Re: Question about Lua 5.1.3 and realloc() on life.lua

["Bogdan Marinescu" <bogdan.marinescu@...>]

>  Umm, reinventing the wheel?
Not really. Just giving my statistics module something to chew on :)
It appeared to be easier than trying to understand how to get/compute
fragmentation information from dlmalloc. And it still appears so.
mallinfo() doesn't quite tell me the whole story.
>  But if you really, really want to write your own allocator for
>  Lua, here are my suggestions: use quick-fit with a relatively
>  coarse quantization (*) and tune the fixed bin sizes to the
>  observed allocations. Use lazy coalescing for the fixed bin sizes
>  and immediate coalescing for the medium sizes. Round up large
>  sizes to pages and let the OS manage them. Block splitting should
>  look for size 2n+k at least.
Errr. What pages? What OS? :) There aren't any of these strange
animals on my ARM (no MMU, in case you're wondering). As for the fixed
bin sizes, I tried it, and 32 seems to be a value that satisfies most
applications. Not life.lua though :)
Thanks for the tips, I might try them eventually.