Ross Harvey wrote:
> Now, supposedly, the DTYPE_PIPE code reuses the same buffer bits to get
> better cache performance. Does that comment just represent wishful
> thinking?

The implementation of new pipes contains some explicit code to
be cache friendly, like resetting the ring buffer position pointer to zero
if buffer is emptied.

> I like the fact that DTYPE_PIPE uses unwired buffer space, and
> doesn't compete for nmbclusters (a common bottleneck) -- but it's a bit
> discouraging to think it only helps performance on older CPUs. Still,
> probably worth doing.

It's also usefull if the processor would be busy. During tests, my
machine was mainly idle. The couple-concurrent-bw_pipe
test was most likely limited by context switch time.
 
> I would guess that patching those buffer sizes makes no difference on
> your fast CPU?

Didn't try.

Jaromir
-- 
Jaromir Dolecek <jdolecek@NetBSD.org>      http://www.ics.muni.cz/~dolecek/
NetBSD - just plain best OS! -=<>=- Got spare MCA cards or docs? Hand me them!