Re: [PATCH 0/4] Reclaim page capture v4

[KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>]

> > I tested your patch in my desktop.
> > The test is just kernel compile with single thread.
> > My system environment is as follows.
> > 
> > model name	: Intel(R) Core(TM)2 Quad CPU    Q6600  @ 2.40GHz
> > MemTotal:        2065856 kB
> > 
> > When I tested vanilla, compile time is as follows.
> > 
> > 2433.53user 187.96system 42:05.99elapsed 103%CPU (0avgtext+0avgdata
> > 0maxresident)k
> > 588752inputs+4503408outputs (127major+55456246minor)pagefaults 0swaps
> > 
> > When I tested your patch, as follows.
> > 
> > 2489.63user 202.41system 44:47.71elapsed 100%CPU (0avgtext+0avgdata
> > 0maxresident)k
> > 538608inputs+4503928outputs (130major+55531561minor)pagefaults 0swaps
> > 
> > Regresstion almost is above 2 minutes.
> > Do you think It is a trivial?
> > 
> > I know your patch is good to allocate hugepage.
> > But, I think many users don't need it, including embedded system and
> > desktop users yet.
> > 
> > So I suggest you made it enable optionally.
> 
> Hmmm.  I would not expect to see any significant impact for this kind of
> workload as we should not be triggering capture for the lower order
> allocations at all.  Something screwey must be occuring.  I will go and
> reproduce this here and see if I can figure out just what causes this.

yup.
I also think this is significant regression.

if this is reproduced, that patch shouldn't be merged to -mm, IMHO.




--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>