Using compiler flags in Linux to milk your software performance

edited January 2009 in Science & Tech
I wrote this article a while back and thought I would share it here. It details the concept of making use of compiler optimizations for individual processors under Linux or *BSD. Using this method you can gain increased responsiveness and lower execution times for most of your software including the kernel itself. You can speed your hardware up with overclocking and you can speed your software up with processor optimizations. Enjoy

http://nerdnos.net/index.php?guides&4

Comments

  • SnarkasmSnarkasm Madison, WI Icrontian
    edited January 2009
    Excellent article. Great to see somebody of your experience helping out through writing.
  • edited January 2009
    At NerdNOS we come from a world of antiquated server systems. You can forget about overclocking anything on these ancient things. We spent a bunch of time learning what made them run faster. I did a TON of research and testing in 06 involving recompiling FreeBSD and it's kernels when I was building an Intrusion Prevention System on an Appro AMD64 server. The difference on that machine was when I started, my GigE bridge would drop about 20% of the packets on inspection. When I was finished tweaking compiler optimizations and other small variables I got the packet loss to less than 0.001 percent. SourceFire sells a similar system for $30k+ based on similar technology but theirs requires much beefier hardware. My system handled more traffic than their model on lesser hardware. I did this with compiler optimizations in the kernel, in the packet capturing engine, in the logging and response engine, and in the drivers for the GigE broadcom cards.

    I'm sure a desktop user could use this for other things with similar results like Xserver or KDE binaries or whatever desktop tools they use.
Sign In or Register to comment.