizeof (*list) * count); 884 list = tmpbuf (sizeof (*list) * count);
219 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 885 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
220 886
221 EXTEND (SP, count); 887 EXTEND (SP, count);
222 for (i = 0; i < count; ++i) 888 for (i = 0; i < count; ++i)
223 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 889 PUSH_CLOBJ (stash_platform, list [i]);
224 890
225void 891void
226context_from_type (FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 892context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
227 PPCODE: 893 PPCODE:
894 CONTEXT_NOTIFY_CALLBACK;
228 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (0, type, 0, 0, &res)); 895 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
229 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 896 XPUSH_CLOBJ_CONTEXT;
230 897
231void 898void
232context (FUTURE properties, FUTURE devices, FUTURE notify = 0) 899context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
233 PPCODE: 900 PPCODE:
234 /* der Gipfel der Kunst */ 901 cl_uint device_count;
902 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
903
904 CONTEXT_NOTIFY_CALLBACK;
905 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
906 XPUSH_CLOBJ_CONTEXT;
235 907
236void 908void
237wait_for_events (...) 909wait_for_events (...)
238 CODE: 910 CODE:
239 EVENT_LIST (0, items); 911 EVENT_LIST (0);
240 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 912 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
241 913
242PROTOTYPES: DISABLE 914PROTOTYPES: DISABLE
243 915
244MODULE = OpenCL PACKAGE = OpenCL::Platform 916MODULE = OpenCL PACKAGE = OpenCL::Platform
245 917
246void 918void
247info (OpenCL::Platform self, cl_platform_info name) 919info (OpenCL::Platform self, cl_platform_info name)
248 PPCODE: 920 PPCODE:
249 INFO (Platform) 921 INFO (Platform)
922
923void
924unload_compiler (OpenCL::Platform self)
925 CODE:
926#if CL_VERSION_1_2
927 clUnloadPlatformCompiler (self);
928#endif
250 929
251#BEGIN:platform 930#BEGIN:platform
252 931
253void 932void
254profile (OpenCL::Platform self) 933profile (OpenCL::Platform self)
280 list = tmpbuf (sizeof (*list) * count); 959 list = tmpbuf (sizeof (*list) * count);
281 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 960 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
282 961
283 EXTEND (SP, count); 962 EXTEND (SP, count);
284 for (i = 0; i < count; ++i) 963 for (i = 0; i < count; ++i)
285 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 964 PUSH_CLOBJ (stash_device, list [i]);
286 965
287void 966void
288context (OpenCL::Platform self, FUTURE properties, SV *devices, FUTURE notify = 0) 967context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
289 PPCODE: 968 PPCODE:
969 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
970 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
971
972 cl_uint device_count;
973 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
974
975 CONTEXT_NOTIFY_CALLBACK;
976 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
977 XPUSH_CLOBJ_CONTEXT;
978
979void
980context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
981 PPCODE:
982 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
983 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
984
985 CONTEXT_NOTIFY_CALLBACK;
986 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
987 XPUSH_CLOBJ_CONTEXT;
988
989MODULE = OpenCL PACKAGE = OpenCL::Device
990
991void
992info (OpenCL::Device self, cl_device_info name)
993 PPCODE:
994 INFO (Device)
995
996#if CL_VERSION_1_2
997
998void
999sub_devices (OpenCL::Device self, SV *properties)
1000 PPCODE:
290 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1001 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
291 croak ("OpenCL::Platform argument 'device' must be an arrayref with device objects, in call"); 1002 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
292 1003
293 AV *av = (AV *)SvRV (devices); 1004 properties = SvRV (properties);
294 cl_uint num_devices = av_len (av) + 1; 1005
295 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1006 cl_uint count = av_len ((AV *)properties) + 1;
1007 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1008
296 int i; 1009 int i;
1010 for (i = 0; i < count; ++i)
1011 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
297 1012
298 for (i = num_devices; i--; ) 1013 props [count] = 0;
299 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
300 1014
301 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (0, num_devices, device_list, 0, 0, &res)); 1015 cl_uint num_devices;
302 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1016 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1017 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1018 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
303 1019
304void 1020 EXTEND (SP, num_devices);
305context_from_type (OpenCL::Platform self, FUTURE properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1021 for (i = 0; i < count; ++i)
306 PPCODE: 1022 PUSH_CLOBJ (stash_subdevice, list [i]);
307 cl_context_properties props[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self, 0 };
308 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
309 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
310 1023
311MODULE = OpenCL PACKAGE = OpenCL::Device 1024#endif
312
313void
314info (OpenCL::Device self, cl_device_info name)
315 PPCODE:
316 INFO (Device)
317 1025
318#BEGIN:device 1026#BEGIN:device
319 1027
320void 1028void
321type (OpenCL::Device self) 1029type (OpenCL::Device self)
322 PPCODE: 1030 PPCODE:
323 cl_device_type value [1]; 1031 cl_device_type value [1];
324 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1032 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
325 EXTEND (SP, 1); 1033 EXTEND (SP, 1);
326 const int i = 0; 1034 const int i = 0;
327 PUSHs (sv_2mortal (newSViv (value [i]))); 1035 PUSHs (sv_2mortal (newSVuv (value [i])));
328 1036
329void 1037void
330vendor_id (OpenCL::Device self) 1038vendor_id (OpenCL::Device self)
331 ALIAS: 1039 ALIAS:
332 vendor_id = CL_DEVICE_VENDOR_ID 1040 vendor_id = CL_DEVICE_VENDOR_ID
353 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1061 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
354 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1062 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
355 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1063 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
356 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1064 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
357 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1065 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
358 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1066 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
359 PPCODE: 1067 PPCODE:
360 cl_uint value [1]; 1068 cl_uint value [1];
361 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1069 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
362 EXTEND (SP, 1); 1070 EXTEND (SP, 1);
363 const int i = 0; 1071 const int i = 0;
477 PPCODE: 1185 PPCODE:
478 cl_command_queue_properties value [1]; 1186 cl_command_queue_properties value [1];
479 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
480 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
481 const int i = 0; 1189 const int i = 0;
482 PUSHs (sv_2mortal (newSViv (value [i]))); 1190 PUSHs (sv_2mortal (newSVuv (value [i])));
483 1191
484void 1192void
485platform (OpenCL::Device self) 1193platform (OpenCL::Device self)
486 PPCODE: 1194 PPCODE:
487 cl_platform_id value [1]; 1195 cl_platform_id value [1];
488 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1196 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
489 EXTEND (SP, 1); 1197 EXTEND (SP, 1);
490 const int i = 0; 1198 const int i = 0;
491 { 1199 {
492 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1200 PUSH_CLOBJ (stash_platform, value [i]);
493 } 1201 }
494 1202
495void 1203void
496name (OpenCL::Device self) 1204name (OpenCL::Device self)
497 ALIAS: 1205 ALIAS:
516 cl_device_id value [1]; 1224 cl_device_id value [1];
517 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1225 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
518 EXTEND (SP, 1); 1226 EXTEND (SP, 1);
519 const int i = 0; 1227 const int i = 0;
