Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!njin!princeton!udel!wuarchive!emory!att!bellcore!messy!mo From: mo@messy.bellcore.com (Michael O'Dell) Newsgroups: comp.arch Subject: Re: In defense of the TI990[0] Message-ID: <1990Nov15.160610.19795@bellcore.bellcore.com> Date: 15 Nov 90 16:06:10 GMT References: <13261@encore.Encore.COM> Sender: usenet@bellcore.bellcore.com (Poster of News) Reply-To: mo@messy.UUCP (Michael O'Dell) Distribution: usa, na Organization: Center for Chaotic Repeatabilty Lines: 12 Yes, there were certainly some performance problems, but it is the ONLY machine I know of which understood coroutines and processes at such a deep level. Frankly, the design, coupled with fast caches on-chip, might be a very interesting thing to consider if you want lots of processes and fast context switching. At some point, if you context switch often enough, putting things in registers can cost more than it saves. So, if you postulate an environment with enough processes and enough switching, a 990 style machine with good, fast, local caches might well win. -Mike