Re: 2.6.19-rc5-mm2

[Eric Dumazet <dada1@cosmosbay.com>]

On Tuesday 14 November 2006 10:41, Andrew Morton wrote:
> Presently at
>
> http://userweb.kernel.org/~akpm/2.6.19-rc5-mm2/
>
> and will appear later at
>
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.19-rc5/2.
>6.19-rc5-mm2/

Hi all

I noticed a slowdown (3%) on a io micro-benchmark on my machine with 
2.6.19-rc5-mm2

It appears time-uninline-jiffiesh.patch is sub-optimal at least for current 
compilers (tested gcc-4.0.4 here)

May I suggest :

1) make sure jiffies_to_usecs() is defined before being used in 
timespec_trunc() : Compiler will just optimize away not *needed* code.

OR :

2) Revert to inline versions of four functions jiffies_to_msecs(), 
jiffies_to_usecs(), msecs_to_jiffies() and usecs_to_jiffies() .

IMHO there is litle gain to call a function just to perform so basic 
arithmetics, that sometime compiler can perform at compilation time.

OR 

3) replace
	(jiffies_to_usecs(1) * 1000)
by
	(NSEC_PER_SEC / HZ)

With current patch, timespec_trunc() is not anymore a tail function.

struct timespec timespec_trunc(struct timespec t, unsigned gran)
{
        if (gran <= jiffies_to_usecs(1) * 1000) {

Much better here to have :

	if (gran < SOME_CONSTANT) 


Thank you
Eric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/