diff options
| author | Hristo Gueorguiev <prem.nirved@gmail.com> | 2017-05-03 16:30:45 +0300 |
|---|---|---|
| committer | Hristo Gueorguiev <prem.nirved@gmail.com> | 2017-05-03 16:30:45 +0300 |
| commit | 6bf4115c13962c99d1cdc97f2be92c4922f3fd33 (patch) | |
| tree | 569c512a242caf2ea4465f2eef561933ed937a2f /intern/cycles/kernel/CMakeLists.txt | |
| parent | 6f9c839f444f92c4b0c336a6f5e31cb9660d7dbc (diff) | |
Cycles: Split kernel - sort shaders
Reduce thread divergence in kernel_shader_eval.
Rays are sorted in blocks of 2048 according to shader->id.
On R9 290 Classroom is ~30% faster, and Pabellon Barcelone is ~8% faster.
No sorting for CUDA split kernel.
Reviewers: sergey, maiself
Reviewed By: maiself
Differential Revision: https://developer.blender.org/D2598
Diffstat (limited to 'intern/cycles/kernel/CMakeLists.txt')
| -rw-r--r-- | intern/cycles/kernel/CMakeLists.txt | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/intern/cycles/kernel/CMakeLists.txt b/intern/cycles/kernel/CMakeLists.txt index a92e8bc4aee..9bb0455b9d5 100644 --- a/intern/cycles/kernel/CMakeLists.txt +++ b/intern/cycles/kernel/CMakeLists.txt @@ -21,6 +21,8 @@ set(SRC kernels/opencl/kernel_lamp_emission.cl kernels/opencl/kernel_do_volume.cl kernels/opencl/kernel_indirect_background.cl + kernels/opencl/kernel_shader_setup.cl + kernels/opencl/kernel_shader_sort.cl kernels/opencl/kernel_shader_eval.cl kernels/opencl/kernel_holdout_emission_blurring_pathtermination_ao.cl kernels/opencl/kernel_subsurface_scatter.cl @@ -248,6 +250,8 @@ set(SRC_SPLIT_HEADERS split/kernel_path_init.h split/kernel_queue_enqueue.h split/kernel_scene_intersect.h + split/kernel_shader_setup.h + split/kernel_shader_sort.h split/kernel_shader_eval.h split/kernel_shadow_blocked_ao.h split/kernel_shadow_blocked_dl.h @@ -457,6 +461,8 @@ delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_scene_interse delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_lamp_emission.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_do_volume.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_indirect_background.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) +delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_shader_setup.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) +delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_shader_sort.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_shader_eval.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_holdout_emission_blurring_pathtermination_ao.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "kernels/opencl/kernel_subsurface_scatter.cl" ${CYCLES_INSTALL_PATH}/source/kernel/kernels/opencl) |