Conversation

Notices

1. Embed this notice
   Andrew Kelley (andrewrk@mastodon.social)'s status on Friday, 07-Feb-2025 09:11:38 JST Andrew Kelley

   last night I had an idea for a fast multi-threaded allocator, whipped it up, and at first glance it outperforms glibc.

   need to do some more testing, but it's looking good so far

   it's only 212 lines of code!

   • Embed this notice
     feld (feld@friedcheese.us)'s status on Friday, 07-Feb-2025 09:11:38 JST feld

     in reply to
     @andrewrk I am a complete idiot when it comes to the real technical discussion here, but I've been watching the development of snmalloc from a distance which may very well and up landing in FreeBSD to replace jemalloc. I haven't seen any projects actually adopt it elsewhere. If it would be beneficial for Zig that would be neat. Don't mean to derail your R&D project here, just felt like a suitable place to mention its existence.
     
     https://github.com/microsoft/snmalloc
   • Embed this notice
     Andrew Kelley (andrewrk@mastodon.social)'s status on Friday, 07-Feb-2025 10:59:19 JST Andrew Kelley

     in reply to

     • feld

     @feld thanks, I read the paper just now and it helped me to understand why some measurements were favorable and others not - specifically the pattern of one thread allocating and another thread freeing in my simple design will result in ever-growing RSS

     I know I "should" start with the paper but for me papers are too boring until I've already tried from first principles

   • Embed this notice
     feld (feld@friedcheese.us)'s status on Friday, 07-Feb-2025 10:59:19 JST feld

     in reply to

     • David Chisnall (*Now with 50% more sarcasm!*)
     @andrewrk you can also find David here @david_chisnall