[PATCH 0/2] Improve GTA02 NAND read performance by 41%
micael.h at gmail.com
Wed Oct 22 12:51:00 CEST 2008
On Wed, Oct 22, 2008 at 2:14 AM, Andy Green <andy at openmoko.com> wrote:
> | Even with the improved timings applied on stable-tracking and bs=2048,
> | the "old" kernel from 2.4.24 are still faster.
> | hmm.. I'll need to check this more tomorrow..
> Thanks for looking at it... NAND access on stable-tracking is borked
> anyway, something about ECC stuff blows CRC errors when you mount jffs2.
> ~ Did you turn off ECC support? If not you will likely find a bunch of
> junk in dmesg that can explain the slowdown.
I disabled the NAND debug option as it was spamming dmesg with prints
and making access super slow. (no errors)
I still used the HWECC option as the 2.6.24 kernel is built with.
However I did re-enable the prints for when the HW ECC didn't match to
the calculated one, but none of them could be seen in dmesg and I
could mount the nand-root partition successfully.
A good test could be to disable the ECC stuff.. Also the better
performance with bs=2048 gives some indication that SW is causing some
of the slowdown.
Anyhow, I'll check this more later tonight.
More information about the openmoko-kernel