At 00.33 -0700 0-09-11, Simon Peyton-Jones wrote:
>| How about Hugs' *extremely* fast compilation time? (Which is
>| the main reason for me (and many others) to use Hugs.)
>
>It won't be as fast as Hugs, but it will be much much faster than GHC.
>No figures yet, though. Getting it 'fast enough' is probably the major
>risk area.
Speed may not be as crucial as for Hugs, because Hugs is written mainly for
32-bit computers (and on MacOS, nobody was able to compile it even for
32-bit 68k, but only for 32-bit PPC (RISC) or better), but the new G3/G4
CPU's on the Mac have 64/128-bit CPU's which are much faster. (In effect
the old definition of "supercomputer" of gigaflop/second.) A similar
development is taking place on the DOS/Windows PC's.
It is not really worth even trying making a GHI for the old slop 32-bit
CPU's; for example the new BSD UNIX-based MacOS X won't even run on those
computers.
But one other problem with Hugs is that it is written in a very imperative
C-style, which makes it difficult to implement things like the vents I now
look into. It would be good if the new GHI has some concurrency built into
it, or at least, is more encapsulated with local data, so it's easier to
implement events handling and such.
Hans Aberg
