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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.62 by root, Tue May 1 22:04:56 2012 UTC vs.
Revision 1.77 by root, Sat May 5 14:57:32 2012 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4
5#include "ecb.h"//D
4 6
5#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really 7#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
6#include "xthread.h" 8#include "xthread.h"
7#include "schmorp.h" 9#include "schmorp.h"
8 10
26#if PREFER_1_1 28#if PREFER_1_1
27 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
28#endif 30#endif
29 31
30#ifdef __APPLE__ 32#ifdef __APPLE__
31 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
32#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
33 #include <CL/opencl.h> 38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
34#endif 50#endif
35 51
36#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
37 #undef PREFER_1_1 53 #undef PREFER_1_1
38 #define PREFER_1_1 1 54 #define PREFER_1_1 1
39#endif 55#endif
40 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
41typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 68typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 69typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 70typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 71typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 72typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 73typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 74typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
61 77
62typedef struct mapped * OpenCL__Mapped; 78typedef struct mapped * OpenCL__Mapped;
63 79
64typedef SV *FUTURE;
65
66static HV 80static HV
67 *stash_platform, 81 *stash_platform,
68 *stash_device, 82 *stash_device,
83 *stash_subdevice,
69 *stash_context, 84 *stash_context,
70 *stash_queue, 85 *stash_queue,
71 *stash_program, 86 *stash_program,
72 *stash_kernel, 87 *stash_kernel,
73 *stash_sampler, 88 *stash_sampler,
133} 148}
134 149
135/*****************************************************************************/ 150/*****************************************************************************/
136 151
137/* up to two temporary buffers */ 152/* up to two temporary buffers */
138static void * 153static void * ecb_noinline
139tmpbuf (size_t size) 154tmpbuf (size_t size)
140{ 155{
141 enum { buffers = 3 }; 156 enum { buffers = 4 };
142 static int idx; 157 static int idx;
143 static void *buf [buffers]; 158 static void *buf [buffers];
144 static size_t len [buffers]; 159 static size_t len [buffers];
145 160
146 idx = (idx + 1) % buffers; 161 idx = (idx + 1) % buffers;
153 } 168 }
154 169
155 return buf [idx]; 170 return buf [idx];
156} 171}
157 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
158/*****************************************************************************/ 188/*****************************************************************************/
159 189
160typedef struct 190typedef struct
161{ 191{
162 IV iv; 192 IV iv;
163 const char *name; 193 const char *name;
164 #define const_iv(name) { (IV)CL_ ## name, # name }, 194 #define const_iv(name) { (IV)CL_ ## name, # name },
165} ivstr; 195} ivstr;
196
197typedef struct
198{
199 NV nv;
200 const char *name;
201 #define const_nv(name) { (NV)CL_ ## name, # name },
202} nvstr;
166 203
167static const char * 204static const char *
168iv2str (IV value, const ivstr *base, int count, const char *fallback) 205iv2str (IV value, const ivstr *base, int count, const char *fallback)
169{ 206{
170 int i; 207 int i;
219 if (res) \ 256 if (res) \
220 FAIL (name); 257 FAIL (name);
221 258
222/*****************************************************************************/ 259/*****************************************************************************/
223 260
261static SV *
262new_clobj (HV *stash, IV id)
263{
264 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
265}
266
267#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
268#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
269
270/* cl objects are either \$iv, or [$iv, ...] */
271/* they can be upgraded at runtime to the array form */
272static void * ecb_noinline
273SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
274{
275 // sv_derived_from is quite slow :(
276 if (SvROK (sv) && sv_derived_from (sv, pkg))
277 return (void *)SvIV (SvRV (sv));
278
279 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
280}
281
282// the "no-inherit" version of the above
283static void * ecb_noinline
284SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
285{
286 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
287 return (void *)SvIV (SvRV (sv));
288
289 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
290}
291
292/*****************************************************************************/
293
224static cl_context_properties * 294static cl_context_properties * ecb_noinline
225SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 295SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
226{ 296{
227 if (!sv || !SvOK (sv)) 297 if (!sv || !SvOK (sv))
228 if (extra) 298 if (extra)
229 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 299 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
230 else 300 else
236 int i, len = av_len (av) + 1; 306 int i, len = av_len (av) + 1;
237 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 307 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
238 cl_context_properties *l = p; 308 cl_context_properties *l = p;
239 309
240 if (len & 1) 310 if (len & 1)
241 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 311 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
242 312
243 while (extracount--) 313 while (extracount--)
244 *l++ = *extra++; 314 *l++ = *extra++;
245 315
246 for (i = 0; i < len; i += 2) 316 for (i = 0; i < len; i += 2)
249 SV *p_sv = *av_fetch (av, i + 1, 0); 319 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override 320 cl_context_properties v = SvIV (p_sv); // code below can override
251 321
252 switch (t) 322 switch (t)
253 { 323 {
324 case CL_CONTEXT_PLATFORM:
325 if (SvROK (p_sv))
326 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
327 break;
328
254 case CL_GLX_DISPLAY_KHR: 329 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv)) 330 if (!SvOK (p_sv))
256 { 331 {
257 void *func = glsym ("_glXGetCurrentDisplay"); 332 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func) 333 if (func)
281 *l = 0; 356 *l = 0;
282 357
283 return p; 358 return p;
284 } 359 }
285 360
286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 361 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
287} 362}
288 363
289/*****************************************************************************/ 364// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
290 365// is a pointer (and only on well-behaved systems).
291static SV * 366static void * ecb_noinline
292new_clobj (HV *stash, IV id) 367object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
293{ 368{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash)); 369 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
295} 370 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
371 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
296 372
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id))) 373 AV *av = (AV *)SvRV (arg);
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id))) 374 void **list = 0;
375 cl_uint count = av_len (av) + 1;
299 376
300/* cl objects are either \$iv, or [$iv, ...] */ 377 if (count)
301/* they can be upgraded at runtime to the array form */ 378 {
302static void * 379 list = tmpbuf (sizeof (*list) * count);
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 380 int i;
304{ 381 for (i = 0; i < count; ++i)
305 if (SvROK (sv) && sv_derived_from (sv, pkg)) 382 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
306 return (void *)SvIV (SvRV (sv)); 383 }
307 384
308 croak ("%s: %s is not of type %s", func, svname, pkg); 385 if (!count && !or_undef)
386 croak ("%s: '%s' must contain at least one %s object",
387 cv_get_name (cv), argname, klass);
388
389 *rcount = count;
390 return (void *)list;
309} 391}
310 392
311/*****************************************************************************/ 393/*****************************************************************************/
312/* callback stuff */ 394/* callback stuff */
313 395
335static void (*eq_signal_func)(void *signal_arg, int value); 417static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg; 418static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT; 419static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail; 420static eq_item *eq_head, *eq_tail;
339 421
340static void 422static void ecb_noinline
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 423eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{ 424{
343 eq_item *item = malloc (sizeof (eq_item)); 425 eq_item *item = malloc (sizeof (eq_item));
344 426
345 item->next = 0; 427 item->next = 0;
412} 494}
413 495
414/*****************************************************************************/ 496/*****************************************************************************/
415/* context notify */ 497/* context notify */
416 498
417static void 499static void ecb_noinline
418eq_context_push (void *data1, void *data2, void *data3) 500eq_context_push (void *data1, void *data2, void *data3)
419{ 501{
420 dSP; 502 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0))); 503 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 504 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
444 { \ 526 { \
445 pfn_notify = eq_context_notify; \ 527 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \ 528 user_data = s_get_cv (notify); \
447 } 529 }
448 530
449static SV * 531static SV * ecb_noinline
450new_clobj_context (cl_context ctx, void *user_data) 532new_clobj_context (cl_context ctx, void *user_data)
451{ 533{
452 SV *sv = new_clobj (stash_context, (IV)ctx); 534 SV *sv = new_clobj (stash_context, (IV)ctx);
453 535
454 if (user_data) 536 if (user_data)
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 555static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474 556
475static void CL_CALLBACK 557static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data) 558eq_program_notify (cl_program program, void *user_data)
477{ 559{
560 clRetainProgram (program);
561
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 562 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
563}
564
565typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
566
567static program_callback ecb_noinline
568make_program_callback (SV *notify, void **ruser_data)
569{
570 if (SvOK (notify))
571 {
572 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
573 return eq_program_notify;
574 }
575 else
576 {
577 *ruser_data = 0;
578 return 0;
579 }
479} 580}
480 581
481struct build_args 582struct build_args
482{ 583{
483 cl_program program; 584 cl_program program;
497 else 598 else
498 clReleaseProgram (arg->program); 599 clReleaseProgram (arg->program);
499 600
500 free (arg->options); 601 free (arg->options);
501 free (arg); 602 free (arg);
603
604 return 0;
502} 605}
503 606
504static void 607static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 608build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{ 609{
546 size_t res; 649 size_t res;
547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 650 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
548 return res; 651 return res;
549} 652}
550 653
551static cl_event * 654static cl_event * ecb_noinline
552event_list (SV **items, cl_uint *rcount, cl_event extra) 655event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
553{ 656{
554 cl_uint count = *rcount; 657 cl_uint count = *rcount;
555 658
556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed 659 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
557 count = 0; 660 *rcount = count = 0;
558 661
559 if (!count && !extra) 662 if (!count && !extra)
560 return 0; 663 return 0;
561 664
562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 665 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
563 int i = 0; 666 int i = 0;
564 667
565 while (count--) 668 while (count--)
566 if (SvOK (items [count])) 669 if (SvOK (items [count]))
567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 670 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
568 671
569 if (extra) 672 if (extra)
570 list [i++] = extra; 673 list [i++] = extra;
571 674
572 *rcount = i; 675 *rcount = i;
574 return i ? list : 0; 677 return i ? list : 0;
575} 678}
576 679
577#define EVENT_LIST(skip) \ 680#define EVENT_LIST(skip) \
578 cl_uint event_list_count = items - (skip); \ 681 cl_uint event_list_count = items - (skip); \
579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 682 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
580 683
581#define INFO(class) \ 684#define INFO(class) \
582{ \ 685{ \
583 size_t size; \ 686 size_t size; \
584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 687 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
666 769
667 mapped->ptr = 0; 770 mapped->ptr = 0;
668} 771}
669 772
670static void 773static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count) 774mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{ 775{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event); 776 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
674 cl_event ev; 777 cl_event ev;
675 778
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev)); 779 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677 780
678 clReleaseEvent (mapped->event); 781 clReleaseEvent (mapped->event);
699 eq_signal_arg = (void*)arg; 802 eq_signal_arg = (void*)arg;
700 803
701BOOT: 804BOOT:
702{ 805{
703 HV *stash = gv_stashpv ("OpenCL", 1); 806 HV *stash = gv_stashpv ("OpenCL", 1);
807
704 static const ivstr *civ, const_iv[] = { 808 static const ivstr *civ, const_iv[] = {
705 { sizeof (cl_char ), "SIZEOF_CHAR" }, 809 { sizeof (cl_char ), "SIZEOF_CHAR" },
706 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 810 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
707 { sizeof (cl_short ), "SIZEOF_SHORT" }, 811 { sizeof (cl_short ), "SIZEOF_SHORT" },
708 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 812 { sizeof (cl_ushort), "SIZEOF_USHORT" },
711 { sizeof (cl_long ), "SIZEOF_LONG" }, 815 { sizeof (cl_long ), "SIZEOF_LONG" },
712 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 816 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
713 { sizeof (cl_half ), "SIZEOF_HALF" }, 817 { sizeof (cl_half ), "SIZEOF_HALF" },
714 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 818 { sizeof (cl_float ), "SIZEOF_FLOAT" },
715 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 819 { sizeof (cl_double), "SIZEOF_DOUBLE" },
820 { PREFER_1_1 , "PREFER_1_1" },
716#include "constiv.h" 821#include "constiv.h"
717 }; 822 };
718 823
719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 824 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 825 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721 826
827 static const nvstr *cnv, const_nv[] = {
828#include "constnv.h"
829 };
830
831 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
832 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
833
834 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
835
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 836 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 837 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
838 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 839 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 840 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 841 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 842 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 843 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
769 list = tmpbuf (sizeof (*list) * count); 884 list = tmpbuf (sizeof (*list) * count);
770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 885 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
771 886
772 EXTEND (SP, count); 887 EXTEND (SP, count);
773 for (i = 0; i < count; ++i) 888 for (i = 0; i < count; ++i)
774 PUSH_CLOBJ (stash_platform, (IV)list [i]); 889 PUSH_CLOBJ (stash_platform, list [i]);
775 890
776void 891void
777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 892context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
778 PPCODE: 893 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK; 894 CONTEXT_NOTIFY_CALLBACK;
780 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 895 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
781 XPUSH_CLOBJ_CONTEXT; 896 XPUSH_CLOBJ_CONTEXT;
782 897
783void 898void
784context (FUTURE properties, FUTURE devices, FUTURE notify) 899context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
785 PPCODE: 900 PPCODE:
786 /* der Gipfel der Kunst */ 901 cl_uint device_count;
902 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
903
904 CONTEXT_NOTIFY_CALLBACK;
905 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
906 XPUSH_CLOBJ_CONTEXT;
787 907
788void 908void
789wait_for_events (...) 909wait_for_events (...)
