It was Oct 2, 2000, 09:51, when Austin L. Denyer keyboarded: >I once had to re-write a system that performed around 40 different >calculations depending on the type of data received. Conventional >programming (which was the way the original was written) would have had a >subroutine for each case, with tests to check which one to jump to. Self >modifying code allowed me to write one block of code, with NO jumps or >tests - the code changed itself based on the input data. That was (and for me sometimes still is!) the way to go! I pulled things like that on a Wang/VS system and an IBM micro S/36 that had 128K of memory. It had to furnish services to 40 people. It was programmed in RPG (blagh), but I put assembler into action, stuffed registers and memory addresses all over the place and had a top performance. I also liked the NRT's. Non Requester Terminals, kind of like daemons. I could stick often-used routines in there, like client lookups, and call those from any source. Depending on the caller, the NRT would change itself and deliver what was asked. >My successor HATED it... LOL!!! Sounds soooo familiar! Paul -- The Tao that is seen Is not the true Tao, Until you bring fresh toner. http://nlpagan.net - ICQ 147208 - Registered Linux User 174403 -=PINE 4.21 on Linux Mandrake 7.1=-
