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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.28 by root, Thu Apr 19 13:49:33 2012 UTC vs.
Revision 1.70 by root, Sat May 5 02:33:55 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#include "ecb.h"//D
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
4 31
5#ifdef __APPLE__ 32#ifdef __APPLE__
6 #include <OpenCL/opencl.h> 33 #include <OpenCL/opencl.h>
7#else 34#else
8 #include <CL/opencl.h> 35 #include <CL/opencl.h>
9#endif 36#endif
10 37
38#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1
40 #define PREFER_1_1 1
41#endif
42
11typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
12typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
13typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
14typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
15typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
16typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
17typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
18typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
19typedef cl_mem OpenCL__Image2D;
20typedef cl_mem OpenCL__Image3D;
21typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
22typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
23typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
24typedef cl_mem OpenCL__Image2D_ornull;
25typedef cl_mem OpenCL__Image3D_ornull;
26typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
27typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
28typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
29typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
30typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
31 60
61typedef struct mapped * OpenCL__Mapped;
62
32typedef SV *FUTURE; 63typedef SV *FUTURE;
33 64
65static HV
66 *stash_platform,
67 *stash_device,
68 *stash_subdevice,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
89
34/*****************************************************************************/ 90/*****************************************************************************/
35 91
92// name must include a leading underscore
93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
94// for perl. doh.
95static void *
96glsym (const char *name)
97{
98 void *fun = 0;
99
100 #if defined I_DLFCN && defined RTLD_DEFAULT
101 fun = dlsym (RTLD_DEFAULT, name + 1);
102 if (!fun) fun = dlsym (RTLD_DEFAULT, name);
103
104 if (!fun)
105 {
106 static void *libgl;
107 static const char *glso[] = {
108 "libGL.so.1",
109 "libGL.so.3",
110 "libGL.so.4.0",
111 "libGL.so",
112 "/usr/lib/libGL.so",
113 "/usr/X11R6/lib/libGL.1.dylib"
114 };
115 int i;
116
117 for (i = 0; !libgl && i < sizeof (glso) / sizeof (glso [0]); ++i)
118 {
119 libgl = dlopen (glso [i], RTLD_LAZY);
120 if (libgl)
121 break;
122 }
123
124 if (libgl)
125 {
126 fun = dlsym (libgl, name + 1);
127 if (!fun) fun = dlsym (libgl, name);
128 }
129 }
130 #endif
131
132 return fun;
133}
134
135/*****************************************************************************/
136
36/* up to two temporary buffers */ 137/* up to two temporary buffers */
37static void * 138static void * ecb_noinline
38tmpbuf (size_t size) 139tmpbuf (size_t size)
39{ 140{
40 enum { buffers = 3 }; 141 enum { buffers = 4 };
41 static int idx; 142 static int idx;
42 static void *buf [buffers]; 143 static void *buf [buffers];
43 static size_t len [buffers]; 144 static size_t len [buffers];
44 145
45 idx = ++idx % buffers; 146 idx = (idx + 1) % buffers;
46 147
47 if (len [idx] < size) 148 if (len [idx] < size)
48 { 149 {
49 free (buf [idx]); 150 free (buf [idx]);
50 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
51 buf [idx] = malloc (len [idx]); 152 buf [idx] = malloc (len [idx]);
52 } 153 }
53 154
54 return buf [idx]; 155 return buf [idx];
156}
157
158static const char * ecb_noinline
159cv_get_name (CV *cv)
160{
161 static char fullname [256];
162
163 GV *gv = CvGV (cv); // gv better be nonzero
164
165 HV *stash = GvSTASH (gv);
166 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
167 const char *gvname = GvNAME (gv);
168
169 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
170 return fullname;
55} 171}
56 172
57/*****************************************************************************/ 173/*****************************************************************************/
58 174
59typedef struct 175typedef struct
118 if (res) \ 234 if (res) \
119 FAIL (name); 235 FAIL (name);
120 236
121/*****************************************************************************/ 237/*****************************************************************************/
122 238
123#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
124#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
125
126static void * 239static SV *
240new_clobj (HV *stash, IV id)
241{
242 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
243}
244
245#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
246#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
247
248/* cl objects are either \$iv, or [$iv, ...] */
249/* they can be upgraded at runtime to the array form */
250static void * ecb_noinline
127SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 251SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
128{ 252{
253 // sv_derived_from is quite slow :(
129 if (SvROK (sv) && sv_derived_from (sv, pkg)) 254 if (SvROK (sv) && sv_derived_from (sv, pkg))
130 return (void *)SvIV (SvRV (sv)); 255 return (void *)SvIV (SvRV (sv));
131 256
132 croak ("%s: %s is not of type %s", func, svname, pkg); 257 croak ("%s: %s is not of type %s", func, svname, pkg);
258}
259
260// the "no-inherit" version of the above
261static void * ecb_noinline
262SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash)
263{
264 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
265 return (void *)SvIV (SvRV (sv));
266
267 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash));
133} 268}
134 269
135/*****************************************************************************/ 270/*****************************************************************************/
136 271
137static cl_context_properties * 272static cl_context_properties * ecb_noinline
138SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 273SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
139{ 274{
140 if (!sv || !SvOK (sv)) 275 if (!sv || !SvOK (sv))
141 if (extra) 276 if (extra)
142 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 277 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
149 int i, len = av_len (av) + 1; 284 int i, len = av_len (av) + 1;
150 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 285 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
151 cl_context_properties *l = p; 286 cl_context_properties *l = p;
152 287
153 if (len & 1) 288 if (len & 1)
154 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 289 croak ("%s: %s is not a property list (must contain an even number of elements)", func, svname);
155 290
156 while (extracount--) 291 while (extracount--)
157 *l++ = *extra++; 292 *l++ = *extra++;
158 293
159 for (i = 0; i < len; ++i) 294 for (i = 0; i < len; i += 2)
160 { 295 {
161 cl_context_properties t = SvIV (*av_fetch (av, i, 0)); 296 cl_context_properties t = SvIV (*av_fetch (av, i , 0));
162 cl_context_properties v; 297 SV *p_sv = *av_fetch (av, i + 1, 0);
163 298 cl_context_properties v = SvIV (p_sv); // code below can override
164 ++i;
165 299
166 switch (t) 300 switch (t)
167 { 301 {
302 case CL_CONTEXT_PLATFORM:
303 if (SvROK (p_sv))
304 v = (cl_context_properties)SvCLOBJ (func, svname, p_sv, "OpenCL::Platform");
305 break;
306
307 case CL_GLX_DISPLAY_KHR:
308 if (!SvOK (p_sv))
309 {
310 void *func = glsym ("_glXGetCurrentDisplay");
311 if (func)
312 v = (cl_context_properties)((void *(*)(void))func)();
313 }
314 break;
315
316 case CL_GL_CONTEXT_KHR:
317 if (!SvOK (p_sv))
318 {
319 void *func = glsym ("_glXGetCurrentContext");
320 if (func)
321 v = (cl_context_properties)((void *(*)(void))func)();
322 }
323 break;
324
168 default: 325 default:
169 /* unknown property, treat as int */ 326 /* unknown property, treat as int */
170 v = SvIV (*av_fetch (av, i, 0));
171 break; 327 break;
172 } 328 }
173 329
174 *l++ = t; 330 *l++ = t;
175 *l++ = v; 331 *l++ = v;
181 } 337 }
182 338
183 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 339 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
184} 340}
185 341
342static void * ecb_noinline
343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
344{
345 const char *funcname = cv_get_name (cv);
346
347 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
348 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
349 funcname, argname, or_undef ? "undef or " : "", klass);
350
351 AV *av = (AV *)SvRV (arg);
352 void **list = 0;
353 cl_uint count = av_len (av) + 1;
354
355 if (count)
356 {
357 list = tmpbuf (sizeof (*list) * count);
358 int i;
359 for (i = 0; i < count; ++i)
360 list [i] = SvCLOBJ (funcname, argname, *av_fetch (av, i, 1), klass);
361 }
362
363 if (!count && !or_undef)
364 croak ("%s: '%s' must contain at least one %s object",
365 funcname, argname, klass);
366
367 *rcount = count;
368 return (void *)list;
369}
370
186/*****************************************************************************/ 371/*****************************************************************************/
372/* callback stuff */
373
374/* default context callback, log to stderr */
375static void CL_CALLBACK
376context_default_notify (const char *msg, const void *info, size_t cb, void *data)
377{
378 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
379}
380
381typedef struct
382{
383 int free_cb;
384 void (*push)(void *data1, void *data2, void *data3);
385} eq_vtbl;
386
387typedef struct eq_item
388{
389 struct eq_item *next;
390 eq_vtbl *vtbl;
391 SV *cb;
392 void *data1, *data2, *data3;
393} eq_item;
394
395static void (*eq_signal_func)(void *signal_arg, int value);
396static void *eq_signal_arg;
397static xmutex_t eq_lock = X_MUTEX_INIT;
398static eq_item *eq_head, *eq_tail;
399
400static void ecb_noinline
401eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
402{
403 eq_item *item = malloc (sizeof (eq_item));
404
405 item->next = 0;
406 item->vtbl = vtbl;
407 item->cb = cb;
408 item->data1 = data1;
409 item->data2 = data2;
410 item->data3 = data3;
411
412 X_LOCK (eq_lock);
413
414 *(eq_head ? &eq_tail->next : &eq_head) = item;
415 eq_tail = item;
416
417 X_UNLOCK (eq_lock);
418
419 eq_signal_func (eq_signal_arg, 0);
420}
421
422static eq_item *
423eq_dec (void)
424{
425 eq_item *res;
426
427 X_LOCK (eq_lock);
428
429 res = eq_head;
430 if (res)
431 eq_head = res->next;
432
433 X_UNLOCK (eq_lock);
434
435 return res;
436}
437
438static void
439eq_poll (void)
440{
441 eq_item *item;
442
443 while ((item = eq_dec ()))
444 {
445 ENTER;
446 SAVETMPS;
447
448 dSP;
449 PUSHMARK (SP);
450 EXTEND (SP, 2);
451
452 if (item->vtbl->free_cb)
453 sv_2mortal (item->cb);
454
455 PUTBACK;
456 item->vtbl->push (item->data1, item->data2, item->data3);
457
458 SV *cb = item->cb;
459 free (item);
460
461 call_sv (cb, G_DISCARD | G_VOID);
462
463 FREETMPS;
464 LEAVE;
465 }
466}
467
468static void
469eq_poll_interrupt (pTHX_ void *c_arg, int value)
470{
471 eq_poll ();
472}
473
474/*****************************************************************************/
475/* context notify */
476
477static void ecb_noinline
478eq_context_push (void *data1, void *data2, void *data3)
