lkml.org 
[lkml]   [2008]   [Nov]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [git pull] scheduler updates


On Sat, 8 Nov 2008, Linus Torvalds wrote:
>
> Can we please just remove that idiocy? Or can somebody give a _sane_
> argument for it?

Btw, the only _possible_ sane argument I see is

- two consecutive rdtsc calls
- timing the code in between
- the code in between is not self-serializing

and quite frankly, if that's the case, then it's _that_ code that should
have the barriers, not some generic "[v]get_cycles()".

IOW, the rdtsc_barrier may make sense when you're synchronizing the TSC to
some other hardware event (eg the "tie the TSC to the HPET" kind of
code), but then the barriers are about the code, not about the TSC access
itself.

Linus


\
 
 \ /
  Last update: 2008-11-08 19:41    [from the cache]
©2003-2011 Jasper Spaans