[EMAIL PROTECTED] (Dean Kent) writes: > Hopefully not far off topic, but this recently published article > indicates that IBM may merge i, p and z series processors using > POWER6. At one time I recall seeing an official IBM statement > saying that this would never happen because of the unique > requirements for (at that time) S/390 Architecture. Any thoughts on > whether this is pure speculation, or if it is a real possibilty due > to economics, advances in technology or perhaps a desire to make > zSeries less costly and broaden the market? Could this result in > affordable desktop mainframes (so we don't need to worry about > whether IBM will ever allow zOS to run on Hercules)? ;-) > > http://www.realworldtech.com/page.cfm?ArticleID=RWT121905001634
can you say fort knox? 1980 there was effort afoot to convert much of the internal microprocessors, controllers to 801. the follow-on to the 4341 was going to be an 801. the issue at the time was that there was a huge number of different microprocessors all with different architectures and programming. the convergence to 801 was to eliminate a lot of the different variety. the issue for the low and mid-range 370 at the time was that they they implemented 370 in microcode with something like an avg. of 10:1 microcode instructions per 370 instruction (not all that different from the current generation of software implementing mainframe architecture on intel platforms). this was something of the ecps boost for vm on the 148 & 4341 ... moving high use kernel paths into microcode ... with a resulting 10:1 performance boost (and much greater for some involving instruction simulation and not having to save/restore registers as part of context switch between user mode and kernel mode) ... a couple more detailed discussion of ecps: http://www.garlic.com/~lynn/94.html#21 370 ECPS VM microcode assist http://www.garlic.com/~lynn/94.html#27 370 ECPS VM microcode assist http://www.garlic.com/~lynn/94.html#28 370 ECPS VM microcode assist at least for the 4341, the project was aborted. the issue being that chip silicon was progressing to the point where it was possible to directly implement 370 architecture in hardware (as opposed to having much simpler hardware architecture that in turn was progremmed to implement more complex 370 architecture). as a result the 4341 following became a a direct 370 silicon implementation rather than 801 chip with microprogramming implementing 370. note that 370 instructions running close to hardware speed was already coming close to happening with high-end machines like 3033. another aspect of this presented itself with amdahl's initial hypervisor implementation. one of the issues for the high-end machines were that they tended to be horizontal microcode which was significantly harder program than veritical microcode and/or 370. even tho 370 instructions were coming close to running at hardware speed, the high-end machines were still (horizontal) microcode and had reputation for being extremely hard to program. i gave a presentation at baybunch about the ecps experience ... that while there was a 10:1 performance gain moving kernel code directly into microcode (on 148 & 4341) ... there was even large performance pickup by not having to context switch into the vm kernel (save/restoring registers, etc). amdahl had previously created "macro" code for their high-end machines .... basically part of the hardware context ... but using a subset of 370 programming (rather than the much more difficult to program horizontal microcode). several people then implemented a subset of virtual machine function as hypervisor support ... with the implementation being done using "macrocode". ibm eventually responded with pr/sm (on 3090) and then expanded it to multiple with LPARs (logical partitions). however, this was a much more difficult undertaking ... requiring implementation done directly in the 3090 microcode. and as previously noted in some recent posts in other threads in this n.g. ... virtualization is the new, new thing this season. http://www.garlic.com/~lynn/2005u.html#36 Mainframe Applications and Records Keeping? http://www.garlic.com/~lynn/2005u.html#37 Mainframe Applications and Records Keeping? an 801 from that period that did survive was romp ... which was research, office products effort to build a displaywriter follow-on. when that was killed, it was decided to retarget the machine to the unix workstation market ... and that company that had done the port for the ibm/pc pc/ix was hired to do a similar port for romp, which was called aix. the followon to romp was rios, or power. misc. past 801, romp, rios, power, etc postings http://www.garlic.com/~lynn/subtopic.html#801 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
