qianfengz Goto Github PK

config-tools

Tools to auto-create and process the configs used for generating asm kernels as input of the asm igemm dynamic generator

data-tool

gpubuffersmatch

Used to check whether the outputs (which are usually float values) from two gpu kernels match by examining the data in the two output buffers (excluding NaN and Inf data) and calculating the Root Mean Square Difference of them

igemmgen

Code generator for implicit gemm algorithm

reducetensortest

Examples to use miopenReduceTensor and compare with corresponding implementation using cudnnReduceTensor

xformers

Hackable and optimized Transformers building blocks, supporting a composable construction.