[daip] [forwarded from Stephen White] DDT

[Patrick P. Murphy]

I plan on putting these on the DDT page soon, unless any of you think I
should ask for clarification...

------- start of forwarded message (RFC 934 encapsulation) -------
Content-Length: 1412
From: Stephen White <white at astro.umd.edu>
To: pmurphy at NRAO.EDU
Cc: white at astro.umd.edu
Subject: DDT
Date: Thu, 18 May 2000 17:06:46 -0400 (EDT)
Message-Id: <200005182106.RAA11698 at carina.astro.umd.edu>


Hi Pat,

Since I have some numbers not on the DDT page, here are my most recent
(best) benchmarks on a variety of machines. They are all Sun machines
running Solaris 2.6 or 2.7, all using version 31DEC99 (SOL, not SUL, binaries
compiled with Sunsoft compilers) and all using the /tmp trick:

Ultra 10/440 MHz, 384 MB:      22.2 (EIDE disk)
Ultra  5/360 MHz, 256 MB:      14.0 (EIDE disk)
Ultra 10/300 MHz, 128 MB:      13.4 (EIDE disk)
Ultra  5/333 MHz, 128 MB:      10.3 (EIDE disk)
Ultra 1/170E, 167 MHz, 384 MB: 10.6 (fast external SCSI disk, /tmp on an 
                                     internal SCSI disk)

I also tried implementing Chris Flatters' recommendations for tuning
Solaris on my 1/170 and it did bring the benchmark without using the /tmp
trick up to within 5% of the value using /tmp: the difference is probably 
again that /tmp is on a different drive and hence you don't get collisions
writing to scratch files on the same disk as the output files. However,
I have a largish /tmp and have never yet had any problems using /tmp.

regards,

Stephen White

- --------------------------------------------------------------------------
 Department of Astronomy * University of Maryland * College Park MD 20742
              Phone: (301) 405 1547 * FAX: (301) 314 9067
                      E-mail: white at astro.umd.edu
- --------------------------------------------------------------------------


------- end -------