[MLton] OO in SML (was: mGTK and MLton)

Stephen Weeks MLton@mlton.org
Thu, 18 Nov 2004 15:29:05 -0800

> I still haven't heard anything about the feasability of SSE2 and/or
> access to the 32*32->64 MUL instruction. I am still interested in
> these, but I think adding them is beyond my current understanding of
> MLton. Especially since SSE2 has a distinct register set.

I agree that SSE2 would be hard.  If you have to have it, easiest for
now would be to use the FFI if possible, and leave it to a C compiler
to do the SSE2 stuff.  Which is I guess what you're doing.

Adding the optimization to do 32*32->64 MUL should be much easier, but
I'm not sure if any of us has the time to do it right now.  You could
at least do better than doing all the conversions and the multiply in
SML by doing a single C call via

	_import "mul": Int32.int * Int32.int -> Int64.int;

It is likely that we will work on a port of MLton to x86-64 next year.
We will be working on the code generator, so at that point we will
certainly revisit these issues.