On 27/11/11 01:08, Dave Budenberg wrote:
> (I note that the guidance at http://www.yaffs.net/yaffs-memory-footprint
> seems significantly different from what we are seeing.)

That calculation is for the old YAFFS1 code.

I made a worked example for YAFFS2 on eCos a couple of years ago, based 
on updated information provided by Charles at the time. The result can 
be seen at 
http://www.ecoscentric.com/ecospro/doc/html/ref/yaffs-mem-reqs.html .

> 1. Is what I am seeing appropriate for the configuration we are using?
> 2. Is it realistic to achieve 50kB code size & around 10kB RAM using Yaffs?

I haven't looked into the code size recently, but that sort of RAM 
footprint will only be possible with a very small number of files.

If you're using YAFFS2, you may be able to squeeze the code size a bit 
by conditioning out the code specific to YAFFS1.

> 3. If Yaffs can’t meet our resource requirements, what alternative file
> systems might be appropriate?

YAFFS is designed for raw NAND, but an SD card is several steps removed 
from that and intended to be used by a regular desktop filesystem (vfat, 
ext2, etc). Provided you trust that your SD card has a good flash 
translation layer providing wear levelling, one of the desktop 
journalling filesystems might be worth looking into - though I fear that 
that many will have similarly onerous RAM requirements.

Ross