520 { 1228 {
521 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1229 PUSH_CLOBJ (stash_device, value [i]);
522 } 1230 }
523 1231
524void 1232void
525partition_types_ext (OpenCL::Device self) 1233partition_types_ext (OpenCL::Device self)
526 ALIAS: 1234 ALIAS:
537 for (i = 0; i < n; ++i) 1245 for (i = 0; i < n; ++i)
538 PUSHs (sv_2mortal (newSVuv (value [i]))); 1246 PUSHs (sv_2mortal (newSVuv (value [i])));
539 1247
540#END:device 1248#END:device
541 1249
1250MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1251
1252#if CL_VERSION_1_2
1253
1254void
1255DESTROY (OpenCL::SubDevice self)
1256 CODE:
1257 clReleaseDevice (self);
1258
1259#endif
1260
542MODULE = OpenCL PACKAGE = OpenCL::Context 1261MODULE = OpenCL PACKAGE = OpenCL::Context
543 1262
544void 1263void
545DESTROY (OpenCL::Context context) 1264DESTROY (OpenCL::Context self)
546 CODE: 1265 CODE:
547 clReleaseContext (context); 1266 clReleaseContext (self);
548 1267
549void 1268void
550info (OpenCL::Context self, cl_context_info name) 1269info (OpenCL::Context self, cl_context_info name)
551 PPCODE: 1270 PPCODE:
552 INFO (Context) 1271 INFO (Context)
553 1272
554void 1273void
555queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1274queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
556 PPCODE: 1275 PPCODE:
557 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1276 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
558 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1277 XPUSH_CLOBJ (stash_queue, queue);
559 1278
560void 1279void
561user_event (OpenCL::Context self) 1280user_event (OpenCL::Context self)
562 PPCODE: 1281 PPCODE:
563 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1282 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
564 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1283 XPUSH_CLOBJ (stash_userevent, ev);
565 1284
566void 1285void
567buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1286buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
568 PPCODE: 1287 PPCODE:
569 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1288 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
570 croak ("clCreateBuffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1289 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
571 1290
572 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1291 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
573 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1292 XPUSH_CLOBJ (stash_bufferobj, mem);
574 1293
575void 1294void
576buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1295buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
577 PPCODE: 1296 PPCODE:
578 STRLEN len; 1297 STRLEN len;
579 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1298 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
580 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1299 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
581 croak ("clCreateBuffer: have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1300 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
582 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1301 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
583 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1302 XPUSH_CLOBJ (stash_bufferobj, mem);
584 1303
1304#if CL_VERSION_1_2
1305
585void 1306void
586image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1307image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
587 PPCODE: 1308 PPCODE:
588 STRLEN len; 1309 STRLEN len;
589 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1310 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
590 const cl_image_format format = { channel_order, channel_type }; 1311 const cl_image_format format = { channel_order, channel_type };
1312 const cl_image_desc desc = {
1313 type,
1314 width, height, depth,
1315 array_size, row_pitch, slice_pitch,
1316 num_mip_level, num_samples,
1317 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1318 };
591 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
592 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1320 HV *stash = stash_image;
1321 switch (type)
1322 {
1323 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1324 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1325 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1326 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1327 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1328 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1329 }
1330 XPUSH_CLOBJ (stash, mem);
593 1331
1332#endif
1333
594void 1334void
595image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1335image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
596 PPCODE: 1336 PPCODE:
597 STRLEN len; 1337 STRLEN len;
598 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1338 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
599 const cl_image_format format = { channel_order, channel_type }; 1339 const cl_image_format format = { channel_order, channel_type };
1340#if PREFER_1_1
1341 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1342#else
1343 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1344 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1345#endif
1346 XPUSH_CLOBJ (stash_image2d, mem);
1347
1348void
1349image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1350 PPCODE:
1351 STRLEN len;
1352 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1353 const cl_image_format format = { channel_order, channel_type };
1354#if PREFER_1_1
600 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1355 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
601 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1356#else
1357 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1358 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1359#endif
1360 XPUSH_CLOBJ (stash_image3d, mem);
1361
1362#if cl_apple_gl_sharing || cl_khr_gl_sharing
1363
1364void
1365gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1366 PPCODE:
1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1368 XPUSH_CLOBJ (stash_bufferobj, mem);
1369
1370void
1371gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1372 PPCODE:
1373 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1374 XPUSH_CLOBJ (stash_image2d, mem);
1375
1376#if CL_VERSION_1_2
1377
1378void
1379gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1380 ALIAS:
1381 PPCODE:
1382 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1383 cl_gl_object_type type;
1384 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1385 HV *stash = stash_memory;
1386 switch (type)
1387 {
1388 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1389 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1390 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1391 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1392 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1393 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1394 }
1395 XPUSH_CLOBJ (stash, mem);
1396
1397#endif
1398
1399void
1400gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1401 PPCODE:
1402#if PREFER_1_1
1403 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1404#else
1405 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1406#endif
1407 XPUSH_CLOBJ (stash_image2d, mem);
1408
1409void
1410gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1411 PPCODE:
1412#if PREFER_1_1
1413 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1414#else
1415 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1416#endif
1417 XPUSH_CLOBJ (stash_image3d, mem);
1418
1419#endif
602 1420
603void 1421void
604supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1422supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
605 PPCODE: 1423 PPCODE:
606{ 1424{
607 cl_uint count; 1425 cl_uint count;
608 cl_image_format *list; 1426 cl_image_format *list;
609 int i; 1427 int i;
610 1428
611 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count)); 1429 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, 0, 0, &count));
612 Newx (list, count, cl_image_format); 1430 Newx (list, count, cl_image_format);
613 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0)); 1431 NEED_SUCCESS (GetSupportedImageFormats, (self, flags, image_type, count, list, 0));
614 1432
615 EXTEND (SP, count); 1433 EXTEND (SP, count);
