Subject: Re: about untar performance on -current
To: Jared D. McNeill <jmcneill@invisible.yi.org>
From: Frank van der Linden <fvdl@wasabisystems.com>
List: tech-kern
Date: 04/12/2001 13:29:50
On Thu, Apr 12, 2001 at 08:03:23AM -0300, Jared D. McNeill wrote:
> Do you think you could try the ffs_dirpref() patch I posted here yesterday with 
> this and post numbers for that (with your patch) as well?

Funny you should mention that.. I just ran a simple test. I extracted
gcc-2.95.2.tar.gz (12M file) on the P5-200, 64M, with:

wd0: drive supports 16-sector PIO transfers, LBA addressing
wd0: 7339 MB, 15907 cyl, 15 head, 63 sec, 512 bytes/sect x 15032115 sectors
wd0: 32-bit data port
wd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 2 (Ultra/33)
wd0(pciide0:0:0): using PIO mode 4, DMA mode 2 (using DMA data transfers)

The command was 'time tar xzf gcc-2.95.2.tgz'

1.5:			10.768u 9.270s 0:49.86 40.1%    0+0k 802+9919io 58pf+0w
1.5T:			10.753u 10.174s 0:56.66 36.9%   0+0k 376+1381io 31pf+0w
1.5T+dirpref:		10.585u 10.279s 0:48.91 42.6%   0+0k 85+865io 20pf+0w
1.5T+enami:		10.464u 9.849s 0:43.05 47.1%    0+0k 333+1629io 31pf+0w
1.5T+enami+dirpref:	10.846u 9.194s 0:30.59 65.4%    0+0k 93+942io 31pf+0w

So the patches seem to make a good team together, almost cutting 50% off
the execution time.

Only the 'rm -rf' time is still worse:

1.5:			0.060u 0.435s 0:07.25 6.7%      0+0k 209+489io 19pf+0w
1.5T:			0.038u 0.661s 0:15.06 4.5%      0+0k 59+1107io 8pf+0w
1.5T+enami+dirpref:	0.094u 0.637s 0:11.72 6.1%      0+0k 32+1313io 8pf+0w

- Frank

-- 
Frank van der Linden                           fvdl@wasabisystems.com
======================================================================
Quality NetBSD CDs, Support & Service.   http://www.wasabisystems.com/