| gpu.c: create_kernel_vars: only take ppcg_kernel |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: localize_bounds: only take ppcg_kernel |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: find_array_index: take ppcg_kernel instead of gpu_gen |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: remove_local_accesses: pass in relevant schedule instead of computing it |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: remove_local_accesses: take gpu_prog instead of gpu_gen |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: group_writes: fix error handling |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu_group_references: isolate all required schedule information in structure |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: add functions for adding/detecting synchronization nodes in schedule tree |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: set_schedule_modulo: allow smaller number of band members than identifiers |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: generate the AST from a single schedule tree |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: create_kernel_vars: improve error handling |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu_group_references: take into account "shared" mark |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu_group_references: distinguish between shared_schedule and copy_schedule |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu_group.c: compute_group_bounds_core: improve error handling |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: create_kernel: obtain number of block dimensions from marked band |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: read_sizes: split off read_tile_sizes |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: mark_outer_tilable: mark point band as to be mapped to threads |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: read grid and block sizes as soon as kernel object is created |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: store the number of (parallel) tile dimensions in ppcg_kernel |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu_group.c: check_scalar_live_ranges: isolate dependence on gpu_gen |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: use isl_schedule_node_map_descendant to insert kernel marks |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: insert context node in final schedule tree |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: band_select_outer_band: extract out n_outer_coincidence |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: move any_force_private from gpu_gen to ppcg_kernel |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| polybench_test.sh: handle --verbose option |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| ppcg_scop: replace union map schedule by schedule tree |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu: avoid copying in data from local arrays |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| gpu.c: transform_expr: dereference pointers to scalars |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| ppcg_scop_from_pet_scop: use pet_scop_collect_must_kills |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
| cuda.c: extract out declare_device_array |
skimo
as Sven Verdoolaege
|
More...
|
about 12 years ago
|