616 for (i = 0; i < count; ++i) 1434 for (i = 0; i < count; ++i)
617 { 1435 {
618 AV *av = newAV (); 1436 AV *av = newAV ();
623} 1441}
624 1442
625void 1443void
626sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1444sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
627 PPCODE: 1445 PPCODE:
628 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1446 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
629 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1447 XPUSH_CLOBJ (stash_sampler, sampler);
630 1448
631void 1449void
632program_with_source (OpenCL::Context self, SV *program) 1450program_with_source (OpenCL::Context self, SV *program)
633 PPCODE: 1451 PPCODE:
634 STRLEN len; 1452 STRLEN len;
635 size_t len2; 1453 size_t len2;
636 const char *ptr = SvPVbyte (program, len); 1454 const char *ptr = SvPVbyte (program, len);
637 1455
638 len2 = len; 1456 len2 = len;
639 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1457 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
640 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1458 XPUSH_CLOBJ (stash_program, prog);
1459
1460void
1461program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1462 PPCODE:
1463 cl_uint device_count;
1464 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1465
1466 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1467 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1468
1469 binaries = SvRV (binaries);
1470
1471 if (device_count != av_len ((AV *)binaries) + 1)
1472 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1473
1474 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1475 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1476 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1477
1478 int i;
1479 for (i = 0; i < device_count; ++i)
1480 {
1481 STRLEN len;
1482 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1483 length_list [i] = len;
1484 }
1485
1486 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1487 length_list, binary_list,
1488 GIMME_V == G_ARRAY ? status_list : 0, &res));
1489
1490 EXTEND (SP, 2);
1491 PUSH_CLOBJ (stash_program, prog);
1492
1493 if (GIMME_V == G_ARRAY)
1494 {
1495 AV *av = newAV ();
1496 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1497
1498 for (i = device_count; i--; )
1499 av_store (av, i, newSViv (status_list [i]));
1500 }
1501
1502#if CL_VERSION_1_2
1503
1504void
1505program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1506 PPCODE:
1507 cl_uint device_count;
1508 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1509
1510 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514void
1515link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1516 CODE:
1517 cl_uint device_count = 0;
1518 cl_device_id *device_list = 0;
1519
1520 if (SvOK (devices))
1521 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1522
1523 cl_uint program_count;
1524 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1525
1526 void *user_data;
1527 program_callback pfn_notify = make_program_callback (notify, &user_data);
1528
1529 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1530 program_count, program_list, pfn_notify, user_data, &res));
1531
1532 XPUSH_CLOBJ (stash_program, prog);
1533
1534#endif
641 1535
642#BEGIN:context 1536#BEGIN:context
643 1537
644void 1538void
645reference_count (OpenCL::Context self) 1539reference_count (OpenCL::Context self)
662 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1556 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
663 int i, n = size / sizeof (*value); 1557 int i, n = size / sizeof (*value);
664 EXTEND (SP, n); 1558 EXTEND (SP, n);
665 for (i = 0; i < n; ++i) 1559 for (i = 0; i < n; ++i)
666 { 1560 {
667 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1561 PUSH_CLOBJ (stash_device, value [i]);
668 } 1562 }
669 1563
670void 1564void
671properties (OpenCL::Context self) 1565properties (OpenCL::Context self)
672 PPCODE: 1566 PPCODE:
687DESTROY (OpenCL::Queue self) 1581DESTROY (OpenCL::Queue self)
688 CODE: 1582 CODE:
689 clReleaseCommandQueue (self); 1583 clReleaseCommandQueue (self);
690 1584
691void 1585void
692enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1586read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1587 ALIAS:
1588 enqueue_read_buffer = 0
693 PPCODE: 1589 PPCODE:
694 cl_event ev = 0;
695 EVENT_LIST (6, items - 6); 1590 EVENT_LIST (6);
696 1591
697 SvUPGRADE (data, SVt_PV); 1592 SvUPGRADE (data, SVt_PV);
698 SvGROW (data, len); 1593 SvGROW (data, len);
699 SvPOK_only (data); 1594 SvPOK_only (data);
700 SvCUR_set (data, len); 1595 SvCUR_set (data, len);
1596
1597 cl_event ev = 0;
701 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1598 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
702 1599
703 if (ev) 1600 if (ev)
704 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1601 XPUSH_CLOBJ (stash_event, ev);
705 1602
706void 1603void
707enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1604write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1605 ALIAS:
1606 enqueue_write_buffer = 0
708 PPCODE: 1607 PPCODE:
709 cl_event ev = 0; 1608 EVENT_LIST (5);
1609
710 STRLEN len; 1610 STRLEN len;
711 char *ptr = SvPVbyte (data, len); 1611 char *ptr = SvPVbyte (data, len);
712 EVENT_LIST (5, items - 5);
713 1612
1613 cl_event ev = 0;
714 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
715 1615
716 if (ev) 1616 if (ev)
717 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1617 XPUSH_CLOBJ (stash_event, ev);
718 1618
1619#if CL_VERSION_1_2
1620
719void 1621void
720enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1622fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1623 ALIAS:
1624 enqueue_fill_buffer = 0
721 PPCODE: 1625 PPCODE:
1626 EVENT_LIST (5);
1627
1628 STRLEN len;
1629 char *ptr = SvPVbyte (data, len);
1630
722 cl_event ev = 0; 1631 cl_event ev = 0;
1632 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1633
1634 if (ev)
1635 XPUSH_CLOBJ (stash_event, ev);
1636
1637void
1638fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1639 ALIAS:
1640 enqueue_fill_image = 0
1641 PPCODE:
1642 EVENT_LIST (12);
1643
1644 const size_t origin [3] = { x, y, z };
1645 const size_t region [3] = { width, height, depth };
1646
1647 const cl_float c_f [4] = { r, g, b, a };
1648 const cl_uint c_u [4] = { r, g, b, a };
1649 const cl_int c_s [4] = { r, g, b, a };
1650 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1651 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1652 cl_image_format format;
1653 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1654 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1655 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1656 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1657
1658 cl_event ev = 0;
1659 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1660 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1661
1662 if (ev)
1663 XPUSH_CLOBJ (stash_event, ev);
1664
1665#endif
1666
1667void
1668copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1669 ALIAS:
1670 enqueue_copy_buffer = 0
1671 PPCODE:
723 EVENT_LIST (6, items - 6); 1672 EVENT_LIST (6);
724 1673
1674 cl_event ev = 0;
725 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1675 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
726 1676
727 if (ev) 1677 if (ev)
728 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1678 XPUSH_CLOBJ (stash_event, ev);
729 1679
730void 1680void
731enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1681read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1682 ALIAS:
1683 enqueue_read_buffer_rect = 0
732 PPCODE: 1684 PPCODE:
733 cl_event ev = 0; 1685 EVENT_LIST (17);
1686
734 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1687 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
735 const size_t host_origin[3] = { host_x, host_y, host_z }; 1688 const size_t host_origin[3] = { host_x, host_y, host_z };
736 const size_t region[3] = { width, height, depth }; 1689 const size_t region[3] = { width, height, depth };
737 EVENT_LIST (17, items - 17);
738 1690
739 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
740 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
741 1693
742 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
752 1704
753 SvUPGRADE (data, SVt_PV); 1705 SvUPGRADE (data, SVt_PV);
754 SvGROW (data, len); 1706 SvGROW (data, len);
755 SvPOK_only (data); 1707 SvPOK_only (data);
