Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!nrl-cmf!ames!pasteur!ucbvax!tut.cis.ohio-state.edu!rutgers!att!homxb!houxs!beyer
From: beyer@houxs.ATT.COM (J.BEYER)
Newsgroups: comp.arch
Subject: Re: delay lines for memory
Summary: other hard-to-program memory technologies...
Message-ID: <1257@houxs.ATT.COM>
Date: 17 Mar 89 18:36:25 GMT
References: <21976@lll-winken.LLNL.GOV> <28411@ucbvax.BERKELEY.EDU>
Organization: AT&T BL Holmdel NJ USA
Lines: 19


In addition to delay line memory, some machines used magnetic drum memories
(e.g., IBM 650). The trouble with delay line memory and most drum memory is
that you have to wait until the address you want comes along out of the memory.
By Murphy's law, the one you want just went by, so you have to wait until it
comes around again on the guitar.

The larger the memory, the longer it is and the longer you have to wait.
There are ways around this, sort-of. What was done to the 650 was that
each instruction had 2 addresses, one where the data came from or went to,
and one, the address where the next instruction was to come from. A clever
assembler was used that knew the execution times of the instructions, and
arranged to place the next instruction at a location that would just be
coming around as the current one finished. Fun, huh!

-- 
Jean-David Beyer
A.T.&T., Holmdel, New Jersey, 07733
houxs!beyer