Path: utzoo!attcan!uunet!husc6!uwvax!vanvleck!uwmcsd1!bbn!bbn.com!cosell From: cosell@bbn.com (Bernie Cosell) Newsgroups: comp.misc Subject: Re: Info wanted on eniac computers Message-ID: <25948@bbn.COM> Date: 19 Jun 88 03:15:39 GMT References: <198@marque.mu.edu> <17496@glacier.STANFORD.EDU> <11111@agate.BERKELEY.EDU> Sender: news@bbn.COM Reply-To: cosell@bbn.com (Bernie Cosell) Organization: Bolt Beranek and Newman Inc., Cambridge MA Lines: 27 In article <11111@agate.BERKELEY.EDU> weemba@garnet.berkeley.edu (Obnoxious Math Grad Student) writes: }How old is the 650 you ask? It used a drum for main memory. This was }before "core". It stored numbers in *decimal* form. Optimization tricks }consisted of laying out your drum references very carefully. } }I've actually talked about the 650 to some geezers who used the real }thing. They were so happy.... Ah yes, the 650 -- reminiscences from the geezer end of the spectrum. One cute thing you left out (did your simulator get this right???) was the drum rotation optimization. Each instruction had a "next address" field to give the explict location of the next instruction to execute. Getting things laid out on the drum just right was a VERY big deal. The original assembler, SAP, was quickly supplanted by an improved assembler for the beast: SAP was the "Symbolic assembly program", but SOAP was the "symbolic *optimizing* assembly program". The trick of the "O" part was to put the next instruction at _just_ the right place on the drum so that when you finished the current instruction the next was _just_ coming up on the heads. What a neat/ugly optimization problem. By the time the 650 died, there was a multi-pass (to say the least) fortran for the thing. I haven't seen my 650 manuals in decades now... I wonder what I did with them... Probably right next to my 407 board-wiring manuals.... __ / ) Bernie Cosell /--< _ __ __ o _ BBN Labs, Cambridge, MA 02238 /___/_(<_/ (_/) )_(_(<_ cosell@bbn.com