Skip to content

Performance regression caused by read lock in brgemm #323

@zhczhong

Description

@zhczhong

According to the perf test from @yifeizh2, the read lock will bring 10-20% performance regression

  data type bs  hidden_list  without lock with lock  
bf16 128 16x512 0.021597 0.01967 109.79%
bf16 128 512x256 0.025172 0.029239 86.09%
bf16 128 256x128 0.015969 0.026034 61.34%
bf16 128 512x1024 0.046416 0.055459 83.69%
bf16 128 1024x1024 0.082655 0.089712 92.13%
bf16 128 1024x512 0.066204 0.066523 99.52%
bf16 128 512x256 0.025992 0.029516 88.06%

Metadata

Metadata

Labels

Type

No type

Projects

No projects

Relationships

None yet

Development

No branches or pull requests

Issue actions