[email protected] (Shmuel Metz , Seymour J.) writes: > Certainly for the ES/9000 and 43xx processors; I'm not sure about, > e.g., the 370/145.
late 70s there was start of effort to move the large variety of internal microprocessors to 801/risc (iliad chips) ... this included the follow-ons to 4331/4344 (i.e. 4361/4381), the as/400 follow-on to the s/38 ... and a lot of other internal microprocessors. various issues cropped up with iliad chips ... and the effort was abandoned ... 4361/4381 doing their own custom cisc chip, crash project to do cisc chip for as/400 (decade later, as/400 did move to varient of 801/risc power/pc chip), etc. in the wake of abandoning that effort, some number of 801/risc engineers leave and show up on risc efforts at other vendors. i contributed some to the whitepaper that killed the effort for 4381. low/mid range were veritical microcode processors simulating 370 ... somewhat akin to current day Hercules effort on intel processors. The idea was to move to common 801/risc for microprocessors ... minimizing the duplication of effort around the corporation developing new chips and associated (microcode) programming environment. The whitepaper claims were that cisc technology had gotten to the stage where much of 370 instructions could be implemented directly in circuits (rather than emulated in microcode). That even with higher mip rate of 801/risc, there was still approx. 10:1 instruction emulation overhead (needed 20mip microprocessor to get 2mip 370) ... while cisc chip might only be 3-5 mips ... quite a bit of that could be 370 instructions nearly native in the chip. small piece from that whitepaper: - The 4341MG1 is about twice the performance of a 3148. Yet the 4341MG1's cycle time is only about 1.4 times faster than the 3148's. The rest of the performance improvement comes from applying more circuits to the design. - The 4341MG2 is about 1.6 times faster than the 4341MG1. Yet the 4341MG2's cycle time is 1.4 times faster than the 4341MG1's. Once again, performance can be attained through more circuitry, not just faster circuitry. - The 3031 is about 1.2 times faster than the 3158-3. Yet the 3031's cycle time is the same as the 3158-3's. ... snip ... previous reference to benchmark with 4341MG1 slightly faster than 3031 http://www.garlic.com/~lynn/2010m.html#42 IBM 3883 Manuals the 3031 reference is slight obfuscation. the 158-3 was single (horizontal microcode processor) engine shared between the 370 microcode and the integrated channel microcode. the 3031 was 158-3 with two processor engines ... one dedicated to running 370 microcode (w/o the integrated channel microcode) and one dedicated to the "303x channel director) running the integrated channel microcode (w/o the 370 microcode). recent reference to 158 engine with integrated channel microcode was used for 303x channel director for all 303x processors (i.e. 3031 was 158-3 repackaged to use channel director, 3032 was 168-3 repackaged to use channel director, and 3033 started out as 168-3 wiring diagram using 20% faster chip with channel director) http://www.garlic.com/~lynn/2010m.html#15 History of Hard-coded Offsets when 4361/4381 did come out ... there was some expectation that it would continue the explosion in mid-range sales that started with 4331/4341 (at the end of the 70s) ... however by then, the mid-range market was starting to move to workstations and large PCs (servers). a couple recent posts discussing the explosion in the mid-range market ... and then mid-range moving to workstations and large PCs: http://www.garlic.com/~lynn/2010m.html#25 Idiotic programming style edicts http://www.garlic.com/~lynn/2010m.html#32 Idiotic programming style edicts http://www.garlic.com/~lynn/2010m.html#36 A Bright Future for Big Iron? http://www.garlic.com/~lynn/2010m.html#41 IBM 3883 Manuals http://www.garlic.com/~lynn/2010m.html#43 IBM 3883 Manuals -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- 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
