Re: FW: how and where to {-# specialize #-} ?

Something similar was proposed by the gcc folk at their recent conference. The idea is that when you gcc -O4 your .c files, you get .o files containing an intermediate typed SSA form. Upon linking the program, gcc invokes a 'smart linker' which runs another pass on the whole program. This allows for all kinds of cross-module inlining, specialisation etc.

It's interesting that they are just *proposing* this. From the dagstuhl presentation http://www.dagstuhl.de/00381 in 2000, I understood that the SGI Pro64 compiler infrastructure already did this (i.e. stored a similar intermediate code structure in the .o files). I also understood that the Pro64 compiler sources were to be GPL'd. Cheers, Alan