def get_inst_task_num_ratio()

in cluster-trace-gpu-v2020/analysis/utils.py [0:0]


def get_inst_task_num_ratio(dfa, inst_num_list=[2, 8, 20, 64, 100, 256, 512]):
    total_num_task, total_num_inst = len(dfa), sum(dfa['inst_num'])
    data_df = []
    for i in inst_num_list:
        temp_df = dfa[dfa['inst_num'] >= i]
        task_num_ratio = len(temp_df) / total_num_task
        inst_num_ratio = sum(temp_df['inst_num']) / total_num_inst
        data_df.append([task_num_ratio, inst_num_ratio])
    out_df = pd.DataFrame(data_df, columns=['num_task_ratio','num_inst_ratio'])
    out_df = out_df.T.rename(columns=dict(zip(range(len(inst_num_list)), inst_num_list)))
    return out_df