socha@drivax.UUCP (01/22/87)
> In article <2221@jade.BERKELEY.EDU>, srm@iris.berkeley.edu (Richard Mateosian) writes: > > Overheard at the Levco booth: The 68020 can supposedly be > > microcoded so as to run identically to a 80386. Someone > > even suggested that at 25 Mhz 68020 would do such at 3 MIPS. > > That's ridiculous. SO TRUE, SO TRUE. > I think that the MMU instructions would be tough for the 68020 to be > handled alone but don't forget that IBM has been shipping a XT370 and > a AT370 which use a 68000+8087+"other hardware" with new microcode > to support a 370 "architecture" and run a variant of VM/CMS. About when the XT370 appeared, there was an article in IBM's Journal of R&D. It discussed how to implement an XT370 and discussed about 6 methods. The second last one (I think) said approximately the following: 1: Lets take a chip like the 68000 and re-microcode it to do the kernel of the '370 instructions (the "other hardware"!?) 2: But, all those in-frequently used instructions can be done in macro-code on a standard 68000! (Same basic technique used in the Micro-VAX) 3: an Intel '8087 is very close to IBM floating point so a slightly modified version would be used (I'm not sure if this was in article.) AND there are the three big LSI chips in the XT370. BUT, let us consider the 68000 itself. When examined you see that between the micro-code and ALU/registers there is a rather large area of random logic. This logic exists for things like the condition code registers AND, to take microcode instructions and generate control bits for the ALU. (This area grew larger in the 68010 and is huge in the 68020!) 68 bits leave the microcode ROM but over 100 (123?) enter the ALU. This random logic does lots of work. Therefore I maintain that it had to change (i.e. be re-designed) for the XT370 "other hardware". The '386 is very different therefore if nothing else, this random logic would have to be thrown out and re-done from scratch. A much greater effort than just re-microcoding the 68000. > This is not something I would like to do though. Of course IBM can > throw oodles of money and hundreds of people and get a "solution", Actually it was done at Motorola and Intel! Motorola did the '370 and Intel did the floating point chip (just look at the 'bat' and lower case 'i' on the chips) > I'll bet Motorola is REAL interested in running their chip as a 80386 :-) :-) GRINS ALL AROUND :-) but, Too bad its nearly impossible -:( -- -- UUCP:...!amdahl!drivax!socha WAT Iron'75 "Everything should be made as simple as possible but not simpler." A. Einstein