Path: utzoo!attcan!uunet!crdgw1!crdos1!davidsen From: davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) Newsgroups: comp.arch Subject: Re: Tektronix shutdown & move away from 88k's?? Message-ID: <2817@crdos1.crd.ge.COM> Date: 31 Oct 90 13:29:15 GMT References: <1536@ftc.framentec.fr> <1990Oct19.120218.9450@canterbury.ac.nz> <15497@hydra.gatech.EDU> <2176@lupine.NCD.COM> <42310@mips.mips.COM> <42488@mips.mips.COM> Reply-To: davidsen@crdos1.crd.ge.com (bill davidsen) Organization: GE Corp R&D Center, Schenectady NY Lines: 26 In article pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: | As to other rumours, I read recently, again on Byte or on the net, that | the 88k has enough spare die to also be able to store some interesting | chunk of microprogram. This to me makes much less sense, but who knows? Loadable control store is a great idea, and can really improve the performance of a program. It becomes awkward in a shared CPU environment, where a bad implementation could mean a context switch includes a change of instruction set! If it were managed like a non-sharable device, it would be doable, but I have doubts about the number of people who could make use of the ability, and it sort of implies access to kernel mode or a way to lock certain status bits (makes things truely complex). In all the years we have had VAXen, and have had hackers of both the brilliant and demented types, and only one (as far as I know) person actually added iseful and working instructions. The FFT1 and FFT2 instructions do give a 40-50% reduction in CPU time for certain problems, but they represent such a small percentage of use that I doubt anyone ever loads them. They are hopefully preserved in his thesis for future need. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) The Twin Peaks Haloween costume: stark naked in a body bag