Performance logging and find bottlenecks

Enable performance logging

monolish provides a logger to measure the execution time of each function at runtime and find bottlenecks. Functions on elements cannot be logged because the processing time is too short.

To enable logging, call the following function.

monolish::util::set_log_level(3);
monolish::util::set_log_filename(filename);

The default log level is zero, so normally the logger does not work.

By changing the log level, the detail of the log is determined.

0: No logging
1: Linear solver only
2: Linear solver and BLAS functions
3: Linear solver, BLAS functions, and Util functions

set_log_filename specifies the name of the output file. If the output file name is not specified, it will be output to the standard output.

The output log is in yml format.

Install monolish-log-viewer

The monolish project has developed a tool to analyze and visualize the output log files. The program is available in monolish/python/.

It has been uploaded to the PyPI server and can be installed using the pip command.

If pip is not available, use the following command.

apt install python-3-pip

Install the logger using pip. Execute the following command:

python3 -m pip install monolish-log-viewer

How to use monolish-log-viewer

monolish_log_viewer outputs HTML files from log files in yml format output by monolish. To generate log.html from log.yml, execute the following command.

monolish_log_viewer log.yml log.html

NOTE:

The sum of the func. and util. times may not match the solver time, because the solver, func. and util. functions are timed separately.