756 SvCUR_set (data, len); 1708 SvCUR_set (data, len);
1709
1710 cl_event ev = 0;
757 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1711 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
758 1712
759 if (ev) 1713 if (ev)
760 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1714 XPUSH_CLOBJ (stash_event, ev);
761 1715
762void 1716void
763enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1717write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1718 ALIAS:
1719 enqueue_write_buffer_rect = 0
764 PPCODE: 1720 PPCODE:
765 cl_event ev = 0; 1721 EVENT_LIST (17);
1722
766 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1723 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
767 const size_t host_origin[3] = { host_x, host_y, host_z }; 1724 const size_t host_origin[3] = { host_x, host_y, host_z };
768 const size_t region[3] = { width, height, depth }; 1725 const size_t region[3] = { width, height, depth };
769 STRLEN len; 1726 STRLEN len;
770 char *ptr = SvPVbyte (data, len); 1727 char *ptr = SvPVbyte (data, len);
771 EVENT_LIST (17, items - 17);
772 1728
773 if (!buf_row_pitch) 1729 if (!buf_row_pitch)
774 buf_row_pitch = region [0]; 1730 buf_row_pitch = region [0];
775 1731
776 if (!buf_slice_pitch) 1732 if (!buf_slice_pitch)
785 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1741 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
786 1742
787 if (len < min_len) 1743 if (len < min_len)
788 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1744 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
789 1745
1746 cl_event ev = 0;
790 NEED_SUCCESS (EnqueueWriteBufferRect, (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)); 1747 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
791 1748
792 if (ev) 1749 if (ev)
793 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1750 XPUSH_CLOBJ (stash_event, ev);
794 1751
795void 1752void
796enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1753copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1754 ALIAS:
1755 enqueue_copy_buffer_rect = 0
797 PPCODE: 1756 PPCODE:
798 cl_event ev = 0; 1757 EVENT_LIST (16);
1758
799 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
800 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1760 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
801 const size_t region[3] = { width, height, depth }; 1761 const size_t region[3] = { width, height, depth };
802 EVENT_LIST (16, items - 16);
803 1762
1763 cl_event ev = 0;
804 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1764 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
805 1765
806 if (ev) 1766 if (ev)
807 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1767 XPUSH_CLOBJ (stash_event, ev);
808 1768
809void 1769void
810enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1770read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1771 ALIAS:
1772 enqueue_read_image = 0
811 PPCODE: 1773 PPCODE:
812 cl_event ev = 0; 1774 EVENT_LIST (12);
1775
813 const size_t src_origin[3] = { src_x, src_y, src_z }; 1776 const size_t src_origin[3] = { src_x, src_y, src_z };
814 const size_t region[3] = { width, height, depth }; 1777 const size_t region[3] = { width, height, depth };
815 EVENT_LIST (12, items - 12);
816 1778
817 if (!row_pitch) 1779 if (!row_pitch)
818 row_pitch = img_row_pitch (src); 1780 row_pitch = img_row_pitch (src);
819 1781
820 if (depth > 1 && !slice_pitch) 1782 if (depth > 1 && !slice_pitch)
824 1786
825 SvUPGRADE (data, SVt_PV); 1787 SvUPGRADE (data, SVt_PV);
826 SvGROW (data, len); 1788 SvGROW (data, len);
827 SvPOK_only (data); 1789 SvPOK_only (data);
828 SvCUR_set (data, len); 1790 SvCUR_set (data, len);
1791
1792 cl_event ev = 0;
829 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1793 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
830 1794
831 if (ev) 1795 if (ev)
832 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1796 XPUSH_CLOBJ (stash_event, ev);
833 1797
834void 1798void
835enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1799write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1800 ALIAS:
1801 enqueue_write_image = 0
836 PPCODE: 1802 PPCODE:
837 cl_event ev = 0; 1803 EVENT_LIST (12);
1804
838 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1805 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
839 const size_t region[3] = { width, height, depth }; 1806 const size_t region[3] = { width, height, depth };
840 STRLEN len; 1807 STRLEN len;
841 char *ptr = SvPVbyte (data, len); 1808 char *ptr = SvPVbyte (data, len);
842 EVENT_LIST (12, items - 12);
843 1809
844 if (!row_pitch) 1810 if (!row_pitch)
845 row_pitch = img_row_pitch (dst); 1811 row_pitch = img_row_pitch (dst);
846 1812
847 if (depth > 1 && !slice_pitch) 1813 if (depth > 1 && !slice_pitch)
850 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1816 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
851 1817
852 if (len < min_len) 1818 if (len < min_len)
853 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1819 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
854 1820
1821 cl_event ev = 0;
855 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1822 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
856 1823
857 if (ev) 1824 if (ev)
858 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1825 XPUSH_CLOBJ (stash_event, ev);
859 1826
860void 1827void
861enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1828copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1829 ALIAS:
1830 enqueue_copy_image = 0
862 PPCODE: 1831 PPCODE:
863 cl_event ev = 0; 1832 EVENT_LIST (12);
1833
864 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
865 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1835 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
866 const size_t region[3] = { width, height, depth }; 1836 const size_t region[3] = { width, height, depth };
867 EVENT_LIST (12, items - 12);
868 1837
1838 cl_event ev = 0;
869 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1839 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
870 1840
871 if (ev) 1841 if (ev)
872 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1842 XPUSH_CLOBJ (stash_event, ev);
873 1843
874void 1844void
875enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1845copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1846 ALIAS:
1847 enqueue_copy_image_to_buffer = 0
876 PPCODE: 1848 PPCODE:
1849 EVENT_LIST (10);
1850
1851 const size_t src_origin[3] = { src_x, src_y, src_z };
1852 const size_t region [3] = { width, height, depth };
1853
877 cl_event ev = 0; 1854 cl_event ev = 0;
1855 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1856
1857 if (ev)
1858 XPUSH_CLOBJ (stash_event, ev);
1859
1860void
1861copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1862 ALIAS:
1863 enqueue_copy_buffer_to_image = 0
1864 PPCODE:
1865 EVENT_LIST (10);
1866
878 const size_t src_origin[3] = { src_x, src_y, src_z }; 1867 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1868 const size_t region [3] = { width, height, depth };
1869
1870 cl_event ev = 0;
1871 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1872
1873 if (ev)
1874 XPUSH_CLOBJ (stash_event, ev);
1875
1876void
1877map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1878 ALIAS:
1879 enqueue_map_buffer = 0
1880 PPCODE:
1881 EVENT_LIST (6);
1882
1883 size_t cb = SvIV (cb_);
1884
1885 if (!SvOK (cb_))
1886 {
1887 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1888 cb -= offset;
1889 }
1890
1891 cl_event ev;
1892 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1893 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1894
1895void
1896map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1897 ALIAS:
1898 enqueue_map_image = 0
1899 PPCODE:
1900 size_t width = SvIV (width_);
1901 if (!SvOK (width_))
1902 {
1903 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1904 width -= x;
1905 }
1906
1907 size_t height = SvIV (width_);
1908 if (!SvOK (height_))
1909 {
1910 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1911 height -= y;
1912 }
1913
1914 size_t depth = SvIV (width_);
1915 if (!SvOK (depth_))
1916 {
1917 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1918 depth -= z;
1919
1920 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1921 if (!depth)
1922 depth = 1;
1923 }
1924
1925 const size_t origin[3] = { x, y, z };
