[Cerowrt-devel] juhosg rocks
dave.taht at gmail.com
Tue Mar 27 16:53:34 EDT 2012
There are a lot of people involved in openwrt, and one of the more
invisible (doesn't do irc) is juhosg.
(This might account for his productivity, I don't know!)
Among many other things, Gabor is responsible for much of the ar71xx
architecture that is in cerowrt.
This is the commit that landed today. My hat is off to Gabor...
(and darn it, I have to do a new build now)
Author: juhosg <juhosg at 3c298f89-4303-0410-b956-a3cf2f4a3e73>
Date: Tue Mar 27 19:38:17 2012 +0000
ar71xx: boost SPI flash read performance
page read speed
old new delta
DB120 929 KiB/s 2597 KiB/s +179.55%
TL-WR1043ND v1 754 KiB/s 2166 KiB/s +187.27%
TL-WR703N v1 745 KiB/s 2176 KiB/s +192.08%
TL-MR3220 v1 752 KiB/s 2154 KiB/s +186.44%
TL-WR2543ND v1 564 KiB/s 2130 KiB/s +277.66%
TL-WR741ND v2 525 KiB/s 1767 KiB/s +236.57%
ALFA-AP96 702 KiB/s 1977 KiB/s +181.62%
WNDR3700 697 KiB/s 1965 KiB/s +181.92%
My last known 'boot to a working internet connection' was about 126
seconds, when last I measured it.
30 seconds would be good... and while I hope that the fix above knocks a
few seconds off the boot,
doing better would take extensive profiling that I do not have time for in
I also note that I've filed some new bugs in the bugtracker regarding some
regressions in linux 3.3 that have been fixed and will be showing up in a
stabler kernel release RSN. I'm not sure if #346 addresses my long standing
issues with multicast routing and pim or not, nor if #347 accounts for a
running out of memory problem I've seen after some extreme benchmarking
while using 4 over 6 tunneling... but we'll see.
I keep hoping to find the time to figure out why QFQ hangs, but so far, no
/me goes back to making sure bug 113 is dead, dead, dead.
US Tel: 1-239-829-5608
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Cerowrt-devel