Re: /tmp as tmpfs and consequence for imaging software

[bastien ROUCARIES <roucaries.bastien@gmail.com>]

Le Monday 14 November 2011 00:14:18, Josselin Mouette a écrit :
> Le dimanche 13 novembre 2011 à 23:20 +0100, Bastien ROUCARIES a écrit :
> > No it does not work like you said. We know the matrix structure, not
> > the kernel. We map and unmap manually. Doing as you said is
> > inneficient and trash a lot cache and so on.
> 
> This is getting insane. Please learn how to use madvise and
> posix_fadvise and let the kernel deal with paging. The kernel knows
> everything about the underlying hardware; the application does not.

Yes I do it with  MADV_DONTNEED when needed. And the kernel do not know when to use MADV_DONTNEED on 16MB block and as you said I 
have performance enhancement (but we should take in account system call cost). 

> 
> > Memory is used as a cache. This is not broken. This a valid use.
> 
> By paging the data manually to the disk, the only thing you are
> achieving is duplicating it at the time of reading/writing. Please learn
> how to use memory pages and stop telling people you know better when it
> is obvious you don’t know what you are talking about.

Thanks