Tune aldebaran HHS changes - Githubissues

ROCm / rocBLAS

Next generation BLAS implementation for ROCm platform

https://rocm.docs.amd.com/projects/rocBLAS/en/latest/

Other

344 stars 166 forks source link

Tune aldebaran HHS changes #1324

Closed aferoz21 closed 1 year ago

aferoz21 commented 1 year ago

resolves #SWDEV-350792

Summary of proposed changes:

TN, NT problem sizes of HHS tuned.
rocBLAS-Bench verification with hpl init and rocBLAS-test validation done.
Several macro tile and depthU tried to improve the efficiency.