NVIDIA / Fuser

A Fusion Code Generator for NVIDIA GPUs (commonly known as "nvFuser")
Other
271 stars 53 forks source link

add knob controls unroll on top of vectorization in inner reduction #3393

Closed liqiangxl closed 1 week ago

liqiangxl commented 1 week ago

Feature Added: Knob for Unrolling on Top of Vectorization in Inner Reduction

Introduced a new knob to control unrolling in conjunction with vectorization for inner reduction loops. Note: The knob is currently not integrated into the heuristics and must be manually enabled for use. Added a test case to validate the functionality and ensure the knob's proper behavior.

liqiangxl commented 1 week ago

!test

liqiangxl commented 1 week ago

!test

liqiangxl commented 1 week ago

!test