/iGEMMgen

Code generator for implicit gemm algorithm

Primary LanguagePythonMIT LicenseMIT

iGEMMgen

Code generator for implicit gemm algorithm

Generate

Please install numpy before use this tool.

# generate code based on tunable configuration
python3 igemm_codegen.py config/igemm_v4r1_dynamic.config

The output file will result in out directory. result in a assembly file *.s, a codeobject *.hsaco and a host driver executable conv_driver.exe. This executable accept same cmdline argument as MIOpenDriver. For a quick start, can use script/v4r1_origin_conv.sh int the top directory to launch the driver with several tensor descriptors.

more description to be added