On Tuesday 15 March 2011, Philippe De Muyter wrote:
On Mon, Mar 14, 2011 at 06:13:16PM +0100, Arnd Bergmann wrote: Here are the raw results :
tmp179:~ # for i in `seq 1 10`; do echo $i; ./flashbench --open-au --open-au-nr= $i --erasesize=$[1024 * 1024] /dev/sdb --random; done ... 4 1MiB 5.53M/s 512KiB 5.19M/s 256KiB 4.52M/s 128KiB 4.65M/s 64KiB 5.04M/s 32KiB 4.5M/s 16KiB 4.3M/s 5 1MiB 5.22M/s 512KiB 5.2M/s 256KiB 4.6M/s 128KiB 4.73M/s 64KiB 4.22M/s 32KiB 3.8M/s 16KiB 3.25M/s 6 1MiB 5.27M/s 512KiB 5.08M/s 256KiB 4.11M/s 128KiB 3.28M/s 64KiB 4.05M/s 32KiB 3.52M/s 16KiB 3M/s
It's not completely clear from the data, but I think the difference between 4 and 5 is significant, so I'd mark this card down as having four 1MB segments open, with a question mark. It may be doing something slightly smarter, because I would expect a much sharper drop in performance.
tmp179:~ # ./flashbench --findfat --fat-nr=10 --erasesize=$[1024 * 1024] /dev/sdb --random 1MiB 4.76M/s 5.53M/s 5.53M/s 5.53M/s 5.53M/s 5.53M/s 5.53M/s 5.53M/s 5.53M/s 5.53M/s 512KiB 5.02M/s 4.71M/s 4.42M/s 4.77M/s 4.77M/s 4.77M/s 4.77M/s 4.77M/s 4.9M/s 4.77M/s 256KiB 4.38M/s 4.43M/s 4.59M/s 4.46M/s 4.59M/s 4.59M/s 4.59M/s 4.59M/s 4.9M/s 4.9M/s 128KiB 4.3M/s 4.67M/s 4.55M/s 4.55M/s 4.55M/s 4.55M/s 4.55M/s 4.51M/s 4.57M/s 4.65M/s 64KiB 4.73M/s 4.58M/s 4.64M/s 4.62M/s 4.64M/s 4.64M/s 4.63M/s 4.63M/s 4.59M/s 4.62M/s 32KiB 4.45M/s 4.44M/s 4.45M/s 4.45M/s 4.44M/s 4.46M/s 4.45M/s 4.44M/s 4.43M/s 4.63M/s 16KiB 4.35M/s 4.26M/s 4.27M/s 4.27M/s 4.27M/s 4.27M/s 4.2M/s 4.26M/s 4.26M/s 4.26M/s
The first column seems different from all the other ones, so I'd assume that the card is meant for FAT16 with the FAT stored in the first megabyte, which is reasonable for a 2 GB card.
The one number that is currently missing is the page size. Can you rerun the findfat test with --blocksize=512?
I have added the other data from your test to https://wiki.linaro.org/WorkingGroups/KernelConsolidation/Projects/FlashCard... and hope that you are still motivated to do similar tests on other cards. That should be much faster now.
I'll also retest my own two CF cards for comparison.
Thanks for all your help!
Arnd