879 const size_t region[3] = { width, height, depth }; 1926 const size_t region[3] = { width, height, depth };
1927 size_t row_pitch, slice_pitch;
880 EVENT_LIST (10, items - 10); 1928 EVENT_LIST (10);
881 1929
1930 cl_event ev;
1931 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1932
1933 size_t cb = slice_pitch ? slice_pitch * region [2]
1934 : row_pitch ? row_pitch * region [1]
1935 : region [0];
1936
1937 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1938
1939void
1940unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1941 PPCODE:
1942 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1943 if (GIMME_V != G_VOID)
1944 {
1945 clRetainEvent (mapped->event);
1946 XPUSH_CLOBJ (stash_event, mapped->event);
1947 }
1948
1949void
1950task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1951 ALIAS:
1952 enqueue_task = 0
1953 PPCODE:
1954 EVENT_LIST (2);
1955
1956 cl_event ev = 0;
882 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1957 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
883 1958
884 if (ev) 1959 if (ev)
885 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1960 XPUSH_CLOBJ (stash_event, ev);
886 1961
887void 1962void
888enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
889 PPCODE:
890 cl_event ev = 0;
891 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
892 const size_t region[3] = { width, height, depth };
893 EVENT_LIST (10, items - 10);
894
895 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
896
897 if (ev)
898 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
899
900void
901enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
902 PPCODE:
903 cl_event ev = 0;
904 EVENT_LIST (2, items - 2);
905
906 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
907
908 if (ev)
909 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
910
911void
912enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1963nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1964 ALIAS:
1965 enqueue_nd_range_kernel = 0
913 PPCODE: 1966 PPCODE:
914 cl_event ev = 0; 1967 EVENT_LIST (5);
1968
915 size_t *gwo = 0, *gws, *lws = 0; 1969 size_t *gwo = 0, *gws, *lws = 0;
916 int gws_len; 1970 int gws_len;
917 size_t *lists; 1971 size_t *lists;
918 int i; 1972 int i;
919 EVENT_LIST (5, items - 5);
920 1973
921 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1974 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
922 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
923 1976
924 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1977 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
925 1978
926 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1979 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
927 1980
928 gws = lists + gws_len * 0; 1981 gws = lists + gws_len * 0;
929 for (i = 0; i < gws_len; ++i) 1982 for (i = 0; i < gws_len; ++i)
1983 {
930 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1984 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1985 // at least nvidia crashes for 0-sized work group sizes, work around
1986 if (!gws [i])
1987 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1988 }
931 1989
932 if (SvOK (global_work_offset)) 1990 if (SvOK (global_work_offset))
933 { 1991 {
934 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1992 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
935 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1993 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
942 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 2000 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
943 } 2001 }
944 2002
945 if (SvOK (local_work_size)) 2003 if (SvOK (local_work_size))
946 { 2004 {
947 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 2005 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
948 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 2006 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
949 2007
950 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 2008 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
951 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 2009 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
952 2010
953 lws = lists + gws_len * 2; 2011 lws = lists + gws_len * 2;
954 for (i = 0; i < gws_len; ++i) 2012 for (i = 0; i < gws_len; ++i)
2013 {
955 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 2014 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
2015 // at least nvidia crashes for 0-sized work group sizes, work around
2016 if (!lws [i])
2017 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2018 }
956 } 2019 }
957 2020
2021 cl_event ev = 0;
958 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2022 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
959 2023
960 if (ev) 2024 if (ev)
961 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2025 XPUSH_CLOBJ (stash_event, ev);
962 2026
2027#if CL_VERSION_1_2
2028
963void 2029void
964enqueue_marker (OpenCL::Queue self) 2030migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2031 ALIAS:
2032 enqueue_migrate_mem_objects = 0
965 PPCODE: 2033 PPCODE:
2034 EVENT_LIST (3);
2035
2036 cl_uint obj_count;
2037 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2038
2039 cl_event ev = 0;
2040 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2041
2042 if (ev)
2043 XPUSH_CLOBJ (stash_event, ev);
2044
2045#endif
2046
2047#if cl_apple_gl_sharing || cl_khr_gl_sharing
2048
2049void
2050acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
2051 ALIAS:
2052 release_gl_objects = 1
2053 enqueue_acquire_gl_objects = 0
2054 enqueue_release_gl_objects = 1
2055 PPCODE:
2056 EVENT_LIST (2);
2057
2058 cl_uint obj_count;
2059 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2060
966 cl_event ev; 2061 cl_event ev = 0;
967 NEED_SUCCESS (EnqueueMarker, (self, &ev));
968 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
969 2062
2063 if (ix)
2064 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2065 else
2066 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2067
2068 if (ev)
2069 XPUSH_CLOBJ (stash_event, ev);
2070
2071#endif
2072
970void 2073void
971enqueue_wait_for_events (OpenCL::Queue self, ...) 2074wait_for_events (OpenCL::Queue self, ...)
2075 ALIAS:
2076 enqueue_wait_for_events = 0
972 CODE: 2077 CODE:
973 EVENT_LIST (1, items - 1); 2078 EVENT_LIST (1);
2079#if PREFER_1_1
974 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2080 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2081#else
2082 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2083#endif
975 2084
976void 2085void
977enqueue_barrier (OpenCL::Queue self) 2086marker (OpenCL::Queue self, ...)
2087 ALIAS:
2088 enqueue_marker = 0
978 CODE: 2089 PPCODE:
2090 EVENT_LIST (1);
2091 cl_event ev = 0;
2092#if PREFER_1_1
2093 if (!event_list_count)
2094 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2095 else
2096#if CL_VERSION_1_2
2097 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2098#else
2099 {
2100 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2101 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2102 }
2103#endif
2104#else
2105 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2106#endif
2107 if (ev)
2108 XPUSH_CLOBJ (stash_event, ev);
2109
2110void
2111barrier (OpenCL::Queue self, ...)
2112 ALIAS:
2113 enqueue_barrier = 0
2114 PPCODE:
2115 EVENT_LIST (1);
2116 cl_event ev = 0;
2117#if PREFER_1_1
2118 if (!event_list_count && GIMME_V == G_VOID)
979 NEED_SUCCESS (EnqueueBarrier, (self)); 2119 NEED_SUCCESS (EnqueueBarrier, (self));
2120 else
2121#if CL_VERSION_1_2
2122 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2123#else
2124 {
2125 if (event_list_count)
2126 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2127
2128 if (GIMME_V != G_VOID)
2129 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2130 }
2131#endif
2132#else
2133 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2134#endif
2135 if (ev)
2136 XPUSH_CLOBJ (stash_event, ev);
980 2137
981void 2138void
982flush (OpenCL::Queue self) 2139flush (OpenCL::Queue self)
983 CODE: 2140 CODE:
984 NEED_SUCCESS (Flush, (self)); 2141 NEED_SUCCESS (Flush, (self));
1002 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2159 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1003 EXTEND (SP, 1); 2160 EXTEND (SP, 1);
1004 const int i = 0; 2161 const int i = 0;
1005 { 2162 {
1006 NEED_SUCCESS (RetainContext, (value [i])); 2163 NEED_SUCCESS (RetainContext, (value [i]));
1007 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2164 PUSH_CLOBJ (stash_context, value [i]);
1008 } 2165 }
1009 2166
1010void 2167void
1011device (OpenCL::Queue self) 2168device (OpenCL::Queue self)
1012 PPCODE: 2169 PPCODE:
1013 cl_device_id value [1]; 2170 cl_device_id value [1];
1014 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2171 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1015 EXTEND (SP, 1); 2172 EXTEND (SP, 1);
1016 const int i = 0; 2173 const int i = 0;
1017 { 2174 {
1018 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2175 PUSH_CLOBJ (stash_device, value [i]);
1019 } 2176 }
1020 2177
1021void 2178void
1022reference_count (OpenCL::Queue self) 2179reference_count (OpenCL::Queue self)
1023 PPCODE: 2180 PPCODE:
1032 PPCODE: 2189 PPCODE:
1033 cl_command_queue_properties value [1]; 2190 cl_command_queue_properties value [1];
1034 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2191 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1035 EXTEND (SP, 1); 2192 EXTEND (SP, 1);
1036 const int i = 0; 2193 const int i = 0;
1037 PUSHs (sv_2mortal (newSViv (value [i]))); 2194 PUSHs (sv_2mortal (newSVuv (value [i])));
1038 2195
1039#END:command_queue 2196#END:command_queue
1040 2197
1041MODULE = OpenCL PACKAGE = OpenCL::Memory 2198MODULE = OpenCL PACKAGE = OpenCL::Memory
1042 2199
1057 PPCODE: 2214 PPCODE:
1058 cl_mem_object_type value [1]; 2215 cl_mem_object_type value [1];
1059 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2216 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1060 EXTEND (SP, 1); 2217 EXTEND (SP, 1);
1061 const int i = 0; 2218 const int i = 0;
1062 PUSHs (sv_2mortal (newSViv (value [i]))); 2219 PUSHs (sv_2mortal (newSVuv (value [i])));
1063 2220
1064void 2221void
1065flags (OpenCL::Memory self) 2222flags (OpenCL::Memory self)
1066 PPCODE: 2223 PPCODE:
1067 cl_mem_flags value [1]; 2224 cl_mem_flags value [1];
1068 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2225 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1069 EXTEND (SP, 1); 2226 EXTEND (SP, 1);
1070 const int i = 0; 2227 const int i = 0;
1071 PUSHs (sv_2mortal (newSViv (value [i]))); 2228 PUSHs (sv_2mortal (newSVuv (value [i])));
1072 2229
1073void 2230void
1074size (OpenCL::Memory self) 2231size (OpenCL::Memory self)
1075 ALIAS: 2232 ALIAS:
1076 size = CL_MEM_SIZE 2233 size = CL_MEM_SIZE
1110 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2267 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1111 EXTEND (SP, 1); 2268 EXTEND (SP, 1);
1112 const int i = 0; 2269 const int i = 0;
1113 { 2270 {
1114 NEED_SUCCESS (RetainContext, (value [i])); 2271 NEED_SUCCESS (RetainContext, (value [i]));
1115 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2272 PUSH_CLOBJ (stash_context, value [i]);
1116 } 2273 }
1117 2274
1118void 2275void
1119associated_memobject (OpenCL::Memory self) 2276associated_memobject (OpenCL::Memory self)
1120 PPCODE: 2277 PPCODE:
1122 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2279 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1123 EXTEND (SP, 1); 2280 EXTEND (SP, 1);
1124 const int i = 0; 2281 const int i = 0;
1125 { 2282 {
1126 NEED_SUCCESS (RetainMemObject, (value [i])); 2283 NEED_SUCCESS (RetainMemObject, (value [i]));
1127 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2284 PUSH_CLOBJ (stash_memory, value [i]);
1128 } 2285 }
1129 2286
1130#END:mem 2287#END:mem
2288
2289#if cl_apple_gl_sharing || cl_khr_gl_sharing
2290
2291void
2292gl_object_info (OpenCL::Memory self)
2293 PPCODE:
2294 cl_gl_object_type type;
2295 cl_GLuint name;
2296 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
2297 EXTEND (SP, 2);
2298 PUSHs (sv_2mortal (newSVuv (type)));
2299 PUSHs (sv_2mortal (newSVuv (name)));
2300
2301#endif
1131 2302
1132MODULE = OpenCL PACKAGE = OpenCL::BufferObj 2303MODULE = OpenCL PACKAGE = OpenCL::BufferObj
1133 2304
1134void 2305void
1135sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size) 2306sub_buffer_region (OpenCL::BufferObj self, cl_mem_flags flags, size_t origin, size_t size)
1138 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2309 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1139 2310
1140 cl_buffer_region crdata = { origin, size }; 2311 cl_buffer_region crdata = { origin, size };
1141 2312
1142 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2313 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1143 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2314 XPUSH_CLOBJ (stash_buffer, mem);
1144 2315
1145MODULE = OpenCL PACKAGE = OpenCL::Image 2316MODULE = OpenCL PACKAGE = OpenCL::Image
1146 2317
1147void 2318void
1148image_info (OpenCL::Image self, cl_image_info name) 2319image_info (OpenCL::Image self, cl_image_info name)
1149 PPCODE: 2320 PPCODE:
1150 INFO (Image) 2321 INFO (Image)
2322
2323void
2324format (OpenCL::Image self)
2325 PPCODE:
2326 cl_image_format format;
2327 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2328 EXTEND (SP, 2);
2329 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2330 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1151 2331
1152#BEGIN:image 2332#BEGIN:image
1153 2333
1154void 2334void
1155element_size (OpenCL::Image self) 2335element_size (OpenCL::Image self)
1167 const int i = 0; 2347 const int i = 0;
1168 PUSHs (sv_2mortal (newSVuv (value [i]))); 2348 PUSHs (sv_2mortal (newSVuv (value [i])));
1169 2349
1170#END:image 2350#END:image
1171 2351
2352#if cl_apple_gl_sharing || cl_khr_gl_sharing
2353
2354#BEGIN:gl_texture
2355
2356void
2357target (OpenCL::Image self)
2358 PPCODE:
2359 cl_GLenum value [1];
2360 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
2361 EXTEND (SP, 1);
2362 const int i = 0;
2363 PUSHs (sv_2mortal (newSVuv (value [i])));
2364
2365void
2366gl_mipmap_level (OpenCL::Image self)
2367 PPCODE:
2368 cl_GLint value [1];
2369 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
2370 EXTEND (SP, 1);
2371 const int i = 0;
2372 PUSHs (sv_2mortal (newSViv (value [i])));
2373
2374#END:gl_texture
2375
2376#endif
2377
1172MODULE = OpenCL PACKAGE = OpenCL::Sampler 2378MODULE = OpenCL PACKAGE = OpenCL::Sampler
1173 2379
1174void 2380void
1175DESTROY (OpenCL::Sampler self) 2381DESTROY (OpenCL::Sampler self)
1176 CODE: 2382 CODE:
1199 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2405 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1200 EXTEND (SP, 1); 2406 EXTEND (SP, 1);
1201 const int i = 0; 2407 const int i = 0;
1202 { 2408 {
1203 NEED_SUCCESS (RetainContext, (value [i])); 2409 NEED_SUCCESS (RetainContext, (value [i]));
1204 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2410 PUSH_CLOBJ (stash_context, value [i]);
1205 } 2411 }
1206 2412
1207void 2413void
1208normalized_coords (OpenCL::Sampler self) 2414normalized_coords (OpenCL::Sampler self)
1209 PPCODE: 2415 PPCODE:
1210 cl_addressing_mode value [1]; 2416 cl_addressing_mode value [1];
1211 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2417 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1212 EXTEND (SP, 1); 2418 EXTEND (SP, 1);
1213 const int i = 0; 2419 const int i = 0;
1214 PUSHs (sv_2mortal (newSViv (value [i]))); 2420 PUSHs (sv_2mortal (newSVuv (value [i])));
1215 2421
1216void 2422void
1217addressing_mode (OpenCL::Sampler self) 2423addressing_mode (OpenCL::Sampler self)
1218 PPCODE: 2424 PPCODE:
1219 cl_filter_mode value [1]; 2425 cl_filter_mode value [1];
1220 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2426 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1221 EXTEND (SP, 1); 2427 EXTEND (SP, 1);
1222 const int i = 0; 2428 const int i = 0;
1223 PUSHs (sv_2mortal (newSViv (value [i]))); 2429 PUSHs (sv_2mortal (newSVuv (value [i])));
1224 2430
1225void 2431void
1226filter_mode (OpenCL::Sampler self) 2432filter_mode (OpenCL::Sampler self)
1227 PPCODE: 2433 PPCODE:
1228 cl_bool value [1]; 2434 cl_bool value [1];
1239DESTROY (OpenCL::Program self) 2445DESTROY (OpenCL::Program self)
1240 CODE: 2446 CODE:
1241 clReleaseProgram (self); 2447 clReleaseProgram (self);
1242 2448
1243void 2449void
1244build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2450build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2451 ALIAS:
2452 build_async = 1
1245 CODE: 2453 CODE:
2454 cl_uint device_count = 0;
2455 cl_device_id *device_list = 0;
2456
