johnl@esegue.segue.boston.ma.us (John R. Levine) (10/20/89)
In article <2398@gmu90x.UUCP> rmiller@gmu90x.UUCP (Richard Miller) writes: >Why would IBM not refer to the 801 as a RISC in an article published 6 >years ago and call the America (which, by almost every metric ever >seriously considered is less RISCy) a RISC at its initial announcement? >I can't think of any reason other than marketing hype. I can. The 801 project started before the Berkeley one (it takes forever for IBM to publish anything, first they have to file all the patents and figure out what they want to keep secret) and at the time the term RISC hadn't yet been coined, much less become a standard term for a style of architecture design. Furthermore, the 801 project was research, not a product, and it did a whole lot of things other than stripping down the instruction set. Berkeley took the mediocre compiler technology of PCC as a given and designed their chip to work well given PCC's limitations. The 801 project had some of the best compiler people around, and they did a lot of work tuning their state-of-the-art optimizing compiler and their instruction set to each other. That's why the 801 and its descendants don't use register windows. They had no trouble doing register allocation in the compiler, and their language was based on PL/I which has nested scopes, something that register windows don't address well. Their compiler also has provision for extensive runtime checking of things like array bounds. I have seen some of its object code and it is very impressive. The compiler has since been retargeted to the 370, the ROMP, and no doubt other CPUs, and has had front ends added for C and I believe Pascal. (That is the compiler that comes with AIX, which is why AIX does so well on dhrystones.) The ROMP processor used in the RT PC was influenced by the 801, but I have heard that the 801 people thought that the ROMP missed the boat in many ways. They sure were horrified to learn that the first released compiler on the ROMP was a PCC port that I originally did. The America will doubtless do better. The 801 also apparently had fiber-optic I/O buses and other hardware innovations that have received much less press than the software aspects. Calling it a RISC, while accurate, misses the 801 project's goal of creating a high performance low cost system that pushed the state of the art in more than one direction. I only wish they had published more earlier so we could have taken better advantage of it. Or perhaps that's just as well, every hot new compiler that does register allocation by graph coloring appears to me to infringe IBM patents, and when IBM decides to enforce their patent rights, things in the compiler biz will get mighty exciting. -- John R. Levine, Segue Software, POB 349, Cambridge MA 02238, +1 617 864 9650 johnl@esegue.segue.boston.ma.us, {ima|lotus|spdcc}!esegue!johnl Massachusetts has over 100,000 unlicensed drivers. -The Globe