lentferj's projects / dragonfly.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: b263df5) | patch
jme(4): Disable TXCSUM by default to improve bulk data transmit performance
authorSepherosa Ziehau <sephe@dragonflybsd.org>
Sat, 21 Mar 2009 06:11:20 +0000 (14:11 +0800)
committerSepherosa Ziehau <sephe@dragonflybsd.org>
Sat, 21 Mar 2009 06:35:46 +0000 (14:35 +0800)
commit3d2aeb15c4275b09d608f9c4fef075e6d9bb5e6a
tree2ee5f894c0c2957cbaab5a0b07004a1156bcb5d1tree | snapshot
parentb263df52a14b5b918f5e76f1f376cffb8af0e1dfcommit | diff
jme(4): Disable TXCSUM by default to improve bulk data transmit performance

It looks like when the hardware does the TX checksum in the internal TX
FIFO, the DMA read logic is kinda changed; the bursty read effect seems
to get discounted.

After the TX csum offloading is disabled, the device bulk data transmit
speed finally reach gigE line speed and the tiny packets transmit
performance is only reduced marginally (-6Kpps~-8Kpps).
sys/dev/netif/jme/if_jme.c diff | blob | blame | history
lentferj's DragonFly repo