479{
480 dSP;
481 PUSHs (sv_2mortal (newSVpv (data1, 0)));
482 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
483 PUTBACK;
484
485 free (data1);
486 free (data2);
487}
488
489static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
490
491static void CL_CALLBACK
492eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
493{
494 void *pvt_copy = malloc (cb);
495 memcpy (pvt_copy, pvt, cb);
496 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
497}
498
499#define CONTEXT_NOTIFY_CALLBACK \
500 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
501 void *user_data = 0; \
502 \
503 if (SvOK (notify)) \
504 { \
505 pfn_notify = eq_context_notify; \
506 user_data = s_get_cv (notify); \
507 }
508
509static SV * ecb_noinline
510new_clobj_context (cl_context ctx, void *user_data)
511{
512 SV *sv = new_clobj (stash_context, (IV)ctx);
513
514 if (user_data)
515 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
516
517 return sv;
518}
519
520#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
521
522/*****************************************************************************/
523/* build/compile/link notify */
524
525static void
526eq_program_push (void *data1, void *data2, void *data3)
527{
528 dSP;
529 PUSH_CLOBJ (stash_program, data1);
530 PUTBACK;
531}
532
533static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
534
535static void CL_CALLBACK
536eq_program_notify (cl_program program, void *user_data)
537{
538 clRetainProgram (program);
539
540 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
541}
542
543typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
544
545static program_callback ecb_noinline
546make_program_callback (SV *notify, void **ruser_data)
547{
548 if (SvOK (notify))
549 {
550 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
551 return eq_program_notify;
552 }
553 else
554 {
555 *ruser_data = 0;
556 return 0;
557 }
558}
559
560struct build_args
561{
562 cl_program program;
563 char *options;
564 void *user_data;
565 cl_uint num_devices;
566};
567
568X_THREAD_PROC (build_program_thread)
569{
570 struct build_args *arg = thr_arg;
571
572 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
573
574 if (arg->user_data)
575 eq_program_notify (arg->program, arg->user_data);
576 else
577 clReleaseProgram (arg->program);
578
579 free (arg->options);
580 free (arg);
581
582 return 0;
583}
584
585static void
586build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
587{
588 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
589
590 arg->program = program;
591 arg->options = strdup (options);
592 arg->user_data = user_data;
593 arg->num_devices = num_devices;
594 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
595
596 xthread_t id;
597 thread_create (&id, build_program_thread, arg);
598}
599
600/*****************************************************************************/
601/* event objects */
602
603static void
604eq_event_push (void *data1, void *data2, void *data3)
605{
606 dSP;
607 PUSH_CLOBJ (stash_event, data1);
608 PUSHs (sv_2mortal (newSViv ((IV)data2)));
609 PUTBACK;
610}
611
612static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
613
614static void CL_CALLBACK
615eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
616{
617 clRetainEvent (event);
618 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
619}
620
621/*****************************************************************************/
622/* utilities for XS code */
187 623
188static size_t 624static size_t
189img_row_pitch (cl_mem img) 625img_row_pitch (cl_mem img)
190{ 626{
191 size_t res; 627 size_t res;
192 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 628 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
193 return res; 629 return res;
194} 630}
195 631
196static cl_event * 632static cl_event * ecb_noinline
197event_list (SV **items, int count) 633event_list (SV **items, cl_uint *rcount, cl_event extra)
198{ 634{
199 if (!count) 635 cl_uint count = *rcount;
636
637 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
638 *rcount = count = 0;
639
640 if (!count && !extra)
200 return 0; 641 return 0;
201 642
202 cl_event *list = tmpbuf (sizeof (cl_event) * count); 643 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
644 int i = 0;
203 645
204 while (count--) 646 while (count--)
647 if (SvOK (items [count]))
205 list [count] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 648 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
206 649
650 if (extra)
651 list [i++] = extra;
652
653 *rcount = i;
654
207 return list; 655 return i ? list : 0;
208} 656}
209 657
210#define EVENT_LIST(items,count) \ 658#define EVENT_LIST(skip) \
211 cl_uint event_list_count = (count); \ 659 cl_uint event_list_count = items - (skip); \
212 cl_event *event_list_ptr = event_list (&ST (items), event_list_count) 660 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
213 661
214#define INFO(class) \ 662#define INFO(class) \
215{ \ 663{ \
216 size_t size; \ 664 size_t size; \
217 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 665 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
221 SvCUR_set (sv, size); \ 669 SvCUR_set (sv, size); \
222 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 670 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
223 XPUSHs (sv); \ 671 XPUSHs (sv); \
224} 672}
225 673
674/*****************************************************************************/
675/* mapped_xxx */
676
677static OpenCL__Mapped
678SvMAPPED (SV *sv)
679{
680 // no typechecking atm., keep your fingers crossed
681 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
682}
683
684struct mapped
685{
686 cl_command_queue queue;
687 cl_mem memobj;
688 void *ptr;
689 size_t cb;
690 cl_event event;
691 size_t row_pitch;
692 size_t slice_pitch;
693};
694
695static SV *
696mapped_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)
697{
698 SV *data = newSV (0);
699 SvUPGRADE (data, SVt_PVMG);
700
701 OpenCL__Mapped mapped;
702 New (0, mapped, 1, struct mapped);
703
704 clRetainCommandQueue (queue);
705
706 mapped->queue = queue;
707 mapped->memobj = memobj;
708 mapped->ptr = ptr;
709 mapped->cb = cb;
710 mapped->event = ev;
711 mapped->row_pitch = row_pitch;
712 mapped->slice_pitch = slice_pitch;
713
714 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
715
716 if (SvLEN (data))
717 Safefree (data);
718
719 SvPVX (data) = (char *)ptr;
720 SvCUR_set (data, cb);
721 SvLEN_set (data, 0);
722 SvPOK_only (data);
723
724 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
725
726 if (!(flags & CL_MAP_WRITE))
727 SvREADONLY_on (data);
728
729 return obj;
730}
731
732static void
733mapped_detach (SV *sv, OpenCL__Mapped mapped)
734{
735 SV *data = SvRV (sv);
736
737 // the next check checks both before AND after detach, where SvPVX should STILL be 0
738 if (SvPVX (data) != (char *)mapped->ptr)
739 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
740 else
741 {
742 SvREADONLY_off (data);
743 SvCUR_set (data, 0);
744 SvPVX (data) = 0;
745 SvOK_off (data);
746 }
747
748 mapped->ptr = 0;
749}
750
751static void
752mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
753{
754 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
755 cl_event ev;
756
757 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
758
759 clReleaseEvent (mapped->event);
760 mapped->event = ev;
761
762 mapped_detach (self, mapped);
763}
764
765/*****************************************************************************/
766
226MODULE = OpenCL PACKAGE = OpenCL 767MODULE = OpenCL PACKAGE = OpenCL
227 768
228PROTOTYPES: ENABLE 769PROTOTYPES: ENABLE
770
771void
772poll ()
773 CODE:
774 eq_poll ();
775
776void
777_eq_initialise (IV func, IV arg)
778 CODE:
779 eq_signal_func = (void (*)(void *, int))func;
780 eq_signal_arg = (void*)arg;
229 781
230BOOT: 782BOOT:
231{ 783{
232 HV *stash = gv_stashpv ("OpenCL", 1); 784 HV *stash = gv_stashpv ("OpenCL", 1);
233 static const ivstr *civ, const_iv[] = { 785 static const ivstr *civ, const_iv[] = {
242 { sizeof (cl_half ), "SIZEOF_HALF" }, 794 { sizeof (cl_half ), "SIZEOF_HALF" },
243 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 795 { sizeof (cl_float ), "SIZEOF_FLOAT" },
244 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 796 { sizeof (cl_double), "SIZEOF_DOUBLE" },
245#include "constiv.h" 797#include "constiv.h"
246 }; 798 };
799
247 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 800 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
248 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 801 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
802
803 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
804 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
805 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
806 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
807 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
808 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
809 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
810 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
811 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
812 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
813 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
814 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
815 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
816 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
817 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
818 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
819 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
820 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
821 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
822 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
823 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
824 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
825 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
826
827 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
249} 828}
250 829
251cl_int 830cl_int
252errno () 831errno ()
253 CODE: 832 CODE:
254 errno = res; 833 RETVAL = res;
834 OUTPUT:
835 RETVAL
255 836
256const char * 837const char *
257err2str (cl_int err) 838err2str (cl_int err = res)
258 839
259const char * 840const char *
260enum2str (cl_uint value) 841enum2str (cl_uint value)
261 842
262void 843void
270 list = tmpbuf (sizeof (*list) * count); 851 list = tmpbuf (sizeof (*list) * count);
271 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 852 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
272 853
273 EXTEND (SP, count); 854 EXTEND (SP, count);
274 for (i = 0; i < count; ++i) 855 for (i = 0; i < count; ++i)
275 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 856 PUSH_CLOBJ (stash_platform, list [i]);
276 857
277void 858void
278context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 859context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
279 PPCODE: 860 PPCODE:
861 CONTEXT_NOTIFY_CALLBACK;
280 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 862 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
281 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 863 XPUSH_CLOBJ_CONTEXT;
282 864
283void 865void
284context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 866context (FUTURE properties, FUTURE devices, FUTURE notify)
285 PPCODE: 867 PPCODE:
286 /* der Gipfel der Kunst */ 868 /* der Gipfel der Kunst */
287 869
288void 870void
289wait_for_events (...) 871wait_for_events (...)