790 CODE: 910 CODE:
791 EVENT_LIST (0); 911 EVENT_LIST (0);
842 EXTEND (SP, count); 962 EXTEND (SP, count);
843 for (i = 0; i < count; ++i) 963 for (i = 0; i < count; ++i)
844 PUSH_CLOBJ (stash_device, list [i]); 964 PUSH_CLOBJ (stash_device, list [i]);
845 965
846void 966void
847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 967context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
848 PPCODE: 968 PPCODE:
969 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
970 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
971
972 cl_uint device_count;
973 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
974
975 CONTEXT_NOTIFY_CALLBACK;
976 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
977 XPUSH_CLOBJ_CONTEXT;
978
979void
980context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
981 PPCODE:
982 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
983 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
984
985 CONTEXT_NOTIFY_CALLBACK;
986 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
987 XPUSH_CLOBJ_CONTEXT;
988
989MODULE = OpenCL PACKAGE = OpenCL::Device
990
991void
992info (OpenCL::Device self, cl_device_info name)
993 PPCODE:
994 INFO (Device)
995
996#if CL_VERSION_1_2
997
998void
999sub_devices (OpenCL::Device self, SV *properties)
1000 PPCODE:
849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1001 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
850 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1002 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
851 1003
852 AV *av = (AV *)SvRV (devices); 1004 properties = SvRV (properties);
853 cl_uint num_devices = av_len (av) + 1; 1005
854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1006 cl_uint count = av_len ((AV *)properties) + 1;
1007 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
855 1008
856 int i; 1009 int i;
857 for (i = num_devices; i--; ) 1010 for (i = 0; i < count; ++i)
858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device"); 1011 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
859 1012
860 CONTEXT_NOTIFY_CALLBACK; 1013 props [count] = 0;
861 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res));
862 XPUSH_CLOBJ_CONTEXT;
863 1014
864void 1015 cl_uint num_devices;
865context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 1016 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
866 PPCODE: 1017 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 1018 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
868 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
869 1019
870 CONTEXT_NOTIFY_CALLBACK; 1020 EXTEND (SP, num_devices);
871 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res)); 1021 for (i = 0; i < count; ++i)
872 XPUSH_CLOBJ_CONTEXT; 1022 PUSH_CLOBJ (stash_subdevice, list [i]);
873 1023
874MODULE = OpenCL PACKAGE = OpenCL::Device 1024#endif
875
876void
877info (OpenCL::Device self, cl_device_info name)
878 PPCODE:
879 INFO (Device)
880 1025
881#BEGIN:device 1026#BEGIN:device
882 1027
883void 1028void
884type (OpenCL::Device self) 1029type (OpenCL::Device self)
1100 for (i = 0; i < n; ++i) 1245 for (i = 0; i < n; ++i)
1101 PUSHs (sv_2mortal (newSVuv (value [i]))); 1246 PUSHs (sv_2mortal (newSVuv (value [i])));
1102 1247
1103#END:device 1248#END:device
1104 1249
1250MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1251
1252#if CL_VERSION_1_2
1253
1254void
1255DESTROY (OpenCL::SubDevice self)
1256 CODE:
1257 clReleaseDevice (self);
1258
1259#endif
1260
1105MODULE = OpenCL PACKAGE = OpenCL::Context 1261MODULE = OpenCL PACKAGE = OpenCL::Context
1106 1262
1107void 1263void
1108DESTROY (OpenCL::Context context) 1264DESTROY (OpenCL::Context self)
1109 CODE: 1265 CODE:
1110 clReleaseContext (context); 1266 clReleaseContext (self);
1111 1267
1112void 1268void
1113info (OpenCL::Context self, cl_context_info name) 1269info (OpenCL::Context self, cl_context_info name)
1114 PPCODE: 1270 PPCODE:
1115 INFO (Context) 1271 INFO (Context)
1156 const cl_image_desc desc = { 1312 const cl_image_desc desc = {
1157 type, 1313 type,
1158 width, height, depth, 1314 width, height, depth,
1159 array_size, row_pitch, slice_pitch, 1315 array_size, row_pitch, slice_pitch,
1160 num_mip_level, num_samples, 1316 num_mip_level, num_samples,
1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1317 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1162 }; 1318 };
1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1319 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1164 HV *stash = stash_image; 1320 HV *stash = stash_image;
1165 switch (type) 1321 switch (type)
1166 { 1322 {
1299 1455
1300 len2 = len; 1456 len2 = len;
1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1457 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1302 XPUSH_CLOBJ (stash_program, prog); 1458 XPUSH_CLOBJ (stash_program, prog);
1303 1459
1460void
1461program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1462 PPCODE:
1463 cl_uint device_count;
1464 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1465
1466 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1467 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1468
1469 binaries = SvRV (binaries);
1470
1471 if (device_count != av_len ((AV *)binaries) + 1)
1472 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1473
1474 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1475 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1476 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1477
1478 int i;
1479 for (i = 0; i < device_count; ++i)
1480 {
1481 STRLEN len;
1482 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1483 length_list [i] = len;
1484 }
1485
1486 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1487 length_list, binary_list,
1488 GIMME_V == G_ARRAY ? status_list : 0, &res));
1489
1490 EXTEND (SP, 2);
1491 PUSH_CLOBJ (stash_program, prog);
1492
1493 if (GIMME_V == G_ARRAY)
1494 {
1495 AV *av = newAV ();
1496 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1497
1498 for (i = device_count; i--; )
1499 av_store (av, i, newSViv (status_list [i]));
1500 }
1501
1502#if CL_VERSION_1_2
