[ WIP ] [ hgemm ] Use zero-padding in non-8-divisible M @open sesame 07/01 14:08

nnstreamer / nntrainer

NNtrainer is Software Framework for Training Neural Network Models on Devices.

Apache License 2.0

134 stars 71 forks source link

[ WIP ] [ hgemm ] Use zero-padding in non-8-divisible M @open sesame 07/01 14:08 #2655

Open skykongkong8 opened 2 days ago

skykongkong8 commented 2 days ago

Using zero padding is not an optimal solution, but still gives better performance.
Fine-grained tiling for optimization will be added in the near future

TC : 1:1:577:1024 x 1:1:4096:1024 fp16 : 55 ms fp32 : 74 ms

Self evaluation:

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

taos-ci commented 2 days ago

:memo: TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2655. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/.