Re: mec(4) TX improvement and mbuf statistics

Izumi Tsutsui <[email protected]>
Newsgroups gmane.os.netbsd.ports.sgimips
Message-ID <[email protected]>
[email protected] wrote:

> Here are my results with your patch (and the cpuflags you suggested, on
> both machines)

> On a R10K O2:
 :
> mec0 TX pkts queued total                    296540      346 misc
 :
> mec0 TX pkts copied to new mbufs              15138       17 misc

> On a R5K O2:
 :
> mec0 TX pkts queued total                    119472      113 misc
 :
> mec0 TX pkts copied to new mbufs               2259        2 misc

Hmm, a bit higher rates of packets are copied to new mbufs,
but I think it's still worth to handle TX packets with
the new strategy.
If there is no objection, I'll commit it soon.

I've also put a patch for netbsd-4:
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/mec-fix-netbsd-4-20080822.diff

This also includes other related changes in -current:
sgimips/sgimips/bus.c		1.50
sgimips/mace/if_mec.c		1.9, 1.12, 1.14, 1.17, 1.18, 1.20,
				1.21(patch), 1.22-1.25, 1.26(patch),
				1.27(patch), 1.28, 1.29, 1.30(patch)
sgimips/mace/if_mecreg.h	1.4
sgimips/conf/files.sgimips	1.44

I'm not sure if it's worth to pullup all these changes,
but any reports/comments would be appreciated.
---
Izumi Tsutsui
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.