Conversation

Notices

1. Embed this notice
   A myriad of Qyriad (qyriad@chaos.social)'s status on Sunday, 04-Aug-2024 23:50:47 JST A myriad of Qyriad

   Honestly one of the more interesting things we're finding with Zig is just that there's no global allocator at all. Things that need to allocate conventionally take an allocator parameter and the application code decides what allocator to use

   It does have the problem of needing to pass a helper param around and around in any wrappings APIs, which of course will bubble up if you realize late that some function deep in the call stack needs to allocate

   But it's a really interesting design choice

   • Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Sunday, 04-Aug-2024 23:51:16 JST Rich Felker

     in reply to

     @Qyriad There are all sorts of things that are slow because of gratuitous fine grained allocation and you could just pass them a bump or obstack like allocator where free is a nop, and nuke the whole thing when the aggregate task is done.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     A myriad of Qyriad (qyriad@chaos.social)'s status on Sunday, 04-Aug-2024 23:51:17 JST A myriad of Qyriad

     in reply to

     *most* of the time you're going to want to use the same allocator for everything, but occasionally you might e.g. know that there's a reasonably small upper bound on the memory usage of something and use a stack-backed allocator for a few specific functions

     This sounds *extremely* niche to us though and I'm not sure when it would actually be beneficial