Re: [PATCH] linux-2.5.44_vsyscall_A1

[Jeff Garzik <jgarzik@pobox.com>]

john stultz wrote:
> All,
> 	Here is the next rev of the i386 vsyscall gettimeofday port I sent
> earlier. I've added a fix for the linker trouble in the last one, as
> well as wrapped the whole thing in CONFIG_VSYSCALL. 
> 
> 	This patch implements gettimeofday in a user readable page, allowing
> for calls to gettimeofday to execute completely in userspace, giving a
> significant performance boost. 
>         
> 	Changes to glibc are unnecessary, because users that want to use the
> vsyscall can do so by LD_PRELOADING a library which alias gettimeofday
> before executing their application. This will not affect any other
> application and allows the backward compatibility issue to be ignored.
> I've created an example library (to be attached in a following email)
> and ran a quick performance test w/ and w/o the preloaded library,
> giving the following results:
>         
> Normal gettimeofday
> gettimeofday ( 1403307us / 1000000runs ) = 1.403306us
> vsyscall LD_PRELOAD gettimeofday
> gettimeofday ( 285423us / 1000000runs ) = 0.285423us
> 
> Since this code uses the TSC for calculating time of day, this patch
> will not help systems that suffer from TSC skew (ie: many NUMA systems,
> etc). However, for UP and SMP boxes this is a pretty major win.
> Alternative methods to use the cyclone/HPET registers for NUMA boxes are
> also feasible in the future. 
> 
> Please let me know your thoughts about 2.5 integration. 



In terms of implementation, I think it's way too x86-specific...  some 
of the vsyscall infrastructure can be more generic, making it easier for 
other arches to implement the same functionality.  Also use of TSC isn't 
a terribly good idea...

Overall the idea has existed for a while, and I think it's a good one... 
  gettimeofday(2) is very definitely a performance bottleneck in 
databases, in apache [if nothing's changed in that area...], etc.

	Jeff