Prerequisite

MPI implementation such as openmpi or MPICH

Install

Install from source

pip install -r requirements.txt
pip install .

Or Install from pip

pip install recorder-utils

Usage

recorder2csv /path/to/your_trace_folder/   /path/to/output.csv/

A csv trace file will be generated to the output file you specified

Test

Use the test file under test directory, the output will be in the data directory

Example

Under Recorder-util directory

python3 test/test_convert_text.py

API

RecorderReader:

• GM : RecorderMetadata
• funcs : string[], utf-8
• records : *PyRecord[], range(): 0 - total_rank-1
• LMs : LocalMetadata[], range(): 0 - total_rank-1

RecorderMetadata:

• total_ranks : c_int
• start_ts : c_double
• time_resolution : c_double
• ts_buffer_elements : c_int
• ts_compression_algo : c_int

LocalMetadata:

• total_records : c_size_t
• num_files : int
• filemap : set()
• function_count : int[], share the index with Recorder.funcs

PyRecord:

• tstart : c_double
• tend : c_double
• level : c_ubyte
• func_id : c_ubyte
• tid : c_int
• arg_count : c_ubyte

args_to_strs : string[], utf-8

Intervals

Current intervals are taken from POSIX and MPIIO layers. They are stored in the dataframe that return by the function build_offset_intervals. The intervals contain I/O records and corresponding file_id, rank, function, offset, size, start time, and end time.