[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Fwd: Re: Re: Speeding up the mounting of JFFS Device



David:

More correspondence.

Caveat: Even though I wrote the node cache code.
I do not believe this is the right long term solution.
I was in a hurry.

Bye!
Dan
---- Begin Included Message ----

From: "Dan" <dmcdon@xxxxxxx.com>
Sent: Tue, 31 Oct 2000 16:46:33 -0700
To: afong@xxxxxxx.org
Subject: Re: Re: Speeding up the mounting of JFFS Device


> Hi Dan,

> It seems that my testing yesterday were somewhat invalid. Anyways, 
> I've been gathering some statistics and the node cache really makes
> things alot nicer.

> Thanks!
> Amy

> Without node cache:

[..snip]

> /dev/mtdblock1 (2 megs - 640k)
> real	0m1.384s user	0m0.000s sys	0m1.380s
> real	0m1.385s user	0m0.000s sys	0m1.380s

> /dev/mtdblock2 (14 megs)
> real	0m13.941s user	0m0.010s sys	0m13.850s
> real	0m13.861s user	0m0.000s sys	0m13.860s

> With node cache:

> /dev/mtdblock1 (2 megs - 640k)
> real	0m4.250s user	0m0.000s sys	0m4.240s
> real	0m1.272s user	0m0.020s sys	0m1.250s

> /dev/mtdblock2 (14 megs)
> real	0m15.895s user	0m0.000s sys	0m15.890s
> real	0m0.218s user	0m0.020s sys	0m0.200s

Interesting results! Those are the types of performance
improvements that I have been experiencing. The first
time mounting with node cache still performs the
full scan_flash() to get the data structures to write.
If you add in the writing time, the node cache takes
a few seconds longer the first time through. The
real gain comes in subsequent mounts as your testing
proves.

During the building of the file system, the return
 from scan_flash() also expect the "delete_list" to
be initialized. This is one element that is currently
missing from the node cache. I need to add this
in. I have been pressed for time lately and haven't
got that done. Fortunately, as long as you're not
deleting files this is set to NULL anyway.

I have been able to add files to node cache partitions
without problems, so far. But it really needs to be fixed.

Again, thanks for your help with my bug. That's one
less I need to uncover later.



--------------------------------
Bye!
Dan McDonald
The Late Night Software Shop 
Sign up today for your Free E-mail at: http://www.canoe.ca/CanoeMail 

---- End Included Message ----





--------------------------------
Bye!
Dan McDonald
The Late Night Software Shop 
Sign up today for your Free E-mail at: http://www.canoe.ca/CanoeMail 

To unsubscribe from this list: send the line "unsubscribe jffs-dev" in
the body of a message to majordomo@xxxxxxx.com