Article 66 of comp.sys.cdc: Path: matra.meer.net!news.spies.com!genmagic!news.sgi.com!news.maxwell.syr.edu!newsfeed.internetmci.com!news.nwrain.net!news From: "phil hystad" Newsgroups: comp.lang.fortran,comp.sys.cdc Subject: Re: Reminiscing About the Cyber 6600 Date: 2 May 1997 23:09:07 GMT Organization: Zoo Lines: 44 Distribution: inet Message-ID: <01bc574e$31a45490$98dc86cd@ntpeh> References: <01bc567d$25cfb000$7a59e1cc@default> NNTP-Posting-Host: p34.bpp1.nwrain.net X-Newsreader: Microsoft Internet News 4.70.1161 Xref: matra.meer.net comp.lang.fortran:50185 comp.sys.cdc:66 Richard Lerner wrote in article <01bc567d$25cfb000$7a59e1cc@default>... > Anybody want to discuss the beloved CDC 6600 and its successors. > > Now there was a manly (sorry) machine. > > Who else would would have thought of setting a register to zero > permanently? > > So c'mon, pull out your CPU 'nerd' cards, and let us relive our youth > vicariously I worked at CDC at Sunnyvale Operations, home of Compiler Products group in the early 1970s. I remember a constant contest between us in the FTN compiler group and others at Larry Labs (Lawrence Livermore Labs) on who could create the fastest and most accurate ATAN function for the CDC 6600. The key in producing such an optimized code was developing the CPM path for the code since the multiple functional units and various register reservation rules produced multiple parallel and stalled execution paths. Also, I remember in early 1975 when the operating systems were unbundled from the hardware -- read that as, you pay extra for the operating system. Some of us are old enough to remember when the operating system was just a free part of the computer, it came with the machine. But, this is where the name operating system was NOS/BE was created merely by changing the copyright statement in the SCOPE 3.4.5 system. And, accordingly, NOS/VE came from Kronos V2.1 (I believe). Yes...register B0 was wired to zero! This was not as bad as the instruction SB0 B0 which was often used as a NOOP instruction since any attempt to modify B0 (even setting it to itself) resulted in a non-operation...EXCEPT...it caused a register reservation wait on B0 even though it was not possible to be modified! How do you tell a 6600 from a 6400 using instructions only...well, they handled executing a zero word instruction differently. The 6600 would essentially hang in a tight infinite loop and the 6400 would result in a Mode 0 error.