search

ROCm / hipBLASLt

hipBLASLt is a library that provides general matrix-matrix operations with a flexible API and extends functionalities beyond a traditional BLAS library
https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/index.html
MIT License
40 stars 57 forks source link

Update optimal value #843

Closed aazz44ss closed 2 weeks ago

aazz44ss commented 2 weeks ago

Add optimal SUS = DU * bytePerElement recalculate optimal LRVW with accurate LDS usage remove MIWaveTile != 1 restriction for 1LDSB