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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.64 by root, Thu May 3 23:30:08 2012 UTC vs.
Revision 1.71 by root, Sat May 5 02:45:38 2012 UTC

40 #define PREFER_1_1 1 40 #define PREFER_1_1 1
41#endif 41#endif
42 42
43typedef cl_platform_id OpenCL__Platform; 43typedef cl_platform_id OpenCL__Platform;
44typedef cl_device_id OpenCL__Device; 44typedef cl_device_id OpenCL__Device;
45typedef cl_device_id OpenCL__SubDevice;
45typedef cl_context OpenCL__Context; 46typedef cl_context OpenCL__Context;
46typedef cl_command_queue OpenCL__Queue; 47typedef cl_command_queue OpenCL__Queue;
47typedef cl_mem OpenCL__Memory; 48typedef cl_mem OpenCL__Memory;
48typedef cl_mem OpenCL__Buffer; 49typedef cl_mem OpenCL__Buffer;
49typedef cl_mem OpenCL__BufferObj; 50typedef cl_mem OpenCL__BufferObj;
57typedef cl_event OpenCL__Event; 58typedef cl_event OpenCL__Event;
58typedef cl_event OpenCL__UserEvent; 59typedef cl_event OpenCL__UserEvent;
59 60
60typedef struct mapped * OpenCL__Mapped; 61typedef struct mapped * OpenCL__Mapped;
61 62
62typedef SV *FUTURE;
63
64static HV 63static HV
65 *stash_platform, 64 *stash_platform,
66 *stash_device, 65 *stash_device,
66 *stash_subdevice,
67 *stash_context, 67 *stash_context,
68 *stash_queue, 68 *stash_queue,
69 *stash_program, 69 *stash_program,
70 *stash_kernel, 70 *stash_kernel,
71 *stash_sampler, 71 *stash_sampler,
151 } 151 }
152 152
153 return buf [idx]; 153 return buf [idx];
154} 154}
155 155
156static const char * ecb_noinline
157cv_get_name (CV *cv)
158{
159 static char fullname [256];
160
161 GV *gv = CvGV (cv); // gv better be nonzero
162
163 HV *stash = GvSTASH (gv);
164 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
165 const char *gvname = GvNAME (gv);
166
167 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
168 return fullname;
169}
170
156/*****************************************************************************/ 171/*****************************************************************************/
157 172
158typedef struct 173typedef struct
159{ 174{
160 IV iv; 175 IV iv;
217 if (res) \ 232 if (res) \
218 FAIL (name); 233 FAIL (name);
219 234
220/*****************************************************************************/ 235/*****************************************************************************/
221 236
237static SV *
238new_clobj (HV *stash, IV id)
239{
240 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
241}
242
243#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
244#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
245
246/* cl objects are either \$iv, or [$iv, ...] */
247/* they can be upgraded at runtime to the array form */
248static void * ecb_noinline
249SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
250{
251 // sv_derived_from is quite slow :(
252 if (SvROK (sv) && sv_derived_from (sv, pkg))
253 return (void *)SvIV (SvRV (sv));
254
255 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
256}
257
258// the "no-inherit" version of the above
259static void * ecb_noinline
260SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
261{
262 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
263 return (void *)SvIV (SvRV (sv));
264
265 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
266}
267
268/*****************************************************************************/
269
222static cl_context_properties * ecb_noinline 270static cl_context_properties * ecb_noinline
223SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 271SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
224{ 272{
225 if (!sv || !SvOK (sv)) 273 if (!sv || !SvOK (sv))
226 if (extra) 274 if (extra)
227 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 275 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
228 else 276 else
234 int i, len = av_len (av) + 1; 282 int i, len = av_len (av) + 1;
235 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 283 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
236 cl_context_properties *l = p; 284 cl_context_properties *l = p;
237 285
238 if (len & 1) 286 if (len & 1)
239 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 287 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
240 288
241 while (extracount--) 289 while (extracount--)
242 *l++ = *extra++; 290 *l++ = *extra++;
243 291
244 for (i = 0; i < len; i += 2) 292 for (i = 0; i < len; i += 2)
247 SV *p_sv = *av_fetch (av, i + 1, 0); 295 SV *p_sv = *av_fetch (av, i + 1, 0);
248 cl_context_properties v = SvIV (p_sv); // code below can override 296 cl_context_properties v = SvIV (p_sv); // code below can override
249 297
250 switch (t) 298 switch (t)
251 { 299 {
300 case CL_CONTEXT_PLATFORM:
301 if (SvROK (p_sv))
302 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
303 break;
304
252 case CL_GLX_DISPLAY_KHR: 305 case CL_GLX_DISPLAY_KHR:
253 if (!SvOK (p_sv)) 306 if (!SvOK (p_sv))
254 { 307 {
255 void *func = glsym ("_glXGetCurrentDisplay"); 308 void *func = glsym ("_glXGetCurrentDisplay");
256 if (func) 309 if (func)
279 *l = 0; 332 *l = 0;
280 333
281 return p; 334 return p;
282 } 335 }
283 336
284 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 337 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
285} 338}
286 339
287/*****************************************************************************/ 340// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
288 341// is a pointer (and only on well-behaved systems).
289static SV *
290new_clobj (HV *stash, IV id)
291{
292 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
293}
294
295#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
296#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
297
298/* cl objects are either \$iv, or [$iv, ...] */
299/* they can be upgraded at runtime to the array form */
300static void * ecb_noinline 342static void * ecb_noinline
301SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 343object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
302{ 344{
303 // sv_derived_from is quite slow :( 345 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
304 if (SvROK (sv) && sv_derived_from (sv, pkg)) 346 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
305 return (void *)SvIV (SvRV (sv)); 347 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
306 348
307 croak ("%s: %s is not of type %s", func, svname, pkg); 349 AV *av = (AV *)SvRV (arg);
308} 350 void **list = 0;
351 cl_uint count = av_len (av) + 1;
309 352
310// the "no-inherit" version of the above 353 if (count)
311static void * ecb_noinline 354 {
312SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash) 355 list = tmpbuf (sizeof (*list) * count);
313{ 356 int i;
314 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash) 357 for (i = 0; i < count; ++i)
315 return (void *)SvIV (SvRV (sv)); 358 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
359 }
316 360
317 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash)); 361 if (!count && !or_undef)
362 croak ("%s: '%s' must contain at least one %s object",
363 cv_get_name (cv), argname, klass);
364
365 *rcount = count;
366 return (void *)list;
318} 367}
319 368
320/*****************************************************************************/ 369/*****************************************************************************/
321/* callback stuff */ 370/* callback stuff */
322 371
482static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 531static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
483 532
484static void CL_CALLBACK 533static void CL_CALLBACK
485eq_program_notify (cl_program program, void *user_data) 534eq_program_notify (cl_program program, void *user_data)
486{ 535{
536 clRetainProgram (program);
537
487 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 538 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
539}
540
541typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
542
543static program_callback ecb_noinline
544make_program_callback (SV *notify, void **ruser_data)
545{
546 if (SvOK (notify))
547 {
548 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
549 return eq_program_notify;
550 }
551 else
552 {
553 *ruser_data = 0;
554 return 0;
555 }
488} 556}
489 557
490struct build_args 558struct build_args
491{ 559{
492 cl_program program; 560 cl_program program;
558 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 626 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
559 return res; 627 return res;
560} 628}
561 629
562static cl_event * ecb_noinline 630static cl_event * ecb_noinline
563event_list (SV **items, cl_uint *rcount, cl_event extra) 631event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
564{ 632{
565 cl_uint count = *rcount; 633 cl_uint count = *rcount;
566 634
567 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed 635 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
568 *rcount = count = 0; 636 *rcount = count = 0;
573 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 641 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
574 int i = 0; 642 int i = 0;
575 643
576 while (count--) 644 while (count--)
577 if (SvOK (items [count])) 645 if (SvOK (items [count]))
578 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 646 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
579 647
580 if (extra) 648 if (extra)
581 list [i++] = extra; 649 list [i++] = extra;
582 650
583 *rcount = i; 651 *rcount = i;
585 return i ? list : 0; 653 return i ? list : 0;
586} 654}
587 655
588#define EVENT_LIST(skip) \ 656#define EVENT_LIST(skip) \
589 cl_uint event_list_count = items - (skip); \ 657 cl_uint event_list_count = items - (skip); \
590 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 658 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
591 659
592#define INFO(class) \ 660#define INFO(class) \
593{ \ 661{ \
594 size_t size; \ 662 size_t size; \
595 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 663 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
677 745
678 mapped->ptr = 0; 746 mapped->ptr = 0;
679} 747}
680 748
681static void 749static void
682mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count) 750mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
683{ 751{
684 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event); 752 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
685 cl_event ev; 753 cl_event ev;
686 754
687 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev)); 755 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
688 756
689 clReleaseEvent (mapped->event); 757 clReleaseEvent (mapped->event);
730 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 798 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
731 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 799 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
732 800
733 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 801 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
734 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 802 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
803 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
735 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 804 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
736 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 805 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
737 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 806 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
738 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 807 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
739 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 808 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
780 list = tmpbuf (sizeof (*list) * count); 849 list = tmpbuf (sizeof (*list) * count);
781 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 850 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
782 851
783 EXTEND (SP, count); 852 EXTEND (SP, count);
784 for (i = 0; i < count; ++i) 853 for (i = 0; i < count; ++i)
785 PUSH_CLOBJ (stash_platform, (IV)list [i]); 854 PUSH_CLOBJ (stash_platform, list [i]);
786 855
787void 856void
788context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 857context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
789 PPCODE: 858 PPCODE:
790 CONTEXT_NOTIFY_CALLBACK; 859 CONTEXT_NOTIFY_CALLBACK;
791 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res)); 860 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
792 XPUSH_CLOBJ_CONTEXT; 861 XPUSH_CLOBJ_CONTEXT;
793 862
794void 863void
795context (FUTURE properties, FUTURE devices, FUTURE notify) 864context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
796 PPCODE: 865 PPCODE:
797 /* der Gipfel der Kunst */ 866 cl_uint device_count;
867 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
868
869 CONTEXT_NOTIFY_CALLBACK;
870 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
871 XPUSH_CLOBJ_CONTEXT;
798 872
799void 873void
800wait_for_events (...) 874wait_for_events (...)
801 CODE: 875 CODE:
802 EVENT_LIST (0); 876 EVENT_LIST (0);
853 EXTEND (SP, count); 927 EXTEND (SP, count);
854 for (i = 0; i < count; ++i) 928 for (i = 0; i < count; ++i)
855 PUSH_CLOBJ (stash_device, list [i]); 929 PUSH_CLOBJ (stash_device, list [i]);
856 930
857void 931void
858context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef) 932context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
859 PPCODE: 933 PPCODE:
860 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
861 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call");
862
863 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 934 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
864 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2); 935 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
865 936
866 AV *av = (AV *)SvRV (devices); 937 cl_uint device_count;
867 cl_uint num_devices = av_len (av) + 1; 938 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
868 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices);
869
870 int i;
871 for (i = num_devices; i--; )
872 device_list [i] = SvCLOBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
873 939
874 CONTEXT_NOTIFY_CALLBACK; 940 CONTEXT_NOTIFY_CALLBACK;
875 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res)); 941 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
876 XPUSH_CLOBJ_CONTEXT; 942 XPUSH_CLOBJ_CONTEXT;
877 943
878void 944void
879context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 945context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
880 PPCODE: 946 PPCODE:
881 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 947 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
882 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2); 948 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
883 949
884 CONTEXT_NOTIFY_CALLBACK; 950 CONTEXT_NOTIFY_CALLBACK;
885 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res)); 951 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
886 XPUSH_CLOBJ_CONTEXT; 952 XPUSH_CLOBJ_CONTEXT;
887 953
889 955
890void 956void
891info (OpenCL::Device self, cl_device_info name) 957info (OpenCL::Device self, cl_device_info name)
892 PPCODE: 958 PPCODE:
893 INFO (Device) 959 INFO (Device)
960
961#if CL_VERSION_1_2
962
963void
964sub_devices (OpenCL::Device self, SV *properties)
965 PPCODE:
966 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
967 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
968
969 properties = SvRV (properties);
970
971 cl_uint count = av_len ((AV *)properties) + 1;
972 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
973
974 int i;
975 for (i = 0; i < count; ++i)
976 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
977
978 props [count] = 0;
979
980 cl_uint num_devices;
981 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
982 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
983 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
984
985 EXTEND (SP, num_devices);
986 for (i = 0; i < count; ++i)
987 PUSH_CLOBJ (stash_subdevice, list [i]);
988
989#endif
894 990
895#BEGIN:device 991#BEGIN:device
896 992
897void 993void
898type (OpenCL::Device self) 994type (OpenCL::Device self)
1114 for (i = 0; i < n; ++i) 1210 for (i = 0; i < n; ++i)
1115 PUSHs (sv_2mortal (newSVuv (value [i]))); 1211 PUSHs (sv_2mortal (newSVuv (value [i])));
1116 1212
1117#END:device 1213#END:device
1118 1214
1215MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1216
1217#if CL_VERSION_1_2
1218
1219void
1220DESTROY (OpenCL::SubDevice self)
1221 CODE:
1222 clReleaseDevice (self);
1223
1224#endif
1225
1119MODULE = OpenCL PACKAGE = OpenCL::Context 1226MODULE = OpenCL PACKAGE = OpenCL::Context
1120 1227
1121void 1228void
1122DESTROY (OpenCL::Context context) 1229DESTROY (OpenCL::Context self)
1123 CODE: 1230 CODE:
1124 clReleaseContext (context); 1231 clReleaseContext (self);
1125 1232
1126void 1233void
1127info (OpenCL::Context self, cl_context_info name) 1234info (OpenCL::Context self, cl_context_info name)
1128 PPCODE: 1235 PPCODE:
1129 INFO (Context) 1236 INFO (Context)
1170 const cl_image_desc desc = { 1277 const cl_image_desc desc = {
1171 type, 1278 type,
1172 width, height, depth, 1279 width, height, depth,
1173 array_size, row_pitch, slice_pitch, 1280 array_size, row_pitch, slice_pitch,
1174 num_mip_level, num_samples, 1281 num_mip_level, num_samples,
1175 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1282 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1176 }; 1283 };
1177 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1284 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1178 HV *stash = stash_image; 1285 HV *stash = stash_image;
1179 switch (type) 1286 switch (type)
1180 { 1287 {
1316 XPUSH_CLOBJ (stash_program, prog); 1423 XPUSH_CLOBJ (stash_program, prog);
1317 1424
1318void 1425void