1503
1504void
1505program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1506 PPCODE:
1507 cl_uint device_count;
1508 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1509
1510 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514void
1515link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1516 CODE:
1517 cl_uint device_count = 0;
1518 cl_device_id *device_list = 0;
1519
1520 if (SvOK (devices))
1521 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1522
1523 cl_uint program_count;
1524 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1525
1526 void *user_data;
1527 program_callback pfn_notify = make_program_callback (notify, &user_data);
1528
1529 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1530 program_count, program_list, pfn_notify, user_data, &res));
1531
1532 XPUSH_CLOBJ (stash_program, prog);
1533
1534#endif
1535
1304#BEGIN:context 1536#BEGIN:context
1305 1537
1306void 1538void
1307reference_count (OpenCL::Context self) 1539reference_count (OpenCL::Context self)
1308 ALIAS: 1540 ALIAS:
1353void 1585void
1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1586read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS: 1587 ALIAS:
1356 enqueue_read_buffer = 0 1588 enqueue_read_buffer = 0
1357 PPCODE: 1589 PPCODE:
1358 cl_event ev = 0;
1359 EVENT_LIST (6); 1590 EVENT_LIST (6);
1360 1591
1361 SvUPGRADE (data, SVt_PV); 1592 SvUPGRADE (data, SVt_PV);
1362 SvGROW (data, len); 1593 SvGROW (data, len);
1363 SvPOK_only (data); 1594 SvPOK_only (data);
1364 SvCUR_set (data, len); 1595 SvCUR_set (data, len);
1596
1597 cl_event ev = 0;
1365 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1598 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1366 1599
1367 if (ev) 1600 if (ev)
1368 XPUSH_CLOBJ (stash_event, ev); 1601 XPUSH_CLOBJ (stash_event, ev);
1369 1602
1370void 1603void
1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1604write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS: 1605 ALIAS:
1373 enqueue_write_buffer = 0 1606 enqueue_write_buffer = 0
1374 PPCODE: 1607 PPCODE:
1375 cl_event ev = 0; 1608 EVENT_LIST (5);
1609
1376 STRLEN len; 1610 STRLEN len;
1377 char *ptr = SvPVbyte (data, len); 1611 char *ptr = SvPVbyte (data, len);
1378 EVENT_LIST (5);
1379 1612
1613 cl_event ev = 0;
1380 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1614 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1381 1615
1382 if (ev) 1616 if (ev)
1383 XPUSH_CLOBJ (stash_event, ev); 1617 XPUSH_CLOBJ (stash_event, ev);
1384 1618
1387void 1621void
1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1622fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS: 1623 ALIAS:
1390 enqueue_fill_buffer = 0 1624 enqueue_fill_buffer = 0
1391 PPCODE: 1625 PPCODE:
1392 cl_event ev = 0; 1626 EVENT_LIST (5);
1627
1393 STRLEN len; 1628 STRLEN len;
1394 char *ptr = SvPVbyte (data, len); 1629 char *ptr = SvPVbyte (data, len);
1395 EVENT_LIST (5);
1396 1630
1631 cl_event ev = 0;
1397 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1632 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1398 1633
1399 if (ev) 1634 if (ev)
1400 XPUSH_CLOBJ (stash_event, ev); 1635 XPUSH_CLOBJ (stash_event, ev);
1401 1636
1402void 1637void
1403fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...) 1638fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1404 ALIAS: 1639 ALIAS:
1405 enqueue_fill_image = 0 1640 enqueue_fill_image = 0
1406 PPCODE: 1641 PPCODE:
1407 cl_event ev = 0; 1642 EVENT_LIST (12);
1408 STRLEN len; 1643
1409 const size_t origin [3] = { x, y, z }; 1644 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth }; 1645 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412 1646
1413 const cl_float c_f [4] = { r, g, b, a }; 1647 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a }; 1648 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a }; 1649 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1650 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1653 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1654 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1421 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1655 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1422 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1656 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423 1657
1658 cl_event ev = 0;
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1659 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1425 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1660 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426 1661
1427 if (ev) 1662 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev); 1663 XPUSH_CLOBJ (stash_event, ev);
1432void 1667void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1668copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS: 1669 ALIAS:
1435 enqueue_copy_buffer = 0 1670 enqueue_copy_buffer = 0
1436 PPCODE: 1671 PPCODE:
1672 EVENT_LIST (6);
1673
1437 cl_event ev = 0; 1674 cl_event ev = 0;
1438 EVENT_LIST (6);
1439
1440 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1675 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1676
1442 if (ev) 1677 if (ev)
1443 XPUSH_CLOBJ (stash_event, ev); 1678 XPUSH_CLOBJ (stash_event, ev);
1444 1679
1445void 1680void
1446read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1681read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1447 ALIAS: 1682 ALIAS:
1448 enqueue_read_buffer_rect = 0 1683 enqueue_read_buffer_rect = 0
1449 PPCODE: 1684 PPCODE:
1450 cl_event ev = 0; 1685 EVENT_LIST (17);
1686
1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1687 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1452 const size_t host_origin[3] = { host_x, host_y, host_z }; 1688 const size_t host_origin[3] = { host_x, host_y, host_z };
1453 const size_t region[3] = { width, height, depth }; 1689 const size_t region[3] = { width, height, depth };
1454 EVENT_LIST (17);
1455 1690
1456 if (!buf_row_pitch) 1691 if (!buf_row_pitch)
1457 buf_row_pitch = region [0]; 1692 buf_row_pitch = region [0];
1458 1693
1459 if (!buf_slice_pitch) 1694 if (!buf_slice_pitch)
1469 1704
1470 SvUPGRADE (data, SVt_PV); 1705 SvUPGRADE (data, SVt_PV);
1471 SvGROW (data, len); 1706 SvGROW (data, len);
1472 SvPOK_only (data); 1707 SvPOK_only (data);
1473 SvCUR_set (data, len); 1708 SvCUR_set (data, len);
1709
1710 cl_event ev = 0;
1474 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1711 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1475 1712
1476 if (ev) 1713 if (ev)
1477 XPUSH_CLOBJ (stash_event, ev); 1714 XPUSH_CLOBJ (stash_event, ev);
1478 1715
1479void 1716void
1480write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1717write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1481 ALIAS: 1718 ALIAS:
1482 enqueue_write_buffer_rect = 0 1719 enqueue_write_buffer_rect = 0
1483 PPCODE: 1720 PPCODE:
1484 cl_event ev = 0; 1721 EVENT_LIST (17);
1722
1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1723 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1486 const size_t host_origin[3] = { host_x, host_y, host_z }; 1724 const size_t host_origin[3] = { host_x, host_y, host_z };
1487 const size_t region[3] = { width, height, depth }; 1725 const size_t region[3] = { width, height, depth };
1488 STRLEN len; 1726 STRLEN len;
1489 char *ptr = SvPVbyte (data, len); 1727 char *ptr = SvPVbyte (data, len);
1490 EVENT_LIST (17);
1491 1728
1492 if (!buf_row_pitch) 1729 if (!buf_row_pitch)
1493 buf_row_pitch = region [0]; 1730 buf_row_pitch = region [0];
1494 1731
1495 if (!buf_slice_pitch) 1732 if (!buf_slice_pitch)
1504 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1741 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1505 1742
1506 if (len < min_len) 1743 if (len < min_len)
1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1744 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1508 1745
1746 cl_event ev = 0;
1509 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1747 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1510 1748
1511 if (ev) 1749 if (ev)
1512 XPUSH_CLOBJ (stash_event, ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1513 1751
1514void 1752void
1515copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1753copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1516 ALIAS: 1754 ALIAS:
1517 enqueue_copy_buffer_rect = 0 1755 enqueue_copy_buffer_rect = 0
1518 PPCODE: 1756 PPCODE:
1519 cl_event ev = 0; 1757 EVENT_LIST (16);
1758
1520 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1521 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1760 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1522 const size_t region[3] = { width, height, depth }; 1761 const size_t region[3] = { width, height, depth };
1523 EVENT_LIST (16);
1524 1762
1763 cl_event ev = 0;
1525 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1764 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1526 1765
1527 if (ev) 1766 if (ev)
1528 XPUSH_CLOBJ (stash_event, ev); 1767 XPUSH_CLOBJ (stash_event, ev);
1529 1768
1530void 1769void
1531read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1770read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1532 ALIAS: 1771 ALIAS:
1533 enqueue_read_image = 0 1772 enqueue_read_image = 0
1534 PPCODE: 1773 PPCODE:
1535 cl_event ev = 0; 1774 EVENT_LIST (12);
1775
1536 const size_t src_origin[3] = { src_x, src_y, src_z }; 1776 const size_t src_origin[3] = { src_x, src_y, src_z };
1537 const size_t region[3] = { width, height, depth }; 1777 const size_t region[3] = { width, height, depth };
1538 EVENT_LIST (12);
1539 1778
1540 if (!row_pitch) 1779 if (!row_pitch)
1541 row_pitch = img_row_pitch (src); 1780 row_pitch = img_row_pitch (src);
1542 1781
1543 if (depth > 1 && !slice_pitch) 1782 if (depth > 1 && !slice_pitch)
1547 1786
1548 SvUPGRADE (data, SVt_PV); 1787 SvUPGRADE (data, SVt_PV);
1549 SvGROW (data, len); 1788 SvGROW (data, len);
1550 SvPOK_only (data); 1789 SvPOK_only (data);
1551 SvCUR_set (data, len); 1790 SvCUR_set (data, len);
1791
1792 cl_event ev = 0;
1552 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1793 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1553 1794
1554 if (ev) 1795 if (ev)
1555 XPUSH_CLOBJ (stash_event, ev); 1796 XPUSH_CLOBJ (stash_event, ev);
1556 1797
1557void 1798void
1558write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1799write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1559 ALIAS: 1800 ALIAS:
1560 enqueue_write_image = 0 1801 enqueue_write_image = 0
1561 PPCODE: 1802 PPCODE:
1562 cl_event ev = 0; 1803 EVENT_LIST (12);
1804
1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1805 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1564 const size_t region[3] = { width, height, depth }; 1806 const size_t region[3] = { width, height, depth };
1565 STRLEN len; 1807 STRLEN len;
1566 char *ptr = SvPVbyte (data, len); 1808 char *ptr = SvPVbyte (data, len);
1567 EVENT_LIST (12);
1568 1809
1569 if (!row_pitch) 1810 if (!row_pitch)
1570 row_pitch = img_row_pitch (dst); 1811 row_pitch = img_row_pitch (dst);
1571 1812
1572 if (depth > 1 && !slice_pitch) 1813 if (depth > 1 && !slice_pitch)
1575 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1816 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1576 1817
1577 if (len < min_len) 1818 if (len < min_len)
1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1819 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1579 1820
1821 cl_event ev = 0;
1580 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1822 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1581 1823
1582 if (ev) 1824 if (ev)
1583 XPUSH_CLOBJ (stash_event, ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1584 1826
1585void 1827void
1586copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1828copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1587 ALIAS: 1829 ALIAS:
1588 enqueue_copy_image = 0 1830 enqueue_copy_image = 0
1589 PPCODE: 1831 PPCODE:
1590 cl_event ev = 0; 1832 EVENT_LIST (12);
1833
1591 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1592 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1835 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1593 const size_t region[3] = { width, height, depth }; 1836 const size_t region[3] = { width, height, depth };
1594 EVENT_LIST (12);
1595 1837
1838 cl_event ev = 0;
1596 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1839 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1597 1840
1598 if (ev) 1841 if (ev)
1599 XPUSH_CLOBJ (stash_event, ev); 1842 XPUSH_CLOBJ (stash_event, ev);
1600 1843
1601void 1844void
1602copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1845copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1603 ALIAS: 1846 ALIAS:
1604 enqueue_copy_image_to_buffer = 0 1847 enqueue_copy_image_to_buffer = 0
1605 PPCODE: 1848 PPCODE:
1606 cl_event ev = 0; 1849 EVENT_LIST (10);
1850
1607 const size_t src_origin[3] = { src_x, src_y, src_z }; 1851 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth }; 1852 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610 1853
1854 cl_event ev = 0;
1611 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1855 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612 1856
1613 if (ev) 1857 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev); 1858 XPUSH_CLOBJ (stash_event, ev);
1615 1859
1616void 1860void
1617copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1861copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1618 ALIAS: 1862 ALIAS:
1619 enqueue_copy_buffer_to_image = 0 1863 enqueue_copy_buffer_to_image = 0
1620 PPCODE: 1864 PPCODE:
1621 cl_event ev = 0; 1865 EVENT_LIST (10);
1866
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1867 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth }; 1868 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625 1869
1870 cl_event ev = 0;
1626 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1871 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627 1872
1628 if (ev) 1873 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev); 1874 XPUSH_CLOBJ (stash_event, ev);
1630 1875
1631void 1876void
1632map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, size_t cb = 0, ...) 1877map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1633 ALIAS: 1878 ALIAS:
1634 enqueue_map_buffer = 0 1879 enqueue_map_buffer = 0
1635 PPCODE: 1880 PPCODE:
1881 EVENT_LIST (6);
1882
1883 size_t cb = SvIV (cb_);
1884
1885 if (!SvOK (cb_))
1886 {
1887 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1888 cb -= offset;
1889 }
1890
1636 cl_event ev; 1891 cl_event ev;
1637 EVENT_LIST (6);
1638
1639 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res)); 1892 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1640 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0)); 1893 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1641 1894
1642void 1895void
1643map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, size_t width = 0, size_t height = 0, size_t depth = 0, ...) 1896map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1644 ALIAS: 1897 ALIAS:
1645 enqueue_map_image = 0 1898 enqueue_map_image = 0
1646 PPCODE: 1899 PPCODE:
1647 cl_event ev; 1900 size_t width = SvIV (width_);
1901 if (!SvOK (width_))
1902 {
1903 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1904 width -= x;
1905 }
1906
1907 size_t height = SvIV (width_);
1908 if (!SvOK (height_))
1909 {
1910 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1911 height -= y;
1912 }
1913
1914 size_t depth = SvIV (width_);
1915 if (!SvOK (depth_))
1916 {
1917 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1918 depth -= z;
1919
1920 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1921 if (!depth)
1922 depth = 1;
1923 }
1924
1648 const size_t origin[3] = { x, y, z }; 1925 const size_t origin[3] = { x, y, z };
1649 const size_t region[3] = { width, height, depth }; 1926 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch; 1927 size_t row_pitch, slice_pitch;
1651 EVENT_LIST (10); 1928 EVENT_LIST (10);
1652 1929
1930 cl_event ev;
1653 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res)); 1931 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1654 1932
1655 size_t cb = slice_pitch ? slice_pitch * region [2] 1933 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1] 1934 : row_pitch ? row_pitch * region [1]
1657 : region [0]; 1935 : region [0];
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch)); 1937 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1660 1938
1661void 1939void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1940unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE: 1941 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2); 1942 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID) 1943 if (GIMME_V != G_VOID)
1666 { 1944 {
1667 clRetainEvent (mapped->event); 1945 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event); 1946 XPUSH_CLOBJ (stash_event, mapped->event);
1669 } 1947 }
1671void 1949void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1950task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS: 1951 ALIAS:
1674 enqueue_task = 0 1952 enqueue_task = 0
1675 PPCODE: 1953 PPCODE:
1954 EVENT_LIST (2);
1955
1676 cl_event ev = 0; 1956 cl_event ev = 0;
1677 EVENT_LIST (2);
1678
1679 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1957 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1680 1958
1681 if (ev) 1959 if (ev)
1682 XPUSH_CLOBJ (stash_event, ev); 1960 XPUSH_CLOBJ (stash_event, ev);
1683 1961
1684void 1962void
1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1963nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1686 ALIAS: 1964 ALIAS:
1687 enqueue_nd_range_kernel = 0 1965 enqueue_nd_range_kernel = 0
1688 PPCODE: 1966 PPCODE:
1689 cl_event ev = 0; 1967 EVENT_LIST (5);
1968
1690 size_t *gwo = 0, *gws, *lws = 0; 1969 size_t *gwo = 0, *gws, *lws = 0;
1691 int gws_len; 1970 int gws_len;
1692 size_t *lists; 1971 size_t *lists;
1693 int i; 1972 int i;
1694 EVENT_LIST (5);
1695 1973
1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1974 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1698 1976
1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1977 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1738 if (!lws [i]) 2016 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 2017 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 } 2018 }
1741 } 2019 }
1742 2020
2021 cl_event ev = 0;
1743 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2022 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1744 2023
1745 if (ev) 2024 if (ev)
1746 XPUSH_CLOBJ (stash_event, ev); 2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#if CL_VERSION_1_2
2028
2029void
2030migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2031 ALIAS:
2032 enqueue_migrate_mem_objects = 0
2033 PPCODE:
2034 EVENT_LIST (3);
2035
2036 cl_uint obj_count;
2037 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2038
2039 cl_event ev = 0;
2040 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2041
2042 if (ev)
2043 XPUSH_CLOBJ (stash_event, ev);
2044
2045#endif
1747 2046
1748#if cl_apple_gl_sharing || cl_khr_gl_sharing 2047#if cl_apple_gl_sharing || cl_khr_gl_sharing
1749 2048
1750void 2049void
1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2050acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1752 ALIAS: 2051 ALIAS:
1753 release_gl_objects = 1 2052 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0 2053 enqueue_acquire_gl_objects = 0
1755 enqueue_release_gl_objects = 1 2054 enqueue_release_gl_objects = 1
1756 PPCODE: 2055 PPCODE:
1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2056 EVENT_LIST (2);
1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2057
2058 cl_uint obj_count;
2059 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1759 2060
1760 cl_event ev = 0; 2061 cl_event ev = 0;
1761 EVENT_LIST (2);
1762 AV *av = (AV *)SvRV (objects);
1763 cl_uint num_objects = av_len (av) + 1;
1764 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1765 int i;
1766
1767 for (i = num_objects; i--; )
1768 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1769 2062
1770 if (ix) 2063 if (ix)
1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2064 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1772 else 2065 else
1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2066 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1774 2067
1775 if (ev) 2068 if (ev)
1776 XPUSH_CLOBJ (stash_event, ev); 2069 XPUSH_CLOBJ (stash_event, ev);
1777 2070
1778#endif 2071#endif
1792void 2085void
1793marker (OpenCL::Queue self, ...) 2086marker (OpenCL::Queue self, ...)
1794 ALIAS: 2087 ALIAS:
1795 enqueue_marker = 0 2088 enqueue_marker = 0
1796 PPCODE: 2089 PPCODE:
2090 EVENT_LIST (1);
1797 cl_event ev = 0; 2091 cl_event ev = 0;
1798 EVENT_LIST (1);
1799#if PREFER_1_1 2092#if PREFER_1_1
1800 if (!event_list_count) 2093 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2094 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else 2095 else
1803#if CL_VERSION_1_2 2096#if CL_VERSION_1_2
1817void 2110void
1818barrier (OpenCL::Queue self, ...) 2111barrier (OpenCL::Queue self, ...)
1819 ALIAS: 2112 ALIAS:
1820 enqueue_barrier = 0 2113 enqueue_barrier = 0
1821 PPCODE: 2114 PPCODE:
2115 EVENT_LIST (1);
1822 cl_event ev = 0; 2116 cl_event ev = 0;
1823 EVENT_LIST (1);
1824#if PREFER_1_1 2117#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID) 2118 if (!event_list_count && GIMME_V == G_VOID)
1826 NEED_SUCCESS (EnqueueBarrier, (self)); 2119 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else 2120 else
1828#if CL_VERSION_1_2 2121#if CL_VERSION_1_2
1912void 2205void
1913info (OpenCL::Memory self, cl_mem_info name) 2206info (OpenCL::Memory self, cl_mem_info name)
1914 PPCODE: 2207 PPCODE:
1915 INFO (MemObject) 2208 INFO (MemObject)
1916 2209
2210void
2211destructor_callback (OpenCL::Memory self, SV *notify)
2212 PPCODE:
2213
1917#BEGIN:mem 2214#BEGIN:mem
1918 2215
1919void 2216void
1920type (OpenCL::Memory self) 2217type (OpenCL::Memory self)
1921 PPCODE: 2218 PPCODE:
2156void 2453void
2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2454build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS: 2455 ALIAS:
2159 build_async = 1 2456 build_async = 1
2160 CODE: 2457 CODE:
2161 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2162 void *user_data = 0;
2163 cl_uint num_devices = 0; 2458 cl_uint device_count = 0;
2164 cl_device_id *device_list = 0; 2459 cl_device_id *device_list = 0;
2165 2460
2166 if (SvOK (devices)) 2461 if (SvOK (devices))
2167 { 2462 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2463
2464 void *user_data;
2465 program_callback pfn_notify = make_program_callback (notify, &user_data);
2466
2467 if (ix)
2468 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2469 else
2470 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2471
2472#if CL_VERSION_1_2
2473
2474void
2475compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2476 CODE:
2477 cl_uint device_count = 0;
2478 cl_device_id *device_list = 0;
2479
2480 if (SvOK (devices))
2481 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2482
2483 cl_uint header_count = 0;
2484 cl_program *header_list = 0;
2485 const char **header_name = 0;
2486
2487 if (SvOK (headers))
2488 {
2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2489 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2490 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2170 2491
2171 AV *av = (AV *)SvRV (devices); 2492 HV *hv = (HV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173 2493
2174 if (num_devices) 2494 header_count = hv_iterinit (hv);
2495 header_list = tmpbuf (sizeof (*header_list) * header_count);
2496 header_name = tmpbuf (sizeof (*header_name) * header_count);
2497
2498 HE *he;
2499 int i = 0;
2500 while (he = hv_iternext (hv))
2175 { 2501 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2502 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2503 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2177 int count; 2504 ++i;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 } 2505 }
2181 } 2506 }
2182 2507
2183 if (SvOK (notify)) 2508 void *user_data;
2184 { 2509 program_callback pfn_notify = make_program_callback (notify, &user_data);
2185 NEED_SUCCESS (RetainProgram, (self));
2186 pfn_notify = eq_program_notify;
2187 user_data = SvREFCNT_inc (s_get_cv (notify));
2188 }
2189 2510
2190 if (ix)
2191 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2192 else
2193 NEED_SUCCESS (BuildProgram, (self, num_devices, device_list, SvPVbyte_nolen (options), pfn_notify, user_data)); 2511 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2512 header_count, header_list, header_name, pfn_notify, user_data));
2513
2514#endif
2194 2515
2195void 2516void
2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2517build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2197 PPCODE: 2518 PPCODE:
2198 size_t size; 2519 size_t size;
2226 char *value = tmpbuf (size); 2547 char *value = tmpbuf (size);
2227 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2548 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2228 EXTEND (SP, 1); 2549 EXTEND (SP, 1);
2229 const int i = 0; 2550 const int i = 0;
2230 PUSHs (sv_2mortal (newSVpv (value, 0))); 2551 PUSHs (sv_2mortal (newSVpv (value, 0)));
2552
2553void
2554binary_type (OpenCL::Program self, OpenCL::Device device)
2555 PPCODE:
2556 cl_program_binary_type value [1];
2557 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2558 EXTEND (SP, 1);
2559 const int i = 0;
2560 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2231 2561
2232#END:program_build 2562#END:program_build
2233 2563
2234void 2564void
2235kernel (OpenCL::Program program, SV *function) 2565kernel (OpenCL::Program program, SV *function)
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2732 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2733 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404 2734
2405 case 'z': nullarg = 1; size = SvIV (sv); break; 2735 case 'z': nullarg = 1; size = SvIV (sv); break;
2406 2736
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2737 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2408 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2738 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2409 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2739 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410 2740
2411 default: 2741 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2742 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 } 2743 }
2414 2744
2485 CODE: 2815 CODE:
2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2816 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2487 2817
2488void 2818void
2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value) 2819set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
2493 CODE: 2820 CODE:
2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0); 2821 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2495 2822
2496void 2823void
2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2824set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2614 const int i = 0; 2941 const int i = 0;
2615 PUSHs (sv_2mortal (newSVuv (value [i]))); 2942 PUSHs (sv_2mortal (newSVuv (value [i])));
2616 2943
2617#END:kernel_work_group 2944#END:kernel_work_group
2618 2945
2946#if CL_VERSION_1_2
2947
2948void
2949arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2950 PPCODE:
2951 size_t size;
2952 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2953 SV *sv = sv_2mortal (newSV (size));
2954 SvUPGRADE (sv, SVt_PV);
2955 SvPOK_only (sv);
2956 SvCUR_set (sv, size);
2957 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2958 XPUSHs (sv);
2959
2960#BEGIN:kernel_arg
2961
2962void
2963arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2964 PPCODE:
2965 cl_kernel_arg_address_qualifier value [1];
2966 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2967 EXTEND (SP, 1);
2968 const int i = 0;
2969 PUSHs (sv_2mortal (newSVuv (value [i])));
2970
2971void
2972arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2973 PPCODE:
2974 cl_kernel_arg_access_qualifier value [1];
2975 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2976 EXTEND (SP, 1);
2977 const int i = 0;
2978 PUSHs (sv_2mortal (newSVuv (value [i])));
2979
2980void
2981arg_type_name (OpenCL::Kernel self, cl_uint idx)
2982 ALIAS:
2983 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2984 arg_name = CL_KERNEL_ARG_NAME
2985 PPCODE:
2986 size_t size;
2987 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2988 char *value = tmpbuf (size);
2989 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2990 EXTEND (SP, 1);
2991 const int i = 0;
2992 PUSHs (sv_2mortal (newSVpv (value, 0)));
2993
2994void
2995arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2996 PPCODE:
2997 cl_kernel_arg_type_qualifier value [1];
2998 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2999 EXTEND (SP, 1);
3000 const int i = 0;
3001 PUSHs (sv_2mortal (newSVuv (value [i])));
3002
3003#END:kernel_arg
3004
3005#endif
3006
2619MODULE = OpenCL PACKAGE = OpenCL::Event 3007MODULE = OpenCL PACKAGE = OpenCL::Event
2620 3008
2621void 3009void
2622DESTROY (OpenCL::Event self) 3010DESTROY (OpenCL::Event self)
2623 CODE: 3011 CODE:
2732 Safefree (mapped); 3120 Safefree (mapped);
2733 3121
2734void 3122void
2735unmap (OpenCL::Mapped self, ...) 3123unmap (OpenCL::Mapped self, ...)
2736 CODE: 3124 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1); 3125 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2738 3126
2739bool 3127bool
2740mapped (OpenCL::Mapped self) 3128mapped (OpenCL::Mapped self)
2741 CODE: 3129 CODE:
2742 RETVAL = !!self->ptr; 3130 RETVAL = !!self->ptr;
2769ptr (OpenCL::Mapped self) 3157ptr (OpenCL::Mapped self)
2770 CODE: 3158 CODE:
2771 RETVAL = PTR2IV (self->ptr); 3159 RETVAL = PTR2IV (self->ptr);
2772 OUTPUT: 3160 OUTPUT:
2773 RETVAL 3161 RETVAL
3162
3163void
3164set (OpenCL::Mapped self, size_t offset, SV *data)
3165 CODE:
3166 STRLEN len;
3167 const char *ptr = SvPVbyte (data, len);
3168
3169 if (offset + len > self->cb)
3170 croak ("OpenCL::Mapped::set out of bound condition detected");
3171
3172 memcpy (offset + (char *)self->ptr, ptr, len);
2774 3173
2775MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer 3174MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2776 3175
2777MODULE = OpenCL PACKAGE = OpenCL::MappedImage 3176MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2778 3177

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines