On 04/12/2012 01:10 AM, Peter Easthope wrote:
Any tips about Flashbench? Googling "site:debian.org flashbench"
suggests it finds little interest in the Debian community.
Thanks, ... Peter E.
I got some odd results, I'm still trying to figure it all out & I'd like
some help too, I'm really just sorting out my confused notes here as
while I was trying to do this a couple of weeks ago my thinking jelly
got ouchie, this link has helped a bit,
http://blogofterje.wordpress.com/2012/01/14/optimizing-fs-on-sd-card/
one problem is I don't really see much difference in the benchmark
results, probably because my benchmarking script is so rubbish the drive
isn't the bottleneck.
after installing git & build-essential I did
mkdir ~/git
cd ~/git/
git clone git://git.linaro.org/people/arnd/flashbench.git
cd ~/git/flashbench/
ls
#vim ~/git/flashbench/README
make
# simple I know but..
# find your target drive & take a note of the factory defined start
sector as this is hopefully aligned on an erase block
fdisk -cul
# so my 32GB SanDisk Ultra SDHC card looked like
Disk /dev/sdc: 31.9 GB, 31914983424 bytes
255 heads, 63 sectors/track, 3880 cylinders, total 62333952 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000
Device Boot Start End Blocks Id System
/dev/sdc1 8192 62333951 31162880 c W95 FAT32 (LBA)
# my SanDisk USB pen drive & my Kingston SD card had a start sector of
2048 but my 120GB OCZ Agility3 had no partition table on it at all
# Define variable for the target drive so commands can be pasted into a
terminal window (replace X with your drive letter)
trg=/dev/sdX
# Backup the partition table for the target drive
mkdir -p ~/backup/
umount ${trg}*
dd if=${trg}* of=~/backup/${trg}_mbr_512.ddimg bs=512 count=1
# fill the start of the drive with random noise as apparently unused
drives sometimes give odd numbers THIS WILL BLANK THE DRIVE, DON'T DO
THIS IF THERE IS ANYTHING ON THE DRIVE YOU WANT, IN FACT YOU NEED TO
UNCOMMENT THE DD COMMAND IF YOU WANT TO ACTUALLY DO IT
umount ${trg}*
#dd if=/dev/urandom of=${trg} bs=1M count=128
#Lets run flashbench, this is on my 32GB SanDisk Ultra SDHC card
cd ~/git/flashbench/
./flashbench -a ${trg}
align 8589934592 pre 1.8ms on 2.02ms post 1.82ms diff
211µs
align 4294967296 pre 1.81ms on 1.97ms post 1.76ms diff
179µs
align 2147483648 pre 1.84ms on 2.03ms post 1.83ms diff
193µs
align 1073741824 pre 1.82ms on 1.95ms post 1.85ms diff
117µs
align 536870912 pre 1.78ms on 1.99ms post 1.87ms diff 162µs
align 268435456 pre 1.84ms on 2ms post 1.83ms diff 167µs
align 134217728 pre 1.8ms on 1.99ms post 1.78ms diff 203µs
align 67108864 pre 1.8ms on 1.92ms post 1.83ms diff 104µs
align 33554432 pre 1.78ms on 1.88ms post 1.8ms diff 91.3µs
align 16777216 pre 1.79ms on 1.93ms post 1.8ms diff 134µs
align 8388608 pre 1.78ms on 1.85ms post 1.73ms diff 102µs
align 4194304 pre 1.79ms on 1.84ms post 1.79ms diff 54.6µs
align 2097152 pre 1.8ms on 1.83ms post 1.79ms diff 32.1µs
align 1048576 pre 1.82ms on 1.81ms post 1.82ms diff -9261ns
align 524288 pre 1.85ms on 1.84ms post 1.83ms diff 7.86µs
align 262144 pre 1.79ms on 1.82ms post 1.84ms diff 3.73µs
align 131072 pre 1.8ms on 1.78ms post 1.83ms diff -30388n
align 65536 pre 1.8ms on 1.79ms post 1.79ms diff -4769ns
align 32768 pre 1.84ms on 1.79ms post 1.76ms diff -13269n
# to help you spot where you are in the output 134217728 = 128MB &
524288 = 512k; so as the difference halves from ~100µs or more to 55µs
at 4194304 = 4MB if I'm reading the README correctly this would imply
that the erase block size on this card is 8MB
One thing I don't understand is, if that is the case, how come the
origonal partition started at 4MB
8192 * 512 = 4194304
Irrespective I now start the partition for this SD card at 16384 that
way, it's allied on a 4MB & an 8MB boundary, I haven't yet figured out
what open AUs are and what to do with them, I got slightly disheartened
when on my OCZ SSD drive it didn't seem to make any difference to speed
where I started the partition, I need to try again with a CF or SD card,
maybe it was a buffering thing.
That's where I've got so far, not sure I've helped you but my note are
now a little more ordered.
Let us know how you get on
BstRgetc
--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4f8a6cf4.9060...@homeurl.co.uk