290 CODE: 872 CODE:
291 EVENT_LIST (0, items); 873 EVENT_LIST (0);
292 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 874 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
293 875
294PROTOTYPES: DISABLE 876PROTOTYPES: DISABLE
295 877
296MODULE = OpenCL PACKAGE = OpenCL::Platform 878MODULE = OpenCL PACKAGE = OpenCL::Platform
297 879
298void 880void
299info (OpenCL::Platform self, cl_platform_info name) 881info (OpenCL::Platform self, cl_platform_info name)
300 PPCODE: 882 PPCODE:
301 INFO (Platform) 883 INFO (Platform)
884
885void
886unload_compiler (OpenCL::Platform self)
887 CODE:
888#if CL_VERSION_1_2
889 clUnloadPlatformCompiler (self);
890#endif
302 891
303#BEGIN:platform 892#BEGIN:platform
304 893
305void 894void
306profile (OpenCL::Platform self) 895profile (OpenCL::Platform self)
332 list = tmpbuf (sizeof (*list) * count); 921 list = tmpbuf (sizeof (*list) * count);
333 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 922 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
334 923
335 EXTEND (SP, count); 924 EXTEND (SP, count);
336 for (i = 0; i < count; ++i) 925 for (i = 0; i < count; ++i)
337 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 926 PUSH_CLOBJ (stash_device, list [i]);
338 927
339void 928void
340context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 929context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
341 PPCODE: 930 PPCODE:
342 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 931 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
343 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 932 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2);
344 933
345 AV *av = (AV *)SvRV (devices); 934 cl_uint device_count = 0;
346 cl_uint num_devices = av_len (av) + 1; 935 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
347 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
348 int i;
349 936
350 for (i = num_devices; i--; ) 937 CONTEXT_NOTIFY_CALLBACK;
351 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
352
353 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 938 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
354 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 939 XPUSH_CLOBJ_CONTEXT;
355 940
356void 941void
357context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 942context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
358 PPCODE: 943 PPCODE:
359 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 944 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
360 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 945 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
946
947 CONTEXT_NOTIFY_CALLBACK;
361 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 948 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
362 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 949 XPUSH_CLOBJ_CONTEXT;
363 950
364MODULE = OpenCL PACKAGE = OpenCL::Device 951MODULE = OpenCL PACKAGE = OpenCL::Device
365 952
366void 953void
367info (OpenCL::Device self, cl_device_info name) 954info (OpenCL::Device self, cl_device_info name)
368 PPCODE: 955 PPCODE:
369 INFO (Device) 956 INFO (Device)
957
958#if CL_VERSION_1_2
959
960void
961sub_devices (OpenCL::Device self, SV *properties)
962 PPCODE:
963 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
964 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
965
966 properties = SvRV (properties);
967
968 cl_uint count = av_len ((AV *)properties) + 1;
969 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
970
971 int i;
972 for (i = 0; i < count; ++i)
973 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
974
975 props [count] = 0;
976
977 cl_uint num_devices;
978 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
979 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
980 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
981
982 EXTEND (SP, num_devices);
983 for (i = 0; i < count; ++i)
984 PUSH_CLOBJ (stash_subdevice, list [i]);
985
986#endif
370 987
371#BEGIN:device 988#BEGIN:device
372 989
373void 990void
374type (OpenCL::Device self) 991type (OpenCL::Device self)
375 PPCODE: 992 PPCODE:
376 cl_device_type value [1]; 993 cl_device_type value [1];
377 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 994 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
378 EXTEND (SP, 1); 995 EXTEND (SP, 1);
379 const int i = 0; 996 const int i = 0;
380 PUSHs (sv_2mortal (newSViv (value [i]))); 997 PUSHs (sv_2mortal (newSVuv (value [i])));
381 998
382void 999void
383vendor_id (OpenCL::Device self) 1000vendor_id (OpenCL::Device self)
384 ALIAS: 1001 ALIAS:
385 vendor_id = CL_DEVICE_VENDOR_ID 1002 vendor_id = CL_DEVICE_VENDOR_ID
406 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1023 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
407 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1024 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
408 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1025 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
409 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1026 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
410 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1027 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
411 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1028 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
412 PPCODE: 1029 PPCODE:
413 cl_uint value [1]; 1030 cl_uint value [1];
414 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1031 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
415 EXTEND (SP, 1); 1032 EXTEND (SP, 1);
416 const int i = 0; 1033 const int i = 0;
530 PPCODE: 1147 PPCODE:
531 cl_command_queue_properties value [1]; 1148 cl_command_queue_properties value [1];
532 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1149 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
533 EXTEND (SP, 1); 1150 EXTEND (SP, 1);
534 const int i = 0; 1151 const int i = 0;
535 PUSHs (sv_2mortal (newSViv (value [i]))); 1152 PUSHs (sv_2mortal (newSVuv (value [i])));
536 1153
537void 1154void
538platform (OpenCL::Device self) 1155platform (OpenCL::Device self)
539 PPCODE: 1156 PPCODE:
540 cl_platform_id value [1]; 1157 cl_platform_id value [1];
541 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1158 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
542 EXTEND (SP, 1); 1159 EXTEND (SP, 1);
543 const int i = 0; 1160 const int i = 0;
544 { 1161 {
545 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i])); 1162 PUSH_CLOBJ (stash_platform, value [i]);
546 } 1163 }
547 1164
548void 1165void
549name (OpenCL::Device self) 1166name (OpenCL::Device self)
550 ALIAS: 1167 ALIAS:
569 cl_device_id value [1]; 1186 cl_device_id value [1];
570 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1187 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
571 EXTEND (SP, 1); 1188 EXTEND (SP, 1);
572 const int i = 0; 1189 const int i = 0;
573 { 1190 {
574 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1191 PUSH_CLOBJ (stash_device, value [i]);
575 } 1192 }
576 1193
577void 1194void
578partition_types_ext (OpenCL::Device self) 1195partition_types_ext (OpenCL::Device self)
579 ALIAS: 1196 ALIAS:
590 for (i = 0; i < n; ++i) 1207 for (i = 0; i < n; ++i)
591 PUSHs (sv_2mortal (newSVuv (value [i]))); 1208 PUSHs (sv_2mortal (newSVuv (value [i])));
592 1209
593#END:device 1210#END:device
594 1211
1212MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1213
1214#if CL_VERSION_1_2
1215
1216void
1217DESTROY (OpenCL::SubDevice self)
1218 CODE:
1219 clReleaseDevice (self);
1220
1221#endif
1222
595MODULE = OpenCL PACKAGE = OpenCL::Context 1223MODULE = OpenCL PACKAGE = OpenCL::Context
596 1224
597void 1225void
598DESTROY (OpenCL::Context context) 1226DESTROY (OpenCL::Context self)
599 CODE: 1227 CODE:
600 clReleaseContext (context); 1228 clReleaseContext (self);
601 1229
602void 1230void
603info (OpenCL::Context self, cl_context_info name) 1231info (OpenCL::Context self, cl_context_info name)
604 PPCODE: 1232 PPCODE:
605 INFO (Context) 1233 INFO (Context)
606 1234
607void 1235void
608queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1236queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
609 PPCODE: 1237 PPCODE:
610 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1238 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
611 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1239 XPUSH_CLOBJ (stash_queue, queue);
612 1240
613void 1241void
614user_event (OpenCL::Context self) 1242user_event (OpenCL::Context self)
615 PPCODE: 1243 PPCODE:
616 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1244 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
617 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1245 XPUSH_CLOBJ (stash_userevent, ev);
618 1246
619void 1247void
620buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1248buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
621 PPCODE: 1249 PPCODE:
622 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1250 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
623 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1251 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
624 1252
625 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1253 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
626 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1254 XPUSH_CLOBJ (stash_bufferobj, mem);
627 1255
628void 1256void
629buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1257buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
630 PPCODE: 1258 PPCODE:
631 STRLEN len; 1259 STRLEN len;
632 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1260 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
633 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1261 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
634 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1262 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
635 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1263 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
636 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1264 XPUSH_CLOBJ (stash_bufferobj, mem);
637 1265
1266#if CL_VERSION_1_2
1267
638void 1268void
639image2d (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) 1269image (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)
640 PPCODE: 1270 PPCODE:
641 STRLEN len; 1271 STRLEN len;
642 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1272 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
643 const cl_image_format format = { channel_order, channel_type }; 1273 const cl_image_format format = { channel_order, channel_type };
1274 const cl_image_desc desc = {
1275 type,
1276 width, height, depth,
1277 array_size, row_pitch, slice_pitch,
1278 num_mip_level, num_samples,
1279 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1280 };
644 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
645 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1282 HV *stash = stash_image;
1283 switch (type)
1284 {
1285 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1286 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1287 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1288 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1289 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1290 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1291 }
1292 XPUSH_CLOBJ (stash, mem);
646 1293
1294#endif
1295
647void 1296void
648image3d (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) 1297image2d (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)
649 PPCODE: 1298 PPCODE:
650 STRLEN len; 1299 STRLEN len;
651 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1300 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
652 const cl_image_format format = { channel_order, channel_type }; 1301 const cl_image_format format = { channel_order, channel_type };
1302#if PREFER_1_1
1303 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1304#else
1305 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1306 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1307#endif
1308 XPUSH_CLOBJ (stash_image2d, mem);
