>http://www-cs-faculty.stanford.edu/~uno/taocp.html#vol4

I think Mike's point is that Professor Knuth is still alive and actively 
writing at Stanford.  When I was given a copy of Volume 1 back in 1972, I 
expected Professor Knuth to be some 60-ish professor, complete with a tweed 
suit, pipe and leather elbow patches.  I was very surprised to learn that he 
was just 10 years older than myself.

I have an interesting story to tell regarding Volume 1 of The Art of Computer 
Programming.   Buried within this tome is an algorithm for storage allocation 
and de-allocation.  The whole idea is to allocate storage in chunks that are a 
power of 2 in length (32, 64, 128, etc.).  Knuth calls this the "Buddy Storage 
Algorithm" since by knowing the address of a chunk, you know it's size.  And 
you can reclaim storage by locating the "buddy" chunk and checking a single bit 
to see if it is free or in-use.  The algorithm is very efficient and did not 
require searching linked-list structures like the "Best Fit" or "First Fit" 
algorithms that were in common use.

Back in 1976 I was working at a major Canadian Bank.  We were running an online 
retail banking application in over 1000 branches across Canada using CICS and a 
IBM S/370-168.  Very marginal for performance.  We were constantly tuning to 
squeeze every last CPU cycle out of the application.  We used the STROBE field 
developed program to profile our app and discovered that a very large portion 
of our CPU cycles were being spent doing CICS GETMAIN and FREEMAIN.

A colleague of mine did some "midnight engineering" to see what would happen if 
we replaced the CICS management program DFHSCP with our own variant using 
Knuth's Buddy Storage Algorithm.  Management did not want to encourage this, 
because we were trying to go "vanilla" with our CICS software and not employ 
any local mods.  But when Angus was able to show a DOUBLING of performance as a 
consequence of the change, management seized upon this, since only by getting 
past our performance and capacity problems could we enhance the application to 
do some things that would give the bank a competitive advantage.

The Bank stuck with this modified Storage Control Program for several years, 
until Hursley made some big changes in this area, probably with CICS/VS 1.7 or 
later.

John

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to