Re: [Haskell-cafe] Re: Compiler backend question

On Tue, Jan 01, 2008 at 06:44:46PM +0100, Achim Schneider wrote:

Peter Verswyvelen wrote:

...

Another question regarding the backend: a cool feature of the Microsoft Visual C++ (MVC) compiler is its ability to perform "LTCG" (link-time-code-generation), performing whole program optimization. It something like this possible with Haskell / (or the GCC backend?). Would it be possible to feed all the generated C code of the GHC compiler into the MVC compiler, to generate one big MVC / LTCG generated executable? It would be interesting to see how much the whole program optimization approach (which can do analysis on the program as if it was a single module) would improve performance...

jhc does that. Or rather, it doesn't compile Haskell modules separately but throws them together (presumably) in the first pass.

It optimizes everything seperately to a level that is analogous to ghc core (which I call, oddly enough, jhc core). it then collects everything together and translates it into grin, which doesn't have an exact equivalent in ghc but is equivalent to a first order monadic strict functional language and analogous enough to the SSA intermediate form of imperative compilers that one would find it quite comfy were one used to that. I would like to move jhc to more of a 'link-time-code-generation' model though if I understand what you mean, right now jhc does a full monolithic compilation which is pretty resource intensive. I would like to switch to more of an 'incremental whole program compilation' (no, that isn't a contridiction) whereby individual files will compile to .o files, but then before the final linking a special pass will be done to generate the appropriate specialized "glue" for putting the .o files together, this should give a nice middle ground between fully monolithic and the limitations of abiding by the standard C linker. John -- John Meacham - ⑆repetae.net⑆john⑈