On Fri, 27 Oct 2000, Jeff V. Merkey wrote:
>
> Linux has lots of n-sqared linear list searches all over the place, and
> there's a ton of spots I've seen it go linear by doing fine grained
> manipulation of lock_kernel() [like in BLOCK.C in NWFS for sending async
> IO to ll_rw_block()]. I could see where there would be many spots
> where playing with this would cause problems.
>
> 2.5 will be better.
fs/locks.c is one hell of a sick puppy. Nothing new about that. I'm kinda
curious about "n-squared" searches in other places, though - mind showing
them?
BTW, what spinlocks get contention in variant without BKL? And what about
comparison between the BKL and non-BKL versions? If it's something like
BKL no BKL
4-way 50 20
8-way 30 30
- something is certainly wrong, but restoring the BKL is _not_ a win.
I didn't look into recent changes in fs/locks.c, but I have quite problem
inventing a scenario when _adding_ BKL (without reverting other changes)
might give an absolute improvement. Well, I see a couple of really perverted
scenarios, but... Seriously, folks, could you compare the 4 variants above
and gather the contention data for the -test9 on your loads? That would help
a lot.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Oct 31 2000 - 21:00:20 EST