View Single Post
Posts: 2,152 | Thanked: 1,490 times | Joined on Jan 2006 @ Czech Republic
#161
Hmm, did a bit non-scientific test of filesystem performance for both systems

ext2 on mmc

Code:
Nokia770-26:~# time tar >/dev/null cf - /usr
tar: Removing leading '/' from member names
tar: Names longer than 100 chars not supported.
tar: Names longer than 100 chars not supported.
tar: Names longer than 100 chars not supported.
tar: Names longer than 100 chars not supported.
real    1m 23.95s
user    0m 3.23s
sys     0m 12.28s
jffs2 on internal flash

Code:
Nokia770-26:~# time tar >/dev/null cf - /usr
tar: Removing leading '/' from member names
tar: Names longer than 100 chars not supported.
tar: Names longer than 100 chars not supported.
tar: Names longer than 100 chars not supported.
tar: Names longer than 100 chars not supported.
real    1m 11.50s
user    0m 3.44s
sys     0m 54.65s
This reads everything under /usr. Looks like jffs2 is a bit faster but look also on sys time. While with mmc sys time is only 12s (i.e the cpu is mostly idle and waits for mmc i/o) with jffs2 it is 54s (out of 71s) so the CPU is mostly maxed out while reading. This means that when you want your nokia to actually do something cpu intensive while reading (xvid playback, any application startup), using MMC is probably better.