[MLton-devel] teaching a compiler course using MLton

Matthew Fluet Matthew Fluet <fluet@CS.Cornell.EDU>
Wed, 18 Sep 2002 16:04:04 -0400 (EDT)

> I'm worried though about how feasible this strategy is given the
> complexity of the backend and the detailed assumptions it makes about
> object layout (representations), chunks, etc.

With all the recent additions to the RSSA and MachineIL semantics, this is
probably a little harder than it was a year ago.  The MachineIL is pretty
explicit about moving stack pointer around, having precomputed object
headers, making the distinction between pseudo-regs and stack slots,
having explicit GC calls, etc.

RSSA probably isn't much better, but relaxes some of those distinction.

The chunk concept doesn't apply to the x86-codegen; we take top-level SSA
functions one at a time and compile them to assembly.  

>   -- simple first-order optimizations -- loop-invariance,  common
>               subexpression, useless argument, ...

If you end with these types of optimizations, you could compile to the
SSA-IL, although that would likely be a parallel move from a similar IL.

This SF.NET email is sponsored by: AMD - Your access to the experts
on Hammer Technology! Open Source & Linux Developers, register now
for the AMD Developer Symposium. Code: EX8664
MLton-devel mailing list