[HPCM] HPCM 1.11 교육 자료

HPCM R&D WIKI page

- 주소: https://rndwiki-pro.its.hpecorp.net/display/HPCM/Rackmap

※ HPE network 에서 접속 가능

 

AMD Diagnostic Tools

- ROCm: 6.0.0

- CrayPE: 23.12

 

(1) Rochpcg: ROCm base HPCG (e.g. rochpcg -local)

ROCHPCG

 

(2) Rochpl: ROCm base HPL (e.g. linpack-rochpl.sh)

 

(3) Babelstream: AMD CPU & GPU (e.g. run_babelstream --platform parrypeak)

 

(4) 기타

# cm health check cwcpuperf

 

"--LOOP" and "--fabric" parameters can be specified with cwcpuperf and fabricperf   "--LOOP": Number of times the diagnostic needs to be run. Default: 1  "--fabric": The interface on which diagnostics runs. Options: ib0, eth1, bond0, hsn1