diff options
| author | Luo Xionghu <xionghu.luo@intel.com> | 2015-04-09 11:37:57 +0800 |
|---|---|---|
| committer | Zhigang Gong <zhigang.gong@intel.com> | 2015-05-04 15:13:57 +0800 |
| commit | bbd5c94e0511a1d9f37e3b7522d6418d960138e1 (patch) | |
| tree | a3243cf2e59357fc187193b92ef3804b3fb0e328 /src/CMakeLists.txt | |
| parent | 9dedb7bb929cc3580292a0db8334537dc2493fd8 (diff) | |
| download | beignet-bbd5c94e0511a1d9f37e3b7522d6418d960138e1.tar.gz | |
Optimization of clEnqueueCopyBufferToImage for 16 aligned case.
We can change the image_channel_order to CL_RGBA and
image_channel_data_type to CL_UNSIGNED_INT32 for some special
case, thus 16 bytes can be read by one work item. Bandwidth is
fully used.
v2: merge patch 3 of initializing region0;
remove k dimension in kernel for 2d image.
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: "Weng, Chuanbo" <chuanbo.weng@intel.com>
Diffstat (limited to 'src/CMakeLists.txt')
| -rw-r--r-- | src/CMakeLists.txt | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index da695324..4e67c71f 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -51,7 +51,7 @@ cl_internal_copy_image_2d_to_2d_array cl_internal_copy_image_1d_array_to_1d_arra cl_internal_copy_image_2d_array_to_2d_array cl_internal_copy_image_2d_array_to_2d cl_internal_copy_image_2d_array_to_3d cl_internal_copy_image_3d_to_2d_array cl_internal_copy_image_2d_to_buffer cl_internal_copy_image_2d_to_buffer_align16 cl_internal_copy_image_3d_to_buffer -cl_internal_copy_buffer_to_image_2d cl_internal_copy_buffer_to_image_3d +cl_internal_copy_buffer_to_image_2d cl_internal_copy_buffer_to_image_2d_align16 cl_internal_copy_buffer_to_image_3d cl_internal_fill_buf_align8 cl_internal_fill_buf_align4 cl_internal_fill_buf_align2 cl_internal_fill_buf_unalign cl_internal_fill_buf_align128 cl_internal_fill_image_1d |