1309
1310void
1311image3d (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)
1312 PPCODE:
1313 STRLEN len;
1314 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1315 const cl_image_format format = { channel_order, channel_type };
1316#if PREFER_1_1
653 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1317 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
654 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1318#else
1319 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1320 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1321#endif
1322 XPUSH_CLOBJ (stash_image3d, mem);
655 1323
656#if cl_apple_gl_sharing || cl_khr_gl_sharing 1324#if cl_apple_gl_sharing || cl_khr_gl_sharing
657 1325
658void 1326void
659gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1327gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
660 PPCODE: 1328 PPCODE:
661 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1329 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
662 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1330 XPUSH_CLOBJ (stash_bufferobj, mem);
1331
1332void
1333gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1334 PPCODE:
1335 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1336 XPUSH_CLOBJ (stash_image2d, mem);
1337
1338#if CL_VERSION_1_2
1339
1340void
1341gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1342 ALIAS:
1343 PPCODE:
1344 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1345 cl_gl_object_type type;
1346 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1347 HV *stash = stash_memory;
1348 switch (type)
1349 {
1350 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1351 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1352 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1353 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1354 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1355 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1356 }
1357 XPUSH_CLOBJ (stash, mem);
1358
1359#endif
663 1360
664void 1361void
665gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1362gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
666 PPCODE: 1363 PPCODE:
1364#if PREFER_1_1
667 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1365 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
668 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1366#else
1367 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1368#endif
1369 XPUSH_CLOBJ (stash_image2d, mem);
669 1370
670void 1371void
671gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1372gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
672 PPCODE: 1373 PPCODE:
1374#if PREFER_1_1
673 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1375 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
674 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1376#else
675 1377 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
676void 1378#endif
677gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1379 XPUSH_CLOBJ (stash_image3d, mem);
678 PPCODE:
679 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
680 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
681 1380
682#endif 1381#endif
683 1382
684void 1383void
685supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1384supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
705 1404
706void 1405void
707sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1406sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
708 PPCODE: 1407 PPCODE:
709 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1408 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
710 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1409 XPUSH_CLOBJ (stash_sampler, sampler);
711 1410
712void 1411void
713program_with_source (OpenCL::Context self, SV *program) 1412program_with_source (OpenCL::Context self, SV *program)
714 PPCODE: 1413 PPCODE:
715 STRLEN len; 1414 STRLEN len;
716 size_t len2; 1415 size_t len2;
717 const char *ptr = SvPVbyte (program, len); 1416 const char *ptr = SvPVbyte (program, len);
718 1417
719 len2 = len; 1418 len2 = len;
720 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1419 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
721 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1420 XPUSH_CLOBJ (stash_program, prog);
1421
1422void
1423program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1424 PPCODE:
1425 cl_uint device_count;
1426 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1427
1428 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1429 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1430
1431 binaries = SvRV (binaries);
1432
1433 if (device_count != av_len ((AV *)binaries) + 1)
1434 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1435
1436 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1437 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1438 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1439
1440 int i;
1441 for (i = 0; i < device_count; ++i)
1442 {
1443 STRLEN len;
1444 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1445 length_list [i] = len;
1446 }
1447
1448 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1449 length_list, binary_list,
1450 GIMME_V == G_ARRAY ? status_list : 0, &res));
1451
1452 EXTEND (SP, 2);
1453 PUSH_CLOBJ (stash_program, prog);
1454
1455 if (GIMME_V == G_ARRAY)
1456 {
1457 AV *av = newAV ();
1458 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1459
1460 for (i = device_count; i--; )
1461 av_store (av, i, newSViv (status_list [i]));
1462 }
1463
1464#if CL_VERSION_1_2
1465
1466void
1467program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1468 PPCODE:
1469 cl_uint device_count = 0;
1470 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1471
1472 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1473
1474 XPUSH_CLOBJ (stash_program, prog);
1475
1476void
1477link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1478 CODE:
1479 cl_uint device_count = 0;
1480 cl_device_id *device_list = 0;
1481
1482 if (SvOK (devices))
1483 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1484
1485 cl_uint program_count;
1486 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1487
1488 void *user_data;
1489 program_callback pfn_notify = make_program_callback (notify, &user_data);
1490
1491 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1492 program_count, program_list, pfn_notify, user_data, &res));
1493
1494 XPUSH_CLOBJ (stash_program, prog);
1495
1496#endif
722 1497
723#BEGIN:context 1498#BEGIN:context
724 1499
725void 1500void
726reference_count (OpenCL::Context self) 1501reference_count (OpenCL::Context self)
743 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1518 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
744 int i, n = size / sizeof (*value); 1519 int i, n = size / sizeof (*value);
745 EXTEND (SP, n); 1520 EXTEND (SP, n);
746 for (i = 0; i < n; ++i) 1521 for (i = 0; i < n; ++i)
747 { 1522 {
748 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 1523 PUSH_CLOBJ (stash_device, value [i]);
749 } 1524 }
750 1525
751void 1526void
752properties (OpenCL::Context self) 1527properties (OpenCL::Context self)
753 PPCODE: 1528 PPCODE:
768DESTROY (OpenCL::Queue self) 1543DESTROY (OpenCL::Queue self)
769 CODE: 1544 CODE:
770 clReleaseCommandQueue (self); 1545 clReleaseCommandQueue (self);
771 1546
772void 1547void
773enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1548read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1549 ALIAS:
1550 enqueue_read_buffer = 0
774 PPCODE: 1551 PPCODE:
775 cl_event ev = 0;
776 EVENT_LIST (6, items - 6); 1552 EVENT_LIST (6);
777 1553
778 SvUPGRADE (data, SVt_PV); 1554 SvUPGRADE (data, SVt_PV);
779 SvGROW (data, len); 1555 SvGROW (data, len);
780 SvPOK_only (data); 1556 SvPOK_only (data);
781 SvCUR_set (data, len); 1557 SvCUR_set (data, len);
1558
1559 cl_event ev = 0;
782 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1560 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
783 1561
784 if (ev) 1562 if (ev)
785 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1563 XPUSH_CLOBJ (stash_event, ev);
786 1564
787void 1565void
788enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1566write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1567 ALIAS:
1568 enqueue_write_buffer = 0
789 PPCODE: 1569 PPCODE:
790 cl_event ev = 0; 1570 EVENT_LIST (5);
1571
791 STRLEN len; 1572 STRLEN len;
792 char *ptr = SvPVbyte (data, len); 1573 char *ptr = SvPVbyte (data, len);
793 EVENT_LIST (5, items - 5);
794 1574
1575 cl_event ev = 0;
795 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
796 1577
797 if (ev) 1578 if (ev)
798 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
799 1580
1581#if CL_VERSION_1_2
1582
800void 1583void
801enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1584fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1585 ALIAS:
1586 enqueue_fill_buffer = 0
802 PPCODE: 1587 PPCODE:
1588 EVENT_LIST (5);
1589
1590 STRLEN len;
1591 char *ptr = SvPVbyte (data, len);
1592
803 cl_event ev = 0; 1593 cl_event ev = 0;
1594 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1595
1596 if (ev)
1597 XPUSH_CLOBJ (stash_event, ev);
1598
1599void
1600fill_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, ...)
1601 ALIAS:
1602 enqueue_fill_image = 0
1603 PPCODE:
1604 EVENT_LIST (12);
1605
1606 const size_t origin [3] = { x, y, z };
1607 const size_t region [3] = { width, height, depth };
1608
1609 const cl_float c_f [4] = { r, g, b, a };
1610 const cl_uint c_u [4] = { r, g, b, a };
1611 const cl_int c_s [4] = { r, g, b, a };
1612 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1613 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1614 cl_image_format format;
1615 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1616 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1617 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1618 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1619
1620 cl_event ev = 0;
1621 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1622 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1623
1624 if (ev)
1625 XPUSH_CLOBJ (stash_event, ev);
1626
1627#endif
1628
1629void
1630copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1631 ALIAS:
1632 enqueue_copy_buffer = 0
1633 PPCODE:
804 EVENT_LIST (6, items - 6); 1634 EVENT_LIST (6);
805 1635
1636 cl_event ev = 0;
806 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1637 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
807 1638
808 if (ev) 1639 if (ev)
809 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1640 XPUSH_CLOBJ (stash_event, ev);
810 1641
811void 1642void
812enqueue_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, ...) 1643read_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, ...)
1644 ALIAS:
1645 enqueue_read_buffer_rect = 0
813 PPCODE: 1646 PPCODE:
814 cl_event ev = 0; 1647 EVENT_LIST (17);
1648
815 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1649 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
816 const size_t host_origin[3] = { host_x, host_y, host_z }; 1650 const size_t host_origin[3] = { host_x, host_y, host_z };
817 const size_t region[3] = { width, height, depth }; 1651 const size_t region[3] = { width, height, depth };
818 EVENT_LIST (17, items - 17);
819 1652
820 if (!buf_row_pitch) 1653 if (!buf_row_pitch)
821 buf_row_pitch = region [0]; 1654 buf_row_pitch = region [0];
822 1655
823 if (!buf_slice_pitch) 1656 if (!buf_slice_pitch)
833 1666
834 SvUPGRADE (data, SVt_PV); 1667 SvUPGRADE (data, SVt_PV);
835 SvGROW (data, len); 1668 SvGROW (data, len);
836 SvPOK_only (data); 1669 SvPOK_only (data);
837 SvCUR_set (data, len); 1670 SvCUR_set (data, len);
1671
1672 cl_event ev = 0;
838 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)); 1673 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));
839 1674
840 if (ev) 1675 if (ev)
841 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1676 XPUSH_CLOBJ (stash_event, ev);
842 1677
843void 1678void
844enqueue_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, ...) 1679write_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, ...)
1680 ALIAS:
1681 enqueue_write_buffer_rect = 0
845 PPCODE: 1682 PPCODE:
846 cl_event ev = 0; 1683 EVENT_LIST (17);
1684
847 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1685 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
848 const size_t host_origin[3] = { host_x, host_y, host_z }; 1686 const size_t host_origin[3] = { host_x, host_y, host_z };
849 const size_t region[3] = { width, height, depth }; 1687 const size_t region[3] = { width, height, depth };
850 STRLEN len; 1688 STRLEN len;
851 char *ptr = SvPVbyte (data, len); 1689 char *ptr = SvPVbyte (data, len);
852 EVENT_LIST (17, items - 17);
853 1690
854 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
855 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
856 1693
857 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
866 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1703 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
867 1704
868 if (len < min_len) 1705 if (len < min_len)
869 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1706 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
870 1707
1708 cl_event ev = 0;
871 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)); 1709 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));
872 1710
873 if (ev) 1711 if (ev)
874 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1712 XPUSH_CLOBJ (stash_event, ev);
875 1713
876void 1714void
877enqueue_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, ...) 1715copy_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, ...)
1716 ALIAS:
1717 enqueue_copy_buffer_rect = 0
878 PPCODE: 1718 PPCODE:
879 cl_event ev = 0; 1719 EVENT_LIST (16);
1720
880 const size_t src_origin[3] = { src_x, src_y, src_z }; 1721 const size_t src_origin[3] = { src_x, src_y, src_z };
881 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1722 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
882 const size_t region[3] = { width, height, depth }; 1723 const size_t region[3] = { width, height, depth };
883 EVENT_LIST (16, items - 16);
884 1724
1725 cl_event ev = 0;
885 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)); 1726 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));
886 1727
887 if (ev) 1728 if (ev)
888 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1729 XPUSH_CLOBJ (stash_event, ev);
889 1730
890void 1731void
891enqueue_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, ...) 1732read_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, ...)
1733 ALIAS:
1734 enqueue_read_image = 0
892 PPCODE: 1735 PPCODE:
893 cl_event ev = 0; 1736 EVENT_LIST (12);
1737
894 const size_t src_origin[3] = { src_x, src_y, src_z }; 1738 const size_t src_origin[3] = { src_x, src_y, src_z };
895 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
896 EVENT_LIST (12, items - 12);
897 1740
898 if (!row_pitch) 1741 if (!row_pitch)
899 row_pitch = img_row_pitch (src); 1742 row_pitch = img_row_pitch (src);
900 1743
901 if (depth > 1 && !slice_pitch) 1744 if (depth > 1 && !slice_pitch)
905 1748
906 SvUPGRADE (data, SVt_PV); 1749 SvUPGRADE (data, SVt_PV);
907 SvGROW (data, len); 1750 SvGROW (data, len);
908 SvPOK_only (data); 1751 SvPOK_only (data);
909 SvCUR_set (data, len); 1752 SvCUR_set (data, len);
1753
1754 cl_event ev = 0;
910 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)); 1755 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));
911 1756
912 if (ev) 1757 if (ev)
913 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1758 XPUSH_CLOBJ (stash_event, ev);
914 1759
915void 1760void
916enqueue_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, ...) 1761write_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, ...)
1762 ALIAS:
1763 enqueue_write_image = 0
917 PPCODE: 1764 PPCODE:
918 cl_event ev = 0; 1765 EVENT_LIST (12);
1766
919 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1767 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
920 const size_t region[3] = { width, height, depth }; 1768 const size_t region[3] = { width, height, depth };
921 STRLEN len; 1769 STRLEN len;
922 char *ptr = SvPVbyte (data, len); 1770 char *ptr = SvPVbyte (data, len);
923 EVENT_LIST (12, items - 12);
924 1771
925 if (!row_pitch) 1772 if (!row_pitch)
926 row_pitch = img_row_pitch (dst); 1773 row_pitch = img_row_pitch (dst);
927 1774
928 if (depth > 1 && !slice_pitch) 1775 if (depth > 1 && !slice_pitch)
931 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1778 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
932 1779
933 if (len < min_len) 1780 if (len < min_len)
934 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1781 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
935 1782
1783 cl_event ev = 0;
936 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)); 1784 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));
937 1785
938 if (ev) 1786 if (ev)
939 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1787 XPUSH_CLOBJ (stash_event, ev);
940 1788
941void 1789void
942enqueue_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, ...) 1790copy_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, ...)
1791 ALIAS:
1792 enqueue_copy_image = 0
943 PPCODE: 1793 PPCODE:
944 cl_event ev = 0; 1794 EVENT_LIST (12);
1795
945 const size_t src_origin[3] = { src_x, src_y, src_z }; 1796 const size_t src_origin[3] = { src_x, src_y, src_z };
946 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1797 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
947 const size_t region[3] = { width, height, depth }; 1798 const size_t region[3] = { width, height, depth };
948 EVENT_LIST (12, items - 12);
949 1799
1800 cl_event ev = 0;
950 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1801 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
951 1802
952 if (ev) 1803 if (ev)
953 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1804 XPUSH_CLOBJ (stash_event, ev);
954 1805
955void 1806void
956enqueue_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, ...) 1807copy_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, ...)
1808 ALIAS:
1809 enqueue_copy_image_to_buffer = 0
957 PPCODE: 1810 PPCODE:
1811 EVENT_LIST (10);
1812
1813 const size_t src_origin[3] = { src_x, src_y, src_z };
1814 const size_t region [3] = { width, height, depth };
1815
958 cl_event ev = 0; 1816 cl_event ev = 0;
1817 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1818
1819 if (ev)
1820 XPUSH_CLOBJ (stash_event, ev);
1821
1822void
1823copy_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, ...)
1824 ALIAS:
1825 enqueue_copy_buffer_to_image = 0
1826 PPCODE:
1827 EVENT_LIST (10);
1828
959 const size_t src_origin[3] = { src_x, src_y, src_z }; 1829 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1830 const size_t region [3] = { width, height, depth };
1831
1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839map_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, ...)
1840 ALIAS:
1841 enqueue_map_buffer = 0
1842 PPCODE:
1843 EVENT_LIST (6);
1844
1845 size_t cb = SvIV (cb_);
1846
1847 if (!SvOK (cb_))
1848 {
1849 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1850 cb -= offset;
1851 }
1852
1853 cl_event ev;
1854 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1855 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1856
1857void
1858map_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, ...)
1859 ALIAS:
1860 enqueue_map_image = 0
1861 PPCODE:
1862 size_t width = SvIV (width_);
1863 if (!SvOK (width_))
1864 {
1865 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1866 width -= x;
1867 }
1868
1869 size_t height = SvIV (width_);
1870 if (!SvOK (height_))
1871 {
1872 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1873 height -= y;
1874 }
1875
1876 size_t depth = SvIV (width_);
1877 if (!SvOK (depth_))
1878 {
1879 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1880 depth -= z;
1881
1882 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1883 if (!depth)
1884 depth = 1;
1885 }
1886
1887 const size_t origin[3] = { x, y, z };
960 const size_t region[3] = { width, height, depth }; 1888 const size_t region[3] = { width, height, depth };
1889 size_t row_pitch, slice_pitch;
961 EVENT_LIST (10, items - 10); 1890 EVENT_LIST (10);
962 1891
1892 cl_event ev;
1893 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));
1894
1895 size_t cb = slice_pitch ? slice_pitch * region [2]
1896 : row_pitch ? row_pitch * region [1]
1897 : region [0];
1898
1899 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1900
1901void
1902unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1903 PPCODE:
1904 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1905 if (GIMME_V != G_VOID)
1906 {
1907 clRetainEvent (mapped->event);
1908 XPUSH_CLOBJ (stash_event, mapped->event);
1909 }
1910
1911void
1912task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1913 ALIAS:
1914 enqueue_task = 0
1915 PPCODE:
1916 EVENT_LIST (2);
1917
1918 cl_event ev = 0;
963 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1919 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
964 1920
965 if (ev) 1921 if (ev)
966 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1922 XPUSH_CLOBJ (stash_event, ev);
967 1923
968void 1924void
969enqueue_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, ...)
970 PPCODE:
971 cl_event ev = 0;
972 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
973 const size_t region[3] = { width, height, depth };
974 EVENT_LIST (10, items - 10);
975
976 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
977
978 if (ev)
979 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
980
981void
982enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
983 PPCODE:
984 cl_event ev = 0;
985 EVENT_LIST (2, items - 2);
986
987 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
988
989 if (ev)
990 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
991
992void
993enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1925nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1926 ALIAS:
1927 enqueue_nd_range_kernel = 0
994 PPCODE: 1928 PPCODE:
995 cl_event ev = 0; 1929 EVENT_LIST (5);
1930
996 size_t *gwo = 0, *gws, *lws = 0; 1931 size_t *gwo = 0, *gws, *lws = 0;
997 int gws_len; 1932 int gws_len;
998 size_t *lists; 1933 size_t *lists;
999 int i; 1934 int i;
1000 EVENT_LIST (5, items - 5);
1001 1935
1002 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1936 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1003 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1937 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1004 1938
1005 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1939 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1006 1940
1007 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1941 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1008 1942
1009 gws = lists + gws_len * 0; 1943 gws = lists + gws_len * 0;
1010 for (i = 0; i < gws_len; ++i) 1944 for (i = 0; i < gws_len; ++i)
1945 {
1011 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1946 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1947 // at least nvidia crashes for 0-sized work group sizes, work around
1948 if (!gws [i])
1949 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1950 }
1012 1951
1013 if (SvOK (global_work_offset)) 1952 if (SvOK (global_work_offset))
1014 { 1953 {
1015 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1954 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1016 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1955 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1023 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]); 1962 gwo [i] = SvIV (AvARRAY (SvRV (global_work_offset))[i]);
1024 } 1963 }
1025 1964
1026 if (SvOK (local_work_size)) 1965 if (SvOK (local_work_size))
1027 { 1966 {
1028 if (SvOK (local_work_size) && !SvROK (local_work_size) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1967 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1029 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1968 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1030 1969
1031 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1970 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1032 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1971 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1033 1972
1034 lws = lists + gws_len * 2; 1973 lws = lists + gws_len * 2;
1035 for (i = 0; i < gws_len; ++i) 1974 for (i = 0; i < gws_len; ++i)
1975 {
1036 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1976 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1977 // at least nvidia crashes for 0-sized work group sizes, work around
1978 if (!lws [i])
1979 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1980 }
1037 } 1981 }
1038 1982
1983 cl_event ev = 0;
1039 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1984 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1040 1985
1041 if (ev) 1986 if (ev)
1042 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1987 XPUSH_CLOBJ (stash_event, ev);
1988
1989#if CL_VERSION_1_2
1990
1991void
1992migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1993 ALIAS:
1994 enqueue_migrate_mem_objects = 0
1995 PPCODE:
1996 EVENT_LIST (3);
1997
1998 cl_uint obj_count;
1999 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2000
2001 cl_event ev = 0;
2002 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2003
2004 if (ev)
2005 XPUSH_CLOBJ (stash_event, ev);
2006
2007#endif
1043 2008
1044#if cl_apple_gl_sharing || cl_khr_gl_sharing 2009#if cl_apple_gl_sharing || cl_khr_gl_sharing
1045 2010
1046void 2011void
1047enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2012acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1048 ALIAS: 2013 ALIAS:
2014 release_gl_objects = 1
2015 enqueue_acquire_gl_objects = 0
1049 enqueue_release_gl_objects = 1 2016 enqueue_release_gl_objects = 1
1050 CODE: 2017 PPCODE:
1051 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2018 EVENT_LIST (2);
1052 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2019
2020 cl_uint obj_count;
2021 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1053 2022
1054 cl_event ev = 0; 2023 cl_event ev = 0;
1055 EVENT_LIST (2, items - 2);
1056 AV *av = (AV *)SvRV (objects);
1057 cl_uint num_objects = av_len (av) + 1;
1058 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1059 int i;
1060
1061 for (i = num_objects; i--; )
1062 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1063 2024
1064 if (ix) 2025 if (ix)
1065 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2026 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1066 else 2027 else
1067 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2028 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1068 2029
1069 if (ev) 2030 if (ev)
1070 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2031 XPUSH_CLOBJ (stash_event, ev);
1071 2032
1072#endif 2033#endif
1073 2034
1074void 2035void
1075enqueue_marker (OpenCL::Queue self)
1076 PPCODE:
1077 cl_event ev;
1078 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1079 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1080
1081void
1082enqueue_wait_for_events (OpenCL::Queue self, ...) 2036wait_for_events (OpenCL::Queue self, ...)
2037 ALIAS:
2038 enqueue_wait_for_events = 0
1083 CODE: 2039 CODE:
1084 EVENT_LIST (1, items - 1); 2040 EVENT_LIST (1);
2041#if PREFER_1_1
1085 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2042 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2043#else
2044 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
2045#endif
1086 2046
1087void 2047void
1088enqueue_barrier (OpenCL::Queue self) 2048marker (OpenCL::Queue self, ...)
2049 ALIAS:
2050 enqueue_marker = 0
1089 CODE: 2051 PPCODE:
2052 EVENT_LIST (1);
2053 cl_event ev = 0;
2054#if PREFER_1_1
2055 if (!event_list_count)
2056 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2057 else
2058#if CL_VERSION_1_2
2059 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2060#else
2061 {
2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2063 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2064 }
2065#endif
2066#else
2067 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2068#endif
2069 if (ev)
2070 XPUSH_CLOBJ (stash_event, ev);
2071
2072void
2073barrier (OpenCL::Queue self, ...)
2074 ALIAS:
2075 enqueue_barrier = 0
2076 PPCODE:
2077 EVENT_LIST (1);
2078 cl_event ev = 0;
2079#if PREFER_1_1
2080 if (!event_list_count && GIMME_V == G_VOID)
1090 NEED_SUCCESS (EnqueueBarrier, (self)); 2081 NEED_SUCCESS (EnqueueBarrier, (self));
2082 else
2083#if CL_VERSION_1_2
2084 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2085#else
2086 {
2087 if (event_list_count)
2088 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2089
2090 if (GIMME_V != G_VOID)
2091 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2092 }
2093#endif
2094#else
2095 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2096#endif
2097 if (ev)
2098 XPUSH_CLOBJ (stash_event, ev);
1091 2099
1092void 2100void
1093flush (OpenCL::Queue self) 2101flush (OpenCL::Queue self)
1094 CODE: 2102 CODE:
1095 NEED_SUCCESS (Flush, (self)); 2103 NEED_SUCCESS (Flush, (self));
1113 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2121 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1114 EXTEND (SP, 1); 2122 EXTEND (SP, 1);
1115 const int i = 0; 2123 const int i = 0;
1116 { 2124 {
1117 NEED_SUCCESS (RetainContext, (value [i])); 2125 NEED_SUCCESS (RetainContext, (value [i]));
1118 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2126 PUSH_CLOBJ (stash_context, value [i]);
1119 } 2127 }
1120 2128
1121void 2129void
1122device (OpenCL::Queue self) 2130device (OpenCL::Queue self)
1123 PPCODE: 2131 PPCODE:
1124 cl_device_id value [1]; 2132 cl_device_id value [1];
1125 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2133 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1126 EXTEND (SP, 1); 2134 EXTEND (SP, 1);
1127 const int i = 0; 2135 const int i = 0;
1128 { 2136 {
1129 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2137 PUSH_CLOBJ (stash_device, value [i]);
1130 } 2138 }
1131 2139
1132void 2140void
1133reference_count (OpenCL::Queue self) 2141reference_count (OpenCL::Queue self)
1134 PPCODE: 2142 PPCODE:
1143 PPCODE: 2151 PPCODE:
1144 cl_command_queue_properties value [1]; 2152 cl_command_queue_properties value [1];
1145 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2153 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1146 EXTEND (SP, 1); 2154 EXTEND (SP, 1);
1147 const int i = 0; 2155 const int i = 0;
1148 PUSHs (sv_2mortal (newSViv (value [i]))); 2156 PUSHs (sv_2mortal (newSVuv (value [i])));
1149 2157
1150#END:command_queue 2158#END:command_queue
1151 2159
1152MODULE = OpenCL PACKAGE = OpenCL::Memory 2160MODULE = OpenCL PACKAGE = OpenCL::Memory
1153 2161
1168 PPCODE: 2176 PPCODE:
1169 cl_mem_object_type value [1]; 2177 cl_mem_object_type value [1];
1170 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2178 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1171 EXTEND (SP, 1); 2179 EXTEND (SP, 1);
1172 const int i = 0; 2180 const int i = 0;
1173 PUSHs (sv_2mortal (newSViv (value [i]))); 2181 PUSHs (sv_2mortal (newSVuv (value [i])));
1174 2182
1175void 2183void
1176flags (OpenCL::Memory self) 2184flags (OpenCL::Memory self)
1177 PPCODE: 2185 PPCODE:
1178 cl_mem_flags value [1]; 2186 cl_mem_flags value [1];
1179 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2187 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1180 EXTEND (SP, 1); 2188 EXTEND (SP, 1);
1181 const int i = 0; 2189 const int i = 0;
1182 PUSHs (sv_2mortal (newSViv (value [i]))); 2190 PUSHs (sv_2mortal (newSVuv (value [i])));
1183 2191
1184void 2192void
1185size (OpenCL::Memory self) 2193size (OpenCL::Memory self)
1186 ALIAS: 2194 ALIAS:
1187 size = CL_MEM_SIZE 2195 size = CL_MEM_SIZE
1221 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2229 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1222 EXTEND (SP, 1); 2230 EXTEND (SP, 1);
1223 const int i = 0; 2231 const int i = 0;
1224 { 2232 {
1225 NEED_SUCCESS (RetainContext, (value [i])); 2233 NEED_SUCCESS (RetainContext, (value [i]));
1226 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2234 PUSH_CLOBJ (stash_context, value [i]);
1227 } 2235 }
1228 2236
1229void 2237void
1230associated_memobject (OpenCL::Memory self) 2238associated_memobject (OpenCL::Memory self)
1231 PPCODE: 2239 PPCODE:
1233 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2241 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1234 EXTEND (SP, 1); 2242 EXTEND (SP, 1);
1235 const int i = 0; 2243 const int i = 0;
1236 { 2244 {
1237 NEED_SUCCESS (RetainMemObject, (value [i])); 2245 NEED_SUCCESS (RetainMemObject, (value [i]));
1238 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2246 PUSH_CLOBJ (stash_memory, value [i]);
1239 } 2247 }
1240 2248
1241#END:mem 2249#END:mem
1242 2250
1243#if cl_apple_gl_sharing || cl_khr_gl_sharing 2251#if cl_apple_gl_sharing || cl_khr_gl_sharing
1245void 2253void
1246gl_object_info (OpenCL::Memory self) 2254gl_object_info (OpenCL::Memory self)
1247 PPCODE: 2255 PPCODE:
1248 cl_gl_object_type type; 2256 cl_gl_object_type type;
1249 cl_GLuint name; 2257 cl_GLuint name;
1250 NEED_SUCCESS (clGetGLObjectInfo, (self, &type, &name)); 2258 NEED_SUCCESS (GetGLObjectInfo, (self, &type, &name));
1251 EXTEND (SP, 2); 2259 EXTEND (SP, 2);
1252 PUSHs (sv_2mortal (newSVuv (type))); 2260 PUSHs (sv_2mortal (newSVuv (type)));
1253 PUSHs (sv_2mortal (newSVuv (name))); 2261 PUSHs (sv_2mortal (newSVuv (name)));
1254 2262
1255#endif 2263#endif
1263 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2271 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1264 2272
1265 cl_buffer_region crdata = { origin, size }; 2273 cl_buffer_region crdata = { origin, size };
1266 2274
1267 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2275 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1268 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2276 XPUSH_CLOBJ (stash_buffer, mem);
1269 2277
1270MODULE = OpenCL PACKAGE = OpenCL::Image 2278MODULE = OpenCL PACKAGE = OpenCL::Image
1271 2279
1272void 2280void
1273image_info (OpenCL::Image self, cl_image_info name) 2281image_info (OpenCL::Image self, cl_image_info name)
1274 PPCODE: 2282 PPCODE:
1275 INFO (Image) 2283 INFO (Image)
2284
2285void
2286format (OpenCL::Image self)
2287 PPCODE:
2288 cl_image_format format;
2289 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2290 EXTEND (SP, 2);
2291 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2292 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1276 2293
1277#BEGIN:image 2294#BEGIN:image
1278 2295
1279void 2296void
1280element_size (OpenCL::Image self) 2297element_size (OpenCL::Image self)
1300 2317
1301void 2318void
1302target (OpenCL::Image self) 2319target (OpenCL::Image self)
1303 PPCODE: 2320 PPCODE:
1304 cl_GLenum value [1]; 2321 cl_GLenum value [1];
1305 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0)); 2322 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_TEXTURE_TARGET, sizeof (value), value, 0));
1306 EXTEND (SP, 1); 2323 EXTEND (SP, 1);
1307 const int i = 0; 2324 const int i = 0;
1308 PUSHs (sv_2mortal (newSVuv (value [i]))); 2325 PUSHs (sv_2mortal (newSVuv (value [i])));
1309 2326
1310void 2327void
1311gl_mipmap_level (OpenCL::Image self) 2328gl_mipmap_level (OpenCL::Image self)
1312 PPCODE: 2329 PPCODE:
1313 cl_GLint value [1]; 2330 cl_GLint value [1];
1314 NEED_SUCCESS (GetGlTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0)); 2331 NEED_SUCCESS (GetGLTextureInfo, (self, CL_GL_MIPMAP_LEVEL, sizeof (value), value, 0));
1315 EXTEND (SP, 1); 2332 EXTEND (SP, 1);
1316 const int i = 0; 2333 const int i = 0;
1317 PUSHs (sv_2mortal (newSViv (value [i]))); 2334 PUSHs (sv_2mortal (newSViv (value [i])));
1318 2335
1319#END:gl_texture 2336#END:gl_texture
1350 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2367 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1351 EXTEND (SP, 1); 2368 EXTEND (SP, 1);
1352 const int i = 0; 2369 const int i = 0;
1353 { 2370 {
1354 NEED_SUCCESS (RetainContext, (value [i])); 2371 NEED_SUCCESS (RetainContext, (value [i]));
1355 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2372 PUSH_CLOBJ (stash_context, value [i]);
1356 } 2373 }
1357 2374
1358void 2375void
1359normalized_coords (OpenCL::Sampler self) 2376normalized_coords (OpenCL::Sampler self)
1360 PPCODE: 2377 PPCODE:
1361 cl_addressing_mode value [1]; 2378 cl_addressing_mode value [1];
1362 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2379 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1363 EXTEND (SP, 1); 2380 EXTEND (SP, 1);
1364 const int i = 0; 2381 const int i = 0;
1365 PUSHs (sv_2mortal (newSViv (value [i]))); 2382 PUSHs (sv_2mortal (newSVuv (value [i])));
1366 2383
1367void 2384void
1368addressing_mode (OpenCL::Sampler self) 2385addressing_mode (OpenCL::Sampler self)
1369 PPCODE: 2386 PPCODE:
1370 cl_filter_mode value [1]; 2387 cl_filter_mode value [1];
1371 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2388 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1372 EXTEND (SP, 1); 2389 EXTEND (SP, 1);
1373 const int i = 0; 2390 const int i = 0;
1374 PUSHs (sv_2mortal (newSViv (value [i]))); 2391 PUSHs (sv_2mortal (newSVuv (value [i])));
1375 2392
1376void 2393void
1377filter_mode (OpenCL::Sampler self) 2394filter_mode (OpenCL::Sampler self)
1378 PPCODE: 2395 PPCODE:
1379 cl_bool value [1]; 2396 cl_bool value [1];
1390DESTROY (OpenCL::Program self) 2407DESTROY (OpenCL::Program self)
1391 CODE: 2408 CODE:
1392 clReleaseProgram (self); 2409 clReleaseProgram (self);
1393 2410
1394void 2411void
1395build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2412build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2413 ALIAS:
2414 build_async = 1
1396 CODE: 2415 CODE:
2416 cl_uint device_count = 0;
2417 cl_device_id *device_list = 0;
2418
2419 if (SvOK (devices))
2420 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2421
2422 void *user_data;
2423 program_callback pfn_notify = make_program_callback (notify, &user_data);
2424
2425 if (ix)
2426 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2427 else
2428 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2429
2430#if CL_VERSION_1_2
2431
2432void
2433compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2434 CODE:
2435 cl_uint device_count = 0;
2436 cl_device_id *device_list = 0;
2437
2438 if (SvOK (devices))
2439 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2440
2441 cl_uint header_count = 0;
2442 cl_program *header_list = 0;
2443 const char **header_name = 0;
2444
2445 if (SvOK (headers))
2446 {
2447 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2448 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2449
2450 HV *hv = (HV *)SvRV (devices);
2451
2452 header_count = hv_iterinit (hv);
2453 header_list = tmpbuf (sizeof (*header_list) * header_count);
2454 header_name = tmpbuf (sizeof (*header_name) * header_count);
2455
2456 HE *he;
2457 int i = 0;
2458 while (he = hv_iternext (hv))
2459 {
2460 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2461 header_list [i] = SvCLOBJ ("clCompileProgram", "headers", HeVAL (he), "OpenCL::Program");
2462 ++i;
2463 }
2464 }
2465
2466 void *user_data;
2467 program_callback pfn_notify = make_program_callback (notify, &user_data);
2468
1397 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2469 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2470 header_count, header_list, header_name, pfn_notify, user_data));
2471
2472#endif
1398 2473
1399void 2474void
1400build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2475build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1401 PPCODE: 2476 PPCODE:
1402 size_t size; 2477 size_t size;
1437 2512
1438void 2513void
1439kernel (OpenCL::Program program, SV *function) 2514kernel (OpenCL::Program program, SV *function)
1440 PPCODE: 2515 PPCODE:
1441 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2516 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1442 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2517 XPUSH_CLOBJ (stash_kernel, kernel);
2518
2519void
2520kernels_in_program (OpenCL::Program program)
2521 PPCODE:
2522 cl_uint num_kernels;
2523 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2524 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2525 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2526
2527 int i;
2528 EXTEND (SP, num_kernels);
2529 for (i = 0; i < num_kernels; ++i)
2530 PUSH_CLOBJ (stash_kernel, kernels [i]);
1443 2531
1444void 2532void
1445info (OpenCL::Program self, cl_program_info name) 2533info (OpenCL::Program self, cl_program_info name)
1446 PPCODE: 2534 PPCODE:
1447 INFO (Program) 2535 INFO (Program)
1465 { 2553 {
1466 SV *sv = sv_2mortal (newSV (sizes [i])); 2554 SV *sv = sv_2mortal (newSV (sizes [i]));
1467 SvUPGRADE (sv, SVt_PV); 2555 SvUPGRADE (sv, SVt_PV);
1468 SvPOK_only (sv); 2556 SvPOK_only (sv);
1469 SvCUR_set (sv, sizes [i]); 2557 SvCUR_set (sv, sizes [i]);
1470 ptrs [i] = SvPVX (sv); 2558 ptrs [i] = (void *)SvPVX (sv);
1471 PUSHs (sv); 2559 PUSHs (sv);
1472 } 2560 }
1473 2561
1474 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size)); 2562 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_BINARIES , sizeof (*ptrs ) * n, ptrs , &size));
1475 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY; 2563 if (size != sizeof (*ptrs) * n) XSRETURN_EMPTY;
1495 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2583 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1496 EXTEND (SP, 1); 2584 EXTEND (SP, 1);
1497 const int i = 0; 2585 const int i = 0;
1498 { 2586 {
1499 NEED_SUCCESS (RetainContext, (value [i])); 2587 NEED_SUCCESS (RetainContext, (value [i]));
1500 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2588 PUSH_CLOBJ (stash_context, value [i]);
1501 } 2589 }
1502 2590
1503void 2591void
1504devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
1505 PPCODE: 2593 PPCODE:
1509 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1510 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
1511 EXTEND (SP, n); 2599 EXTEND (SP, n);
1512 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
1513 { 2601 {
1514 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i])); 2602 PUSH_CLOBJ (stash_device, value [i]);
1515 } 2603 }
1516 2604
1517void 2605void
1518source (OpenCL::Program self) 2606source (OpenCL::Program self)
1519 PPCODE: 2607 PPCODE:
1545DESTROY (OpenCL::Kernel self) 2633DESTROY (OpenCL::Kernel self)
1546 CODE: 2634 CODE:
1547 clReleaseKernel (self); 2635 clReleaseKernel (self);
1548 2636
1549void 2637void
2638setf (OpenCL::Kernel self, const char *format, ...)