2457 if (SvOK (devices))
2458 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2459
2460 void *user_data;
2461 program_callback pfn_notify = make_program_callback (notify, &user_data);
2462
2463 if (ix)
2464 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2465 else
2466 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2467
2468#if CL_VERSION_1_2
2469
2470void
2471compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2472 CODE:
2473 cl_uint device_count = 0;
2474 cl_device_id *device_list = 0;
2475
2476 if (SvOK (devices))
2477 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2478
2479 cl_uint header_count = 0;
2480 cl_program *header_list = 0;
2481 const char **header_name = 0;
2482
2483 if (SvOK (headers))
2484 {
2485 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2486 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2487
2488 HV *hv = (HV *)SvRV (devices);
2489
2490 header_count = hv_iterinit (hv);
2491 header_list = tmpbuf (sizeof (*header_list) * header_count);
2492 header_name = tmpbuf (sizeof (*header_name) * header_count);
2493
2494 HE *he;
2495 int i = 0;
2496 while (he = hv_iternext (hv))
2497 {
2498 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2499 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2500 ++i;
2501 }
2502 }
2503
2504 void *user_data;
2505 program_callback pfn_notify = make_program_callback (notify, &user_data);
2506
1246 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2507 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2508 header_count, header_list, header_name, pfn_notify, user_data));
2509
2510#endif
1247 2511
1248void 2512void
1249build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2513build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1250 PPCODE: 2514 PPCODE:
1251 size_t size; 2515 size_t size;
1252 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size)); 2516 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, 0, 0, &size));
1253 SV *sv = sv_2mortal (newSV (size)); 2517 SV *sv = sv_2mortal (newSV (size));
1254 SvUPGRADE (sv, SVt_PV); 2518 SvUPGRADE (sv, SVt_PV);
1255 SvPOK_only (sv); 2519 SvPOK_only (sv);
1256 SvCUR_set (sv, size); 2520 SvCUR_set (sv, size);
1257 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0)); 2521 NEED_SUCCESS (GetProgramBuildInfo, (self, device, name, size, SvPVX (sv), 0));
1258 XPUSHs (sv); 2522 XPUSHs (sv);
1259 2523
1260#BEGIN:program_build 2524#BEGIN:program_build
1261 2525
1262void 2526void
1280 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2544 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1281 EXTEND (SP, 1); 2545 EXTEND (SP, 1);
1282 const int i = 0; 2546 const int i = 0;
1283 PUSHs (sv_2mortal (newSVpv (value, 0))); 2547 PUSHs (sv_2mortal (newSVpv (value, 0)));
1284 2548
2549void
2550binary_type (OpenCL::Program self, OpenCL::Device device)
2551 PPCODE:
2552 cl_program_binary_type value [1];
2553 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2554 EXTEND (SP, 1);
2555 const int i = 0;
2556 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2557
1285#END:program_build 2558#END:program_build
1286 2559
1287void 2560void
1288kernel (OpenCL::Program program, SV *function) 2561kernel (OpenCL::Program program, SV *function)
1289 PPCODE: 2562 PPCODE:
1290 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2563 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1291 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2564 XPUSH_CLOBJ (stash_kernel, kernel);
2565
2566void
2567kernels_in_program (OpenCL::Program program)
2568 PPCODE:
2569 cl_uint num_kernels;
2570 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2571 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2572 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2573
2574 int i;
2575 EXTEND (SP, num_kernels);
2576 for (i = 0; i < num_kernels; ++i)
2577 PUSH_CLOBJ (stash_kernel, kernels [i]);
1292 2578
1293void 2579void
1294info (OpenCL::Program self, cl_program_info name) 2580info (OpenCL::Program self, cl_program_info name)
1295 PPCODE: 2581 PPCODE:
1296 INFO (Program) 2582 INFO (Program)
1314 { 2600 {
1315 SV *sv = sv_2mortal (newSV (sizes [i])); 2601 SV *sv = sv_2mortal (newSV (sizes [i]));
1316 SvUPGRADE (sv, SVt_PV); 2602 SvUPGRADE (sv, SVt_PV);
1317 SvPOK_only (sv); 2603 SvPOK_only (sv);
1318 SvCUR_set (sv, sizes [i]); 2604 SvCUR_set (sv, sizes [i]);
1319 ptrs [i] = SvPVX (sv); 2605 ptrs [i] = (void *)SvPVX (sv);
1320 PUSHs (sv); 2606 PUSHs (sv);
1321 } 2607 }
1322 2608
1323 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2609 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1324 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2610 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1344 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2630 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1345 EXTEND (SP, 1); 2631 EXTEND (SP, 1);
1346 const int i = 0; 2632 const int i = 0;
1347 { 2633 {
1348 NEED_SUCCESS (RetainContext, (value [i])); 2634 NEED_SUCCESS (RetainContext, (value [i]));
1349 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2635 PUSH_CLOBJ (stash_context, value [i]);
1350 } 2636 }
1351 2637
1352void 2638void
1353devices (OpenCL::Program self) 2639devices (OpenCL::Program self)
1354 PPCODE: 2640 PPCODE:
1358 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2644 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1359 int i, n = size / sizeof (*value); 2645 int i, n = size / sizeof (*value);
1360 EXTEND (SP, n); 2646 EXTEND (SP, n);
1361 for (i = 0; i < n; ++i) 2647 for (i = 0; i < n; ++i)
1362 { 2648 {
1363 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2649 PUSH_CLOBJ (stash_device, value [i]);
1364 } 2650 }
1365 2651
1366void 2652void
1367source (OpenCL::Program self) 2653source (OpenCL::Program self)
1368 PPCODE: 2654 PPCODE:
1394DESTROY (OpenCL::Kernel self) 2680DESTROY (OpenCL::Kernel self)
1395 CODE: 2681 CODE:
1396 clReleaseKernel (self); 2682 clReleaseKernel (self);
1397 2683
1398void 2684void
2685setf (OpenCL::Kernel self, const char *format, ...)
2686 CODE:
2687 int i;
2688 for (i = 2; ; ++i)
2689 {
2690 while (*format == ' ')
2691 ++format;
2692
2693 char type = *format++;
2694
2695 if (!type)
2696 break;
2697
2698 if (i >= items)
2699 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2700
2701 SV *sv = ST (i);
2702
2703 union
2704 {
2705 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2706 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2707 cl_half ch; cl_float cf; cl_double cd;
2708 cl_mem cm;
2709 cl_sampler ca;
2710 size_t cz;
2711 cl_event ce;
2712 } arg;
2713 size_t size;
2714 int nullarg = 0;
2715
2716 switch (type)
2717 {
2718 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2719 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2720 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2721 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2722 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2723 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2724 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2725 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2726
2727 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2728 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2729 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2730
2731 case 'z': nullarg = 1; size = SvIV (sv); break;
2732
2733 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2734 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2735 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2736
2737 default:
2738 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2739 }
2740
2741 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2742 if (res)
2743 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2744 }
2745
2746 if (i != items)
2747 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2748
2749void
1399set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2750set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1400 CODE: 2751 CODE:
1401 clSetKernelArg (self, idx, sizeof (value), &value); 2752 clSetKernelArg (self, idx, sizeof (value), &value);
1402 2753
1403void 2754void
1451 clSetKernelArg (self, idx, sizeof (value), &value); 2802 clSetKernelArg (self, idx, sizeof (value), &value);
1452 2803
1453void 2804void
1454set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2805set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1455 CODE: 2806 CODE:
