Re: [PATCH] zero^H^H^H^Hsingle copy pipe

John Fremlin (chief@bandits.org)
07 May 2001 18:01:35 +0100


[Stuff about NetBSD pipes snipped]

I'm testing out Manfred's patch for zero copy pipes, and haven't
crashed it yet.

My hardware is a AMD K6-2 (stepping 1) on an ALi M1541 with 320 Mb -
one quite slow 64 Mb stick and one fast 256 Mb stick.

The lmbench bw_pipe showed a performance improvement of about 30% from
45 (+- 2) Mb/s to 59.5 (+-0.6) Mb/s.

The lmbench (2beta3) lat_pipe showed a performance improvement of
about 20% from a latency of about 27 (+- 1) usec to about 22.4 (+-.6)
usec. There was one outlyer amoung the 10 non zc pipe runs - 25
usec. For zc, the first run was always about 25 usec and after that
very stable around 22 usec.

FWIW the system time from "time" when running (bw_pipe;lat_pipe) 10
times in a row *increased* by 50%, from sys 0m18.740s to sys
0m31.660s.

Script:
for i in 1 2 3 4 5 6 7 8 9 10; do ./bw_pipe; ./lat_pipe; done

Non zero copy:

real 0m49.602s
user 0m10.170s
sys 0m18.740s

Zero copy run 1:

real 0m47.901s
user 0m10.390s
sys 0m31.660s

Zero copy run 2:

real 0m47.492s
user 0m10.600s
sys 0m31.340s

-- 

http://ape.n3.net - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/