improvements for template compilations and parallelization of compilation

[Hicham Mouline]

Hello,

Did or will any work start or end on these aspects in g++:
   . with template metaprogramming libraries such as boost::mpl. 
boost::fusion, being used more and more, and translation units taking longer 
and more memory to compile, is there something compiler writers can do to 
improve this times? The facilities provided provided by such libraries are 
great for average users such as myself, but it is most of the time beyond my 
comprehension and skill to actually write code optimized so that compilation 
does not explode in memory and time.
   . Can/Is compilation of a single translation unit be parallelized at all?

regards,