Path: utzoo!attcan!utgpu!bnr-vpa!bnr-fos!bnr-public!schow From: schow@bnr-public.uucp (Stanley Chow) Newsgroups: comp.arch Subject: Re: Independent Architecture Complilers Message-ID: <414@bnr-fos.UUCP> Date: 17 Apr 89 19:05:57 GMT References: <2889@rti.UUCP> Sender: news@bnr-fos.UUCP Reply-To: schow@bnr-public.UUCP (Stanley Chow) Distribution: comp Organization: Bell-Northern Research, Ottawa, Canada Lines: 28 In article jac@paul.rutgers.edu (J. A. Chandross) writes: >tassos@rti.UUCP (Tassos Markas) writes: >> I'm looking for independent architecture compilers. >> This compiler should accept a high level langauage (I would prefer C) and >> produce microcode that can be easily retargeted for any system architecture. > >Microcode compilation is a very hard problem for non-traditional architectures. >(I know, I've tried.) If you have something fairly straightforward, ie a >non-VLIW you may be able to use one of the existing retargetable microcode >compilers. Otherwise, write it from scratch. It's a lot easier than coercing >something designed for a relatively simple machine into working for your >machine. > I agree with Chandross that in many cases, it is better to write the micro-code from scratch. Considering the amount of micro-code in a machine (especially the really timetime critical bits), it is often quicker to rewrite the micro-code as opposed to writing (or porting) a truly optimizing compiler to a new architecture. If you have a lot of *critical* micro-code, (by a lot, I mean tens of K words), you should question the need. [Remember, this is a proponent of micro-code talking.] Stanley Chow ..!utgpu!bnr-vpa!bnr-fos!schow%bnr-public (613) 763-2831 Clever disclaimers are hard to come by, I will save them for the articles that need them. For now: I speak only for myself.