[boost] Re: Re:Re: serialization library #17

From: Neal D. Becker (ndbecker2_at_[hidden])
Date: 2004-03-02 13:35:13

Jeff Flinn wrote:

[...]
>> >>I have 1 more question.  What about an archive that uses mmap?  This
>> >>should be very efficient for saving/restoring large structures in
>> >>binary format. Is this a good idea?
> 
> This topic has come up on the spirit list. Originally I was a proponent of
> mmap approach based on experiencing major performance improvements with
> earlier version of Windows particularly Windows CE. Today I've seen much
> less difference in performance. I've been told that Windows 2K/Xp using
> NTFS memorymaps files under the covers, and unless your actively adressing
> into the memory range there's little to be gained, atleast on Windows.
> Perhaps other os do this as well?
> 

I'm interested in linux myself.  In restore, a big potential speed gain is
possible if you don't actually use all the data.  In that case it is never
actually read from the disk.

For save, I'm speculating that it is possible for the OS to optimize saving
of unmodified pages using sparse files, but that maybe a bit farfetched.