[bitc-dev] LLVM, CLR, or C

Jonathan S. Shapiro shap at eros-os.org
Wed Mar 17 14:58:50 PDT 2010


I've been updating myself on the current state of LLVM, and it looks like
real progress has been made on the GC front. Given this, and the fact that
the LLVM path would be easier than a re-target to CLR, I don't see a
currently compelling reason to take the diversion on CLR. It's still worth
thinking ahead about which things CLR cannot express, though, because CLR is
clearly an eventual target. We can simulate unboxed semantics for unions
that need that, but SIZEOF might be a concern.

In the very short term, however, I think we should continue with Boehm GC
and C emission. Getting the surface syntax of the language stabilized is the
thing that blocks almost everything else, so anything that *can* wait until
after that probably *should* wait.

Given the way we currently emit C code, it doesn't look like a very big leap
to shift to emitting Henderson-style accurate-collectible C code if Boehm
ends up getting in the way for some reason.

shap
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.coyotos.org/pipermail/bitc-dev/attachments/20100317/41796851/attachment.html 


More information about the bitc-dev mailing list