Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: WAPBL on multicore, 1 (ONE!) process...

> >
> Ah yeah, that was the PR I was referring to also. But I always thought
> that one of the test runs were done without WAPBL. I now see all of them
> have it on. It would be interesting to know the runtime of the build
> without -o log on that PR, just for comparison. It's probably something
> like 15 minutes...

Quoting from the PR addition I wrote:

  I've repeated this experiment on a non-wapbl file system on the same
  system, but there the difference was in the "sync loop" disfavour:

  5799.240u 2232.222s 24:57.93 536.1%     0+0k 494+710112io 32612pf+0w
  5881.837u 2435.640s 27:14.19 508.9%     0+0k 0+720501io 32646pf+0w

  and the number of IOs is comparable.

So... not 15 minutes, but also not over 2 hours, which was the
state with WAPBL before adding that unpublished (?) patch of
Simon and turning off disk cache flushing.


- Håvard

Home | Main Index | Thread Index | Old Index