From:	Ton van Daelen <tvd #*at*# msi.com>
Date:	Mon, 09 Aug 1999 10:01:21 -0700
Subject:	hashing algorithm, resolution

Some of you asked me to summarize the resolution to my problem. Here it
goes...

The strings I was using as keys were all quite similar, the only difference
being
a 4-6 digit string somewhere at the end. SGI's hsearch algorithm apparently
put many of them in the same bucket: ie MOL_12345, MOL_21345, etc were all
hashed to the same index, resulting in many collisions. The documentation
talked
about ways to recompile with a different hashing algorithm, but source was
nowhere
to be found.

I am now using the Tcl hash library, and it is at least 100 times faster.

Regards - Ton

ps. many thanks to Wolf-Dietrich Ihlenfeldt for helping out

At 09:46 PM 7/29/99 -0600, you wrote:
>On Jul 29, 11:11, Ton van Daelen wrote:
>> Subject: CCL:hashing algorithm
>> Can anyone point me to an efficient hashing algorithm? I tried using SGI's
>> hsearch. It works, but setting up the hash table gets really slow for
tables
>> 20,000 rows.
>> Thanks - Ton

  Ton van Daelen, Ph.D.    Product Manager, WebLab2
           
       __o                                        E: tvd - at - msi.com
     _`\<,_          Molecular Simulations Inc.   P: -1-858-799-5329
    (*)/ (*)         9685 Scranton Road           F: -1-858-458-0136
  /\/\/\/\/\/\       San Diego, CA 92121          W: http://www.msi.com

The Cerius2 software developer's kit is now available free-of-charge
Visit http://www.msi.com/sdk/ to sign up
for a copy

SDK training courses coming up: San Diego, July 15-16
http://www.msi.com/about/events/training/training_schedule.html

Similar Messages
07/29/1999:  hashing algorithm
04/05/1999:  Cerius2 software developer's kit training workshops
04/06/1998:  Cerius2 Software Developer's Kit training course
02/18/1998:  Cerius2 SDK workshops, West coast US
11/13/1997:  Cerius2 SDK training workshop, Dec 1-2, Philadelphia


Raw Message Text