Skip to content

dnn(OpenCL): fix conv BASIC workgroup#20774

Merged
opencv-pushbot merged 1 commit intoopencv:3.4from
alalek:dnn_ocl_fix_conv_basic_workgroup
Sep 30, 2021
Merged

dnn(OpenCL): fix conv BASIC workgroup#20774
opencv-pushbot merged 1 commit intoopencv:3.4from
alalek:dnn_ocl_fix_conv_basic_workgroup

Conversation

@alalek
Copy link
Copy Markdown
Member

@alalek alalek commented Sep 29, 2021 •
edited
Loading

Avoid inefficient automatic scheduling of 3D task layout with LWS=1x1x256
Also there is no efficient selection of 3D LWS (GWS must be divisible to LVS).

Finally, task is reworked to process as 1D array.

relates #20655

force_builders=Custom,Linux AVX2,Linux OpenCL,Custom Win
build_image:Custom=ubuntu:18.04
buildworker:Custom=linux-5
test_opencl:Custom=ON

build_image:Linux AVX2=ubuntu:18.04
buildworker:Linux AVX2=linux-3
test_opencl:Linux AVX2=ON

buildworker:Custom Win=windows-3
build_image:Custom Win=msvs2019
test_opencl:Custom Win=ON

@alalek
Copy link
Copy Markdown
Member Author

alalek commented Sep 30, 2021

👍

@opencv-pushbot opencv-pushbot merged commit 4d587c3 into opencv:3.4 Sep 30, 2021
This was referenced Oct 1, 2021
Merged
Merged
@alalek alalek mentioned this pull request Oct 15, 2021
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

@alalek alalek

Labels

bug category: dnn category: ocl

Projects

None yet

Milestone

3.4.16

Development

Successfully merging this pull request may close these issues.

2 participants

@alalek @opencv-pushbot