JFFS3 & performance

[Artem B. Bityuckiy]

On Fri, 17 Dec 2004, Joakim Tjernlund wrote:

> > 
> > > Another idea, calculate the CRC backwards(from the end of the buffer to the beginning).
> > > That give better L1 cache behaviour.
> > >
> > Why (may be some URL)? For all archs? 
> 
> example, when writing a buffer to flash you first CRC it from top to bottom.
> The bottom part of the buffer is then newer than the top of then buffer in the
> L1 cache or the top part has already been evicted from the L1 cache depending
> of the size of the buffer and size of L1 cache.
> 
> Now you start again from the top of the buffer to actually write the buffer
> to flash and you probaly need to read the data from RAM into L1 cache again.
> 
> Reading from flash will cause similar effects.
> 
> If you calculate CRC backwards you will avoid trashing the cache needlessly.
> 
>  Jocke
> 
Ok, thanks, good idea, I'll put it to the TODO file too.
 

--
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.