rb@cci632.UUCP (07/11/86)
In article <874@bu-cs.UUCP> bzs@bu-cs.UUCP (Barry Shein) writes: > >> It was always my feelings that, if a CPU manufacturer were to write the >> language compilers first, THEN generate a CPU design to run it, we'd all >> be alot happier. Yes, that sounds like a CISC design. > >Just for the record, wasn't the WE32x00 (BellMac32) designed by AT&T >to run UNIX and hence, one would presume, C? > >I honestly don't know much about it, but I thought that was the >claim. Then of course there were the Burroughs Cobol machines, >the Lisp machines etc etc but I presume this is all tacit to >the group. One could go one step further and extend the "CPU design" into consideration of the Entire System. This would include things like coding style, applications, operating systems, and "Type of General Purpose". An example from my own experience is the CCI 6/32. Before even the start of the CPU design, goals were set. The system had to excel at C/UNIX/DataBase applications, and be capable of serving many users. In addition, the applications were to use small, modular subroutines rather than Macros. Also because there were so many users, drives, and tasks, context switching had to be very fast. Ironically, the end result was very RISCish. The "calls" and "context switch" were implemented in hardware rather than micro-code. I believe "effective addresses" were also in hardware. The instruction set looks very VAXy, but the "cycles per instruction" is very low. The reasons? There were too many "primitives" to fit in micro-code. The result? Makes that took 8 hours on a medium load VAX take about 15 minutes :-). This is overstatement, but it is very fast.