[ubi][wl] the W/R performance of my wear-leveling unit

[ubi][wl] the W/R performance of my wear-leveling unit

[xiaochuan-xu]

Hi,

I've compared the current WL unit algorithm with 
my new WL unit one on the W/R time.


modprobe nandsim + ubi + ubifs.
And then copy a file (9338201 Bytes) to 

##time for i in `seq 1 20`; do sudo dd if=~/tmp/file of=/mnt/ubifs/file
count=1024&> /dev/null; sync; done

in the first time. the time of currenty algorithm and mine is as
following:

current UBI gives:

real	0m3.145s
user	0m0.124s
sys	0m0.700s

mine:
real	0m2.539s
user	0m0.140s
sys	0m0.772s


some more experiment share:

-- 
yours Sincerely,

xiaochuan-xu (许小川)

Re: [ubi][wl] the W/R performance of my wear-leveling unit

[Artem Bityutskiy]

On Fri, 2008-07-11 at 15:42 +0800, xiaochuan-xu wrote:
> Hi,
> 
> I've compared the current WL unit algorithm with 
> my new WL unit one on the W/R time.
> 
> 
> modprobe nandsim + ubi + ubifs.
> And then copy a file (9338201 Bytes) to 
> 
> ##time for i in `seq 1 20`; do sudo dd if=~/tmp/file of=/mnt/ubifs/file
> count=1024&> /dev/null; sync; done
> 
> in the first time. the time of currenty algorithm and mine is as
> following:
> 
> current UBI gives:
> 
> real	0m3.145s
> user	0m0.124s
> sys	0m0.700s
> 
> mine:
> real	0m2.539s
> user	0m0.140s
> sys	0m0.772s

Looks promising. I'll look at this again but a bit later.

Could you please think how your patch may be split on several separate
independent patches? 

-- 
Best regards,
Artem Bityutskiy (Битюцкий Артём)