Re: [Lse-tech] lockmeter results comparing 2.4.17, 2.5.3, and 2.5.5

[Hanna Linder <hannal@us.ibm.com>]

--On Wednesday, February 27, 2002 16:48:07 -0500 Alexander Viro <viro@math.psu.edu> wrote:
> 
>> > looks a little distressing - the hold times on inode_lock by prune_icache
>> > look bad in terms of latency (contention is still low, but people are still
>> > waiting on it for a very long time). Is this a transient thing, or do people
>> > think this is going to be a problem?
>> 
>> inode_lock hold times are a problem for other reasons.
> 
> ed mm/vmscan.c <<EOF
> /shrink_icache_memory/s/priority/1/
> w
> q
> EOF
> 
> and repeat the tests.  Unreferenced inodes == useless inodes.  Aging is
> already taken care of in dcache and anything that had fallen through
> is fair game.
> 

I applied your patch and reran the tests. Looks like you solved the problem:

  SPINLOCKS         HOLD            WAIT
  UTIL  CON    MEAN(  MAX )   MEAN(  MAX )(% CPU)     TOTAL NOWAIT SPIN RJECT  NAME

        7.1%  0.7us(  19ms)  7.7us(  17ms)( 2.6%) 779799309 92.9%  7.1% 0.00%  *TOTAL*

 0.16% 0.29%  0.6us(  91us)  2.2us(  46us)(0.00%)   5495642 99.7% 0.29%    0%  inode_lock

 0.90% 0.47%  1.4us(  19ms)  280us(  17ms)(0.10%)  12681192 99.5% 0.47%    0%  kernel_flag

The results are again stored at http://lse.sf.net/locking . 

Hanna