Re: AIM7 40% regression with 2.6.26-rc1

["Zhang, Yanmin" <yanmin_zhang@linux.intel.com>]

On Wed, 2008-05-07 at 20:34 -0700, Linus Torvalds wrote:
> 
> On Thu, 8 May 2008, Zhang, Yanmin wrote:
> > 
> > On Tue, 2008-05-06 at 10:23 -0600, Matthew Wilcox wrote:
> > > On Tue, May 06, 2008 at 06:09:34AM -0600, Matthew Wilcox wrote:
> > > > So the only likely things I can see are:
> > > > 
> > > >  - file locks
> > > >  - fasync
> > > 
> > > I've wanted to fix file locks for a while.  Here's a first attempt.
> > > It was done quickly, so I concede that it may well have bugs in it.
> > > I found (and fixed) one with LTP.
> > > 
> > > It takes *no account* of nfsd, nor remote filesystems.  We need to have
> > > a serious discussion about their requirements.
> >
> > I tested it on 8-core stoakley. aim7 result becomes 23% worse than the one of
> > pure 2.6.26-rc1.
> 
> Ouch. That's really odd. The BKL->spinlock conversion looks really 
> obvious, so it shouldn't be that noticeably slower.
> 
> The *one* difference is that the BKL has the whole "you can take it 
> recursively and you can sleep without dropping it because the scheduler 
> will drop it for you" thing. The spinlock conversion changed all of that 
> into explicit "drop and retake" locks, and maybe that causes some issues. 
> 
> But 23% worse? That sounds really odd/extreme.
> 
> Can you do a oprofile run or something?
I collected oprofile data. It looks not useful, as cpu idle is more than 50%.

samples  %        app name                 symbol name
270157    9.4450  multitask                add_long
266419    9.3143  multitask                add_int
238934    8.3534  multitask                add_double
187184    6.5442  multitask                mul_double
159448    5.5745  multitask                add_float
156312    5.4649  multitask                sieve
148081    5.1771  multitask                mul_float
127192    4.4468  multitask                add_short
80480     2.8137  multitask                string_rtns_1
57520     2.0110  vmlinux                  clear_page_c
53935     1.8856  multitask                div_long
48753     1.7045  libc-2.6.90.so           strncat
40825     1.4273  multitask                array_rtns
32807     1.1470  vmlinux                  __copy_user_nocache
31995     1.1186  multitask                div_int
31143     1.0888  multitask                div_float
28821     1.0076  multitask                div_double
26400     0.9230  vmlinux                  find_lock_page
26159     0.9146  vmlinux                  unmap_vmas
25249     0.8827  multitask                div_short
21509     0.7520  vmlinux                  native_read_tsc
18865     0.6595  vmlinux                  copy_user_generic_string
17993     0.6291  vmlinux                  copy_page_c
16367     0.5722  vmlinux                  system_call
14616     0.5110  libc-2.6.90.so           msort_with_tmp
13630     0.4765  vmlinux                  native_sched_clock
12952     0.4528  vmlinux                  copy_page_range
12817     0.4481  libc-2.6.90.so           strcat
12708     0.4443  vmlinux                  calc_delta_mine
12611     0.4409  libc-2.6.90.so           memset
11631     0.4066  bash                     (no symbols)
9991      0.3493  vmlinux                  update_curr
9328      0.3261  vmlinux                  unlock_page