> Self-hosted compilers are interesting from the standpoint of "eat your own > dogfood". For example, I wonder if OpenMP implementations would be better > if compiler authors were forced to parallelize the guts of their compilers > with OpenMP :-) > > On the other hand, one should choose the right tool for the right purpose. > Compilers are different applications from solving PDEs. One would never > write a compiler in Fortran, at least since C has existed, because Fortran > is not a good language for system software, and a compiler often depends > upon those features.
It's worth noting that, in developing Chapel, we thought a lot about how to make it a good language to write a compiler (with parallel analysis passes) in, and I believe that, modulo our current issues with records and strings, that it generally would be. While I agree that I wouldn't want to write 'chpl' in Fortran, Chapel is not Fortran. If we had the funding and mandate to rewrite 'chpl' in Chapel, I'd do it in a heartbeat. But I think it's unlikely that anyone outside of the "Is your PL as macho as mine?" circles would care much... We do eat our own dogfood in the sense that we have 74+k lines of Chapel in our internal and standard modules (counted using wc -l, not SLOC, so that's slightly inflated, but still... a lot of Chapel code). -Brad ------------------------------------------------------------------------------ _______________________________________________ Chapel-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/chapel-developers