2639 CODE:
2640 int i;
2641 for (i = 2; ; ++i)
2642 {
2643 while (*format == ' ')
2644 ++format;
2645
2646 char type = *format++;
2647
2648 if (!type)
2649 break;
2650
2651 if (i >= items)
2652 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2653
2654 SV *sv = ST (i);
2655
2656 union
2657 {
2658 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2659 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2660 cl_half ch; cl_float cf; cl_double cd;
2661 cl_mem cm;
2662 cl_sampler ca;
2663 size_t cz;
2664 cl_event ce;
2665 } arg;
2666 size_t size;
2667 int nullarg = 0;
2668
2669 switch (type)
2670 {
2671 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2672 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2673 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2674 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2675 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2676 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2677 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2678 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2679
2680 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2681 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2682 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2683
2684 case 'z': nullarg = 1; size = SvIV (sv); break;
2685
2686 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2687 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2688 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2689
2690 default:
2691 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2692 }
2693
2694 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2695 if (res)
2696 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2697 }
2698
2699 if (i != items)
2700 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2701
2702void
1550set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2703set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1551 CODE: 2704 CODE:
1552 clSetKernelArg (self, idx, sizeof (value), &value); 2705 clSetKernelArg (self, idx, sizeof (value), &value);
1553 2706
1554void 2707void
1602 clSetKernelArg (self, idx, sizeof (value), &value); 2755 clSetKernelArg (self, idx, sizeof (value), &value);
1603 2756
1604void 2757void
1605set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2758set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1606 CODE: 2759 CODE:
2760 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2761
2762void
2763set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2764 CODE:
2765 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2766
2767void
2768set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2769 ALIAS:
2770 set_image2d = 0
2771 set_image3d = 0
2772 CODE:
2773 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2774
2775void
2776set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2777 CODE:
1607 clSetKernelArg (self, idx, sizeof (value), &value); 2778 clSetKernelArg (self, idx, sizeof (value), &value);
1608 2779
1609void 2780void
1610set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2781set_local (OpenCL::Kernel self, cl_uint idx, size_t size)
1611 CODE: 2782 CODE:
1612 clSetKernelArg (self, idx, sizeof (value), &value); 2783 clSetKernelArg (self, idx, size, 0);
1613
1614void
1615set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value)
1616 CODE:
1617 clSetKernelArg (self, idx, sizeof (value), &value);
1618
1619void
1620set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1621 CODE:
1622 clSetKernelArg (self, idx, sizeof (value), &value);
1623
1624void
1625set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1626 CODE:
1627 clSetKernelArg (self, idx, sizeof (value), &value);
1628 2784
1629void 2785void
1630set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value) 2786set_event (OpenCL::Kernel self, cl_uint idx, OpenCL::Event value)
1631 CODE: 2787 CODE:
1632 clSetKernelArg (self, idx, sizeof (value), &value); 2788 clSetKernelArg (self, idx, sizeof (value), &value);
1668 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2824 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1669 EXTEND (SP, 1); 2825 EXTEND (SP, 1);
1670 const int i = 0; 2826 const int i = 0;
1671 { 2827 {
1672 NEED_SUCCESS (RetainContext, (value [i])); 2828 NEED_SUCCESS (RetainContext, (value [i]));
1673 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2829 PUSH_CLOBJ (stash_context, value [i]);
1674 } 2830 }
1675 2831
1676void 2832void
1677program (OpenCL::Kernel self) 2833program (OpenCL::Kernel self)
1678 PPCODE: 2834 PPCODE:
1680 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2836 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1681 EXTEND (SP, 1); 2837 EXTEND (SP, 1);
1682 const int i = 0; 2838 const int i = 0;
1683 { 2839 {
1684 NEED_SUCCESS (RetainProgram, (value [i])); 2840 NEED_SUCCESS (RetainProgram, (value [i]));
1685 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2841 PUSH_CLOBJ (stash_program, value [i]);
1686 } 2842 }
1687 2843
1688#END:kernel 2844#END:kernel
1689 2845
1690void 2846void
1737 const int i = 0; 2893 const int i = 0;
1738 PUSHs (sv_2mortal (newSVuv (value [i]))); 2894 PUSHs (sv_2mortal (newSVuv (value [i])));
1739 2895
1740#END:kernel_work_group 2896#END:kernel_work_group
1741 2897
2898#if CL_VERSION_1_2
2899
2900void
2901arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2902 PPCODE:
2903 size_t size;
2904 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2905 SV *sv = sv_2mortal (newSV (size));
2906 SvUPGRADE (sv, SVt_PV);
2907 SvPOK_only (sv);
2908 SvCUR_set (sv, size);
2909 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2910 XPUSHs (sv);
2911
2912#BEGIN:kernel_arg
2913
2914void
2915arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2916 PPCODE:
2917 cl_kernel_arg_address_qualifier value [1];
2918 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2919 EXTEND (SP, 1);
2920 const int i = 0;
2921 PUSHs (sv_2mortal (newSVuv (value [i])));
2922
2923void
2924arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2925 PPCODE:
2926 cl_kernel_arg_access_qualifier value [1];
2927 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2928 EXTEND (SP, 1);
2929 const int i = 0;
2930 PUSHs (sv_2mortal (newSVuv (value [i])));
2931
2932void
2933arg_type_name (OpenCL::Kernel self, cl_uint idx)
2934 ALIAS:
2935 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2936 arg_name = CL_KERNEL_ARG_NAME
2937 PPCODE:
2938 size_t size;
2939 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2940 char *value = tmpbuf (size);
2941 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2942 EXTEND (SP, 1);
2943 const int i = 0;
2944 PUSHs (sv_2mortal (newSVpv (value, 0)));
2945
2946void
2947arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2948 PPCODE:
2949 cl_kernel_arg_type_qualifier value [1];
2950 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2951 EXTEND (SP, 1);
2952 const int i = 0;
2953 PUSHs (sv_2mortal (newSVuv (value [i])));
2954
2955#END:kernel_arg
2956
2957#endif
2958
1742MODULE = OpenCL PACKAGE = OpenCL::Event 2959MODULE = OpenCL PACKAGE = OpenCL::Event
1743 2960
1744void 2961void
1745DESTROY (OpenCL::Event self) 2962DESTROY (OpenCL::Event self)
1746 CODE: 2963 CODE:
1748 2965
1749void 2966void
1750wait (OpenCL::Event self) 2967wait (OpenCL::Event self)
1751 CODE: 2968 CODE:
1752 clWaitForEvents (1, &self); 2969 clWaitForEvents (1, &self);
2970
2971void
2972cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2973 CODE:
2974 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1753 2975
1754void 2976void
1755info (OpenCL::Event self, cl_event_info name) 2977info (OpenCL::Event self, cl_event_info name)
1756 PPCODE: 2978 PPCODE:
1757 INFO (Event) 2979 INFO (Event)
1765 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2987 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1766 EXTEND (SP, 1); 2988 EXTEND (SP, 1);
1767 const int i = 0; 2989 const int i = 0;
1768 { 2990 {
1769 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2991 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1770 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2992 PUSH_CLOBJ (stash_queue, value [i]);
1771 } 2993 }
1772 2994
1773void 2995void
1774command_type (OpenCL::Event self) 2996command_type (OpenCL::Event self)
1775 PPCODE: 2997 PPCODE:
1798 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 3020 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1799 EXTEND (SP, 1); 3021 EXTEND (SP, 1);
1800 const int i = 0; 3022 const int i = 0;
1801 { 3023 {
1802 NEED_SUCCESS (RetainContext, (value [i])); 3024 NEED_SUCCESS (RetainContext, (value [i]));
1803 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 3025 PUSH_CLOBJ (stash_context, value [i]);
1804 } 3026 }
1805 3027
1806#END:event 3028#END:event
1807 3029
1808void 3030void
1833void 3055void
1834set_status (OpenCL::UserEvent self, cl_int execution_status) 3056set_status (OpenCL::UserEvent self, cl_int execution_status)
1835 CODE: 3057 CODE:
1836 clSetUserEventStatus (self, execution_status); 3058 clSetUserEventStatus (self, execution_status);
1837 3059
3060MODULE = OpenCL PACKAGE = OpenCL::Mapped
3061
3062void
3063DESTROY (SV *self)
3064 CODE:
3065 OpenCL__Mapped mapped = SvMAPPED (self);
3066
3067 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3068 mapped_detach (self, mapped);
3069
3070 clReleaseCommandQueue (mapped->queue);
3071 clReleaseEvent (mapped->event);
3072 Safefree (mapped);
3073
3074void
3075unmap (OpenCL::Mapped self, ...)
3076 CODE:
3077 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
3078
3079bool
3080mapped (OpenCL::Mapped self)
3081 CODE:
3082 RETVAL = !!self->ptr;
3083 OUTPUT:
3084 RETVAL
3085
3086void
3087wait (OpenCL::Mapped self)
3088 PPCODE:
3089 if (self->event)
3090 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3091
3092void
3093event (OpenCL::Mapped self)
3094 PPCODE:
3095 if (!self->event)
3096 XSRETURN_UNDEF;
3097
3098 clRetainEvent (self->event);
3099 XPUSH_CLOBJ (stash_event, self->event);
3100
3101size_t
3102size (OpenCL::Mapped self)
3103 CODE:
3104 RETVAL = self->cb;
3105 OUTPUT:
3106 RETVAL
3107
3108IV
3109ptr (OpenCL::Mapped self)
3110 CODE:
3111 RETVAL = PTR2IV (self->ptr);
3112 OUTPUT:
3113 RETVAL
3114
3115void
3116set (OpenCL::Mapped self, size_t offset, SV *data)
3117 CODE:
3118 STRLEN len;
3119 const char *ptr = SvPVbyte (data, len);
3120
3121 if (offset + len > self->cb)
3122 croak ("OpenCL::Mapped::set out of bound condition detected");
3123
3124 memcpy (offset + (char *)self->ptr, ptr, len);
3125
3126MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3127
3128MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3129
3130IV
3131row_pitch (OpenCL::Mapped self)
3132 ALIAS:
3133 slice_pitch = 1
3134 CODE:
3135 RETVAL = ix ? self->slice_pitch : self->row_pitch;
3136 OUTPUT:
3137 RETVAL
3138
3139
3140

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines