[Vala] plans to speed up compilation

[Adam Dingle <adam yorba org>]

I work at Yorba where we develop Shotwell, a photo manager currently 
consisting of 45,000 lines of Vala code.  Each time I build on my 
machine (an 2.4 GHz Intel Core 2 Quad) 'valac --ccode' takes 10 seconds 
to run, even if I've changed just 1 line of Vala source.  (I'm talking 
*only* about the time it takes to generate C code; we then invoke gcc on 
the generated C code in a separate step.)  10 seconds isn't so tragic, 
but, still, it would be nice for this to be faster.  So I wonder if the 
Vala team is considering changes to improve compilation speed for large 
programs.  I can see two basic approaches:

1) Modify valac to be multithreaded.  As valac runs on my quad-core 
machine, 3 of the 4 cores sit idle.  If valac could use all 4 cores and 
reduce my compile time from 10 seconds to 2.5 seconds I'd be very happy.

2) Hava valac write out some sort of intermediate data which can be 
reused on subsequent compilations if source files haven't changed at 
all.  (Java works like this, sort of: when I build a Java program, javac 
can read metadata from .class files generated in previous compilations, 
I believe.)

Is the Vala team considering either of these approaches?

adam