httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dean gaudet <>
Subject Re: Filtering STATUS report :-)
Date Tue, 15 Aug 2000 13:05:26 GMT
On Tue, 15 Aug 2000, Tony Finch wrote:

> what is reasonable, based on its own semantics."). Secondly, we can
> add intelligence to the set-aside code so that copied TRANSIENT
> buckets can share the same underlying block in the heap, which would
> correspond to BUFF's fixed-size buffer except that it is more flexible
> -- it can store blocks of data that have passed through different
> numbers of filters, for example.

keeping around "free", fixed-sized blocks in a LIFO is totally worth it.  
you can go the next step over what 1.3 does by doing page-aligned,
page-sized allocations on demand.

lock contention on a free list is an issue... but then so is lock
contention on all the alloc stuff currently.

LIFO means you're more likely to stay in cache.

(i called this a pagemill in splim ;)


View raw message