1319program_with_binary (OpenCL::Context self, SV *devices, SV *binaries) 1426program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1320 PPCODE: 1427 PPCODE:
1321 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1428 cl_uint device_count;
1322 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects"); 1429 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1323
1324 devices = SvRV (devices);
1325 1430
1326 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV) 1431 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1327 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings"); 1432 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1328 1433
1329 binaries = SvRV (binaries); 1434 binaries = SvRV (binaries);
1330 1435
1331 if (av_len ((AV *)devices) != av_len ((AV *)binaries)) 1436 if (device_count != av_len ((AV *)binaries) + 1)
1332 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed"); 1437 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1333 1438
1334 int count = av_len ((AV *)devices) + 1;
1335 cl_device_id *device_list = tmpbuf (sizeof (*device_list) * count);
1336 size_t *length_list = tmpbuf (sizeof (*length_list) * count); 1439 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1337 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count); 1440 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1338 cl_int *status_list = tmpbuf (sizeof (*status_list) * count); 1441 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1339 1442
1340 int i; 1443 int i;
1341 for (i = 0; i < count; ++i) 1444 for (i = 0; i < device_count; ++i)
1342 { 1445 {
1343 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1344 STRLEN len; 1446 STRLEN len;
1345 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len); 1447 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1346 length_list [i] = len; 1448 length_list [i] = len;
1347 } 1449 }
1348 1450
1349 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res)); 1451 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1452 length_list, binary_list,
1453 GIMME_V == G_ARRAY ? status_list : 0, &res));
1350 1454
1351 EXTEND (SP, 2); 1455 EXTEND (SP, 2);
1352 PUSH_CLOBJ (stash_program, prog); 1456 PUSH_CLOBJ (stash_program, prog);
1353 1457
1354 if (GIMME_V == G_ARRAY) 1458 if (GIMME_V == G_ARRAY)
1355 { 1459 {
1356 AV *av = newAV (); 1460 AV *av = newAV ();
1357 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1461 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1358 1462
1359 for (i = count; i--; ) 1463 for (i = device_count; i--; )
1360 av_store (av, i, newSViv (status_list [i])); 1464 av_store (av, i, newSViv (status_list [i]));
1361 } 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;
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
1362 1500
1363#BEGIN:context 1501#BEGIN:context
1364 1502
1365void 1503void
1366reference_count (OpenCL::Context self) 1504reference_count (OpenCL::Context self)
1412void 1550void
1413read_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, ...)
1414 ALIAS: 1552 ALIAS:
1415 enqueue_read_buffer = 0 1553 enqueue_read_buffer = 0
1416 PPCODE: 1554 PPCODE:
1417 cl_event ev = 0;
1418 EVENT_LIST (6); 1555 EVENT_LIST (6);
1419 1556
1420 SvUPGRADE (data, SVt_PV); 1557 SvUPGRADE (data, SVt_PV);
1421 SvGROW (data, len); 1558 SvGROW (data, len);
1422 SvPOK_only (data); 1559 SvPOK_only (data);
1423 SvCUR_set (data, len); 1560 SvCUR_set (data, len);
1561
1562 cl_event ev = 0;
1424 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));
1425 1564
1426 if (ev) 1565 if (ev)
1427 XPUSH_CLOBJ (stash_event, ev); 1566 XPUSH_CLOBJ (stash_event, ev);
1428 1567
1429void 1568void
1430write_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, ...)
1431 ALIAS: 1570 ALIAS:
1432 enqueue_write_buffer = 0 1571 enqueue_write_buffer = 0
1433 PPCODE: 1572 PPCODE:
1434 cl_event ev = 0; 1573 EVENT_LIST (5);
1574
1435 STRLEN len; 1575 STRLEN len;
1436 char *ptr = SvPVbyte (data, len); 1576 char *ptr = SvPVbyte (data, len);
1437 EVENT_LIST (5);
1438 1577
1578 cl_event ev = 0;
1439 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));
1440 1580
1441 if (ev) 1581 if (ev)
1442 XPUSH_CLOBJ (stash_event, ev); 1582 XPUSH_CLOBJ (stash_event, ev);
1443 1583
1446void 1586void
1447fill_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, ...)
1448 ALIAS: 1588 ALIAS:
1449 enqueue_fill_buffer = 0 1589 enqueue_fill_buffer = 0
1450 PPCODE: 1590 PPCODE:
1451 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1452 STRLEN len; 1593 STRLEN len;
1453 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1454 EVENT_LIST (5);
1455 1595
1596 cl_event ev = 0;
1456 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));
1457 1598
1458 if (ev) 1599 if (ev)
1459 XPUSH_CLOBJ (stash_event, ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1460 1601
1461void 1602void
1462fill_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, ...)
