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 1.74 by root, Sat May 5 13:56:00 2012 UTC

28#if PREFER_1_1 28#if PREFER_1_1
29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS 29 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
30#endif 30#endif
31 31
32#ifdef __APPLE__ 32#ifdef __APPLE__
33 #include <OpenCL/opencl.h> 33 #define CLHDR(name) <OpenCL/name>
34#else 34#else
35 #define CLHDR(name) <CL/name>
36#endif
37
35 #include <CL/opencl.h> 38#include CLHDR(opencl.h)
39
40#ifndef CL_VERSION_1_2
41 #include CLHDR(cl_d3d9.h)
42#endif
43
44#if _WIN32
45 #include CLHDR(cl_d3d10.h)
46 #if CL_VERSION_1_2
47 #include CLHDR<cl_d3d11.h>
48 #endif
49 #include CLHDR<cl_dx9_media_sharing.h.h>
36#endif 50#endif
37 51
38#ifndef CL_VERSION_1_2 52#ifndef CL_VERSION_1_2
39 #undef PREFER_1_1 53 #undef PREFER_1_1
40 #define PREFER_1_1 1 54 #define PREFER_1_1 1
41#endif 55#endif
42 56
57// make sure all constants we might use are actually defined
58#include "default.h"
59
43typedef cl_platform_id OpenCL__Platform; 60typedef cl_platform_id OpenCL__Platform;
44typedef cl_device_id OpenCL__Device; 61typedef cl_device_id OpenCL__Device;
62typedef cl_device_id OpenCL__SubDevice;
45typedef cl_context OpenCL__Context; 63typedef cl_context OpenCL__Context;
46typedef cl_command_queue OpenCL__Queue; 64typedef cl_command_queue OpenCL__Queue;
47typedef cl_mem OpenCL__Memory; 65typedef cl_mem OpenCL__Memory;
48typedef cl_mem OpenCL__Buffer; 66typedef cl_mem OpenCL__Buffer;
49typedef cl_mem OpenCL__BufferObj; 67typedef cl_mem OpenCL__BufferObj;
57typedef cl_event OpenCL__Event; 75typedef cl_event OpenCL__Event;
58typedef cl_event OpenCL__UserEvent; 76typedef cl_event OpenCL__UserEvent;
59 77
60typedef struct mapped * OpenCL__Mapped; 78typedef struct mapped * OpenCL__Mapped;
61 79
62typedef SV *FUTURE;
63
64static HV 80static HV
65 *stash_platform, 81 *stash_platform,
66 *stash_device, 82 *stash_device,
83 *stash_subdevice,
67 *stash_context, 84 *stash_context,
68 *stash_queue, 85 *stash_queue,
69 *stash_program, 86 *stash_program,
70 *stash_kernel, 87 *stash_kernel,
71 *stash_sampler, 88 *stash_sampler,
151 } 168 }
152 169
153 return buf [idx]; 170 return buf [idx];
154} 171}
155 172
173static const char * ecb_noinline
174cv_get_name (CV *cv)
175{
176 static char fullname [256];
177
178 GV *gv = CvGV (cv); // gv better be nonzero
179
180 HV *stash = GvSTASH (gv);
181 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
182 const char *gvname = GvNAME (gv);
183
184 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
185 return fullname;
186}
187
156/*****************************************************************************/ 188/*****************************************************************************/
157 189
158typedef struct 190typedef struct
159{ 191{
160 IV iv; 192 IV iv;
217 if (res) \ 249 if (res) \
218 FAIL (name); 250 FAIL (name);
219 251
220/*****************************************************************************/ 252/*****************************************************************************/
221 253
254static SV *
255new_clobj (HV *stash, IV id)
256{
257 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
258}
259
260#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
261#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
262
263/* cl objects are either \$iv, or [$iv, ...] */
264/* they can be upgraded at runtime to the array form */
265static void * ecb_noinline
266SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
267{
268 // sv_derived_from is quite slow :(
269 if (SvROK (sv) && sv_derived_from (sv, pkg))
270 return (void *)SvIV (SvRV (sv));
271
272 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
273}
274
275// the "no-inherit" version of the above
276static void * ecb_noinline
277SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
278{
279 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
280 return (void *)SvIV (SvRV (sv));
281
282 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
283}
284
285/*****************************************************************************/
286
222static cl_context_properties * ecb_noinline 287static cl_context_properties * ecb_noinline
223SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 288SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
224{ 289{
225 if (!sv || !SvOK (sv)) 290 if (!sv || !SvOK (sv))
226 if (extra) 291 if (extra)
227 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 292 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
228 else 293 else
234 int i, len = av_len (av) + 1; 299 int i, len = av_len (av) + 1;
235 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 300 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
236 cl_context_properties *l = p; 301 cl_context_properties *l = p;
237 302
238 if (len & 1) 303 if (len & 1)
239 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 304 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
240 305
241 while (extracount--) 306 while (extracount--)
242 *l++ = *extra++; 307 *l++ = *extra++;
243 308
244 for (i = 0; i < len; i += 2) 309 for (i = 0; i < len; i += 2)
247 SV *p_sv = *av_fetch (av, i + 1, 0); 312 SV *p_sv = *av_fetch (av, i + 1, 0);
248 cl_context_properties v = SvIV (p_sv); // code below can override 313 cl_context_properties v = SvIV (p_sv); // code below can override
249 314
250 switch (t) 315 switch (t)
251 { 316 {
317 case CL_CONTEXT_PLATFORM:
318 if (SvROK (p_sv))
319 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
320 break;
321
252 case CL_GLX_DISPLAY_KHR: 322 case CL_GLX_DISPLAY_KHR:
253 if (!SvOK (p_sv)) 323 if (!SvOK (p_sv))
254 { 324 {
255 void *func = glsym ("_glXGetCurrentDisplay"); 325 void *func = glsym ("_glXGetCurrentDisplay");
256 if (func) 326 if (func)
279 *l = 0; 349 *l = 0;
280 350
281 return p; 351 return p;
282 } 352 }
283 353
284 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 354 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
285} 355}
286 356
287/*****************************************************************************/ 357// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
288 358// 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 359static void * ecb_noinline
301SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 360object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
302{ 361{
303 // sv_derived_from is quite slow :( 362 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
304 if (SvROK (sv) && sv_derived_from (sv, pkg)) 363 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
305 return (void *)SvIV (SvRV (sv)); 364 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
306 365
307 croak ("%s: %s is not of type %s", func, svname, pkg); 366 AV *av = (AV *)SvRV (arg);
308} 367 void **list = 0;
368 cl_uint count = av_len (av) + 1;
309 369
310// the "no-inherit" version of the above 370 if (count)
311static void * ecb_noinline 371 {
312SvCLOBJ_ni (const char *func, const char *svname, SV *sv, HV *stash) 372 list = tmpbuf (sizeof (*list) * count);
313{ 373 int i;
314 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash) 374 for (i = 0; i < count; ++i)
315 return (void *)SvIV (SvRV (sv)); 375 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
376 }
316 377
317 croak ("%s: %s is not of type %s", func, svname, HvNAME (stash)); 378 if (!count && !or_undef)
379 croak ("%s: '%s' must contain at least one %s object",
380 cv_get_name (cv), argname, klass);
381
382 *rcount = count;
383 return (void *)list;
318} 384}
319 385
320/*****************************************************************************/ 386/*****************************************************************************/
321/* callback stuff */ 387/* callback stuff */
322 388
482static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 548static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
483 549
484static void CL_CALLBACK 550static void CL_CALLBACK
485eq_program_notify (cl_program program, void *user_data) 551eq_program_notify (cl_program program, void *user_data)
486{ 552{
553 clRetainProgram (program);
554
487 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0); 555 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
556}
557
558typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
559
560static program_callback ecb_noinline
561make_program_callback (SV *notify, void **ruser_data)
562{
563 if (SvOK (notify))
564 {
565 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
566 return eq_program_notify;
567 }
568 else
569 {
570 *ruser_data = 0;
571 return 0;
572 }
488} 573}
489 574
490struct build_args 575struct build_args
491{ 576{
492 cl_program program; 577 cl_program program;
558 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 643 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
559 return res; 644 return res;
560} 645}
561 646
562static cl_event * ecb_noinline 647static cl_event * ecb_noinline
563event_list (SV **items, cl_uint *rcount, cl_event extra) 648event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
564{ 649{
565 cl_uint count = *rcount; 650 cl_uint count = *rcount;
566 651
567 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed 652 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
568 *rcount = count = 0; 653 *rcount = count = 0;
573 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1)); 658 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
574 int i = 0; 659 int i = 0;
575 660
576 while (count--) 661 while (count--)
577 if (SvOK (items [count])) 662 if (SvOK (items [count]))
578 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 663 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
579 664
580 if (extra) 665 if (extra)
581 list [i++] = extra; 666 list [i++] = extra;
582 667
583 *rcount = i; 668 *rcount = i;
585 return i ? list : 0; 670 return i ? list : 0;
586} 671}
587 672
588#define EVENT_LIST(skip) \ 673#define EVENT_LIST(skip) \
589 cl_uint event_list_count = items - (skip); \ 674 cl_uint event_list_count = items - (skip); \
590 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0) 675 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
591 676
592#define INFO(class) \ 677#define INFO(class) \
593{ \ 678{ \
594 size_t size; \ 679 size_t size; \
595 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 680 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
677 762
678 mapped->ptr = 0; 763 mapped->ptr = 0;
679} 764}
680 765
681static void 766static void
682mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count) 767mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
683{ 768{
684 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event); 769 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
685 cl_event ev; 770 cl_event ev;
686 771
687 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev)); 772 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
688 773
689 clReleaseEvent (mapped->event); 774 clReleaseEvent (mapped->event);
722 { sizeof (cl_long ), "SIZEOF_LONG" }, 807 { sizeof (cl_long ), "SIZEOF_LONG" },
723 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 808 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
724 { sizeof (cl_half ), "SIZEOF_HALF" }, 809 { sizeof (cl_half ), "SIZEOF_HALF" },
725 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 810 { sizeof (cl_float ), "SIZEOF_FLOAT" },
726 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 811 { sizeof (cl_double), "SIZEOF_DOUBLE" },
812 { PREFER_1_1 , "PREFER_1_1" },
727#include "constiv.h" 813#include "constiv.h"
728 }; 814 };
729 815
730 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 816 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)); 817 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
732 818
733 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD); 819 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
734 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD); 820 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
821 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
735 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD); 822 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
736 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD); 823 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
737 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD); 824 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
738 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD); 825 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
739 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD); 826 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
780 list = tmpbuf (sizeof (*list) * count); 867 list = tmpbuf (sizeof (*list) * count);
781 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 868 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
782 869
783 EXTEND (SP, count); 870 EXTEND (SP, count);
784 for (i = 0; i < count; ++i) 871 for (i = 0; i < count; ++i)
785 PUSH_CLOBJ (stash_platform, (IV)list [i]); 872 PUSH_CLOBJ (stash_platform, list [i]);
786 873
787void 874void
788context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 875context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
789 PPCODE: 876 PPCODE:
790 CONTEXT_NOTIFY_CALLBACK; 877 CONTEXT_NOTIFY_CALLBACK;
791 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res)); 878 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
792 XPUSH_CLOBJ_CONTEXT; 879 XPUSH_CLOBJ_CONTEXT;
793 880
794void 881void
795context (FUTURE properties, FUTURE devices, FUTURE notify) 882context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
796 PPCODE: 883 PPCODE:
797 /* der Gipfel der Kunst */ 884 cl_uint device_count;
885 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
886
887 CONTEXT_NOTIFY_CALLBACK;
888 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
889 XPUSH_CLOBJ_CONTEXT;
798 890
799void 891void
800wait_for_events (...) 892wait_for_events (...)
801 CODE: 893 CODE:
802 EVENT_LIST (0); 894 EVENT_LIST (0);
853 EXTEND (SP, count); 945 EXTEND (SP, count);
854 for (i = 0; i < count; ++i) 946 for (i = 0; i < count; ++i)
855 PUSH_CLOBJ (stash_device, list [i]); 947 PUSH_CLOBJ (stash_device, list [i]);
856 948
857void 949void
858context (OpenCL::Platform self, SV *properties = 0, SV *devices, SV *notify = &PL_sv_undef) 950context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
859 PPCODE: 951 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 }; 952 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
864 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context", "properties", properties, extra, 2); 953 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
865 954
866 AV *av = (AV *)SvRV (devices); 955 cl_uint device_count;
867 cl_uint num_devices = av_len (av) + 1; 956 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");
873 957
874 CONTEXT_NOTIFY_CALLBACK; 958 CONTEXT_NOTIFY_CALLBACK;
875 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, num_devices, device_list, pfn_notify, user_data, &res)); 959 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
876 XPUSH_CLOBJ_CONTEXT; 960 XPUSH_CLOBJ_CONTEXT;
877 961
878void 962void
879context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 963context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
880 PPCODE: 964 PPCODE:
881 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self }; 965 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); 966 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
883 967
884 CONTEXT_NOTIFY_CALLBACK; 968 CONTEXT_NOTIFY_CALLBACK;
885 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res)); 969 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
886 XPUSH_CLOBJ_CONTEXT; 970 XPUSH_CLOBJ_CONTEXT;
887 971
889 973
890void 974void
891info (OpenCL::Device self, cl_device_info name) 975info (OpenCL::Device self, cl_device_info name)
892 PPCODE: 976 PPCODE:
893 INFO (Device) 977 INFO (Device)
978
979#if CL_VERSION_1_2
980
981void
982sub_devices (OpenCL::Device self, SV *properties)
983 PPCODE:
984 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
985 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
986
987 properties = SvRV (properties);
988
989 cl_uint count = av_len ((AV *)properties) + 1;
990 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
991
992 int i;
993 for (i = 0; i < count; ++i)
994 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
995
996 props [count] = 0;
997
998 cl_uint num_devices;
999 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1000 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1001 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
1002
1003 EXTEND (SP, num_devices);
1004 for (i = 0; i < count; ++i)
1005 PUSH_CLOBJ (stash_subdevice, list [i]);
1006
1007#endif
894 1008
895#BEGIN:device 1009#BEGIN:device
896 1010
897void 1011void
898type (OpenCL::Device self) 1012type (OpenCL::Device self)
1114 for (i = 0; i < n; ++i) 1228 for (i = 0; i < n; ++i)
1115 PUSHs (sv_2mortal (newSVuv (value [i]))); 1229 PUSHs (sv_2mortal (newSVuv (value [i])));
1116 1230
1117#END:device 1231#END:device
1118 1232
1233MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1234
1235#if CL_VERSION_1_2
1236
1237void
1238DESTROY (OpenCL::SubDevice self)
1239 CODE:
1240 clReleaseDevice (self);
1241
1242#endif
1243
1119MODULE = OpenCL PACKAGE = OpenCL::Context 1244MODULE = OpenCL PACKAGE = OpenCL::Context
1120 1245
1121void 1246void
1122DESTROY (OpenCL::Context context) 1247DESTROY (OpenCL::Context self)
1123 CODE: 1248 CODE:
1124 clReleaseContext (context); 1249 clReleaseContext (self);
1125 1250
1126void 1251void
1127info (OpenCL::Context self, cl_context_info name) 1252info (OpenCL::Context self, cl_context_info name)
1128 PPCODE: 1253 PPCODE:
1129 INFO (Context) 1254 INFO (Context)
1170 const cl_image_desc desc = { 1295 const cl_image_desc desc = {
1171 type, 1296 type,
1172 width, height, depth, 1297 width, height, depth,
1173 array_size, row_pitch, slice_pitch, 1298 array_size, row_pitch, slice_pitch,
1174 num_mip_level, num_samples, 1299 num_mip_level, num_samples,
1175 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1300 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1176 }; 1301 };
1177 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1302 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1178 HV *stash = stash_image; 1303 HV *stash = stash_image;
1179 switch (type) 1304 switch (type)
1180 { 1305 {
1316 XPUSH_CLOBJ (stash_program, prog); 1441 XPUSH_CLOBJ (stash_program, prog);
1317 1442
1318void 1443void
1319program_with_binary (OpenCL::Context self, SV *devices, SV *binaries) 1444program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1320 PPCODE: 1445 PPCODE:
1321 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1446 cl_uint device_count;
1322 croak ("OpenCL::Context::program_with_binary: devices must be specified as reference to an array of device objects"); 1447 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1323
1324 devices = SvRV (devices);
1325 1448
1326 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV) 1449 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"); 1450 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1328 1451
1329 binaries = SvRV (binaries); 1452 binaries = SvRV (binaries);
1330 1453
1331 if (av_len ((AV *)devices) != av_len ((AV *)binaries)) 1454 if (device_count != av_len ((AV *)binaries) + 1)
1332 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed"); 1455 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1333 1456
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); 1457 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1337 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * count); 1458 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1338 cl_int *status_list = tmpbuf (sizeof (*status_list) * count); 1459 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1339 1460
1340 int i; 1461 int i;
1341 for (i = 0; i < count; ++i) 1462 for (i = 0; i < device_count; ++i)
1342 { 1463 {
1343 device_list [i] = SvCLOBJ ("OpenCL::Context::program_with_binary", "devices", *av_fetch ((AV *)devices, i, 0), "OpenCL::Device");
1344 STRLEN len; 1464 STRLEN len;
1345 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len); 1465 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1346 length_list [i] = len; 1466 length_list [i] = len;
1347 } 1467 }
1348 1468
1349 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, count, device_list, length_list, binary_list, GIMME_V == G_ARRAY ? status_list : 0, &res)); 1469 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1470 length_list, binary_list,
1471 GIMME_V == G_ARRAY ? status_list : 0, &res));
1350 1472
1351 EXTEND (SP, 2); 1473 EXTEND (SP, 2);
1352 PUSH_CLOBJ (stash_program, prog); 1474 PUSH_CLOBJ (stash_program, prog);
1353 1475
1354 if (GIMME_V == G_ARRAY) 1476 if (GIMME_V == G_ARRAY)
1355 { 1477 {
1356 AV *av = newAV (); 1478 AV *av = newAV ();
1357 PUSHs (sv_2mortal (newRV_noinc ((SV *)av))); 1479 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1358 1480
1359 for (i = count; i--; ) 1481 for (i = device_count; i--; )
1360 av_store (av, i, newSViv (status_list [i])); 1482 av_store (av, i, newSViv (status_list [i]));
1361 } 1483 }
1484
1485#if CL_VERSION_1_2
1486
1487void
1488program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1489 PPCODE:
1490 cl_uint device_count;
1491 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1492
1493 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1494
1495 XPUSH_CLOBJ (stash_program, prog);
1496
1497void
1498link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1499 CODE:
1500 cl_uint device_count = 0;
1501 cl_device_id *device_list = 0;
1502
1503 if (SvOK (devices))
1504 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1505
1506 cl_uint program_count;
1507 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1508
1509 void *user_data;
1510 program_callback pfn_notify = make_program_callback (notify, &user_data);
1511
1512 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1513 program_count, program_list, pfn_notify, user_data, &res));
1514
1515 XPUSH_CLOBJ (stash_program, prog);
1516
1517#endif
1362 1518
1363#BEGIN:context 1519#BEGIN:context
1364 1520
1365void 1521void
1366reference_count (OpenCL::Context self) 1522reference_count (OpenCL::Context self)
1412void 1568void
1413read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1569read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1414 ALIAS: 1570 ALIAS:
1415 enqueue_read_buffer = 0 1571 enqueue_read_buffer = 0
1416 PPCODE: 1572 PPCODE:
1417 cl_event ev = 0;
1418 EVENT_LIST (6); 1573 EVENT_LIST (6);
1419 1574
1420 SvUPGRADE (data, SVt_PV); 1575 SvUPGRADE (data, SVt_PV);
1421 SvGROW (data, len); 1576 SvGROW (data, len);
1422 SvPOK_only (data); 1577 SvPOK_only (data);
1423 SvCUR_set (data, len); 1578 SvCUR_set (data, len);
1579
1580 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)); 1581 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1582
1426 if (ev) 1583 if (ev)
1427 XPUSH_CLOBJ (stash_event, ev); 1584 XPUSH_CLOBJ (stash_event, ev);
1428 1585
1429void 1586void
1430write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1587write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1431 ALIAS: 1588 ALIAS:
1432 enqueue_write_buffer = 0 1589 enqueue_write_buffer = 0
1433 PPCODE: 1590 PPCODE:
1434 cl_event ev = 0; 1591 EVENT_LIST (5);
1592
1435 STRLEN len; 1593 STRLEN len;
1436 char *ptr = SvPVbyte (data, len); 1594 char *ptr = SvPVbyte (data, len);
1437 EVENT_LIST (5);
1438 1595
1596 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)); 1597 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1440 1598
1441 if (ev) 1599 if (ev)
1442 XPUSH_CLOBJ (stash_event, ev); 1600 XPUSH_CLOBJ (stash_event, ev);
1443 1601
1446void 1604void
1447fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1605fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1448 ALIAS: 1606 ALIAS:
1449 enqueue_fill_buffer = 0 1607 enqueue_fill_buffer = 0
1450 PPCODE: 1608 PPCODE:
1451 cl_event ev = 0; 1609 EVENT_LIST (5);
1610
1452 STRLEN len; 1611 STRLEN len;
1453 char *ptr = SvPVbyte (data, len); 1612 char *ptr = SvPVbyte (data, len);
1454 EVENT_LIST (5);
1455 1613
1614 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)); 1615 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1457 1616
1458 if (ev) 1617 if (ev)
1459 XPUSH_CLOBJ (stash_event, ev); 1618 XPUSH_CLOBJ (stash_event, ev);
1460 1619
1461void 1620void
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, ...) 1621fill_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: 1622 ALIAS:
1464 enqueue_fill_image = 0 1623 enqueue_fill_image = 0
1465 PPCODE: 1624 PPCODE:
1466 cl_event ev = 0; 1625 EVENT_LIST (12);
1626
1467 const size_t origin [3] = { x, y, z }; 1627 const size_t origin [3] = { x, y, z };
1468 const size_t region [3] = { width, height, depth }; 1628 const size_t region [3] = { width, height, depth };
1469 EVENT_LIST (12);
1470 1629
1471 const cl_float c_f [4] = { r, g, b, a }; 1630 const cl_float c_f [4] = { r, g, b, a };
1472 const cl_uint c_u [4] = { r, g, b, a }; 1631 const cl_uint c_u [4] = { r, g, b, a };
1473 const cl_int c_s [4] = { r, g, b, a }; 1632 const cl_int c_s [4] = { r, g, b, a };
1474 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1633 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1477 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1636 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1478 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1637 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) 1638 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."); 1639 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1481 1640
1641 cl_event ev = 0;
1482 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]], 1642 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)); 1643 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1484 1644
1485 if (ev) 1645 if (ev)
1486 XPUSH_CLOBJ (stash_event, ev); 1646 XPUSH_CLOBJ (stash_event, ev);
1490void 1650void
1491copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1651copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1492 ALIAS: 1652 ALIAS:
1493 enqueue_copy_buffer = 0 1653 enqueue_copy_buffer = 0
1494 PPCODE: 1654 PPCODE:
1655 EVENT_LIST (6);
1656
1495 cl_event ev = 0; 1657 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)); 1658 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1499 1659
1500 if (ev) 1660 if (ev)
1501 XPUSH_CLOBJ (stash_event, ev); 1661 XPUSH_CLOBJ (stash_event, ev);
1502 1662
1503void 1663void
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, ...) 1664read_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: 1665 ALIAS:
1506 enqueue_read_buffer_rect = 0 1666 enqueue_read_buffer_rect = 0
1507 PPCODE: 1667 PPCODE:
1508 cl_event ev = 0; 1668 EVENT_LIST (17);
1669
1509 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1670 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 }; 1671 const size_t host_origin[3] = { host_x, host_y, host_z };
1511 const size_t region[3] = { width, height, depth }; 1672 const size_t region[3] = { width, height, depth };
1512 EVENT_LIST (17);
1513 1673
1514 if (!buf_row_pitch) 1674 if (!buf_row_pitch)
1515 buf_row_pitch = region [0]; 1675 buf_row_pitch = region [0];
1516 1676
1517 if (!buf_slice_pitch) 1677 if (!buf_slice_pitch)
1527 1687
1528 SvUPGRADE (data, SVt_PV); 1688 SvUPGRADE (data, SVt_PV);
1529 SvGROW (data, len); 1689 SvGROW (data, len);
1530 SvPOK_only (data); 1690 SvPOK_only (data);
1531 SvCUR_set (data, len); 1691 SvCUR_set (data, len);
1692
1693 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)); 1694 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 1695
1534 if (ev) 1696 if (ev)
1535 XPUSH_CLOBJ (stash_event, ev); 1697 XPUSH_CLOBJ (stash_event, ev);
1536 1698
1537void 1699void
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, ...) 1700write_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: 1701 ALIAS:
1540 enqueue_write_buffer_rect = 0 1702 enqueue_write_buffer_rect = 0
1541 PPCODE: 1703 PPCODE:
1542 cl_event ev = 0; 1704 EVENT_LIST (17);
1705
1543 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1706 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 }; 1707 const size_t host_origin[3] = { host_x, host_y, host_z };
1545 const size_t region[3] = { width, height, depth }; 1708 const size_t region[3] = { width, height, depth };
1546 STRLEN len; 1709 STRLEN len;
1547 char *ptr = SvPVbyte (data, len); 1710 char *ptr = SvPVbyte (data, len);
1548 EVENT_LIST (17);
1549 1711
1550 if (!buf_row_pitch) 1712 if (!buf_row_pitch)
1551 buf_row_pitch = region [0]; 1713 buf_row_pitch = region [0];
1552 1714
1553 if (!buf_slice_pitch) 1715 if (!buf_slice_pitch)
1562 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1724 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
1563 1725
1564 if (len < min_len) 1726 if (len < min_len)
1565 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1727 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1566 1728
1729 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)); 1730 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 1731
1569 if (ev) 1732 if (ev)
1570 XPUSH_CLOBJ (stash_event, ev); 1733 XPUSH_CLOBJ (stash_event, ev);
1571 1734
1572void 1735void
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, ...) 1736copy_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: 1737 ALIAS:
1575 enqueue_copy_buffer_rect = 0 1738 enqueue_copy_buffer_rect = 0
1576 PPCODE: 1739 PPCODE:
1577 cl_event ev = 0; 1740 EVENT_LIST (16);
1741
1578 const size_t src_origin[3] = { src_x, src_y, src_z }; 1742 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 }; 1743 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1580 const size_t region[3] = { width, height, depth }; 1744 const size_t region[3] = { width, height, depth };
1581 EVENT_LIST (16);
1582 1745
1746 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)); 1747 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 1748
1585 if (ev) 1749 if (ev)
1586 XPUSH_CLOBJ (stash_event, ev); 1750 XPUSH_CLOBJ (stash_event, ev);
1587 1751
1588void 1752void
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, ...) 1753read_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: 1754 ALIAS:
1591 enqueue_read_image = 0 1755 enqueue_read_image = 0
1592 PPCODE: 1756 PPCODE:
1593 cl_event ev = 0; 1757 EVENT_LIST (12);
1758
1594 const size_t src_origin[3] = { src_x, src_y, src_z }; 1759 const size_t src_origin[3] = { src_x, src_y, src_z };
1595 const size_t region[3] = { width, height, depth }; 1760 const size_t region[3] = { width, height, depth };
1596 EVENT_LIST (12);
1597 1761
1598 if (!row_pitch) 1762 if (!row_pitch)
1599 row_pitch = img_row_pitch (src); 1763 row_pitch = img_row_pitch (src);
1600 1764
1601 if (depth > 1 && !slice_pitch) 1765 if (depth > 1 && !slice_pitch)
1605 1769
1606 SvUPGRADE (data, SVt_PV); 1770 SvUPGRADE (data, SVt_PV);
1607 SvGROW (data, len); 1771 SvGROW (data, len);
1608 SvPOK_only (data); 1772 SvPOK_only (data);
1609 SvCUR_set (data, len); 1773 SvCUR_set (data, len);
1774
1775 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)); 1776 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 1777
1612 if (ev) 1778 if (ev)
1613 XPUSH_CLOBJ (stash_event, ev); 1779 XPUSH_CLOBJ (stash_event, ev);
1614 1780
1615void 1781void
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, ...) 1782write_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: 1783 ALIAS:
1618 enqueue_write_image = 0 1784 enqueue_write_image = 0
1619 PPCODE: 1785 PPCODE:
1620 cl_event ev = 0; 1786 EVENT_LIST (12);
1787
1621 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1788 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1622 const size_t region[3] = { width, height, depth }; 1789 const size_t region[3] = { width, height, depth };
1623 STRLEN len; 1790 STRLEN len;
1624 char *ptr = SvPVbyte (data, len); 1791 char *ptr = SvPVbyte (data, len);
1625 EVENT_LIST (12);
1626 1792
1627 if (!row_pitch) 1793 if (!row_pitch)
1628 row_pitch = img_row_pitch (dst); 1794 row_pitch = img_row_pitch (dst);
1629 1795
1630 if (depth > 1 && !slice_pitch) 1796 if (depth > 1 && !slice_pitch)
1633 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1799 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1634 1800
1635 if (len < min_len) 1801 if (len < min_len)
1636 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1802 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1637 1803
1804 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)); 1805 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 1806
1640 if (ev) 1807 if (ev)
1641 XPUSH_CLOBJ (stash_event, ev); 1808 XPUSH_CLOBJ (stash_event, ev);
1642 1809
1643void 1810void
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, ...) 1811copy_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: 1812 ALIAS:
1646 enqueue_copy_image = 0 1813 enqueue_copy_image = 0
1647 PPCODE: 1814 PPCODE:
1648 cl_event ev = 0; 1815 EVENT_LIST (12);
1816
1649 const size_t src_origin[3] = { src_x, src_y, src_z }; 1817 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 }; 1818 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1651 const size_t region[3] = { width, height, depth }; 1819 const size_t region[3] = { width, height, depth };
1652 EVENT_LIST (12);
1653 1820
1821 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)); 1822 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1655 1823
1656 if (ev) 1824 if (ev)
1657 XPUSH_CLOBJ (stash_event, ev); 1825 XPUSH_CLOBJ (stash_event, ev);
1658 1826
1659void 1827void
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, ...) 1828copy_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: 1829 ALIAS:
1662 enqueue_copy_image_to_buffer = 0 1830 enqueue_copy_image_to_buffer = 0
1663 PPCODE: 1831 PPCODE:
1664 cl_event ev = 0; 1832 EVENT_LIST (10);
1833
1665 const size_t src_origin[3] = { src_x, src_y, src_z }; 1834 const size_t src_origin[3] = { src_x, src_y, src_z };
1666 const size_t region [3] = { width, height, depth }; 1835 const size_t region [3] = { width, height, depth };
1667 EVENT_LIST (10);
1668 1836
1837 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)); 1838 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1670 1839
1671 if (ev) 1840 if (ev)
1672 XPUSH_CLOBJ (stash_event, ev); 1841 XPUSH_CLOBJ (stash_event, ev);
1673 1842
1674void 1843void
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, ...) 1844copy_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: 1845 ALIAS:
1677 enqueue_copy_buffer_to_image = 0 1846 enqueue_copy_buffer_to_image = 0
1678 PPCODE: 1847 PPCODE:
1679 cl_event ev = 0; 1848 EVENT_LIST (10);
1849
1680 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1850 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1681 const size_t region [3] = { width, height, depth }; 1851 const size_t region [3] = { width, height, depth };
1682 EVENT_LIST (10);
1683 1852
1853 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)); 1854 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1685 1855
1686 if (ev) 1856 if (ev)
1687 XPUSH_CLOBJ (stash_event, ev); 1857 XPUSH_CLOBJ (stash_event, ev);
1688 1858
1689void 1859void
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, ...) 1860map_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: 1861 ALIAS:
1692 enqueue_map_buffer = 0 1862 enqueue_map_buffer = 0
1693 PPCODE: 1863 PPCODE:
1694 cl_event ev;
1695 EVENT_LIST (6); 1864 EVENT_LIST (6);
1865
1696 size_t cb = SvIV (cb_); 1866 size_t cb = SvIV (cb_);
1697 1867
1698 if (!SvOK (cb_)) 1868 if (!SvOK (cb_))
1699 { 1869 {
1700 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0)); 1870 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1701 cb -= offset; 1871 cb -= offset;
1702 } 1872 }
1703 1873
1874 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)); 1875 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)); 1876 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1706 1877
1707void 1878void
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, ...) 1879map_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)); 1920 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1750 1921
1751void 1922void
1752unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...) 1923unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1753 PPCODE: 1924 PPCODE:
1754 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2); 1925 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1755 if (GIMME_V != G_VOID) 1926 if (GIMME_V != G_VOID)
1756 { 1927 {
1757 clRetainEvent (mapped->event); 1928 clRetainEvent (mapped->event);
1758 XPUSH_CLOBJ (stash_event, mapped->event); 1929 XPUSH_CLOBJ (stash_event, mapped->event);
1759 } 1930 }
1761void 1932void
1762task (OpenCL::Queue self, OpenCL::Kernel kernel, ...) 1933task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1763 ALIAS: 1934 ALIAS:
1764 enqueue_task = 0 1935 enqueue_task = 0
1765 PPCODE: 1936 PPCODE:
1937 EVENT_LIST (2);
1938
1766 cl_event ev = 0; 1939 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)); 1940 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1770 1941
1771 if (ev) 1942 if (ev)
1772 XPUSH_CLOBJ (stash_event, ev); 1943 XPUSH_CLOBJ (stash_event, ev);
1773 1944
1774void 1945void
1775nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1946nd_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: 1947 ALIAS:
1777 enqueue_nd_range_kernel = 0 1948 enqueue_nd_range_kernel = 0
1778 PPCODE: 1949 PPCODE:
1779 cl_event ev = 0; 1950 EVENT_LIST (5);
1951
1780 size_t *gwo = 0, *gws, *lws = 0; 1952 size_t *gwo = 0, *gws, *lws = 0;
1781 int gws_len; 1953 int gws_len;
1782 size_t *lists; 1954 size_t *lists;
1783 int i; 1955 int i;
1784 EVENT_LIST (5);
1785 1956
1786 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1957 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1787 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1958 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1788 1959
1789 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1960 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1828 if (!lws [i]) 1999 if (!lws [i])
1829 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i); 2000 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1830 } 2001 }
1831 } 2002 }
1832 2003
2004 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)); 2005 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834 2006
1835 if (ev) 2007 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev); 2008 XPUSH_CLOBJ (stash_event, ev);
2009
2010#if CL_VERSION_1_2
2011
2012void
2013migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2014 ALIAS:
2015 enqueue_migrate_mem_objects = 0
2016 PPCODE:
2017 EVENT_LIST (3);
2018
2019 cl_uint obj_count;
2020 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2021
2022 cl_event ev = 0;
2023 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2024
2025 if (ev)
2026 XPUSH_CLOBJ (stash_event, ev);
2027
2028#endif
1837 2029
1838#if cl_apple_gl_sharing || cl_khr_gl_sharing 2030#if cl_apple_gl_sharing || cl_khr_gl_sharing
1839 2031
1840void 2032void
1841acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2033acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1842 ALIAS: 2034 ALIAS:
1843 release_gl_objects = 1 2035 release_gl_objects = 1
1844 enqueue_acquire_gl_objects = 0 2036 enqueue_acquire_gl_objects = 0
1845 enqueue_release_gl_objects = 1 2037 enqueue_release_gl_objects = 1
1846 PPCODE: 2038 PPCODE:
1847 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2039 EVENT_LIST (2);
1848 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2040
2041 cl_uint obj_count;
2042 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1849 2043
1850 cl_event ev = 0; 2044 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 2045
1860 if (ix) 2046 if (ix)
1861 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2047 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1862 else 2048 else
1863 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2049 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1864 2050
1865 if (ev) 2051 if (ev)
1866 XPUSH_CLOBJ (stash_event, ev); 2052 XPUSH_CLOBJ (stash_event, ev);
1867 2053
1868#endif 2054#endif
1882void 2068void
1883marker (OpenCL::Queue self, ...) 2069marker (OpenCL::Queue self, ...)
1884 ALIAS: 2070 ALIAS:
1885 enqueue_marker = 0 2071 enqueue_marker = 0
1886 PPCODE: 2072 PPCODE:
2073 EVENT_LIST (1);
1887 cl_event ev = 0; 2074 cl_event ev = 0;
1888 EVENT_LIST (1);
1889#if PREFER_1_1 2075#if PREFER_1_1
1890 if (!event_list_count) 2076 if (!event_list_count)
1891 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 2077 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1892 else 2078 else
1893#if CL_VERSION_1_2 2079#if CL_VERSION_1_2
1907void 2093void
1908barrier (OpenCL::Queue self, ...) 2094barrier (OpenCL::Queue self, ...)
1909 ALIAS: 2095 ALIAS:
1910 enqueue_barrier = 0 2096 enqueue_barrier = 0
1911 PPCODE: 2097 PPCODE:
2098 EVENT_LIST (1);
1912 cl_event ev = 0; 2099 cl_event ev = 0;
1913 EVENT_LIST (1);
1914#if PREFER_1_1 2100#if PREFER_1_1
1915 if (!event_list_count && GIMME_V == G_VOID) 2101 if (!event_list_count && GIMME_V == G_VOID)
1916 NEED_SUCCESS (EnqueueBarrier, (self)); 2102 NEED_SUCCESS (EnqueueBarrier, (self));
1917 else 2103 else
1918#if CL_VERSION_1_2 2104#if CL_VERSION_1_2
2246void 2432void
2247build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef) 2433build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2248 ALIAS: 2434 ALIAS:
2249 build_async = 1 2435 build_async = 1
2250 CODE: 2436 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; 2437 cl_uint device_count = 0;
2254 cl_device_id *device_list = 0; 2438 cl_device_id *device_list = 0;
2255 2439
2256 if (SvOK (devices)) 2440 if (SvOK (devices))
2257 { 2441 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2442
2443 void *user_data;
2444 program_callback pfn_notify = make_program_callback (notify, &user_data);
2445
2446 if (ix)
2447 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2448 else
2449 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2450
2451#if CL_VERSION_1_2
2452
2453void
2454compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2455 CODE:
2456 cl_uint device_count = 0;
2457 cl_device_id *device_list = 0;
2458
2459 if (SvOK (devices))
2460 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2461
2462 cl_uint header_count = 0;
2463 cl_program *header_list = 0;
2464 const char **header_name = 0;
2465
2466 if (SvOK (headers))
2467 {
2258 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 2468 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2259 croak ("clProgramBuild: devices must be undef or an array of OpenCL::Device objects."); 2469 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2260 2470
2261 AV *av = (AV *)SvRV (devices); 2471 HV *hv = (HV *)SvRV (devices);
2262 num_devices = av_len (av) + 1;
2263 2472
2264 if (num_devices) 2473 header_count = hv_iterinit (hv);
2474 header_list = tmpbuf (sizeof (*header_list) * header_count);
2475 header_name = tmpbuf (sizeof (*header_name) * header_count);
2476
2477 HE *he;
2478 int i = 0;
2479 while (he = hv_iternext (hv))
2265 { 2480 {
2266 device_list = tmpbuf (sizeof (*device_list) * num_devices); 2481 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2482 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2267 int count; 2483 ++i;
2268 for (count = 0; count < num_devices; ++count)
2269 device_list [count] = SvCLOBJ ("clBuildProgram", "devices", *av_fetch (av, count, 1), "OpenCL::Device");
2270 } 2484 }
2271 } 2485 }
2272 2486
2273 if (SvOK (notify)) 2487 void *user_data;
2274 { 2488 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 2489
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)); 2490 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2491 header_count, header_list, header_name, pfn_notify, user_data));
2492
2493#endif
2284 2494
2285void 2495void
2286build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2496build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
2287 PPCODE: 2497 PPCODE:
2288 size_t size; 2498 size_t size;
2316 char *value = tmpbuf (size); 2526 char *value = tmpbuf (size);
2317 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2527 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
2318 EXTEND (SP, 1); 2528 EXTEND (SP, 1);
2319 const int i = 0; 2529 const int i = 0;
2320 PUSHs (sv_2mortal (newSVpv (value, 0))); 2530 PUSHs (sv_2mortal (newSVpv (value, 0)));
2531
2532void
2533binary_type (OpenCL::Program self, OpenCL::Device device)
2534 PPCODE:
2535 cl_program_binary_type value [1];
2536 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2537 EXTEND (SP, 1);
2538 const int i = 0;
2539 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2321 2540
2322#END:program_build 2541#END:program_build
2323 2542
2324void 2543void
2325kernel (OpenCL::Program program, SV *function) 2544kernel (OpenCL::Program program, SV *function)
2492 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break; 2711 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2493 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break; 2712 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2494 2713
2495 case 'z': nullarg = 1; size = SvIV (sv); break; 2714 case 'z': nullarg = 1; size = SvIV (sv); break;
2496 2715
2497 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break; 2716 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; 2717 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; 2718 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2500 2719
2501 default: 2720 default:
2502 croak ("OpenCL::Kernel::setf format character '%c' not supported", type); 2721 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2503 } 2722 }
2504 2723
2704 const int i = 0; 2923 const int i = 0;
2705 PUSHs (sv_2mortal (newSVuv (value [i]))); 2924 PUSHs (sv_2mortal (newSVuv (value [i])));
2706 2925
2707#END:kernel_work_group 2926#END:kernel_work_group
2708 2927
2928#if CL_VERSION_1_2
2929
2930void
2931arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2932 PPCODE:
2933 size_t size;
2934 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2935 SV *sv = sv_2mortal (newSV (size));
2936 SvUPGRADE (sv, SVt_PV);
2937 SvPOK_only (sv);
2938 SvCUR_set (sv, size);
2939 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2940 XPUSHs (sv);
2941
2942#BEGIN:kernel_arg
2943
2944void
2945arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2946 PPCODE:
2947 cl_kernel_arg_address_qualifier value [1];
2948 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ADDRESS_QUALIFIER, sizeof (value), value, 0));
2949 EXTEND (SP, 1);
2950 const int i = 0;
2951 PUSHs (sv_2mortal (newSVuv (value [i])));
2952
2953void
2954arg_access_qualifier (OpenCL::Kernel self, cl_uint idx)
2955 PPCODE:
2956 cl_kernel_arg_access_qualifier value [1];
2957 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_ACCESS_QUALIFIER, sizeof (value), value, 0));
2958 EXTEND (SP, 1);
2959 const int i = 0;
2960 PUSHs (sv_2mortal (newSVuv (value [i])));
2961
2962void
2963arg_type_name (OpenCL::Kernel self, cl_uint idx)
2964 ALIAS:
2965 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2966 arg_name = CL_KERNEL_ARG_NAME
2967 PPCODE:
2968 size_t size;
2969 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2970 char *value = tmpbuf (size);
2971 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2972 EXTEND (SP, 1);
2973 const int i = 0;
2974 PUSHs (sv_2mortal (newSVpv (value, 0)));
2975
2976void
2977arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2978 PPCODE:
2979 cl_kernel_arg_type_qualifier value [1];
2980 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2981 EXTEND (SP, 1);
2982 const int i = 0;
2983 PUSHs (sv_2mortal (newSVuv (value [i])));
2984
2985#END:kernel_arg
2986
2987#endif
2988
2709MODULE = OpenCL PACKAGE = OpenCL::Event 2989MODULE = OpenCL PACKAGE = OpenCL::Event
2710 2990
2711void 2991void
2712DESTROY (OpenCL::Event self) 2992DESTROY (OpenCL::Event self)
2713 CODE: 2993 CODE:
2822 Safefree (mapped); 3102 Safefree (mapped);
2823 3103
2824void 3104void
2825unmap (OpenCL::Mapped self, ...) 3105unmap (OpenCL::Mapped self, ...)
2826 CODE: 3106 CODE:
2827 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1); 3107 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
2828 3108
2829bool 3109bool
2830mapped (OpenCL::Mapped self) 3110mapped (OpenCL::Mapped self)
2831 CODE: 3111 CODE:
2832 RETVAL = !!self->ptr; 3112 RETVAL = !!self->ptr;

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines