Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 10 Feb 2015 21:41:38 +0100
From:      Wolfgang Zenker <wolfgang@lyxys.ka.sub.org>
To:        "freebsd-mips@freebsd.org" <freebsd-mips@freebsd.org>
Subject:   Re: IPv6 throughput only half of IPv4 on ERL
Message-ID:  <20150210204138.GA688@lyxys.ka.sub.org>
In-Reply-To: <CAJ-VmoncgB__VtVryu67PeLU8-t18K41NdbJgvtKwDWvtqKbBA@mail.gmail.com>
References:  <20150208130034.GA80972@lyxys.ka.sub.org> <CAJ-VmoncgB__VtVryu67PeLU8-t18K41NdbJgvtKwDWvtqKbBA@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
* Adrian Chadd <adrian@freebsd.org> [150208 17:17]:
> everything's bigger in ipv6?

true, but its bigger on other platforms as well :-)

> Well, try kldload hwpmc and see if pmcstat works for you. I haven't
> tried it on the Octeon hardware. That'll be useful for chasing down
> why it's chewing CPU.

I built a kernel with device hwpmc built in, but so far not much success
getting it to do anything except killing the machine.
In dmesg I find this line:
hwpmc: SOFT/16/64/0x67<INT,USR,SYS,REA,WRI> OCTEON/2/64/0x1ff<INT,USR,SYS,EDG,THR,REA,WRI,INV,QUA>
so hwpmc seems to be indeed to be available.

As I have no clue so far how to use pmcstat, I tried one of the example
commands from the man page (pmcstat -S instructions -O /tmp/sample.out)
which resulted in
cpu:1-Trap cause = 2 (TLB miss (load or instr. fetch) - kernel mode)
panic: trap
cpuid = 1
followed by automatic reboot. Maybe hwpmc is not usable on this platform
so far.

Wolfgang



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150210204138.GA688>