[MLton] Possible contributions to optimization

[Kristopher Micinski]

On Tue, Mar 4, 2008 at 5:08 AM, Matthew Fluet <fluet at tti-c.org> wrote:
>
>  Hi Kristopher,
>
>  If you are interested in contributing to MLton, you should subscribe to
>  the MLton development list (mlton at mlton.org; link available from
>  http://mlton.org/Contact).
>
>  The wiki contains a number of pages that point towards various aspects of
>  development:
>    http://mlton.org/Development -- overview of development pages
>    http://mlton.org/CompilerOverview -- overview of compiler structure,
>      including links to various intermediate languages and optimization
>      passes (themselves with links to the corresponding source files).
>    http://mlton.org/Projects -- short list of suggested projects
>    http://mlton.org/Talk -- a very brief overview talk of the compiler
>    http://mlton.org/References#Weeks06 -- a longer talk about the compiler
>
>  Based on your self-description, I would probably suggest thinking about
>  implementing a loop-invariant code motion optimization pass in the SSA IL.
>  Because MLton uses a defunctionalization technique
>  (http://mlton.org/References#CejtinEtAl00) to turn higher-order functions
>  into first-order functions, the SSA IL looks very much like one you would
>  see in a first-order language compiler (like one for C).  This means that
>  a lot of "traditional" compiler optimizations can be implemented in a
>  fairly straightforward manner.  Nearly every compiler textbook will have a
>  description of loop-invariant code motion, and there is a good chance that
>  it would have some measurable impact on code.  It should also be fairly
>  easy to adapt to the SSA IL.
>
>  Some starting points would be to familiarize yourself with the SSA IL,
>  familiarize yourself with some of the other SSA IL optimizations
>  (http://mlton.org/SSASimplify; the CommonSubexp optimization is fairly
>  straightforward and should be familar from a compiler textbook).  You can
>  use the "-keep-pass passName" and "-diag-pass passName" compiler options
>  to dump the program as it appears before and after a pass and to dump
>  various diagnostics generated by the pass.  Also, the "-keep ssa" compiler
>  option will dump the final SSA IL program during the compilation.
>  Browsing through some of the dumped .ssa files will help you see what the
>  intermediate code looks like.
>
>  And, of course, feel free to ask questions on the development list.
>
>  -Matthew

Yes, I intend to by Appel's book, and have already read the Dragon
book and "Modern Compiler Design" (which also includes functional
language compilation, though only a small section). I plan to read the
code in MLton, but want to know if it would be worth implementing a
backend for LLVM (which can then optimize even more with generated
SSA), or if this is just an off the wall idea.

Sorry, didn't have much time, but I plan to subscribe to the developer list.

Thanks all,
    -- Kristopher Micinski