Re: [gentoo-amd64] Re: tmpfs help

[Richard Freeman]

Volker Armin Hemmann wrote:
> On Mittwoch, 13. Februar 2008, Duncan wrote:
>>> removed lots of irrelevant 'my hardware is so cool' stuff'.
>
> You forget some (little) things. Not everything can be swapped out.

How is this relevant?  Some memory is locked or effectively locked. 
That limits buffers/cache in the disk-based fs case, and swap in the 
tmpfs case.  I don't see that there is much of a difference.

 > Swap is
> extremly slow AND it is much worse to swapout/swapin programm code that
> should be run, instead of fetching some files from disk while the programm
> runs.

You keep asserting this.  Why should fetching a page from swap be slower 
than fetching the equivalent space from a file?  In fact, if mmap is 
used to read a file the same code probably is used for both.

And the kernel is not likely to swap out recently-run program pages.  It 
isn't going to keep a bunch of compiled binaries in memory for 10 
minutes while it swaps in and out the compiler - it would swap in and 
out source files and binary files - just as with a disk-based 
filesystem. If for some reason part of the compiler is swapped out, it 
would be rewritten to disk every time it got swapped out - only the 
first time.  I'm sure the kernel would notice that the pages were never 
written since the last swap and would just flush them from memory 
without another write.  However, this should still only happen on idle 
pages, and swapping those out is probably better than any other use of 
memory (and those pages would get swapped out even with a disk-based 
filesystem in use).

I'm of course interested in benchmarks.  In theory swap should be faster 
than files.  However, implementation does matter and maybe there is some 
flaw that makes things slower.  My assertion is that a swap-based 
solution gives the kernel more freedom to optimize memory use than a 
disk-based solution.  So, this should be a fundamentally-superior solution.
-- 
gentoo-amd64@l.g.o mailing list