1463 ALIAS: 1604 ALIAS:
1464 enqueue_fill_image = 0 1605 enqueue_fill_image = 0
1465 PPCODE: 1606 PPCODE:
1466 cl_event ev = 0; 1607 EVENT_LIST (12);
1608
1467 const size_t origin [3] = { x, y, z }; 1609 const size_t origin [3] = { x, y, z };
1468 const size_t region [3] = { width, height, depth }; 1610 const size_t region [3] = { width, height, depth };
1469 EVENT_LIST (12);
1470 1611
1471 const cl_float c_f [4] = { r, g, b, a }; 1612 const cl_float c_f [4] = { r, g, b, a };
1472 const cl_uint c_u [4] = { r, g, b, a }; 1613 const cl_uint c_u [4] = { r, g, b, a };
1473 const cl_int c_s [4] = { r, g, b, a }; 1614 const cl_int c_s [4] = { r, g, b, a };
1474 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1615 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1477 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1618 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1478 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1619 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1479 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)
1480 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.");
1481 1622
1623 cl_event ev = 0;
1482 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]],
1483 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));
1484 1626
1485 if (ev) 1627 if (ev)
1486 XPUSH_CLOBJ (stash_event, ev); 1628 XPUSH_CLOBJ (stash_event, ev);
1490void 1632void
1491copy_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, ...)
1492 ALIAS: 1634 ALIAS:
1493 enqueue_copy_buffer = 0 1635 enqueue_copy_buffer = 0
1494 PPCODE: 1636 PPCODE:
1637 EVENT_LIST (6);
1638
1495 cl_event ev = 0; 1639 cl_event ev = 0;
1496 EVENT_LIST (6);
1497
1498 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));
1499 1641
1500 if (ev) 1642 if (ev)
1501 XPUSH_CLOBJ (stash_event, ev); 1643 XPUSH_CLOBJ (stash_event, ev);
1502 1644
1503void 1645void
1504read_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, ...)
1505 ALIAS: 1647 ALIAS:
1506 enqueue_read_buffer_rect = 0 1648 enqueue_read_buffer_rect = 0
1507 PPCODE: 1649 PPCODE:
1508 cl_event ev = 0; 1650 EVENT_LIST (17);
1651
1509 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 };
1510 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 };
1511 const size_t region[3] = { width, height, depth }; 1654 const size_t region[3] = { width, height, depth };
1512 EVENT_LIST (17);
1513 1655
1514 if (!buf_row_pitch) 1656 if (!buf_row_pitch)
1515 buf_row_pitch = region [0]; 1657 buf_row_pitch = region [0];
1516 1658
1517 if (!buf_slice_pitch) 1659 if (!buf_slice_pitch)
1527 1669
1528 SvUPGRADE (data, SVt_PV); 1670 SvUPGRADE (data, SVt_PV);
1529 SvGROW (data, len); 1671 SvGROW (data, len);
1530 SvPOK_only (data); 1672 SvPOK_only (data);
1531 SvCUR_set (data, len); 1673 SvCUR_set (data, len);
1674
1675 cl_event ev = 0;
1532 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));
1533 1677
1534 if (ev) 1678 if (ev)
1535 XPUSH_CLOBJ (stash_event, ev); 1679 XPUSH_CLOBJ (stash_event, ev);
1536 1680
1537void 1681void
1538write_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, ...)
1539 ALIAS: 1683 ALIAS:
1540 enqueue_write_buffer_rect = 0 1684 enqueue_write_buffer_rect = 0
1541 PPCODE: 1685 PPCODE:
1542 cl_event ev = 0; 1686 EVENT_LIST (17);
1687
1543 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 };
1544 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 };
1545 const size_t region[3] = { width, height, depth }; 1690 const size_t region[3] = { width, height, depth };
1546 STRLEN len; 1691 STRLEN len;
1547 char *ptr = SvPVbyte (data, len); 1692 char *ptr = SvPVbyte (data, len);
1548 EVENT_LIST (17);
1549 1693
1550 if (!buf_row_pitch) 1694 if (!buf_row_pitch)
1551 buf_row_pitch = region [0]; 1695 buf_row_pitch = region [0];
1552 1696
1553 if (!buf_slice_pitch) 1697 if (!buf_slice_pitch)
1562 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];
1563 1707
1564 if (len < min_len) 1708 if (len < min_len)
1565 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1709 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1566 1710
1711 cl_event ev = 0;
1567 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));
1568 1713
1569 if (ev) 1714 if (ev)
1570 XPUSH_CLOBJ (stash_event, ev); 1715 XPUSH_CLOBJ (stash_event, ev);
1571 1716
1572void 1717void
1573copy_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, ...)
1574 ALIAS: 1719 ALIAS:
1575 enqueue_copy_buffer_rect = 0 1720 enqueue_copy_buffer_rect = 0
1576 PPCODE: 1721 PPCODE:
1577 cl_event ev = 0; 1722 EVENT_LIST (16);
1723
1578 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 };
1579 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 };
1580 const size_t region[3] = { width, height, depth }; 1726 const size_t region[3] = { width, height, depth };
1581 EVENT_LIST (16);
1582 1727
1728 cl_event ev = 0;
1583 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));
1584 1730
1585 if (ev) 1731 if (ev)
1586 XPUSH_CLOBJ (stash_event, ev); 1732 XPUSH_CLOBJ (stash_event, ev);
1587 1733
1588void 1734void
1589read_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, ...)
1590 ALIAS: 1736 ALIAS:
1591 enqueue_read_image = 0 1737 enqueue_read_image = 0
1592 PPCODE: 1738 PPCODE:
1593 cl_event ev = 0; 1739 EVENT_LIST (12);
1740
1594 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 };
1595 const size_t region[3] = { width, height, depth }; 1742 const size_t region[3] = { width, height, depth };
1596 EVENT_LIST (12);
1597 1743
1598 if (!row_pitch) 1744 if (!row_pitch)
1599 row_pitch = img_row_pitch (src); 1745 row_pitch = img_row_pitch (src);
1600 1746
1601 if (depth > 1 && !slice_pitch) 1747 if (depth > 1 && !slice_pitch)
1605 1751
1606 SvUPGRADE (data, SVt_PV); 1752 SvUPGRADE (data, SVt_PV);
1607 SvGROW (data, len); 1753 SvGROW (data, len);
1608 SvPOK_only (data); 1754 SvPOK_only (data);
1609 SvCUR_set (data, len); 1755 SvCUR_set (data, len);
1756
1757 cl_event ev = 0;
1610 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));
1611 1759
1612 if (ev) 1760 if (ev)
1613 XPUSH_CLOBJ (stash_event, ev); 1761 XPUSH_CLOBJ (stash_event, ev);
1614 1762
1615void 1763void
1616write_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, ...)
1617 ALIAS: 1765 ALIAS:
1618 enqueue_write_image = 0 1766 enqueue_write_image = 0
1619 PPCODE: 1767 PPCODE:
1620 cl_event ev = 0; 1768 EVENT_LIST (12);
1769
1621 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 };
1622 const size_t region[3] = { width, height, depth }; 1771 const size_t region[3] = { width, height, depth };
1623 STRLEN len; 1772 STRLEN len;
1624 char *ptr = SvPVbyte (data, len); 1773 char *ptr = SvPVbyte (data, len);
1625 EVENT_LIST (12);
1626 1774
1627 if (!row_pitch) 1775 if (!row_pitch)
1628 row_pitch = img_row_pitch (dst); 1776 row_pitch = img_row_pitch (dst);
1629 1777
1630 if (depth > 1 && !slice_pitch) 1778 if (depth > 1 && !slice_pitch)
1633 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;
1634 1782
1635 if (len < min_len) 1783 if (len < min_len)
1636 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1784 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1637 1785
1786 cl_event ev = 0;
1638 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));
1639 1788
1640 if (ev) 1789 if (ev)
1641 XPUSH_CLOBJ (stash_event, ev); 1790 XPUSH_CLOBJ (stash_event, ev);
1642 1791
1643void 1792void
1644copy_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, ...)
1645 ALIAS: 1794 ALIAS:
1646 enqueue_copy_image = 0 1795 enqueue_copy_image = 0
1647 PPCODE: 1796 PPCODE:
1648 cl_event ev = 0; 1797 EVENT_LIST (12);
1798
1649 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 };
1650 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 };
1651 const size_t region[3] = { width, height, depth }; 1801 const size_t region[3] = { width, height, depth };
1652 EVENT_LIST (12);
1653 1802
1803 cl_event ev = 0;
1654 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));
1655 1805
1656 if (ev) 1806 if (ev)
1657 XPUSH_CLOBJ (stash_event, ev); 1807 XPUSH_CLOBJ (stash_event, ev);
1658 1808
1659void 1809void
1660copy_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, ...)
1661 ALIAS: 1811 ALIAS:
1662 enqueue_copy_image_to_buffer = 0 1812 enqueue_copy_image_to_buffer = 0
1663 PPCODE: 1813 PPCODE:
1664 cl_event ev = 0; 1814 EVENT_LIST (10);
1815
1665 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 };
1666 const size_t region [3] = { width, height, depth }; 1817 const size_t region [3] = { width, height, depth };
1667 EVENT_LIST (10);
1668 1818
1819 cl_event ev = 0;
1669 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));
1670 1821
1671 if (ev) 1822 if (ev)
1672 XPUSH_CLOBJ (stash_event, ev); 1823 XPUSH_CLOBJ (stash_event, ev);
1673 1824
1674void 1825void
1675copy_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, ...)
1676 ALIAS: 1827 ALIAS:
1677 enqueue_copy_buffer_to_image = 0 1828 enqueue_copy_buffer_to_image = 0
1678 PPCODE: 1829 PPCODE:
1679 cl_event ev = 0; 1830 EVENT_LIST (10);
1831
1680 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 };
1681 const size_t region [3] = { width, height, depth }; 1833 const size_t region [3] = { width, height, depth };
1682 EVENT_LIST (10);
1683 1834
1835 cl_event ev = 0;
1684 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));
1685 1837
1686 if (ev) 1838 if (ev)
1687 XPUSH_CLOBJ (stash_event, ev); 1839 XPUSH_CLOBJ (stash_event, ev);
1688 1840
1689void 1841void
1690map_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, ...) 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, ...)
1691 ALIAS: 1843 ALIAS:
1692 enqueue_map_buffer = 0 1844 enqueue_map_buffer = 0
1693 PPCODE: 1845 PPCODE:
1694 cl_event ev;
1695 EVENT_LIST (6); 1846 EVENT_LIST (6);
1847
1696 size_t cb = SvIV (cb_); 1848 size_t cb = SvIV (cb_);
1697 1849
1698 if (!SvOK (cb_)) 1850 if (!SvOK (cb_))
1699 { 1851 {
1700 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0)); 1852 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1701 cb -= offset; 1853 cb -= offset;
1702 } 1854 }
1703 1855
1856 cl_event ev;
1704 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));
1705 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));
1706 1859
1707void 1860void
1708map_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, ...) 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, ...)
1749 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch)); 1902 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1750 1903
1751void 1904void
1752unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1905unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1753 PPCODE: 1906 PPCODE:
1754 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2); 1907 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1755 if (GIMME_V != G_VOID) 1908 if (GIMME_V != G_VOID)
1756 { 1909 {
1757 clRetainEvent (mapped->event); 1910 clRetainEvent (mapped->event);
1758 XPUSH_CLOBJ (stash_event, mapped->event); 1911 XPUSH_CLOBJ (stash_event, mapped->event);
1759 } 1912 }
1761void 1914void
1762task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1915task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1763 ALIAS: 1916 ALIAS:
1764 enqueue_task = 0 1917 enqueue_task = 0
1765 PPCODE: 1918 PPCODE:
1919 EVENT_LIST (2);
1920
1766 cl_event ev = 0; 1921 cl_event ev = 0;
1767 EVENT_LIST (2);
1768
1769 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));
1770 1923
1771 if (ev) 1924 if (ev)
1772 XPUSH_CLOBJ (stash_event, ev); 1925 XPUSH_CLOBJ (stash_event, ev);
1773 1926
1774void 1927void
1775nd_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, ...)
1776 ALIAS: 1929 ALIAS:
1777 enqueue_nd_range_kernel = 0 1930 enqueue_nd_range_kernel = 0
1778 PPCODE: 1931 PPCODE:
1779 cl_event ev = 0; 1932 EVENT_LIST (5);
1933
1780 size_t *gwo = 0, *gws, *lws = 0; 1934 size_t *gwo = 0, *gws, *lws = 0;
1781 int gws_len; 1935 int gws_len;
1782 size_t *lists; 1936 size_t *lists;
1783 int i; 1937 int i;
1784 EVENT_LIST (5);
1785 1938
1786 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)
1787 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1940 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1788 1941
1789 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1942 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1828 if (!lws [i]) 1981 if (!lws [i])
1829 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);
1830 } 1983 }
1831 } 1984 }
1832 1985
1986 cl_event ev = 0;
1833 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));
1834 1988
1835 if (ev) 1989 if (ev)
1836 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
1837 2011
1838#if cl_apple_gl_sharing || cl_khr_gl_sharing 2012#if cl_apple_gl_sharing || cl_khr_gl_sharing
1839 2013
1840void 2014void
1841acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2015acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1842 ALIAS: 2016 ALIAS:
1843 release_gl_objects = 1 2017 release_gl_objects = 1
1844 enqueue_acquire_gl_objects = 0 2018 enqueue_acquire_gl_objects = 0
1845 enqueue_release_gl_objects = 1 2019 enqueue_release_gl_objects = 1
1846 PPCODE: 2020 PPCODE:
1847 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2021 EVENT_LIST (2);
1848 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);
1849 2025
1850 cl_event ev = 0; 2026 cl_event ev = 0;
1851 EVENT_LIST (2);
1852 AV *av = (AV *)SvRV (objects);
1853 cl_uint num_objects = av_len (av) + 1;
1854 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1855 int i;
1856
1857 for (i = num_objects; i--; )
1858 object_list [i] = SvCLOBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1859 2027
1860 if (ix) 2028 if (ix)
1861 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));
1862 else 2030 else
1863 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));
1864 2032
1865 if (ev) 2033 if (ev)
1866 XPUSH_CLOBJ (stash_event, ev); 2034 XPUSH_CLOBJ (stash_event, ev);
1867 2035
1868#endif 2036#endif
1882void 2050void
1883marker (OpenCL::Queue self, ...) 2051marker (OpenCL::Queue self, ...)
1884 ALIAS: 2052 ALIAS:
1885 enqueue_marker = 0 2053 enqueue_marker = 0
1886 PPCODE: 2054 PPCODE:
2055 EVENT_LIST (1);
1887 cl_event ev = 0; 2056 cl_event ev = 0;
1888 EVENT_LIST (1);
1889#if PREFER_1_1 2057#if PREFER_1_1
1890 if (!event_list_count) 2058 if (!event_list_count)
1891 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2059 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1892 else 2060 else
1893#if CL_VERSION_1_2 2061#if CL_VERSION_1_2
1907void 2075void
1908barrier (OpenCL::Queue self, ...) 2076barrier (OpenCL::Queue self, ...)
1909 ALIAS: 2077 ALIAS:
1910 enqueue_barrier = 0 2078 enqueue_barrier = 0
1911 PPCODE: 2079 PPCODE:
2080 EVENT_LIST (1);
1912 cl_event ev = 0; 2081 cl_event ev = 0;
1913 EVENT_LIST (1);
1914#if PREFER_1_1 2082#if PREFER_1_1
1915 if (!event_list_count && GIMME_V == G_VOID) 2083 if (!event_list_count && GIMME_V == G_VOID)
1916 NEED_SUCCESS (EnqueueBarrier, (self)); 2084 NEED_SUCCESS (EnqueueBarrier, (self));
1917 else 2085 else
1918#if CL_VERSION_1_2 2086#if CL_VERSION_1_2
2246void 2414void
2247build (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)
2248 ALIAS: 2416 ALIAS:
2249 build_async = 1 2417 build_async = 1
2250 CODE: 2418 CODE:
2251 void (CL_CALLBACK *pfn_notify)(cl_program program, void *user_data) = 0;
2252 void *user_data = 0;
2253 cl_uint num_devices = 0; 2419 cl_uint device_count = 0;
2254 cl_device_id *device_list = 0; 2420 cl_device_id *device_list = 0;
2255 2421
2256 if (SvOK (devices)) 2422 if (SvOK (devices))
2257 { 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 {
2258 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2450 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2259 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");
2260 2452
2261 AV *av = (AV *)SvRV (devices); 2453 HV *hv = (HV *)SvRV (devices);
2262 num_devices = av_len (av) + 1;
2263 2454
2264 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))
2265 { 2462 {
2266 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2463 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2464 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2267 int count; 2465 ++i;
2268 for (count = 0; count < num_devices; ++count)
2269 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2270 } 2466 }
2271 } 2467 }
2272 2468
2273 if (SvOK (notify)) 2469 void *user_data;
2274 { 2470 program_callback pfn_notify = make_program_callback (notify, &user_data);
2275 NEED_SUCCESS (RetainProgram, (self));
2276 pfn_notify = eq_program_notify;
2277 user_data = SvREFCNT_inc (s_get_cv (notify));
2278 }
2279 2471
2280 if (ix)
2281 build_program_async (self, num_devices, device_list, SvPVbyte_nolen (options), user_data);
2282 else
2283 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
2284 2476
2285void 2477void
2286build_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)
2287 PPCODE: 2479 PPCODE:
2288 size_t size; 2480 size_t size;
2492 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2684 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2493 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2685 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2494 2686
2495 case 'z': nullarg = 1; size = SvIV (sv); break; 2687 case 'z': nullarg = 1; size = SvIV (sv); break;
2496 2688
2497 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2689 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2498 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break; 2690 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2499 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break; 2691 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2500 2692
2501 default: 2693 default:
2502 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2694 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2503 } 2695 }
2504 2696
2704 const int i = 0; 2896 const int i = 0;
2705 PUSHs (sv_2mortal (newSVuv (value [i]))); 2897 PUSHs (sv_2mortal (newSVuv (value [i])));
2706 2898
2707#END:kernel_work_group 2899#END:kernel_work_group
2708 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
2709MODULE = OpenCL PACKAGE = OpenCL::Event 2962MODULE = OpenCL PACKAGE = OpenCL::Event
2710 2963
2711void 2964void
2712DESTROY (OpenCL::Event self) 2965DESTROY (OpenCL::Event self)
2713 CODE: 2966 CODE:
2822 Safefree (mapped); 3075 Safefree (mapped);
2823 3076
2824void 3077void
2825unmap (OpenCL::Mapped self, ...) 3078unmap (OpenCL::Mapped self, ...)
2826 CODE: 3079 CODE:
2827 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1); 3080 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2828 3081
2829bool 3082bool
2830mapped (OpenCL::Mapped self) 3083mapped (OpenCL::Mapped self)
2831 CODE: 3084 CODE:
2832 RETVAL = !!self->ptr; 3085 RETVAL = !!self->ptr;

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines