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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.63 by root, Tue May 1 22:25:13 2012 UTC vs.
Revision 1.69 by root, Sat May 5 02:30:27 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
38 #define PREFER_1_1 1 40 #define PREFER_1_1 1
39#endif 41#endif
40 42
41typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
42typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
43typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
44typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
45typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
46typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
47typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
48typedef cl_mem OpenCL__Image; 51typedef cl_mem OpenCL__Image;
49typedef cl_mem OpenCL__Image2D;
50typedef cl_mem OpenCL__Image3D;
51typedef cl_mem OpenCL__Memory_ornull; 52typedef cl_mem OpenCL__Memory_ornull;
52typedef cl_mem OpenCL__Buffer_ornull; 53typedef cl_mem OpenCL__Buffer_ornull;
53typedef cl_mem OpenCL__Image_ornull; 54typedef cl_mem OpenCL__Image_ornull;
54typedef cl_mem OpenCL__Image2D_ornull;
55typedef cl_mem OpenCL__Image3D_ornull;
56typedef cl_sampler OpenCL__Sampler; 55typedef cl_sampler OpenCL__Sampler;
57typedef cl_program OpenCL__Program; 56typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 57typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
64typedef SV *FUTURE; 63typedef SV *FUTURE;
65 64
66static HV 65static HV
67 *stash_platform, 66 *stash_platform,
68 *stash_device, 67 *stash_device,
68 *stash_subdevice,
69 *stash_context, 69 *stash_context,
70 *stash_queue, 70 *stash_queue,
71 *stash_program, 71 *stash_program,
72 *stash_kernel, 72 *stash_kernel,
73 *stash_sampler, 73 *stash_sampler,
133} 133}
134 134
135/*****************************************************************************/ 135/*****************************************************************************/
136 136
137/* up to two temporary buffers */ 137/* up to two temporary buffers */
138static void * 138static void * ecb_noinline
139tmpbuf (size_t size) 139tmpbuf (size_t size)
140{ 140{
141 enum { buffers = 3 }; 141 enum { buffers = 4 };
142 static int idx; 142 static int idx;
143 static void *buf [buffers]; 143 static void *buf [buffers];
144 static size_t len [buffers]; 144 static size_t len [buffers];
145 145
146 idx = (idx + 1) % buffers; 146 idx = (idx + 1) % buffers;
151 len [idx] = ((size + 31) & ~4095) + 4096 - 32; 151 len [idx] = ((size + 31) & ~4095) + 4096 - 32;
152 buf [idx] = malloc (len [idx]); 152 buf [idx] = malloc (len [idx]);
153 } 153 }
154 154
155 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;
156} 171}
157 172
158/*****************************************************************************/ 173/*****************************************************************************/
159 174
160typedef struct 175typedef struct
219 if (res) \ 234 if (res) \
220 FAIL (name); 235 FAIL (name);
221 236
222/*****************************************************************************/ 237/*****************************************************************************/
223 238
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
251SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
252{
253 // sv_derived_from is quite slow :(
254 if (SvROK (sv) && sv_derived_from (sv, pkg))
255 return (void *)SvIV (SvRV (sv));
256
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));
268}
269
270/*****************************************************************************/
271
224static cl_context_properties * 272static cl_context_properties * ecb_noinline
225SvCONTEXTPROPERTIES (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)
226{ 274{
227 if (!sv || !SvOK (sv)) 275 if (!sv || !SvOK (sv))
228 if (extra) 276 if (extra)
229 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
236 int i, len = av_len (av) + 1; 284 int i, len = av_len (av) + 1;
237 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));
238 cl_context_properties *l = p; 286 cl_context_properties *l = p;
239 287
240 if (len & 1) 288 if (len & 1)
241 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);
242 290
243 while (extracount--) 291 while (extracount--)
244 *l++ = *extra++; 292 *l++ = *extra++;
245 293
246 for (i = 0; i < len; i += 2) 294 for (i = 0; i < len; i += 2)
249 SV *p_sv = *av_fetch (av, i + 1, 0); 297 SV *p_sv = *av_fetch (av, i + 1, 0);
250 cl_context_properties v = SvIV (p_sv); // code below can override 298 cl_context_properties v = SvIV (p_sv); // code below can override
251 299
252 switch (t) 300 switch (t)
253 { 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
254 case CL_GLX_DISPLAY_KHR: 307 case CL_GLX_DISPLAY_KHR:
255 if (!SvOK (p_sv)) 308 if (!SvOK (p_sv))
256 { 309 {
257 void *func = glsym ("_glXGetCurrentDisplay"); 310 void *func = glsym ("_glXGetCurrentDisplay");
258 if (func) 311 if (func)
284 } 337 }
285 338
286 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);
287} 340}
288 341
289/*****************************************************************************/ 342static void * ecb_noinline
290 343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
291static SV *
292new_clobj (HV *stash, IV id)
293{ 344{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash)); 345 const char *funcname = cv_get_name (cv);
295}
296 346
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id))) 347 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id))) 348 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
349 funcname, argname, or_undef ? "undef or " : "", klass);
299 350
300/* cl objects are either \$iv, or [$iv, ...] */ 351 AV *av = (AV *)SvRV (arg);
301/* they can be upgraded at runtime to the array form */ 352 void **list = 0;
302static void * 353 cl_uint count = av_len (av) + 1;
303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
304{
305 if (SvROK (sv) && sv_derived_from (sv, pkg))
306 return (void *)SvIV (SvRV (sv));
307 354
308 croak ("%s: %s is not of type %s", func, svname, pkg); 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;
309} 369}
310 370
311/*****************************************************************************/ 371/*****************************************************************************/
312/* callback stuff */ 372/* callback stuff */
313 373
335static void (*eq_signal_func)(void *signal_arg, int value); 395static void (*eq_signal_func)(void *signal_arg, int value);
336static void *eq_signal_arg; 396static void *eq_signal_arg;
337static xmutex_t eq_lock = X_MUTEX_INIT; 397static xmutex_t eq_lock = X_MUTEX_INIT;
338static eq_item *eq_head, *eq_tail; 398static eq_item *eq_head, *eq_tail;
339 399
340static void 400static void ecb_noinline
341eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3) 401eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
342{ 402{
343 eq_item *item = malloc (sizeof (eq_item)); 403 eq_item *item = malloc (sizeof (eq_item));
344 404
345 item->next = 0; 405 item->next = 0;
412} 472}
413 473
414/*****************************************************************************/ 474/*****************************************************************************/
415/* context notify */ 475/* context notify */
416 476
417static void 477static void ecb_noinline
418eq_context_push (void *data1, void *data2, void *data3) 478eq_context_push (void *data1, void *data2, void *data3)
419{ 479{
420 dSP; 480 dSP;
421 PUSHs (sv_2mortal (newSVpv (data1, 0))); 481 PUSHs (sv_2mortal (newSVpv (data1, 0)));
422 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3))); 482 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
444 { \ 504 { \
445 pfn_notify = eq_context_notify; \ 505 pfn_notify = eq_context_notify; \
446 user_data = s_get_cv (notify); \ 506 user_data = s_get_cv (notify); \
447 } 507 }
448 508
449static SV * 509static SV * ecb_noinline
450new_clobj_context (cl_context ctx, void *user_data) 510new_clobj_context (cl_context ctx, void *user_data)
451{ 511{
452 SV *sv = new_clobj (stash_context, (IV)ctx); 512 SV *sv = new_clobj (stash_context, (IV)ctx);
453 513
454 if (user_data) 514 if (user_data)
473static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 533static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
474 534
475static void CL_CALLBACK 535static void CL_CALLBACK
476eq_program_notify (cl_program program, void *user_data) 536eq_program_notify (cl_program program, void *user_data)
477{ 537{
538 clRetainProgram (program);
539
478 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 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 }
479} 558}
480 559
481struct build_args 560struct build_args
482{ 561{
483 cl_program program; 562 cl_program program;
497 else 576 else
498 clReleaseProgram (arg->program); 577 clReleaseProgram (arg->program);
499 578
500 free (arg->options); 579 free (arg->options);
501 free (arg); 580 free (arg);
581
582 return 0;
502} 583}
503 584
504static void 585static void
505build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data) 586build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
506{ 587{
546 size_t res; 627 size_t res;
547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 628 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
548 return res; 629 return res;
549} 630}
550 631
551static cl_event * 632static cl_event * ecb_noinline
552event_list (SV **items, cl_uint *rcount, cl_event extra) 633event_list (SV **items, cl_uint *rcount, cl_event extra)
553{ 634{
554 cl_uint count = *rcount; 635 cl_uint count = *rcount;
555 636
556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed 637 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
557 count = 0; 638 *rcount = count = 0;
558 639
559 if (!count && !extra) 640 if (!count && !extra)
560 return 0; 641 return 0;
561 642
562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 643 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
719 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--)
720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 801 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
721 802
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 803 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 804 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
805 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 806 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 807 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 808 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 809 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 810 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
769 list = tmpbuf (sizeof (*list) * count); 851 list = tmpbuf (sizeof (*list) * count);
770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 852 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
771 853
772 EXTEND (SP, count); 854 EXTEND (SP, count);
773 for (i = 0; i < count; ++i) 855 for (i = 0; i < count; ++i)
774 PUSH_CLOBJ (stash_platform, (IV)list [i]); 856 PUSH_CLOBJ (stash_platform, list [i]);
775 857
776void 858void
777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 859context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
778 PPCODE: 860 PPCODE:
779 CONTEXT_NOTIFY_CALLBACK; 861 CONTEXT_NOTIFY_CALLBACK;
780 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));
781 XPUSH_CLOBJ_CONTEXT; 863 XPUSH_CLOBJ_CONTEXT;
782 864
783void 865void
784context (FUTURE properties, FUTURE devices, FUTURE notify) 866context (FUTURE properties, FUTURE devices, FUTURE notify)
785 PPCODE: 867 PPCODE:
842 EXTEND (SP, count); 924 EXTEND (SP, count);
843 for (i = 0; i < count; ++i) 925 for (i = 0; i < count; ++i)
844 PUSH_CLOBJ (stash_device, list [i]); 926 PUSH_CLOBJ (stash_device, list [i]);
845 927
846void 928void
847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 929context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef)
848 PPCODE: 930 PPCODE:
849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 931 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
850 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);
851 933
852 AV *av = (AV *)SvRV (devices); 934 cl_uint device_count = 0;
853 cl_uint num_devices = av_len (av) + 1; 935 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
854 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
855
856 int i;
857 for (i = num_devices; i--; )
858 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
859 936
860 CONTEXT_NOTIFY_CALLBACK; 937 CONTEXT_NOTIFY_CALLBACK;
861 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, pfn_notify, user_data, &res)); 938 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
862 XPUSH_CLOBJ_CONTEXT; 939 XPUSH_CLOBJ_CONTEXT;
863 940
864void 941void
865context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 942context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
866 PPCODE: 943 PPCODE:
867 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 944 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
868 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);
869 946
870 CONTEXT_NOTIFY_CALLBACK; 947 CONTEXT_NOTIFY_CALLBACK;
871 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));
872 XPUSH_CLOBJ_CONTEXT; 949 XPUSH_CLOBJ_CONTEXT;
873 950
874MODULE = OpenCL PACKAGE = OpenCL::Device 951MODULE = OpenCL PACKAGE = OpenCL::Device
875 952
876void 953void
877info (OpenCL::Device self, cl_device_info name) 954info (OpenCL::Device self, cl_device_info name)
878 PPCODE: 955 PPCODE:
879 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
880 987
881#BEGIN:device 988#BEGIN:device
882 989
883void 990void
884type (OpenCL::Device self) 991type (OpenCL::Device self)
1100 for (i = 0; i < n; ++i) 1207 for (i = 0; i < n; ++i)
1101 PUSHs (sv_2mortal (newSVuv (value [i]))); 1208 PUSHs (sv_2mortal (newSVuv (value [i])));
1102 1209
1103#END:device 1210#END:device
1104 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
1105MODULE = OpenCL PACKAGE = OpenCL::Context 1223MODULE = OpenCL PACKAGE = OpenCL::Context
1106 1224
1107void 1225void
1108DESTROY (OpenCL::Context context) 1226DESTROY (OpenCL::Context self)
1109 CODE: 1227 CODE:
1110 clReleaseContext (context); 1228 clReleaseContext (self);
1111 1229
1112void 1230void
1113info (OpenCL::Context self, cl_context_info name) 1231info (OpenCL::Context self, cl_context_info name)
1114 PPCODE: 1232 PPCODE:
1115 INFO (Context) 1233 INFO (Context)
1299 1417
1300 len2 = len; 1418 len2 = len;
1301 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));
1302 XPUSH_CLOBJ (stash_program, prog); 1420 XPUSH_CLOBJ (stash_program, prog);
1303 1421
1422void
1423program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1424 PPCODE:
1425 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
1426 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects");
1427
1428 devices = SvRV (devices);
1429
1430 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1431 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1432
1433 binaries = SvRV (binaries);
1434
1435 if (av_len ((AV *)devices) != av_len ((AV *)binaries))
1436 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1437
1438 int count = av_len ((AV *)devices) + 1;
1439 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1440 size_t *length_list = tmpbuf (sizeof (*length_list) * count);
1441 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count);
1442 cl_int *status_list = tmpbuf (sizeof (*status_list) * count);
1443
1444 int i;
1445 for (i = 0; i < count; ++i)
1446 {
1447 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1448 STRLEN len;
1449 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1450 length_list [i] = len;
1451 }
1452
1453 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res));
1454
1455 EXTEND (SP, 2);
1456 PUSH_CLOBJ (stash_program, prog);
1457
1458 if (GIMME_V == G_ARRAY)
1459 {
1460 AV *av = newAV ();
1461 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1462
1463 for (i = count; i--; )
1464 av_store (av, i, newSViv (status_list [i]));
1465 }
1466
1467#if CL_VERSION_1_2
1468
1469void
1470program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1471 PPCODE:
1472 cl_uint device_count = 0;
1473 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1474
1475 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1476
1477 XPUSH_CLOBJ (stash_program, prog);
1478
1479void
1480link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1481 CODE:
1482 cl_uint device_count = 0;
1483 cl_device_id *device_list = 0;
1484
1485 if (SvOK (devices))
1486 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1487
1488 cl_uint program_count;
1489 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1490
1491 void *user_data;
1492 program_callback pfn_notify = make_program_callback (notify, &user_data);
1493
1494 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1495 program_count, program_list, pfn_notify, user_data, &res));
1496
1497 XPUSH_CLOBJ (stash_program, prog);
1498
1499#endif
1500
1304#BEGIN:context 1501#BEGIN:context
1305 1502
1306void 1503void
1307reference_count (OpenCL::Context self) 1504reference_count (OpenCL::Context self)
1308 ALIAS: 1505 ALIAS:
1353void 1550void
1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1551read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS: 1552 ALIAS:
1356 enqueue_read_buffer = 0 1553 enqueue_read_buffer = 0
1357 PPCODE: 1554 PPCODE:
1358 cl_event ev = 0;
1359 EVENT_LIST (6); 1555 EVENT_LIST (6);
1360 1556
1361 SvUPGRADE (data, SVt_PV); 1557 SvUPGRADE (data, SVt_PV);
1362 SvGROW (data, len); 1558 SvGROW (data, len);
1363 SvPOK_only (data); 1559 SvPOK_only (data);
1364 SvCUR_set (data, len); 1560 SvCUR_set (data, len);
1561
1562 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)); 1563 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1366 1564
1367 if (ev) 1565 if (ev)
1368 XPUSH_CLOBJ (stash_event, ev); 1566 XPUSH_CLOBJ (stash_event, ev);
1369 1567
1370void 1568void
1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1569write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS: 1570 ALIAS:
1373 enqueue_write_buffer = 0 1571 enqueue_write_buffer = 0
1374 PPCODE: 1572 PPCODE:
1375 cl_event ev = 0; 1573 EVENT_LIST (5);
1574
1376 STRLEN len; 1575 STRLEN len;
1377 char *ptr = SvPVbyte (data, len); 1576 char *ptr = SvPVbyte (data, len);
1378 EVENT_LIST (5);
1379 1577
1578 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)); 1579 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1381 1580
1382 if (ev) 1581 if (ev)
1383 XPUSH_CLOBJ (stash_event, ev); 1582 XPUSH_CLOBJ (stash_event, ev);
1384 1583
1387void 1586void
1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1587fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS: 1588 ALIAS:
1390 enqueue_fill_buffer = 0 1589 enqueue_fill_buffer = 0
1391 PPCODE: 1590 PPCODE:
1392 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1393 STRLEN len; 1593 STRLEN len;
1394 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1395 EVENT_LIST (5);
1396 1595
1596 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)); 1597 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1398 1598
1399 if (ev) 1599 if (ev)
1400 XPUSH_CLOBJ (stash_event, ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1401 1601
1402void 1602void
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, ...) 1603fill_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: 1604 ALIAS:
1405 enqueue_fill_image = 0 1605 enqueue_fill_image = 0
1406 PPCODE: 1606 PPCODE:
1407 cl_event ev = 0; 1607 EVENT_LIST (12);
1408 STRLEN len; 1608
1409 const size_t origin [3] = { x, y, z }; 1609 const size_t origin [3] = { x, y, z };
1410 const size_t region [3] = { width, height, depth }; 1610 const size_t region [3] = { width, height, depth };
1411 EVENT_LIST (12);
1412 1611
1413 const cl_float c_f [4] = { r, g, b, a }; 1612 const cl_float c_f [4] = { r, g, b, a };
1414 const cl_uint c_u [4] = { r, g, b, a }; 1613 const cl_uint c_u [4] = { r, g, b, a };
1415 const cl_int c_s [4] = { r, g, b, a }; 1614 const cl_int c_s [4] = { r, g, b, a };
1416 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1615 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1618 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1619 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) 1620 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."); 1621 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1423 1622
1623 cl_event ev = 0;
1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1624 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)); 1625 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1426 1626
1427 if (ev) 1627 if (ev)
1428 XPUSH_CLOBJ (stash_event, ev); 1628 XPUSH_CLOBJ (stash_event, ev);
1432void 1632void
1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1633copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS: 1634 ALIAS:
1435 enqueue_copy_buffer = 0 1635 enqueue_copy_buffer = 0
1436 PPCODE: 1636 PPCODE:
1637 EVENT_LIST (6);
1638
1437 cl_event ev = 0; 1639 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)); 1640 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1441 1641
1442 if (ev) 1642 if (ev)
1443 XPUSH_CLOBJ (stash_event, ev); 1643 XPUSH_CLOBJ (stash_event, ev);
1444 1644
1445void 1645void
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, ...) 1646read_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: 1647 ALIAS:
1448 enqueue_read_buffer_rect = 0 1648 enqueue_read_buffer_rect = 0
1449 PPCODE: 1649 PPCODE:
1450 cl_event ev = 0; 1650 EVENT_LIST (17);
1651
1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1652 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 }; 1653 const size_t host_origin[3] = { host_x, host_y, host_z };
1453 const size_t region[3] = { width, height, depth }; 1654 const size_t region[3] = { width, height, depth };
1454 EVENT_LIST (17);
1455 1655
1456 if (!buf_row_pitch) 1656 if (!buf_row_pitch)
1457 buf_row_pitch = region [0]; 1657 buf_row_pitch = region [0];
1458 1658
1459 if (!buf_slice_pitch) 1659 if (!buf_slice_pitch)
1469 1669
1470 SvUPGRADE (data, SVt_PV); 1670 SvUPGRADE (data, SVt_PV);
1471 SvGROW (data, len); 1671 SvGROW (data, len);
1472 SvPOK_only (data); 1672 SvPOK_only (data);
1473 SvCUR_set (data, len); 1673 SvCUR_set (data, len);
1674
1675 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)); 1676 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 1677
1476 if (ev) 1678 if (ev)
1477 XPUSH_CLOBJ (stash_event, ev); 1679 XPUSH_CLOBJ (stash_event, ev);
1478 1680
1479void 1681void
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, ...) 1682write_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: 1683 ALIAS:
1482 enqueue_write_buffer_rect = 0 1684 enqueue_write_buffer_rect = 0
1483 PPCODE: 1685 PPCODE:
1484 cl_event ev = 0; 1686 EVENT_LIST (17);
1687
1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1688 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 }; 1689 const size_t host_origin[3] = { host_x, host_y, host_z };
1487 const size_t region[3] = { width, height, depth }; 1690 const size_t region[3] = { width, height, depth };
1488 STRLEN len; 1691 STRLEN len;
1489 char *ptr = SvPVbyte (data, len); 1692 char *ptr = SvPVbyte (data, len);
1490 EVENT_LIST (17);
1491 1693
1492 if (!buf_row_pitch) 1694 if (!buf_row_pitch)
1493 buf_row_pitch = region [0]; 1695 buf_row_pitch = region [0];
1494 1696
1495 if (!buf_slice_pitch) 1697 if (!buf_slice_pitch)
1504 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1706 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1505 1707
1506 if (len < min_len) 1708 if (len < min_len)
1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1709 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1508 1710
1711 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)); 1712 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 1713
1511 if (ev) 1714 if (ev)
1512 XPUSH_CLOBJ (stash_event, ev); 1715 XPUSH_CLOBJ (stash_event, ev);
1513 1716
1514void 1717void
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, ...) 1718copy_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: 1719 ALIAS:
1517 enqueue_copy_buffer_rect = 0 1720 enqueue_copy_buffer_rect = 0
1518 PPCODE: 1721 PPCODE:
1519 cl_event ev = 0; 1722 EVENT_LIST (16);
1723
1520 const size_t src_origin[3] = { src_x, src_y, src_z }; 1724 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 }; 1725 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1522 const size_t region[3] = { width, height, depth }; 1726 const size_t region[3] = { width, height, depth };
1523 EVENT_LIST (16);
1524 1727
1728 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)); 1729 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 1730
1527 if (ev) 1731 if (ev)
1528 XPUSH_CLOBJ (stash_event, ev); 1732 XPUSH_CLOBJ (stash_event, ev);
1529 1733
1530void 1734void
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, ...) 1735read_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: 1736 ALIAS:
1533 enqueue_read_image = 0 1737 enqueue_read_image = 0
1534 PPCODE: 1738 PPCODE:
1535 cl_event ev = 0; 1739 EVENT_LIST (12);
1740
1536 const size_t src_origin[3] = { src_x, src_y, src_z }; 1741 const size_t src_origin[3] = { src_x, src_y, src_z };
1537 const size_t region[3] = { width, height, depth }; 1742 const size_t region[3] = { width, height, depth };
1538 EVENT_LIST (12);
1539 1743
1540 if (!row_pitch) 1744 if (!row_pitch)
1541 row_pitch = img_row_pitch (src); 1745 row_pitch = img_row_pitch (src);
1542 1746
1543 if (depth > 1 && !slice_pitch) 1747 if (depth > 1 && !slice_pitch)
1547 1751
1548 SvUPGRADE (data, SVt_PV); 1752 SvUPGRADE (data, SVt_PV);
1549 SvGROW (data, len); 1753 SvGROW (data, len);
1550 SvPOK_only (data); 1754 SvPOK_only (data);
1551 SvCUR_set (data, len); 1755 SvCUR_set (data, len);
1756
1757 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)); 1758 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 1759
1554 if (ev) 1760 if (ev)
1555 XPUSH_CLOBJ (stash_event, ev); 1761 XPUSH_CLOBJ (stash_event, ev);
1556 1762
1557void 1763void
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, ...) 1764write_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: 1765 ALIAS:
1560 enqueue_write_image = 0 1766 enqueue_write_image = 0
1561 PPCODE: 1767 PPCODE:
1562 cl_event ev = 0; 1768 EVENT_LIST (12);
1769
1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1770 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1564 const size_t region[3] = { width, height, depth }; 1771 const size_t region[3] = { width, height, depth };
1565 STRLEN len; 1772 STRLEN len;
1566 char *ptr = SvPVbyte (data, len); 1773 char *ptr = SvPVbyte (data, len);
1567 EVENT_LIST (12);
1568 1774
1569 if (!row_pitch) 1775 if (!row_pitch)
1570 row_pitch = img_row_pitch (dst); 1776 row_pitch = img_row_pitch (dst);
1571 1777
1572 if (depth > 1 && !slice_pitch) 1778 if (depth > 1 && !slice_pitch)
1575 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1781 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1576 1782
1577 if (len < min_len) 1783 if (len < min_len)
1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1784 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1579 1785
1786 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)); 1787 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 1788
1582 if (ev) 1789 if (ev)
1583 XPUSH_CLOBJ (stash_event, ev); 1790 XPUSH_CLOBJ (stash_event, ev);
1584 1791
1585void 1792void
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, ...) 1793copy_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: 1794 ALIAS:
1588 enqueue_copy_image = 0 1795 enqueue_copy_image = 0
1589 PPCODE: 1796 PPCODE:
1590 cl_event ev = 0; 1797 EVENT_LIST (12);
1798
1591 const size_t src_origin[3] = { src_x, src_y, src_z }; 1799 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 }; 1800 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1593 const size_t region[3] = { width, height, depth }; 1801 const size_t region[3] = { width, height, depth };
1594 EVENT_LIST (12);
1595 1802
1803 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)); 1804 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1597 1805
1598 if (ev) 1806 if (ev)
1599 XPUSH_CLOBJ (stash_event, ev); 1807 XPUSH_CLOBJ (stash_event, ev);
1600 1808
1601void 1809void
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, ...) 1810copy_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: 1811 ALIAS:
1604 enqueue_copy_image_to_buffer = 0 1812 enqueue_copy_image_to_buffer = 0
1605 PPCODE: 1813 PPCODE:
1606 cl_event ev = 0; 1814 EVENT_LIST (10);
1815
1607 const size_t src_origin[3] = { src_x, src_y, src_z }; 1816 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth }; 1817 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610 1818
1819 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)); 1820 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612 1821
1613 if (ev) 1822 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev); 1823 XPUSH_CLOBJ (stash_event, ev);
1615 1824
1616void 1825void
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, ...) 1826copy_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: 1827 ALIAS:
1619 enqueue_copy_buffer_to_image = 0 1828 enqueue_copy_buffer_to_image = 0
1620 PPCODE: 1829 PPCODE:
1621 cl_event ev = 0; 1830 EVENT_LIST (10);
1831
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1832 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth }; 1833 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625 1834
1835 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)); 1836 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627 1837
1628 if (ev) 1838 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev); 1839 XPUSH_CLOBJ (stash_event, ev);
1630 1840
1631void 1841void
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, ...) 1842map_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: 1843 ALIAS:
1634 enqueue_map_buffer = 0 1844 enqueue_map_buffer = 0
1635 PPCODE: 1845 PPCODE:
1846 EVENT_LIST (6);
1847
1848 size_t cb = SvIV (cb_);
1849
1850 if (!SvOK (cb_))
1851 {
1852 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1853 cb -= offset;
1854 }
1855
1636 cl_event ev; 1856 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)); 1857 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)); 1858 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1641 1859
1642void 1860void
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, ...) 1861map_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: 1862 ALIAS:
1645 enqueue_map_image = 0 1863 enqueue_map_image = 0
1646 PPCODE: 1864 PPCODE:
1647 cl_event ev; 1865 size_t width = SvIV (width_);
1866 if (!SvOK (width_))
1867 {
1868 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1869 width -= x;
1870 }
1871
1872 size_t height = SvIV (width_);
1873 if (!SvOK (height_))
1874 {
1875 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1876 height -= y;
1877 }
1878
1879 size_t depth = SvIV (width_);
1880 if (!SvOK (depth_))
1881 {
1882 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1883 depth -= z;
1884
1885 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1886 if (!depth)
1887 depth = 1;
1888 }
1889
1648 const size_t origin[3] = { x, y, z }; 1890 const size_t origin[3] = { x, y, z };
1649 const size_t region[3] = { width, height, depth }; 1891 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch; 1892 size_t row_pitch, slice_pitch;
1651 EVENT_LIST (10); 1893 EVENT_LIST (10);
1652 1894
1895 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)); 1896 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 1897
1655 size_t cb = slice_pitch ? slice_pitch * region [2] 1898 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1] 1899 : row_pitch ? row_pitch * region [1]
1657 : region [0]; 1900 : region [0];
1671void 1914void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1915task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS: 1916 ALIAS:
1674 enqueue_task = 0 1917 enqueue_task = 0
1675 PPCODE: 1918 PPCODE:
1919 EVENT_LIST (2);
1920
1676 cl_event ev = 0; 1921 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)); 1922 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1680 1923
1681 if (ev) 1924 if (ev)
1682 XPUSH_CLOBJ (stash_event, ev); 1925 XPUSH_CLOBJ (stash_event, ev);
1683 1926
1684void 1927void
1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1928nd_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: 1929 ALIAS:
1687 enqueue_nd_range_kernel = 0 1930 enqueue_nd_range_kernel = 0
1688 PPCODE: 1931 PPCODE:
1689 cl_event ev = 0; 1932 EVENT_LIST (5);
1933
1690 size_t *gwo = 0, *gws, *lws = 0; 1934 size_t *gwo = 0, *gws, *lws = 0;
1691 int gws_len; 1935 int gws_len;
1692 size_t *lists; 1936 size_t *lists;
1693 int i; 1937 int i;
1694 EVENT_LIST (5);
1695 1938
1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1939 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1940 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1698 1941
1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1942 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1738 if (!lws [i]) 1981 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 1982 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 } 1983 }
1741 } 1984 }
1742 1985
1986 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)); 1987 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1744 1988
1745 if (ev) 1989 if (ev)
1746 XPUSH_CLOBJ (stash_event, ev); 1990 XPUSH_CLOBJ (stash_event, ev);
1991
1992#if CL_VERSION_1_2
1993
1994void
1995migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
1996 ALIAS:
1997 enqueue_migrate_mem_objects = 0
1998 PPCODE:
1999 EVENT_LIST (3);
2000
2001 cl_uint obj_count;
2002 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2003
2004 cl_event ev = 0;
2005 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2006
2007 if (ev)
2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#endif
1747 2011
1748#if cl_apple_gl_sharing || cl_khr_gl_sharing 2012#if cl_apple_gl_sharing || cl_khr_gl_sharing
1749 2013
1750void 2014void
1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2015acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1752 ALIAS: 2016 ALIAS:
1753 release_gl_objects = 1 2017 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0 2018 enqueue_acquire_gl_objects = 0
1755 enqueue_release_gl_objects = 1 2019 enqueue_release_gl_objects = 1
1756 PPCODE: 2020 PPCODE:
1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2021 EVENT_LIST (2);
1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2022
2023 cl_uint obj_count;
2024 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1759 2025
1760 cl_event ev = 0; 2026 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 2027
1770 if (ix) 2028 if (ix)
1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2029 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1772 else 2030 else
1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2031 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1774 2032
1775 if (ev) 2033 if (ev)
1776 XPUSH_CLOBJ (stash_event, ev); 2034 XPUSH_CLOBJ (stash_event, ev);
1777 2035
1778#endif 2036#endif
1792void 2050void
1793marker (OpenCL::Queue self, ...) 2051marker (OpenCL::Queue self, ...)
1794 ALIAS: 2052 ALIAS:
1795 enqueue_marker = 0 2053 enqueue_marker = 0
1796 PPCODE: 2054 PPCODE:
2055 EVENT_LIST (1);
1797 cl_event ev = 0; 2056 cl_event ev = 0;
1798 EVENT_LIST (1);
1799#if PREFER_1_1 2057#if PREFER_1_1
1800 if (!event_list_count) 2058 if (!event_list_count)
1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2059 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1802 else 2060 else
1803#if CL_VERSION_1_2 2061#if CL_VERSION_1_2
1817void 2075void
1818barrier (OpenCL::Queue self, ...) 2076barrier (OpenCL::Queue self, ...)
1819 ALIAS: 2077 ALIAS:
1820 enqueue_barrier = 0 2078 enqueue_barrier = 0
1821 PPCODE: 2079 PPCODE:
2080 EVENT_LIST (1);
1822 cl_event ev = 0; 2081 cl_event ev = 0;
1823 EVENT_LIST (1);
1824#if PREFER_1_1 2082#if PREFER_1_1
1825 if (!event_list_count && GIMME_V == G_VOID) 2083 if (!event_list_count && GIMME_V == G_VOID)
1826 NEED_SUCCESS (EnqueueBarrier, (self)); 2084 NEED_SUCCESS (EnqueueBarrier, (self));
1827 else 2085 else
1828#if CL_VERSION_1_2 2086#if CL_VERSION_1_2
2156void 2414void
2157build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2415build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2158 ALIAS: 2416 ALIAS:
2159 build_async = 1 2417 build_async = 1
2160 CODE: 2418 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; 2419 cl_uint device_count = 0;
2164 cl_device_id *device_list = 0; 2420 cl_device_id *device_list = 0;
2165 2421
2166 if (SvOK (devices)) 2422 if (SvOK (devices))
2167 { 2423 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2424
2425 void *user_data;
2426 program_callback pfn_notify = make_program_callback (notify, &user_data);
2427
2428 if (ix)
2429 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2430 else
2431 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2432
2433#if CL_VERSION_1_2
2434
2435void
2436compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2437 CODE:
2438 cl_uint device_count = 0;
2439 cl_device_id *device_list = 0;
2440
2441 if (SvOK (devices))
2442 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2443
2444 cl_uint header_count = 0;
2445 cl_program *header_list = 0;
2446 const char **header_name = 0;
2447
2448 if (SvOK (headers))
2449 {
2168 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2450 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2169 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2451 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2170 2452
2171 AV *av = (AV *)SvRV (devices); 2453 HV *hv = (HV *)SvRV (devices);
2172 num_devices = av_len (av) + 1;
2173 2454
2174 if (num_devices) 2455 header_count = hv_iterinit (hv);
2456 header_list = tmpbuf (sizeof (*header_list) * header_count);
2457 header_name = tmpbuf (sizeof (*header_name) * header_count);
2458
2459 HE *he;
2460 int i = 0;
2461 while (he = hv_iternext (hv))
2175 { 2462 {
2176 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2463 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2464 header_list [i] = SvCLOBJ ("clCompileProgram", "headers", HeVAL (he), "OpenCL::Program");
2177 int count; 2465 ++i;
2178 for (count = 0; count < num_devices; ++count)
2179 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2180 } 2466 }
2181 } 2467 }
2182 2468
2183 if (SvOK (notify)) 2469 void *user_data;
2184 { 2470 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 2471
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)); 2472 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2473 header_count, header_list, header_name, pfn_notify, user_data));
2474
2475#endif
2194 2476
2195void 2477void
2196build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2478build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2197 PPCODE: 2479 PPCODE:
2198 size_t size; 2480 size_t size;
2614 const int i = 0; 2896 const int i = 0;
2615 PUSHs (sv_2mortal (newSVuv (value [i]))); 2897 PUSHs (sv_2mortal (newSVuv (value [i])));
2616 2898
2617#END:kernel_work_group 2899#END:kernel_work_group
2618 2900
2901#if CL_VERSION_1_2
2902
2903void
2904arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2905 PPCODE:
2906 size_t size;
2907 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2908 SV *sv = sv_2mortal (newSV (size));
2909 SvUPGRADE (sv, SVt_PV);
2910 SvPOK_only (sv);
2911 SvCUR_set (sv, size);
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2913 XPUSHs (sv);
2914
2915#BEGIN:kernel_arg
2916
2917void
2918arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2919 PPCODE:
2920 cl_kernel_arg_address_qualifier value [1];
2921 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2922 EXTEND (SP, 1);
2923 const int i = 0;
2924 PUSHs (sv_2mortal (newSVuv (value [i])));
2925
2926void
2927arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2928 PPCODE:
2929 cl_kernel_arg_access_qualifier value [1];
2930 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2931 EXTEND (SP, 1);
2932 const int i = 0;
2933 PUSHs (sv_2mortal (newSVuv (value [i])));
2934
2935void
2936arg_type_name (OpenCL::Kernel self, cl_uint idx)
2937 ALIAS:
2938 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2939 arg_name = CL_KERNEL_ARG_NAME
2940 PPCODE:
2941 size_t size;
2942 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2943 char *value = tmpbuf (size);
2944 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2945 EXTEND (SP, 1);
2946 const int i = 0;
2947 PUSHs (sv_2mortal (newSVpv (value, 0)));
2948
2949void
2950arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2951 PPCODE:
2952 cl_kernel_arg_type_qualifier value [1];
2953 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2954 EXTEND (SP, 1);
2955 const int i = 0;
2956 PUSHs (sv_2mortal (newSVuv (value [i])));
2957
2958#END:kernel_arg
2959
2960#endif
2961
2619MODULE = OpenCL PACKAGE = OpenCL::Event 2962MODULE = OpenCL PACKAGE = OpenCL::Event
2620 2963
2621void 2964void
2622DESTROY (OpenCL::Event self) 2965DESTROY (OpenCL::Event self)
2623 CODE: 2966 CODE:
2776set (OpenCL::Mapped self, size_t offset, SV *data) 3119set (OpenCL::Mapped self, size_t offset, SV *data)
2777 CODE: 3120 CODE:
2778 STRLEN len; 3121 STRLEN len;
2779 const char *ptr = SvPVbyte (data, len); 3122 const char *ptr = SvPVbyte (data, len);
2780 3123
2781 if (offset < 0 || offset + len > self->cb) 3124 if (offset + len > self->cb)
2782 croak ("OpenCL::Mapped::set out of bound condition detected"); 3125 croak ("OpenCL::Mapped::set out of bound condition detected");
2783 3126
2784 memcpy (offset + (char *)self->ptr, ptr, len); 3127 memcpy (offset + (char *)self->ptr, ptr, len);
2785 3128
2786MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer 3129MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines