Skip to main content

mastodon - Link to source

During my recent debugging of a `Future`-related memory leak issue, I had occasion to swap a program from using `Future::XS`, to the pure-perl `Future::PP` implementation instead.

As the program runs a very flat stable CPU load, such a change made it easy to see the performance different such a change made.

#perl

A monitoring graph showing "CPU" load over a few days. The graph has two distinct behaviours, hovering over a very stable line around 0.2%, or a very stable 0.24%. The two behaviours are labelled with text arrows, "Future::XS" on the lower 0.2% ones, and "Future::PP" on the higher 0.24% ones. There are a couple of large momentary spikes in the load graph, but these are likely attributed to restart or reload activities and outliers in the overall trend.
#perl
This entry was edited (2 days ago)
in reply to Paul Evans

Paul Evans
mastodon - Link to source
Paul Evans
• •

In absolute terms it's tiny because this program does very little CPU work - it's just a monitoring bridge between a serial port and an HTTP prometheus "metrics" exporter.

But still - a drop from 0.24% with the pureperl to 0.20% with XS, means the XS module does speed the program up by about 16%.

It's nice to be able to so easily see on a graph such a performance improvement by using the XS version, which might otherwise be somewhat hard to measure.

This entry was edited (2 days ago)
⇧