Profiling

Core Technology

Article from Issue 203/2017
Author(s):

We all want our programs to run fast. With profiling, you can locate the code that is slowing you down and fix it.

While hardware becomes faster and faster, the software it runs becomes slower and slower. Although this is a joke, it carries a bit of truth. Quite often, the software uses suboptimal algorithms, introduces costly abstractions, or misses hardware caches. This is okay most of the time, because hardware is forgiving and has enough capacity for many tasks. The rest of the time, you want your code to perform as fast as possible.

Optimizing a program is not the same as rewriting it from scratch the right way. The majority of code is generally not in the "hot path." It's executed only once in awhile during the program's lifetime, and investing optimization efforts into these parts will never pay off. So, you need a tool to tell you which code paths are consuming most of the CPU time.

A technique called "profiling" comes into play here. You build an execution profile of the program (hence the name) that tells you which functions the program runs and how much time they take; then, you look at the profile and decide what you can do about it. While the last step is a bit of an art, the first step just requires a tool that is essential to any developer. Luckily, Linux offers several such tools.

[...]

Use Express-Checkout link below to read the full article (PDF).

Buy this article as PDF

Express-Checkout as PDF
Price $2.95
(incl. VAT)

Buy Linux Magazine

SINGLE ISSUES
 
SUBSCRIPTIONS
 
TABLET & SMARTPHONE APPS
Get it on Google Play

US / Canada

Get it on Google Play

UK / Australia

Related content

  • perf

    The kernel supports performance analysis with built-in tools via the Linux performance counters subsystem. perf is easy to use and offers a detailed view of performance data.

  • Optimizing Linux Performance

    When you're tuning up Linux performance, you need to keep an eye on a number of components and parameters. This article describes some techniques for optimizing performance by tuning TCP/IP, reducing latency, and configuring NUMA-aware systems. You'll also learn about CPU scheduling and memory management tuning.

  • Dance on a Volcano

    The RAW converter Vulkan darktable outpaces its competitors with a modern node-graph-based architecture and massive use of the GPU.

  • Improved Multi-Threading Performance: Google's Perf Tools Version 1.0

    Google Performance Tools, part of Google's open source projects, is now available in version 1.0. Perf Tools is especially useful for optimizing multi-threaded programs.

  • FOSSPicks

    This month Graham dives into LMMS, Name Generator, Heaptrack, Binaural Audio Editor, Meshroom, Palapeli, and much more!

comments powered by Disqus