Elide memcpy where not needed (#1863)
TL;DR: Simple change for ~2% gain on GPU in a typical workload.
Showing
- arbor/backends/gpu/shared_state.cpp 11 additions, 8 deletionsarbor/backends/gpu/shared_state.cpp
- arbor/backends/gpu/shared_state.hpp 4 additions, 1 deletionarbor/backends/gpu/shared_state.hpp
- arbor/backends/multicore/shared_state.cpp 11 additions, 9 deletionsarbor/backends/multicore/shared_state.cpp
- arbor/backends/multicore/shared_state.hpp 4 additions, 1 deletionarbor/backends/multicore/shared_state.hpp
- arbor/fvm_layout.cpp 41 additions, 13 deletionsarbor/fvm_layout.cpp
- arbor/fvm_layout.hpp 5 additions, 0 deletionsarbor/fvm_layout.hpp
- arbor/fvm_lowered_cell_impl.hpp 4 additions, 6 deletionsarbor/fvm_lowered_cell_impl.hpp
- arbor/version.cpp 1 addition, 0 deletionsarbor/version.cpp
- test/unit/test_fvm_lowered.cpp 3 additions, 0 deletionstest/unit/test_fvm_lowered.cpp
- test/unit/test_probe.cpp 12 additions, 0 deletionstest/unit/test_probe.cpp
Please register or sign in to comment