process_time_event.py

import argparse
import pandas as pd

def run(raw_log):
    # file = "/home_local/chunyuan/code/TE-ipex/ipex/memory_time_nanosecond.csv"
    # raw_log = "/home_local/chunyuan/code/TE-ipex/intel_model_zoo/resnet50_latency_log_bf16_20220309143502_instance_0_cores_0-55.log"
    # raw_log = "/home_local/chunyuan/code/TE-ipex/intel_model_zoo/resnet50_latency_log_bf16_20220309150733_instance_0_cores_0-55.log"

    with open(raw_log) as f:
        content = f.read().splitlines()

    read_line = False
    for i, line in enumerate(content):
        if line.startswith("----"):
            break


    rows = []
    index_to_start = i + 1
    for i, line in enumerate(content):
        if i >= index_to_start:
            if len(line.split('|')) == 2:
                row = line.split("|")
                row[1] = row[1].strip()
                rows.append(row)

    df = pd.DataFrame(rows, columns=["event", "TimeStamp(nanoseconds)"])
    print(df.head())

    # df_summary = pd.DataFrame(columns=["start", "end", "diff"])

    # df = pd.read_csv(file)

    df["TimeStamp(nanoseconds)"] = df["TimeStamp(nanoseconds)"].astype(int)

    df['dTimeStamp'] = df['TimeStamp(nanoseconds)'] - df['TimeStamp(nanoseconds)'].shift(1)

    new_df = df.iloc[1::2, :]

    # df_summary["start"] = df[::2, :]

    # print(df.head())
    print(new_df)
    print("sum (in microseconds): ", new_df["dTimeStamp"].sum() / 1000)



if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("-f", "--file_name", default=None, type=str, required=True, help="path to the input onednn log file")
    args = parser.parse_args()

    run(args.file_name)