Generates perf benchmarking parquet from command-line arguments. More...
Go to the source code of this file.
Namespaces | |
namespace | pipeline |
namespace | pipeline.gen_perf_parquet_logs |
Functions | |
pipeline.gen_perf_parquet_logs.parse_args () | |
pipeline.gen_perf_parquet_logs.update_parquet (args) | |
Variables | |
pipeline.gen_perf_parquet_logs.args = parse_args() | |
Generates perf benchmarking parquet from command-line arguments.
perf stat
as arguments. Computes % miss rates for cache levels and TLB, as well as misses per trial. Saves results to a timestamped .parquet
file in a structured batch directory, and also generates a global merged file if needed.zstd
parquet with labeled fields and derived metrics Definition in file gen_perf_parquet_logs.py.