0000000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: > Nowadays (or is it next year's model?), on the z, practically every one. > Nothing will run without the PR/SM hypervisor. > > In the twilight of Sun Microsystems, Sine Nomine ported OpenSolaris > to z. It required z/VM for various assists.
re: http://www.garlic.com/~lynn/2017b.html#30 Virtualization's Past Helps Explain Its Current Importance http://www.garlic.com/~lynn/2017b.html#36 IBM LinuxONE Rockhopper I was involved in doing a lot of the work on ECPS (vm microcode assist) for 138/148 ... and then used in 4300. This is old post with the analysis choosing what to put into ECPS. We were told that 148 had 6kbytes of space for microcode ... and was to choose the top 6kbytes of vm370 kernel code for moving into microcode. 138/148 executed an avg. of 10 native, microcode instructions for every 370 instructions. Kernel code moved into microcode got 10:1 speedup. 6kbyte of vm370 kernel code accounted for 79.55% of kernel CPU use: http://www.garlic.com/~lynn/94.html#21 Later I got approval to give presentation on ECPS development at (user group) BAYBUNCH (held monthly at SLAC). Afterwards the Amdahl attendees cornered me to explain what they were doing on something called HYPERVISOR (basically a subset of vm370 function incorporated totally into the machine). One of the things Amdahl people mentioned was that hi-end processor horizontal microcode was difficult and frequently had long lead time. Ever since 3033, IBM had done been minor microcode features ... that if weren't present, operating systems wouldn't run. To help keep abreast of the IBM's constant flurry of minor microcode features ... they had come up with macrocode ... that was effectively close to 370 machine language ... and much simpler (& faster) to code than horizontal microcode but otherwise function like microcode. The hypervisor was being implemented in this macrocode ... which significantly reduced the development effort. To remain competitive, the 3090 had to come up with their equivalent ... evenutally called PR/SM (and LPAR) .... but was a significantly bigger development effort (having to be done directly in horizontal microcode). This old email talks some about improvement for SIE (part of PR/SM) for trout (aka 3090) compared to 3081 (before they were faced with having to extend to PR/SM to match Amdahl's hypervisor) http://www.garlic.com/~lynn/2006j.html#email810630 http://www.garlic.com/~lynn/2003j.html#email831118 announce 12Feb1985 ... https://www-03.ibm.com/ibm/history/exhibits/mainframe/mainframe_PP3090.html trivia1: Endicott tried to get 370 138/148 announced as every machine ordered shipped with vm370 pre-installed (effectively akin to later PR/SM LPAR), but were overruled by POK/corporate trivia2: above mentions needing two 3370 FBA (even for MVS/XA customers which never supported FBA), for the two 3092 service processors, a pair of 4361s running a modified version of vm370 release 6 (and all service panels done in CMS IOS3270). -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
