From mboxrd@z Thu Jan 1 00:00:00 1970 From: imp@bsdimp.com (Warner Losh) Date: Wed, 15 Mar 2017 11:43:34 -0600 Subject: [TUHS] System Economics (was is Linux "officially branded UNIX") In-Reply-To: <20170315155905.GF2995@mcvoy.com> References: <20170314153815.GA32726@mcvoy.com> <9deec795-ecd6-7924-c10f-b722ee388a0c@kilonet.net> <20170314155718.GH32139@yeono.kjorling.se> <47c38ea0-accb-407b-26c8-6b4877657b21@kilonet.net> <20170315143228.GG25424@yeono.kjorling.se> <20170315155905.GF2995@mcvoy.com> Message-ID: On Wed, Mar 15, 2017 at 9:59 AM, Larry McVoy wrote: > On Wed, Mar 15, 2017 at 11:36:21AM -0400, Arthur Krewat wrote: >> Off-topic: Anyone ever run SunOS on a Sparc-10 or similar platform (670?) >> with two processors? Was it my imagination or did it actually use both >> processors? > > Yeah, SunOS 4.1.4 had some MP work done to it. Pretty sure I posted > about it here and dragged Greg Limes into it. He was involved in that > work. I think it sort of worked up to 4 CPUs but as with all early > kernel threading stuff it worked better when it was a 4 cpus of > userland work, less so when it was 4 cpus of I/O. Solbourne computer produced OS/MP, which was a SMPized version of SunOS 4.1. It scaled to at least 16 CPUs. Userland was 100% compatible with SunOS at the given revision level. My team (the OI group) was part of Solbourne for a while, so we got much of that gear when we were spun out to ParcPlace. Our main build server had 12 CPUs, and it was nice being able to do make -j 16. cfront was, for its time, quite the pig. Now it's lightyears faster than clang, though produces lousy code... OS/MP ran on hardware that was 50MHz SuperSparc CPUs and could be configured up to 14 CPUs and a whopping 256MB of RAM... Warner