[Zlib-devel] Plans for zlib 1.2.5

[Enrico Weigelt]

* Mark Adler <madler at alumni.caltech.edu> schrieb:

> By the way, in case you don't already take advantage of this, 
> if you do decompression in a single call of inflate (which 
> requires knowing the size of the decompressed data a priori), 
> then inflate will not allocate a 32K window buffer, 
> significantly reducing its memory usage.

On an MMU-based system w/ proper on-demand paging, you perhaps
coud use speculatively big anonymous mapping to /dev/zero etc
w/o having to go through malloc(). Maybe (depending on your
target), using a BIG stack buffer could also work good.

Both ideas assume that the (temporarily) allocated address 
space isn't really accessed unless zlib really uses it, so the
unused pages stay virtual.


cu
-- 
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------