> > Ok about the "r" convention I wrongfully used, it is assumed for the integer > register-file. >
Correct. > I am currently using MIPS + soft-float + some other functionalities in > Coprocessor #2 with success (somewhat tested in a functional simulator). About > 10-12 special-purpose instructions are implemented in COP2 plus the > ctc2,cfc2,... etc instructions that are detailed in the MIPS ISA documents. > All > added instructions are faithful to the COP2 convention (6-bit opcode + 26 > implementation specific bits). > OK. > I have used Matt Hiller's patch on binutils, for supporting coprocessor > register > names, which was uploaded at gcc-patch ml circa June 2002. It works, at least > for the things I have used it. Actually the patch doesn't correspond exactly > to > either versions 2.13,2.14 or 2.15 but to some CVS snapshot (I suppose). > Correct. CVS as of the day it was submitted. > My question on the usage of inline assembly regards omitting writing long jump > tables (e.g. with nested switches). This is not usually necessary but it is > when dealing with dynamic code generation. > I'm not sure what the question is here. > Plus: can gcc co-operate with dynamic code generation tools e.g. the GNU > lightning? I have also heard of another tool, dcg (if i spell it correctly). I've not done any of this. -eric