2807 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2808
2809void
2810set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2811 CODE:
2812 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2813
2814void
2815set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2816 CODE:
2817 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2818
2819void
2820set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2821 CODE:
1456 clSetKernelArg (self, idx, sizeof (value), &value); 2822 clSetKernelArg (self, idx, sizeof (value), &value);
1457 2823
1458void 2824void
1459set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2825set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1460 CODE: 2826 CODE:
1461 clSetKernelArg (self, idx, sizeof (value), &value); 2827 clSetKernelArg (self, idx, size, 0);
1462
1463void
1464set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1465 CODE:
1466 clSetKernelArg (self, idx, sizeof (value), &value);
1467
1468void
1469set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1470 CODE:
1471 clSetKernelArg (self, idx, sizeof (value), &value);
1472
1473void
1474set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1475 CODE:
1476 clSetKernelArg (self, idx, sizeof (value), &value);
1477 2828
1478void 2829void
1479set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2830set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1480 CODE: 2831 CODE:
1481 clSetKernelArg (self, idx, sizeof (value), &value); 2832 clSetKernelArg (self, idx, sizeof (value), &value);
1517 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2868 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1518 EXTEND (SP, 1); 2869 EXTEND (SP, 1);
1519 const int i = 0; 2870 const int i = 0;
1520 { 2871 {
1521 NEED_SUCCESS (RetainContext, (value [i])); 2872 NEED_SUCCESS (RetainContext, (value [i]));
1522 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2873 PUSH_CLOBJ (stash_context, value [i]);
1523 } 2874 }
1524 2875
1525void 2876void
1526program (OpenCL::Kernel self) 2877program (OpenCL::Kernel self)
1527 PPCODE: 2878 PPCODE:
1529 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2880 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1530 EXTEND (SP, 1); 2881 EXTEND (SP, 1);
1531 const int i = 0; 2882 const int i = 0;
1532 { 2883 {
1533 NEED_SUCCESS (RetainProgram, (value [i])); 2884 NEED_SUCCESS (RetainProgram, (value [i]));
1534 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2885 PUSH_CLOBJ (stash_program, value [i]);
1535 } 2886 }
1536 2887
1537#END:kernel 2888#END:kernel
1538 2889
1539void 2890void
1586 const int i = 0; 2937 const int i = 0;
1587 PUSHs (sv_2mortal (newSVuv (value [i]))); 2938 PUSHs (sv_2mortal (newSVuv (value [i])));
1588 2939
1589#END:kernel_work_group 2940#END:kernel_work_group
1590 2941
2942#if CL_VERSION_1_2
2943
2944void
2945arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2946 PPCODE:
2947 size_t size;
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2949 SV *sv = sv_2mortal (newSV (size));
2950 SvUPGRADE (sv, SVt_PV);
2951 SvPOK_only (sv);
2952 SvCUR_set (sv, size);
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2954 XPUSHs (sv);
2955
2956#BEGIN:kernel_arg
2957
2958void
2959arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2960 PPCODE:
2961 cl_kernel_arg_address_qualifier value [1];
2962 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2963 EXTEND (SP, 1);
2964 const int i = 0;
2965 PUSHs (sv_2mortal (newSVuv (value [i])));
2966
2967void
2968arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2969 PPCODE:
2970 cl_kernel_arg_access_qualifier value [1];
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVuv (value [i])));
2975
2976void
2977arg_type_name (OpenCL::Kernel self, cl_uint idx)
2978 ALIAS:
2979 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2980 arg_name = CL_KERNEL_ARG_NAME
2981 PPCODE:
2982 size_t size;
2983 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2984 char *value = tmpbuf (size);
2985 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2986 EXTEND (SP, 1);
2987 const int i = 0;
2988 PUSHs (sv_2mortal (newSVpv (value, 0)));
2989
2990void
2991arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2992 PPCODE:
2993 cl_kernel_arg_type_qualifier value [1];
2994 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2995 EXTEND (SP, 1);
2996 const int i = 0;
2997 PUSHs (sv_2mortal (newSVuv (value [i])));
2998
2999#END:kernel_arg
3000
3001#endif
3002
1591MODULE = OpenCL PACKAGE = OpenCL::Event 3003MODULE = OpenCL PACKAGE = OpenCL::Event
1592 3004
1593void 3005void
1594DESTROY (OpenCL::Event self) 3006DESTROY (OpenCL::Event self)
1595 CODE: 3007 CODE:
1597 3009
1598void 3010void
1599wait (OpenCL::Event self) 3011wait (OpenCL::Event self)
1600 CODE: 3012 CODE:
1601 clWaitForEvents (1, &self); 3013 clWaitForEvents (1, &self);
3014
3015void
3016cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
3017 CODE:
3018 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1602 3019
1603void 3020void
1604info (OpenCL::Event self, cl_event_info name) 3021info (OpenCL::Event self, cl_event_info name)
1605 PPCODE: 3022 PPCODE:
1606 INFO (Event) 3023 INFO (Event)
1614 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 3031 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1615 EXTEND (SP, 1); 3032 EXTEND (SP, 1);
1616 const int i = 0; 3033 const int i = 0;
1617 { 3034 {
1618 NEED_SUCCESS (RetainCommandQueue, (value [i])); 3035 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1619 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 3036 PUSH_CLOBJ (stash_queue, value [i]);
1620 } 3037 }
1621 3038
1622void 3039void
1623command_type (OpenCL::Event self) 3040command_type (OpenCL::Event self)
1624 PPCODE: 3041 PPCODE:
1647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3064 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1648 EXTEND (SP, 1); 3065 EXTEND (SP, 1);
1649 const int i = 0; 3066 const int i = 0;
1650 { 3067 {
1651 NEED_SUCCESS (RetainContext, (value [i])); 3068 NEED_SUCCESS (RetainContext, (value [i]));
1652 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 3069 PUSH_CLOBJ (stash_context, value [i]);
1653 } 3070 }
1654 3071
1655#END:event 3072#END:event
1656 3073
1657void 3074void
1682void 3099void
1683set_status (OpenCL::UserEvent self, cl_int execution_status) 3100set_status (OpenCL::UserEvent self, cl_int execution_status)
1684 CODE: 3101 CODE:
1685 clSetUserEventStatus (self, execution_status); 3102 clSetUserEventStatus (self, execution_status);
1686 3103
3104MODULE = OpenCL PACKAGE = OpenCL::Mapped
3105
3106void
3107DESTROY (SV *self)
3108 CODE:
3109 OpenCL__Mapped mapped = SvMAPPED (self);
3110
3111 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3112 mapped_detach (self, mapped);
3113
3114 clReleaseCommandQueue (mapped->queue);
3115 clReleaseEvent (mapped->event);
3116 Safefree (mapped);
3117
3118void
3119unmap (OpenCL::Mapped self, ...)
3120 CODE:
3121 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3122
3123bool
3124mapped (OpenCL::Mapped self)
3125 CODE:
3126 RETVAL = !!self->ptr;
3127 OUTPUT:
3128 RETVAL
3129
3130void
3131wait (OpenCL::Mapped self)
3132 PPCODE:
3133 if (self->event)
3134 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3135
3136void
3137event (OpenCL::Mapped self)
3138 PPCODE:
3139 if (!self->event)
3140 XSRETURN_UNDEF;
3141
3142 clRetainEvent (self->event);
3143 XPUSH_CLOBJ (stash_event, self->event);
3144
3145size_t
3146size (OpenCL::Mapped self)
3147 CODE:
3148 RETVAL = self->cb;
3149 OUTPUT:
3150 RETVAL
3151
3152IV
3153ptr (OpenCL::Mapped self)
3154 CODE:
3155 RETVAL = PTR2IV (self->ptr);
3156 OUTPUT:
3157 RETVAL
3158
3159void
3160set (OpenCL::Mapped self, size_t offset, SV *data)
3161 CODE:
3162 STRLEN len;
3163 const char *ptr = SvPVbyte (data, len);
3164
3165 if (offset + len > self->cb)
3166 croak ("OpenCL::Mapped::set out of bound condition detected");
3167
3168 memcpy (offset + (char *)self->ptr, ptr, len);
3169
3170MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3171
3172MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3173
3174IV
3175row_pitch (OpenCL::Mapped self)
3176 ALIAS:
3177 slice_pitch = 1
3178 CODE:
3179 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3180 OUTPUT:
3181 RETVAL
3182
3183
3184

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines