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

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.53 by root, Wed Apr 25 00:02:25 2012 UTC vs.
Revision 1.62 by root, Tue May 1 22:04:56 2012 UTC

57typedef cl_program OpenCL__Program; 57typedef cl_program OpenCL__Program;
58typedef cl_kernel OpenCL__Kernel; 58typedef cl_kernel OpenCL__Kernel;
59typedef cl_event OpenCL__Event; 59typedef cl_event OpenCL__Event;
60typedef cl_event OpenCL__UserEvent; 60typedef cl_event OpenCL__UserEvent;
61 61
62typedef struct mapped * OpenCL__Mapped;
63
62typedef SV *FUTURE; 64typedef SV *FUTURE;
65
66static HV
67 *stash_platform,
68 *stash_device,
69 *stash_context,
70 *stash_queue,
71 *stash_program,
72 *stash_kernel,
73 *stash_sampler,
74 *stash_event,
75 *stash_userevent,
76 *stash_memory,
77 *stash_buffer,
78 *stash_bufferobj,
79 *stash_image,
80 *stash_image1d,
81 *stash_image1darray,
82 *stash_image1dbuffer,
83 *stash_image2d,
84 *stash_image2darray,
85 *stash_image3d,
86 *stash_mapped,
87 *stash_mappedbuffer,
88 *stash_mappedimage;
63 89
64/*****************************************************************************/ 90/*****************************************************************************/
65 91
66// name must include a leading underscore 92// name must include a leading underscore
67// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 93// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
260 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 286 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname);
261} 287}
262 288
263/*****************************************************************************/ 289/*****************************************************************************/
264 290
265#define NEW_CLOBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr) 291static SV *
266#define PUSH_CLOBJ(class,ptr) PUSHs (NEW_CLOBJ (class, ptr)) 292new_clobj (HV *stash, IV id)
267#define XPUSH_CLOBJ(class,ptr) XPUSHs (NEW_CLOBJ (class, ptr)) 293{
294 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
295}
296
297#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
298#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
268 299
269/* cl objects are either \$iv, or [$iv, ...] */ 300/* cl objects are either \$iv, or [$iv, ...] */
270/* they can be upgraded at runtime to the array form */ 301/* they can be upgraded at runtime to the array form */
271static void * 302static void *
272SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg) 303SvCLOBJ (const char *func, const char *svname, SV *sv, const char *pkg)
416 } 447 }
417 448
418static SV * 449static SV *
419new_clobj_context (cl_context ctx, void *user_data) 450new_clobj_context (cl_context ctx, void *user_data)
420{ 451{
421 SV *sv = NEW_CLOBJ ("OpenCL::Context", ctx); 452 SV *sv = new_clobj (stash_context, (IV)ctx);
422 453
423 if (user_data) 454 if (user_data)
424 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0); 455 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
425 456
426 return sv; 457 return sv;
433 464
434static void 465static void
435eq_program_push (void *data1, void *data2, void *data3) 466eq_program_push (void *data1, void *data2, void *data3)
436{ 467{
437 dSP; 468 dSP;
438 PUSH_CLOBJ ("OpenCL::Program", data1); 469 PUSH_CLOBJ (stash_program, data1);
439 PUTBACK; 470 PUTBACK;
440} 471}
441 472
442static eq_vtbl eq_program_vtbl = { 1, eq_program_push }; 473static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
443 474
490 521
491static void 522static void
492eq_event_push (void *data1, void *data2, void *data3) 523eq_event_push (void *data1, void *data2, void *data3)
493{ 524{
494 dSP; 525 dSP;
495 PUSH_CLOBJ ("OpenCL::Event", data1); 526 PUSH_CLOBJ (stash_event, data1);
496 PUSHs (sv_2mortal (newSViv ((IV)data2))); 527 PUSHs (sv_2mortal (newSViv ((IV)data2)));
497 PUTBACK; 528 PUTBACK;
498} 529}
499 530
500static eq_vtbl eq_event_vtbl = { 1, eq_event_push }; 531static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
505 clRetainEvent (event); 536 clRetainEvent (event);
506 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0); 537 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
507} 538}
508 539
509/*****************************************************************************/ 540/*****************************************************************************/
541/* utilities for XS code */
510 542
511static size_t 543static size_t
512img_row_pitch (cl_mem img) 544img_row_pitch (cl_mem img)
513{ 545{
514 size_t res; 546 size_t res;
515 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 547 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
516 return res; 548 return res;
517} 549}
518 550
519static cl_event * 551static cl_event *
520event_list (SV **items, cl_uint *rcount) 552event_list (SV **items, cl_uint *rcount, cl_event extra)
521{ 553{
522 cl_uint count = *rcount; 554 cl_uint count = *rcount;
523 555
524 if (!count) 556 if (count > 0x7fffffff) // yeha, it's a hack - the caller might have underflowed
557 count = 0;
558
559 if (!count && !extra)
525 return 0; 560 return 0;
526 561
527 cl_event *list = tmpbuf (sizeof (cl_event) * count); 562 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
528 int i = 0; 563 int i = 0;
529 564
530 do 565 while (count--)
531 {
532 --count;
533 if (SvOK (items [count])) 566 if (SvOK (items [count]))
534 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 567 list [i++] = SvCLOBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event");
535 } 568
536 while (count); 569 if (extra)
570 list [i++] = extra;
537 571
538 *rcount = i; 572 *rcount = i;
539 573
540 return i ? list : 0; 574 return i ? list : 0;
541} 575}
542 576
543#define EVENT_LIST(items,count) \ 577#define EVENT_LIST(skip) \
544 cl_uint event_list_count = (count); \ 578 cl_uint event_list_count = items - (skip); \
545 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 579 cl_event *event_list_ptr = event_list (&ST (skip), &event_list_count, 0)
546 580
547#define INFO(class) \ 581#define INFO(class) \
548{ \ 582{ \
549 size_t size; \ 583 size_t size; \
550 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 584 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
553 SvPOK_only (sv); \ 587 SvPOK_only (sv); \
554 SvCUR_set (sv, size); \ 588 SvCUR_set (sv, size); \
555 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 589 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
556 XPUSHs (sv); \ 590 XPUSHs (sv); \
557} 591}
592
593/*****************************************************************************/
594/* mapped_xxx */
595
596static OpenCL__Mapped
597SvMAPPED (SV *sv)
598{
599 // no typechecking atm., keep your fingers crossed
600 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
601}
602
603struct mapped
604{
605 cl_command_queue queue;
606 cl_mem memobj;
607 void *ptr;
608 size_t cb;
609 cl_event event;
610 size_t row_pitch;
611 size_t slice_pitch;
612};
613
614static SV *
615mapped_new (HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags, void *ptr, size_t cb, cl_event ev, size_t row_pitch, size_t slice_pitch)
616{
617 SV *data = newSV (0);
618 SvUPGRADE (data, SVt_PVMG);
619
620 OpenCL__Mapped mapped;
621 New (0, mapped, 1, struct mapped);
622
623 clRetainCommandQueue (queue);
624
625 mapped->queue = queue;
626 mapped->memobj = memobj;
627 mapped->ptr = ptr;
628 mapped->cb = cb;
629 mapped->event = ev;
630 mapped->row_pitch = row_pitch;
631 mapped->slice_pitch = slice_pitch;
632
633 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
634
635 if (SvLEN (data))
636 Safefree (data);
637
638 SvPVX (data) = (char *)ptr;
639 SvCUR_set (data, cb);
640 SvLEN_set (data, 0);
641 SvPOK_only (data);
642
643 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
644
645 if (!(flags & CL_MAP_WRITE))
646 SvREADONLY_on (data);
647
648 return obj;
649}
650
651static void
652mapped_detach (SV *sv, OpenCL__Mapped mapped)
653{
654 SV *data = SvRV (sv);
655
656 // the next check checks both before AND after detach, where SvPVX should STILL be 0
657 if (SvPVX (data) != (char *)mapped->ptr)
658 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
659 else
660 {
661 SvREADONLY_off (data);
662 SvCUR_set (data, 0);
663 SvPVX (data) = 0;
664 SvOK_off (data);
665 }
666
667 mapped->ptr = 0;
668}
669
670static void
671mapped_unmap (SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
672{
673 cl_event *event_list_ptr = event_list (wait_list, &event_list_count, mapped->event);
674 cl_event ev;
675
676 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
677
678 clReleaseEvent (mapped->event);
679 mapped->event = ev;
680
681 mapped_detach (self, mapped);
682}
683
684/*****************************************************************************/
558 685
559MODULE = OpenCL PACKAGE = OpenCL 686MODULE = OpenCL PACKAGE = OpenCL
560 687
561PROTOTYPES: ENABLE 688PROTOTYPES: ENABLE
562 689
590 }; 717 };
591 718
592 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 719 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
593 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 720 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
594 721
722 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
723 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
724 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
725 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
726 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
727 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
728 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
729 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
730 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
731 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
732 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
733 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
734 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
735 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
736 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
737 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
738 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
739 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
740 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
741 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
742 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
743 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
744
595 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt); 745 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
596} 746}
597 747
598cl_int 748cl_int
599errno () 749errno ()
601 RETVAL = res; 751 RETVAL = res;
602 OUTPUT: 752 OUTPUT:
603 RETVAL 753 RETVAL
604 754
605const char * 755const char *
606err2str (cl_int err) 756err2str (cl_int err = res)
607 757
608const char * 758const char *
609enum2str (cl_uint value) 759enum2str (cl_uint value)
610 760
611void 761void
619 list = tmpbuf (sizeof (*list) * count); 769 list = tmpbuf (sizeof (*list) * count);
620 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 770 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
621 771
622 EXTEND (SP, count); 772 EXTEND (SP, count);
623 for (i = 0; i < count; ++i) 773 for (i = 0; i < count; ++i)
624 PUSH_CLOBJ ("OpenCL::Platform", list [i]); 774 PUSH_CLOBJ (stash_platform, (IV)list [i]);
625 775
626void 776void
627context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef) 777context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
628 PPCODE: 778 PPCODE:
629 CONTEXT_NOTIFY_CALLBACK; 779 CONTEXT_NOTIFY_CALLBACK;
636 /* der Gipfel der Kunst */ 786 /* der Gipfel der Kunst */
637 787
638void 788void
639wait_for_events (...) 789wait_for_events (...)
640 CODE: 790 CODE:
641 EVENT_LIST (0, items); 791 EVENT_LIST (0);
642 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr)); 792 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
643 793
644PROTOTYPES: DISABLE 794PROTOTYPES: DISABLE
645 795
646MODULE = OpenCL PACKAGE = OpenCL::Platform 796MODULE = OpenCL PACKAGE = OpenCL::Platform
689 list = tmpbuf (sizeof (*list) * count); 839 list = tmpbuf (sizeof (*list) * count);
690 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 840 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
691 841
692 EXTEND (SP, count); 842 EXTEND (SP, count);
693 for (i = 0; i < count; ++i) 843 for (i = 0; i < count; ++i)
694 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 844 PUSH_CLOBJ (stash_device, list [i]);
695 845
696void 846void
697context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef) 847context (OpenCL::Platform self, cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
698 PPCODE: 848 PPCODE:
699 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 849 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV)
735 PPCODE: 885 PPCODE:
736 cl_device_type value [1]; 886 cl_device_type value [1];
737 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 887 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0));
738 EXTEND (SP, 1); 888 EXTEND (SP, 1);
739 const int i = 0; 889 const int i = 0;
740 PUSHs (sv_2mortal (newSViv (value [i]))); 890 PUSHs (sv_2mortal (newSVuv (value [i])));
741 891
742void 892void
743vendor_id (OpenCL::Device self) 893vendor_id (OpenCL::Device self)
744 ALIAS: 894 ALIAS:
745 vendor_id = CL_DEVICE_VENDOR_ID 895 vendor_id = CL_DEVICE_VENDOR_ID
890 PPCODE: 1040 PPCODE:
891 cl_command_queue_properties value [1]; 1041 cl_command_queue_properties value [1];
892 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1042 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
893 EXTEND (SP, 1); 1043 EXTEND (SP, 1);
894 const int i = 0; 1044 const int i = 0;
895 PUSHs (sv_2mortal (newSViv (value [i]))); 1045 PUSHs (sv_2mortal (newSVuv (value [i])));
896 1046
897void 1047void
898platform (OpenCL::Device self) 1048platform (OpenCL::Device self)
899 PPCODE: 1049 PPCODE:
900 cl_platform_id value [1]; 1050 cl_platform_id value [1];
901 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1051 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
902 EXTEND (SP, 1); 1052 EXTEND (SP, 1);
903 const int i = 0; 1053 const int i = 0;
904 { 1054 {
905 PUSH_CLOBJ ("OpenCL::Platform", value [i]); 1055 PUSH_CLOBJ (stash_platform, value [i]);
906 } 1056 }
907 1057
908void 1058void
909name (OpenCL::Device self) 1059name (OpenCL::Device self)
910 ALIAS: 1060 ALIAS:
929 cl_device_id value [1]; 1079 cl_device_id value [1];
930 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1080 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
931 EXTEND (SP, 1); 1081 EXTEND (SP, 1);
932 const int i = 0; 1082 const int i = 0;
933 { 1083 {
934 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1084 PUSH_CLOBJ (stash_device, value [i]);
935 } 1085 }
936 1086
937void 1087void
938partition_types_ext (OpenCL::Device self) 1088partition_types_ext (OpenCL::Device self)
939 ALIAS: 1089 ALIAS:
966 1116
967void 1117void
968queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1118queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
969 PPCODE: 1119 PPCODE:
970 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1120 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
971 XPUSH_CLOBJ ("OpenCL::Queue", queue); 1121 XPUSH_CLOBJ (stash_queue, queue);
972 1122
973void 1123void
974user_event (OpenCL::Context self) 1124user_event (OpenCL::Context self)
975 PPCODE: 1125 PPCODE:
976 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1126 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
977 XPUSH_CLOBJ ("OpenCL::UserEvent", ev); 1127 XPUSH_CLOBJ (stash_userevent, ev);
978 1128
979void 1129void
980buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1130buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
981 PPCODE: 1131 PPCODE:
982 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1132 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
983 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1133 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
984 1134
985 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1135 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
986 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1136 XPUSH_CLOBJ (stash_bufferobj, mem);
987 1137
988void 1138void
989buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1139buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
990 PPCODE: 1140 PPCODE:
991 STRLEN len; 1141 STRLEN len;
992 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1142 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
993 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1143 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
994 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1144 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
995 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1145 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
996 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1146 XPUSH_CLOBJ (stash_bufferobj, mem);
997 1147
998#if CL_VERSION_1_2 1148#if CL_VERSION_1_2
999 1149
1000void 1150void
1001image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef) 1151image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
1002 PPCODE: 1152 PPCODE:
1003 STRLEN len; 1153 STRLEN len;
1004 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1154 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1005 const cl_image_format format = { channel_order, channel_type }; 1155 const cl_image_format format = { channel_order, channel_type };
1006 const cl_image_desc desc = { 1156 const cl_image_desc desc = {
1009 array_size, row_pitch, slice_pitch, 1159 array_size, row_pitch, slice_pitch,
1010 num_mip_level, num_samples, 1160 num_mip_level, num_samples,
1011 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0 1161 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ ("OpenCL::Context::Image", "data", data, "OpenCL::Buffer") : 0
1012 }; 1162 };
1013 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1163 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1014 char *klass = "OpenCL::Image"; 1164 HV *stash = stash_image;
1015 switch (type) 1165 switch (type)
1016 { 1166 {
1017 case CL_MEM_OBJECT_IMAGE1D_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1167 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1018 case CL_MEM_OBJECT_IMAGE1D: klass = "OpenCL::Image1D"; break; 1168 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1019 case CL_MEM_OBJECT_IMAGE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1169 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1020 case CL_MEM_OBJECT_IMAGE2D: klass = "OpenCL::Image2D"; break; 1170 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1021 case CL_MEM_OBJECT_IMAGE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1171 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1022 case CL_MEM_OBJECT_IMAGE3D: klass = "OpenCL::Image3D"; break; 1172 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1023 } 1173 }
1024 XPUSH_CLOBJ (klass, mem); 1174 XPUSH_CLOBJ (stash, mem);
1025 1175
1026#endif 1176#endif
1027 1177
1028void 1178void
1029image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1179image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
1035 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1185 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1036#else 1186#else
1037 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 }; 1187 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1038 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1188 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1039#endif 1189#endif
1040 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1190 XPUSH_CLOBJ (stash_image2d, mem);
1041 1191
1042void 1192void
1043image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1193image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1044 PPCODE: 1194 PPCODE:
1045 STRLEN len; 1195 STRLEN len;
1049 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1199 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
1050#else 1200#else
1051 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 }; 1201 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1052 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res)); 1202 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1053#endif 1203#endif
1054 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1204 XPUSH_CLOBJ (stash_image3d, mem);
1055 1205
1056#if cl_apple_gl_sharing || cl_khr_gl_sharing 1206#if cl_apple_gl_sharing || cl_khr_gl_sharing
1057 1207
1058void 1208void
1059gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1209gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
1060 PPCODE: 1210 PPCODE:
1061 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1211 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
1062 XPUSH_CLOBJ ("OpenCL::BufferObj", mem); 1212 XPUSH_CLOBJ (stash_bufferobj, mem);
1063 1213
1064void 1214void
1065gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer) 1215gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1066 PPCODE: 1216 PPCODE:
1067 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res)); 1217 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1068 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1218 XPUSH_CLOBJ (stash_image2d, mem);
1069 1219
1070#if CL_VERSION_1_2 1220#if CL_VERSION_1_2
1071 1221
1072void 1222void
1073gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1223gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1074 ALIAS: 1224 ALIAS:
1075 PPCODE: 1225 PPCODE:
1076 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res)); 1226 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1077 cl_gl_object_type type; 1227 cl_gl_object_type type;
1078 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead? 1228 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1079 char *klass = "OpenCL::Memory"; 1229 HV *stash = stash_memory;
1080 switch (type) 1230 switch (type)
1081 { 1231 {
1082 case CL_GL_OBJECT_TEXTURE_BUFFER: klass = "OpenCL::Image1DBuffer"; break; 1232 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1083 case CL_GL_OBJECT_TEXTURE1D: klass = "OpenCL::Image1D"; break; 1233 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1084 case CL_GL_OBJECT_TEXTURE1D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1234 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1085 case CL_GL_OBJECT_TEXTURE2D: klass = "OpenCL::Image2D"; break; 1235 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1086 case CL_GL_OBJECT_TEXTURE2D_ARRAY: klass = "OpenCL::Image2DArray"; break; 1236 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1087 case CL_GL_OBJECT_TEXTURE3D: klass = "OpenCL::Image3D"; break; 1237 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1088 } 1238 }
1089 XPUSH_CLOBJ (klass, mem); 1239 XPUSH_CLOBJ (stash, mem);
1090 1240
1091#endif 1241#endif
1092 1242
1093void 1243void
1094gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1244gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1096#if PREFER_1_1 1246#if PREFER_1_1
1097 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1247 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
1098#else 1248#else
1099 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1249 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1100#endif 1250#endif
1101 XPUSH_CLOBJ ("OpenCL::Image2D", mem); 1251 XPUSH_CLOBJ (stash_image2d, mem);
1102 1252
1103void 1253void
1104gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1254gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1105 PPCODE: 1255 PPCODE:
1106#if PREFER_1_1 1256#if PREFER_1_1
1107 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1257 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
1108#else 1258#else
1109 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res)); 1259 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1110#endif 1260#endif
1111 XPUSH_CLOBJ ("OpenCL::Image3D", mem); 1261 XPUSH_CLOBJ (stash_image3d, mem);
1112 1262
1113#endif 1263#endif
1114 1264
1115void 1265void
1116supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1266supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
1136 1286
1137void 1287void
1138sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1288sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
1139 PPCODE: 1289 PPCODE:
1140 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1290 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
1141 XPUSH_CLOBJ ("OpenCL::Sampler", sampler); 1291 XPUSH_CLOBJ (stash_sampler, sampler);
1142 1292
1143void 1293void
1144program_with_source (OpenCL::Context self, SV *program) 1294program_with_source (OpenCL::Context self, SV *program)
1145 PPCODE: 1295 PPCODE:
1146 STRLEN len; 1296 STRLEN len;
1147 size_t len2; 1297 size_t len2;
1148 const char *ptr = SvPVbyte (program, len); 1298 const char *ptr = SvPVbyte (program, len);
1149 1299
1150 len2 = len; 1300 len2 = len;
1151 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1301 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
1152 XPUSH_CLOBJ ("OpenCL::Program", prog); 1302 XPUSH_CLOBJ (stash_program, prog);
1153 1303
1154#BEGIN:context 1304#BEGIN:context
1155 1305
1156void 1306void
1157reference_count (OpenCL::Context self) 1307reference_count (OpenCL::Context self)
1174 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1324 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
1175 int i, n = size / sizeof (*value); 1325 int i, n = size / sizeof (*value);
1176 EXTEND (SP, n); 1326 EXTEND (SP, n);
1177 for (i = 0; i < n; ++i) 1327 for (i = 0; i < n; ++i)
1178 { 1328 {
1179 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1329 PUSH_CLOBJ (stash_device, value [i]);
1180 } 1330 }
1181 1331
1182void 1332void
1183properties (OpenCL::Context self) 1333properties (OpenCL::Context self)
1184 PPCODE: 1334 PPCODE:
1199DESTROY (OpenCL::Queue self) 1349DESTROY (OpenCL::Queue self)
1200 CODE: 1350 CODE:
1201 clReleaseCommandQueue (self); 1351 clReleaseCommandQueue (self);
1202 1352
1203void 1353void
1204enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1354read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1355 ALIAS:
1356 enqueue_read_buffer = 0
1205 PPCODE: 1357 PPCODE:
1206 cl_event ev = 0; 1358 cl_event ev = 0;
1207 EVENT_LIST (6, items - 6); 1359 EVENT_LIST (6);
1208 1360
1209 SvUPGRADE (data, SVt_PV); 1361 SvUPGRADE (data, SVt_PV);
1210 SvGROW (data, len); 1362 SvGROW (data, len);
1211 SvPOK_only (data); 1363 SvPOK_only (data);
1212 SvCUR_set (data, len); 1364 SvCUR_set (data, len);
1213 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1365 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1214 1366
1215 if (ev) 1367 if (ev)
1216 XPUSH_CLOBJ ("OpenCL::Event", ev); 1368 XPUSH_CLOBJ (stash_event, ev);
1217 1369
1218void 1370void
1219enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1371write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1372 ALIAS:
1373 enqueue_write_buffer = 0
1220 PPCODE: 1374 PPCODE:
1221 cl_event ev = 0; 1375 cl_event ev = 0;
1222 STRLEN len; 1376 STRLEN len;
1223 char *ptr = SvPVbyte (data, len); 1377 char *ptr = SvPVbyte (data, len);
1224 EVENT_LIST (5, items - 5); 1378 EVENT_LIST (5);
1225 1379
1226 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1380 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1227 1381
1228 if (ev) 1382 if (ev)
1229 XPUSH_CLOBJ ("OpenCL::Event", ev); 1383 XPUSH_CLOBJ (stash_event, ev);
1230 1384
1231#if CL_VERSION_1_2 1385#if CL_VERSION_1_2
1232 1386
1233void 1387void
1234enqueue_fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...) 1388fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1389 ALIAS:
1390 enqueue_fill_buffer = 0
1235 PPCODE: 1391 PPCODE:
1236 cl_event ev = 0; 1392 cl_event ev = 0;
1237 STRLEN len; 1393 STRLEN len;
1238 char *ptr = SvPVbyte (data, len); 1394 char *ptr = SvPVbyte (data, len);
1239 EVENT_LIST (5, items - 5); 1395 EVENT_LIST (5);
1240 1396
1241 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1397 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1242 1398
1243 if (ev) 1399 if (ev)
1244 XPUSH_CLOBJ ("OpenCL::Event", ev); 1400 XPUSH_CLOBJ (stash_event, ev);
1245 1401
1246void 1402void
1247enqueue_fill_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, ...) 1403fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1404 ALIAS:
1405 enqueue_fill_image = 0
1248 PPCODE: 1406 PPCODE:
1249 cl_event ev = 0; 1407 cl_event ev = 0;
1250 STRLEN len; 1408 STRLEN len;
1251 const size_t origin [3] = { x, y, z }; 1409 const size_t origin [3] = { x, y, z };
1252 const size_t region [3] = { width, height, depth }; 1410 const size_t region [3] = { width, height, depth };
1253 EVENT_LIST (12, items - 12); 1411 EVENT_LIST (12);
1254 1412
1255 const cl_float c_f [4] = { r, g, b, a }; 1413 const cl_float c_f [4] = { r, g, b, a };
1256 const cl_uint c_u [4] = { r, g, b, a }; 1414 const cl_uint c_u [4] = { r, g, b, a };
1257 const cl_int c_s [4] = { r, g, b, a }; 1415 const cl_int c_s [4] = { r, g, b, a };
1258 const void *c_fus [3] = { &c_f, &c_u, &c_s }; 1416 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1261 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0)); 1419 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1262 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8); 1420 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1263 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type) 1421 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1264 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported."); 1422 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1265 1423
1266 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type]], 1424 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1267 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1425 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1268 1426
1269 if (ev) 1427 if (ev)
1270 XPUSH_CLOBJ ("OpenCL::Event", ev); 1428 XPUSH_CLOBJ (stash_event, ev);
1271 1429
1272#endif 1430#endif
1273 1431
1274void 1432void
1275enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1433copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1434 ALIAS:
1435 enqueue_copy_buffer = 0
1276 PPCODE: 1436 PPCODE:
1277 cl_event ev = 0; 1437 cl_event ev = 0;
1278 EVENT_LIST (6, items - 6); 1438 EVENT_LIST (6);
1279 1439
1280 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1440 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1281 1441
1282 if (ev) 1442 if (ev)
1283 XPUSH_CLOBJ ("OpenCL::Event", ev); 1443 XPUSH_CLOBJ (stash_event, ev);
1284 1444
1285void 1445void
1286enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1446read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1447 ALIAS:
1448 enqueue_read_buffer_rect = 0
1287 PPCODE: 1449 PPCODE:
1288 cl_event ev = 0; 1450 cl_event ev = 0;
1289 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1451 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1290 const size_t host_origin[3] = { host_x, host_y, host_z }; 1452 const size_t host_origin[3] = { host_x, host_y, host_z };
1291 const size_t region[3] = { width, height, depth }; 1453 const size_t region[3] = { width, height, depth };
1292 EVENT_LIST (17, items - 17); 1454 EVENT_LIST (17);
1293 1455
1294 if (!buf_row_pitch) 1456 if (!buf_row_pitch)
1295 buf_row_pitch = region [0]; 1457 buf_row_pitch = region [0];
1296 1458
1297 if (!buf_slice_pitch) 1459 if (!buf_slice_pitch)
1310 SvPOK_only (data); 1472 SvPOK_only (data);
1311 SvCUR_set (data, len); 1473 SvCUR_set (data, len);
1312 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)); 1474 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1313 1475
1314 if (ev) 1476 if (ev)
1315 XPUSH_CLOBJ ("OpenCL::Event", ev); 1477 XPUSH_CLOBJ (stash_event, ev);
1316 1478
1317void 1479void
1318enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1480write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1481 ALIAS:
1482 enqueue_write_buffer_rect = 0
1319 PPCODE: 1483 PPCODE:
1320 cl_event ev = 0; 1484 cl_event ev = 0;
1321 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1485 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
1322 const size_t host_origin[3] = { host_x, host_y, host_z }; 1486 const size_t host_origin[3] = { host_x, host_y, host_z };
1323 const size_t region[3] = { width, height, depth }; 1487 const size_t region[3] = { width, height, depth };
1324 STRLEN len; 1488 STRLEN len;
1325 char *ptr = SvPVbyte (data, len); 1489 char *ptr = SvPVbyte (data, len);
1326 EVENT_LIST (17, items - 17); 1490 EVENT_LIST (17);
1327 1491
1328 if (!buf_row_pitch) 1492 if (!buf_row_pitch)
1329 buf_row_pitch = region [0]; 1493 buf_row_pitch = region [0];
1330 1494
1331 if (!buf_slice_pitch) 1495 if (!buf_slice_pitch)
1343 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1507 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1344 1508
1345 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)); 1509 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1346 1510
1347 if (ev) 1511 if (ev)
1348 XPUSH_CLOBJ ("OpenCL::Event", ev); 1512 XPUSH_CLOBJ (stash_event, ev);
1349 1513
1350void 1514void
1351enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1515copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1516 ALIAS:
1517 enqueue_copy_buffer_rect = 0
1352 PPCODE: 1518 PPCODE:
1353 cl_event ev = 0; 1519 cl_event ev = 0;
1354 const size_t src_origin[3] = { src_x, src_y, src_z }; 1520 const size_t src_origin[3] = { src_x, src_y, src_z };
1355 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1521 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1356 const size_t region[3] = { width, height, depth }; 1522 const size_t region[3] = { width, height, depth };
1357 EVENT_LIST (16, items - 16); 1523 EVENT_LIST (16);
1358 1524
1359 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)); 1525 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1360 1526
1361 if (ev) 1527 if (ev)
1362 XPUSH_CLOBJ ("OpenCL::Event", ev); 1528 XPUSH_CLOBJ (stash_event, ev);
1363 1529
1364void 1530void
1365enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1531read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1532 ALIAS:
1533 enqueue_read_image = 0
1366 PPCODE: 1534 PPCODE:
1367 cl_event ev = 0; 1535 cl_event ev = 0;
1368 const size_t src_origin[3] = { src_x, src_y, src_z }; 1536 const size_t src_origin[3] = { src_x, src_y, src_z };
1369 const size_t region[3] = { width, height, depth }; 1537 const size_t region[3] = { width, height, depth };
1370 EVENT_LIST (12, items - 12); 1538 EVENT_LIST (12);
1371 1539
1372 if (!row_pitch) 1540 if (!row_pitch)
1373 row_pitch = img_row_pitch (src); 1541 row_pitch = img_row_pitch (src);
1374 1542
1375 if (depth > 1 && !slice_pitch) 1543 if (depth > 1 && !slice_pitch)
1382 SvPOK_only (data); 1550 SvPOK_only (data);
1383 SvCUR_set (data, len); 1551 SvCUR_set (data, len);
1384 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)); 1552 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1385 1553
1386 if (ev) 1554 if (ev)
1387 XPUSH_CLOBJ ("OpenCL::Event", ev); 1555 XPUSH_CLOBJ (stash_event, ev);
1388 1556
1389void 1557void
1390enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1558write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1559 ALIAS:
1560 enqueue_write_image = 0
1391 PPCODE: 1561 PPCODE:
1392 cl_event ev = 0; 1562 cl_event ev = 0;
1393 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1563 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1394 const size_t region[3] = { width, height, depth }; 1564 const size_t region[3] = { width, height, depth };
1395 STRLEN len; 1565 STRLEN len;
1396 char *ptr = SvPVbyte (data, len); 1566 char *ptr = SvPVbyte (data, len);
1397 EVENT_LIST (12, items - 12); 1567 EVENT_LIST (12);
1398 1568
1399 if (!row_pitch) 1569 if (!row_pitch)
1400 row_pitch = img_row_pitch (dst); 1570 row_pitch = img_row_pitch (dst);
1401 1571
1402 if (depth > 1 && !slice_pitch) 1572 if (depth > 1 && !slice_pitch)
1408 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1578 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1409 1579
1410 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)); 1580 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1411 1581
1412 if (ev) 1582 if (ev)
1413 XPUSH_CLOBJ ("OpenCL::Event", ev); 1583 XPUSH_CLOBJ (stash_event, ev);
1414 1584
1415void 1585void
1416enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1586copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1587 ALIAS:
1588 enqueue_copy_image = 0
1417 PPCODE: 1589 PPCODE:
1418 cl_event ev = 0; 1590 cl_event ev = 0;
1419 const size_t src_origin[3] = { src_x, src_y, src_z }; 1591 const size_t src_origin[3] = { src_x, src_y, src_z };
1420 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1592 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1421 const size_t region[3] = { width, height, depth }; 1593 const size_t region[3] = { width, height, depth };
1422 EVENT_LIST (12, items - 12); 1594 EVENT_LIST (12);
1423 1595
1424 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1596 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1425 1597
1426 if (ev) 1598 if (ev)
1427 XPUSH_CLOBJ ("OpenCL::Event", ev); 1599 XPUSH_CLOBJ (stash_event, ev);
1428 1600
1429void 1601void
1430enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1602copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1603 ALIAS:
1604 enqueue_copy_image_to_buffer = 0
1431 PPCODE: 1605 PPCODE:
1432 cl_event ev = 0; 1606 cl_event ev = 0;
1433 const size_t src_origin[3] = { src_x, src_y, src_z }; 1607 const size_t src_origin[3] = { src_x, src_y, src_z };
1608 const size_t region [3] = { width, height, depth };
1609 EVENT_LIST (10);
1610
1611 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1612
1613 if (ev)
1614 XPUSH_CLOBJ (stash_event, ev);
1615
1616void
1617copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1618 ALIAS:
1619 enqueue_copy_buffer_to_image = 0
1620 PPCODE:
1621 cl_event ev = 0;
1622 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1623 const size_t region [3] = { width, height, depth };
1624 EVENT_LIST (10);
1625
1626 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1627
1628 if (ev)
1629 XPUSH_CLOBJ (stash_event, ev);
1630
1631void
1632map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, size_t cb = 0, ...)
1633 ALIAS:
1634 enqueue_map_buffer = 0
1635 PPCODE:
1636 cl_event ev;
1637 EVENT_LIST (6);
1638
1639 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1640 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0));
1641
1642void
1643map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, size_t width = 0, size_t height = 0, size_t depth = 0, ...)
1644 ALIAS:
1645 enqueue_map_image = 0
1646 PPCODE:
1647 cl_event ev;
1648 const size_t origin[3] = { x, y, z };
1434 const size_t region[3] = { width, height, depth }; 1649 const size_t region[3] = { width, height, depth };
1650 size_t row_pitch, slice_pitch;
1435 EVENT_LIST (10, items - 10); 1651 EVENT_LIST (10);
1436 1652
1653 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1654
1655 size_t cb = slice_pitch ? slice_pitch * region [2]
1656 : row_pitch ? row_pitch * region [1]
1657 : region [0];
1658
1659 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch));
1660
1661void
1662unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1663 PPCODE:
1664 mapped_unmap (ST (1), mapped, self, &ST (2), items - 2);
1665 if (GIMME_V != G_VOID)
1666 {
1667 clRetainEvent (mapped->event);
1668 XPUSH_CLOBJ (stash_event, mapped->event);
1669 }
1670
1671void
1672task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1673 ALIAS:
1674 enqueue_task = 0
1675 PPCODE:
1676 cl_event ev = 0;
1677 EVENT_LIST (2);
1678
1437 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1679 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1438 1680
1439 if (ev) 1681 if (ev)
1440 XPUSH_CLOBJ ("OpenCL::Event", ev); 1682 XPUSH_CLOBJ (stash_event, ev);
1441 1683
1442void 1684void
1443enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1444 PPCODE:
1445 cl_event ev = 0;
1446 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1447 const size_t region[3] = { width, height, depth };
1448 EVENT_LIST (10, items - 10);
1449
1450 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1451
1452 if (ev)
1453 XPUSH_CLOBJ ("OpenCL::Event", ev);
1454
1455void
1456enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1457 PPCODE:
1458 cl_event ev = 0;
1459 EVENT_LIST (2, items - 2);
1460
1461 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1462
1463 if (ev)
1464 XPUSH_CLOBJ ("OpenCL::Event", ev);
1465
1466void
1467enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1685nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1686 ALIAS:
1687 enqueue_nd_range_kernel = 0
1468 PPCODE: 1688 PPCODE:
1469 cl_event ev = 0; 1689 cl_event ev = 0;
1470 size_t *gwo = 0, *gws, *lws = 0; 1690 size_t *gwo = 0, *gws, *lws = 0;
1471 int gws_len; 1691 int gws_len;
1472 size_t *lists; 1692 size_t *lists;
1473 int i; 1693 int i;
1474 EVENT_LIST (5, items - 5); 1694 EVENT_LIST (5);
1475 1695
1476 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1696 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1477 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1697 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1478 1698
1479 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1699 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1480 1700
1481 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1701 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1482 1702
1483 gws = lists + gws_len * 0; 1703 gws = lists + gws_len * 0;
1484 for (i = 0; i < gws_len; ++i) 1704 for (i = 0; i < gws_len; ++i)
1705 {
1485 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1706 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1707 // at least nvidia crashes for 0-sized work group sizes, work around
1708 if (!gws [i])
1709 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1710 }
1486 1711
1487 if (SvOK (global_work_offset)) 1712 if (SvOK (global_work_offset))
1488 { 1713 {
1489 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1714 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1490 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1715 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1498 } 1723 }
1499 1724
1500 if (SvOK (local_work_size)) 1725 if (SvOK (local_work_size))
1501 { 1726 {
1502 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1727 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1503 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1728 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1504 1729
1505 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1730 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1506 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1731 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1507 1732
1508 lws = lists + gws_len * 2; 1733 lws = lists + gws_len * 2;
1509 for (i = 0; i < gws_len; ++i) 1734 for (i = 0; i < gws_len; ++i)
1735 {
1510 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1736 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1737 // at least nvidia crashes for 0-sized work group sizes, work around
1738 if (!lws [i])
1739 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
1740 }
1511 } 1741 }
1512 1742
1513 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1743 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1514 1744
1515 if (ev) 1745 if (ev)
1516 XPUSH_CLOBJ ("OpenCL::Event", ev); 1746 XPUSH_CLOBJ (stash_event, ev);
1517 1747
1518#if cl_apple_gl_sharing || cl_khr_gl_sharing 1748#if cl_apple_gl_sharing || cl_khr_gl_sharing
1519 1749
1520void 1750void
1521enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 1751acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1522 ALIAS: 1752 ALIAS:
1753 release_gl_objects = 1
1754 enqueue_acquire_gl_objects = 0
1523 enqueue_release_gl_objects = 1 1755 enqueue_release_gl_objects = 1
1524 PPCODE: 1756 PPCODE:
1525 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 1757 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV)
1526 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 1758 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call");
1527 1759
1528 cl_event ev = 0; 1760 cl_event ev = 0;
1529 EVENT_LIST (2, items - 2); 1761 EVENT_LIST (2);
1530 AV *av = (AV *)SvRV (objects); 1762 AV *av = (AV *)SvRV (objects);
1531 cl_uint num_objects = av_len (av) + 1; 1763 cl_uint num_objects = av_len (av) + 1;
1532 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects); 1764 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1533 int i; 1765 int i;
1534 1766
1539 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1771 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1540 else 1772 else
1541 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1773 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1542 1774
1543 if (ev) 1775 if (ev)
1544 XPUSH_CLOBJ ("OpenCL::Event", ev); 1776 XPUSH_CLOBJ (stash_event, ev);
1545 1777
1546#endif 1778#endif
1547 1779
1548void 1780void
1549enqueue_wait_for_events (OpenCL::Queue self, ...) 1781wait_for_events (OpenCL::Queue self, ...)
1782 ALIAS:
1783 enqueue_wait_for_events = 0
1550 CODE: 1784 CODE:
1551 EVENT_LIST (1, items - 1); 1785 EVENT_LIST (1);
1552#if PREFER_1_1 1786#if PREFER_1_1
1553 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 1787 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1554#else 1788#else
1555 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0)); 1789 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1556#endif 1790#endif
1557 1791
1558void 1792void
1559enqueue_marker (OpenCL::Queue self, ...) 1793marker (OpenCL::Queue self, ...)
1794 ALIAS:
1795 enqueue_marker = 0
1560 PPCODE: 1796 PPCODE:
1561 cl_event ev = 0; 1797 cl_event ev = 0;
1562 EVENT_LIST (1, items - 1); 1798 EVENT_LIST (1);
1563#if PREFER_1_1 1799#if PREFER_1_1
1564 if (!event_list_count) 1800 if (!event_list_count)
1565 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0)); 1801 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
1566 else 1802 else
1567#if CL_VERSION_1_2 1803#if CL_VERSION_1_2
1574#endif 1810#endif
1575#else 1811#else
1576 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1812 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1577#endif 1813#endif
1578 if (ev) 1814 if (ev)
1579 XPUSH_CLOBJ ("OpenCL::Event", ev); 1815 XPUSH_CLOBJ (stash_event, ev);
1580 1816
1581void 1817void
1582enqueue_barrier (OpenCL::Queue self, ...) 1818barrier (OpenCL::Queue self, ...)
1819 ALIAS:
1820 enqueue_barrier = 0
1583 PPCODE: 1821 PPCODE:
1584 cl_event ev = 0; 1822 cl_event ev = 0;
1585 EVENT_LIST (1, items - 1); 1823 EVENT_LIST (1);
1586#if PREFER_1_1 1824#if PREFER_1_1
1587 if (!event_list_count && GIMME_V == G_VOID) 1825 if (!event_list_count && GIMME_V == G_VOID)
1588 NEED_SUCCESS (EnqueueBarrier, (self)); 1826 NEED_SUCCESS (EnqueueBarrier, (self));
1589 else 1827 else
1590#if CL_VERSION_1_2 1828#if CL_VERSION_1_2
1600#endif 1838#endif
1601#else 1839#else
1602 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1840 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1603#endif 1841#endif
1604 if (ev) 1842 if (ev)
1605 XPUSH_CLOBJ ("OpenCL::Event", ev); 1843 XPUSH_CLOBJ (stash_event, ev);
1606 1844
1607void 1845void
1608flush (OpenCL::Queue self) 1846flush (OpenCL::Queue self)
1609 CODE: 1847 CODE:
1610 NEED_SUCCESS (Flush, (self)); 1848 NEED_SUCCESS (Flush, (self));
1628 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 1866 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1629 EXTEND (SP, 1); 1867 EXTEND (SP, 1);
1630 const int i = 0; 1868 const int i = 0;
1631 { 1869 {
1632 NEED_SUCCESS (RetainContext, (value [i])); 1870 NEED_SUCCESS (RetainContext, (value [i]));
1633 PUSH_CLOBJ ("OpenCL::Context", value [i]); 1871 PUSH_CLOBJ (stash_context, value [i]);
1634 } 1872 }
1635 1873
1636void 1874void
1637device (OpenCL::Queue self) 1875device (OpenCL::Queue self)
1638 PPCODE: 1876 PPCODE:
1639 cl_device_id value [1]; 1877 cl_device_id value [1];
1640 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 1878 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1641 EXTEND (SP, 1); 1879 EXTEND (SP, 1);
1642 const int i = 0; 1880 const int i = 0;
1643 { 1881 {
1644 PUSH_CLOBJ ("OpenCL::Device", value [i]); 1882 PUSH_CLOBJ (stash_device, value [i]);
1645 } 1883 }
1646 1884
1647void 1885void
1648reference_count (OpenCL::Queue self) 1886reference_count (OpenCL::Queue self)
1649 PPCODE: 1887 PPCODE:
1658 PPCODE: 1896 PPCODE:
1659 cl_command_queue_properties value [1]; 1897 cl_command_queue_properties value [1];
1660 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 1898 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1661 EXTEND (SP, 1); 1899 EXTEND (SP, 1);
1662 const int i = 0; 1900 const int i = 0;
1663 PUSHs (sv_2mortal (newSViv (value [i]))); 1901 PUSHs (sv_2mortal (newSVuv (value [i])));
1664 1902
1665#END:command_queue 1903#END:command_queue
1666 1904
1667MODULE = OpenCL PACKAGE = OpenCL::Memory 1905MODULE = OpenCL PACKAGE = OpenCL::Memory
1668 1906
1683 PPCODE: 1921 PPCODE:
1684 cl_mem_object_type value [1]; 1922 cl_mem_object_type value [1];
1685 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 1923 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0));
1686 EXTEND (SP, 1); 1924 EXTEND (SP, 1);
1687 const int i = 0; 1925 const int i = 0;
1688 PUSHs (sv_2mortal (newSViv (value [i]))); 1926 PUSHs (sv_2mortal (newSVuv (value [i])));
1689 1927
1690void 1928void
1691flags (OpenCL::Memory self) 1929flags (OpenCL::Memory self)
1692 PPCODE: 1930 PPCODE:
1693 cl_mem_flags value [1]; 1931 cl_mem_flags value [1];
1694 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 1932 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1695 EXTEND (SP, 1); 1933 EXTEND (SP, 1);
1696 const int i = 0; 1934 const int i = 0;
1697 PUSHs (sv_2mortal (newSViv (value [i]))); 1935 PUSHs (sv_2mortal (newSVuv (value [i])));
1698 1936
1699void 1937void
1700size (OpenCL::Memory self) 1938size (OpenCL::Memory self)
1701 ALIAS: 1939 ALIAS:
1702 size = CL_MEM_SIZE 1940 size = CL_MEM_SIZE
1736 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 1974 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1737 EXTEND (SP, 1); 1975 EXTEND (SP, 1);
1738 const int i = 0; 1976 const int i = 0;
1739 { 1977 {
1740 NEED_SUCCESS (RetainContext, (value [i])); 1978 NEED_SUCCESS (RetainContext, (value [i]));
1741 PUSH_CLOBJ ("OpenCL::Context", value [i]); 1979 PUSH_CLOBJ (stash_context, value [i]);
1742 } 1980 }
1743 1981
1744void 1982void
1745associated_memobject (OpenCL::Memory self) 1983associated_memobject (OpenCL::Memory self)
1746 PPCODE: 1984 PPCODE:
1748 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 1986 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1749 EXTEND (SP, 1); 1987 EXTEND (SP, 1);
1750 const int i = 0; 1988 const int i = 0;
1751 { 1989 {
1752 NEED_SUCCESS (RetainMemObject, (value [i])); 1990 NEED_SUCCESS (RetainMemObject, (value [i]));
1753 PUSH_CLOBJ ("OpenCL::Memory", value [i]); 1991 PUSH_CLOBJ (stash_memory, value [i]);
1754 } 1992 }
1755 1993
1756#END:mem 1994#END:mem
1757 1995
1758#if cl_apple_gl_sharing || cl_khr_gl_sharing 1996#if cl_apple_gl_sharing || cl_khr_gl_sharing
1778 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2016 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1779 2017
1780 cl_buffer_region crdata = { origin, size }; 2018 cl_buffer_region crdata = { origin, size };
1781 2019
1782 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2020 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1783 XPUSH_CLOBJ ("OpenCL::Buffer", mem); 2021 XPUSH_CLOBJ (stash_buffer, mem);
1784 2022
1785MODULE = OpenCL PACKAGE = OpenCL::Image 2023MODULE = OpenCL PACKAGE = OpenCL::Image
1786 2024
1787void 2025void
1788image_info (OpenCL::Image self, cl_image_info name) 2026image_info (OpenCL::Image self, cl_image_info name)
1874 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2112 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1875 EXTEND (SP, 1); 2113 EXTEND (SP, 1);
1876 const int i = 0; 2114 const int i = 0;
1877 { 2115 {
1878 NEED_SUCCESS (RetainContext, (value [i])); 2116 NEED_SUCCESS (RetainContext, (value [i]));
1879 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2117 PUSH_CLOBJ (stash_context, value [i]);
1880 } 2118 }
1881 2119
1882void 2120void
1883normalized_coords (OpenCL::Sampler self) 2121normalized_coords (OpenCL::Sampler self)
1884 PPCODE: 2122 PPCODE:
1885 cl_addressing_mode value [1]; 2123 cl_addressing_mode value [1];
1886 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0)); 2124 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1887 EXTEND (SP, 1); 2125 EXTEND (SP, 1);
1888 const int i = 0; 2126 const int i = 0;
1889 PUSHs (sv_2mortal (newSViv (value [i]))); 2127 PUSHs (sv_2mortal (newSVuv (value [i])));
1890 2128
1891void 2129void
1892addressing_mode (OpenCL::Sampler self) 2130addressing_mode (OpenCL::Sampler self)
1893 PPCODE: 2131 PPCODE:
1894 cl_filter_mode value [1]; 2132 cl_filter_mode value [1];
1895 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0)); 2133 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1896 EXTEND (SP, 1); 2134 EXTEND (SP, 1);
1897 const int i = 0; 2135 const int i = 0;
1898 PUSHs (sv_2mortal (newSViv (value [i]))); 2136 PUSHs (sv_2mortal (newSVuv (value [i])));
1899 2137
1900void 2138void
1901filter_mode (OpenCL::Sampler self) 2139filter_mode (OpenCL::Sampler self)
1902 PPCODE: 2140 PPCODE:
1903 cl_bool value [1]; 2141 cl_bool value [1];
1995 2233
1996void 2234void
1997kernel (OpenCL::Program program, SV *function) 2235kernel (OpenCL::Program program, SV *function)
1998 PPCODE: 2236 PPCODE:
1999 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2237 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
2000 XPUSH_CLOBJ ("OpenCL::Kernel", kernel); 2238 XPUSH_CLOBJ (stash_kernel, kernel);
2001 2239
2002void 2240void
2003kernels_in_program (OpenCL::Program program) 2241kernels_in_program (OpenCL::Program program)
2004 PPCODE: 2242 PPCODE:
2005 cl_uint num_kernels; 2243 cl_uint num_kernels;
2008 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0)); 2246 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2009 2247
2010 int i; 2248 int i;
2011 EXTEND (SP, num_kernels); 2249 EXTEND (SP, num_kernels);
2012 for (i = 0; i < num_kernels; ++i) 2250 for (i = 0; i < num_kernels; ++i)
2013 PUSH_CLOBJ ("OpenCL::Kernel", kernels [i]); 2251 PUSH_CLOBJ (stash_kernel, kernels [i]);
2014 2252
2015void 2253void
2016info (OpenCL::Program self, cl_program_info name) 2254info (OpenCL::Program self, cl_program_info name)
2017 PPCODE: 2255 PPCODE:
2018 INFO (Program) 2256 INFO (Program)
2066 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2304 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
2067 EXTEND (SP, 1); 2305 EXTEND (SP, 1);
2068 const int i = 0; 2306 const int i = 0;
2069 { 2307 {
2070 NEED_SUCCESS (RetainContext, (value [i])); 2308 NEED_SUCCESS (RetainContext, (value [i]));
2071 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2309 PUSH_CLOBJ (stash_context, value [i]);
2072 } 2310 }
2073 2311
2074void 2312void
2075devices (OpenCL::Program self) 2313devices (OpenCL::Program self)
2076 PPCODE: 2314 PPCODE:
2080 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2318 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
2081 int i, n = size / sizeof (*value); 2319 int i, n = size / sizeof (*value);
2082 EXTEND (SP, n); 2320 EXTEND (SP, n);
2083 for (i = 0; i < n; ++i) 2321 for (i = 0; i < n; ++i)
2084 { 2322 {
2085 PUSH_CLOBJ ("OpenCL::Device", value [i]); 2323 PUSH_CLOBJ (stash_device, value [i]);
2086 } 2324 }
2087 2325
2088void 2326void
2089source (OpenCL::Program self) 2327source (OpenCL::Program self)
2090 PPCODE: 2328 PPCODE:
2116DESTROY (OpenCL::Kernel self) 2354DESTROY (OpenCL::Kernel self)
2117 CODE: 2355 CODE:
2118 clReleaseKernel (self); 2356 clReleaseKernel (self);
2119 2357
2120void 2358void
2359setf (OpenCL::Kernel self, const char *format, ...)
2360 CODE:
2361 int i;
2362 for (i = 2; ; ++i)
2363 {
2364 while (*format == ' ')
2365 ++format;
2366
2367 char type = *format++;
2368
2369 if (!type)
2370 break;
2371
2372 if (i >= items)
2373 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2374
2375 SV *sv = ST (i);
2376
2377 union
2378 {
2379 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2380 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2381 cl_half ch; cl_float cf; cl_double cd;
2382 cl_mem cm;
2383 cl_sampler ca;
2384 size_t cz;
2385 cl_event ce;
2386 } arg;
2387 size_t size;
2388 int nullarg = 0;
2389
2390 switch (type)
2391 {
2392 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2393 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2394 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2395 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2396 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2397 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2398 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2399 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2400
2401 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2402 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2403 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2404
2405 case 'z': nullarg = 1; size = SvIV (sv); break;
2406
2407 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ ("OpenCL::Kernel::setf", "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2408 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2409 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ ("OpenCL::Kernel::setf", "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2410
2411 default:
2412 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2413 }
2414
2415 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2416 if (res)
2417 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2418 }
2419
2420 if (i != items)
2421 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2422
2423void
2121set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2424set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
2122 CODE: 2425 CODE:
2123 clSetKernelArg (self, idx, sizeof (value), &value); 2426 clSetKernelArg (self, idx, sizeof (value), &value);
2124 2427
2125void 2428void
2173 clSetKernelArg (self, idx, sizeof (value), &value); 2476 clSetKernelArg (self, idx, sizeof (value), &value);
2174 2477
2175void 2478void
2176set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2479set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
2177 CODE: 2480 CODE:
2178 clSetKernelArg (self, idx, sizeof (value), &value); 2481 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2179 2482
2180void 2483void
2181set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2484set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
2182 CODE: 2485 CODE:
2183 clSetKernelArg (self, idx, sizeof (value), &value); 2486 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2184 2487
2185void 2488void
2186set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2489set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
2490 ALIAS:
2491 set_image2d = 0
2492 set_image3d = 0
2187 CODE: 2493 CODE:
2188 clSetKernelArg (self, idx, sizeof (value), &value); 2494 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
2189
2190void
2191set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
2192 CODE:
2193 clSetKernelArg (self, idx, sizeof (value), &value);
2194 2495
2195void 2496void
2196set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2497set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
2197 CODE: 2498 CODE:
2198 clSetKernelArg (self, idx, sizeof (value), &value); 2499 clSetKernelArg (self, idx, sizeof (value), &value);
2244 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2545 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
2245 EXTEND (SP, 1); 2546 EXTEND (SP, 1);
2246 const int i = 0; 2547 const int i = 0;
2247 { 2548 {
2248 NEED_SUCCESS (RetainContext, (value [i])); 2549 NEED_SUCCESS (RetainContext, (value [i]));
2249 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2550 PUSH_CLOBJ (stash_context, value [i]);
2250 } 2551 }
2251 2552
2252void 2553void
2253program (OpenCL::Kernel self) 2554program (OpenCL::Kernel self)
2254 PPCODE: 2555 PPCODE:
2256 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2557 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
2257 EXTEND (SP, 1); 2558 EXTEND (SP, 1);
2258 const int i = 0; 2559 const int i = 0;
2259 { 2560 {
2260 NEED_SUCCESS (RetainProgram, (value [i])); 2561 NEED_SUCCESS (RetainProgram, (value [i]));
2261 PUSH_CLOBJ ("OpenCL::Program", value [i]); 2562 PUSH_CLOBJ (stash_program, value [i]);
2262 } 2563 }
2263 2564
2264#END:kernel 2565#END:kernel
2265 2566
2266void 2567void
2346 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2647 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
2347 EXTEND (SP, 1); 2648 EXTEND (SP, 1);
2348 const int i = 0; 2649 const int i = 0;
2349 { 2650 {
2350 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2651 NEED_SUCCESS (RetainCommandQueue, (value [i]));
2351 PUSH_CLOBJ ("OpenCL::Queue", value [i]); 2652 PUSH_CLOBJ (stash_queue, value [i]);
2352 } 2653 }
2353 2654
2354void 2655void
2355command_type (OpenCL::Event self) 2656command_type (OpenCL::Event self)
2356 PPCODE: 2657 PPCODE:
2379 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2680 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
2380 EXTEND (SP, 1); 2681 EXTEND (SP, 1);
2381 const int i = 0; 2682 const int i = 0;
2382 { 2683 {
2383 NEED_SUCCESS (RetainContext, (value [i])); 2684 NEED_SUCCESS (RetainContext, (value [i]));
2384 PUSH_CLOBJ ("OpenCL::Context", value [i]); 2685 PUSH_CLOBJ (stash_context, value [i]);
2385 } 2686 }
2386 2687
2387#END:event 2688#END:event
2388 2689
2389void 2690void
2414void 2715void
2415set_status (OpenCL::UserEvent self, cl_int execution_status) 2716set_status (OpenCL::UserEvent self, cl_int execution_status)
2416 CODE: 2717 CODE:
2417 clSetUserEventStatus (self, execution_status); 2718 clSetUserEventStatus (self, execution_status);
2418 2719
2720MODULE = OpenCL PACKAGE = OpenCL::Mapped
2721
2722void
2723DESTROY (SV *self)
2724 CODE:
2725 OpenCL__Mapped mapped = SvMAPPED (self);
2726
2727 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
2728 mapped_detach (self, mapped);
2729
2730 clReleaseCommandQueue (mapped->queue);
2731 clReleaseEvent (mapped->event);
2732 Safefree (mapped);
2733
2734void
2735unmap (OpenCL::Mapped self, ...)
2736 CODE:
2737 mapped_unmap (ST (0), self, self->queue, &ST (1), items - 1);
2738
2739bool
2740mapped (OpenCL::Mapped self)
2741 CODE:
2742 RETVAL = !!self->ptr;
2743 OUTPUT:
2744 RETVAL
2745
2746void
2747wait (OpenCL::Mapped self)
2748 PPCODE:
2749 if (self->event)
2750 NEED_SUCCESS (WaitForEvents, (1, &self->event));
2751
2752void
2753event (OpenCL::Mapped self)
2754 PPCODE:
2755 if (!self->event)
2756 XSRETURN_UNDEF;
2757
2758 clRetainEvent (self->event);
2759 XPUSH_CLOBJ (stash_event, self->event);
2760
2761size_t
2762size (OpenCL::Mapped self)
2763 CODE:
2764 RETVAL = self->cb;
2765 OUTPUT:
2766 RETVAL
2767
2768IV
2769ptr (OpenCL::Mapped self)
2770 CODE:
2771 RETVAL = PTR2IV (self->ptr);
2772 OUTPUT:
2773 RETVAL
2774
2775MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
2776
2777MODULE = OpenCL PACKAGE = OpenCL::MappedImage
2778
2779IV
2780row_pitch (OpenCL::Mapped self)
2781 ALIAS:
2782 slice_pitch = 1
2783 CODE:
2784 RETVAL = ix ? self->slice_pitch : self->row_pitch;
2785 OUTPUT:
2786 RETVAL
2787
2788
2789

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines