gnupic: Re: [gnupic] LLVM - was Re: [gnupic] gputils development
Subject:
Re: [gnupic] LLVM - was Re: [gnupic] gputils development
From:
Ralph Corderoy ####@####.####
Date:
15 Nov 2008 12:21:17 -0000
Message-Id: <20081115122112.BDD8D4449@blake.inputplus.co.uk>
Hi Walter,
> There have been several attempts at defining standardized
> optimizations. All of them including LLVM work quite well once they
> are written at optimizing code for architectures similar to the
> original target. The problem that most have not solved is making the
> optimizations relevant as the target architectures change. LLVM's
> approach to standardizing IR doesn't solve this problem.
No, but it does allow common optimisations for free, and make
experimentation with architecture-specific ones easier. Other aspects
of LLVM, IIRC, include the advantage of taking sufficient information
from the source all the way through to link time, allowing Link Time
Optimisations, e.g. in theory, discarding an unused global debug flag
and all the pieces of code that check it.
There's lots of interesting papers published on the LLVM site, here's
just a couple.
http://llvm.org/pubs/2005-05-21-PLDI-PoolAlloc.html
http://llvm.org/pubs/2005-06-12-MSP-PointerComp.html
They also have their Clang C compiler as an easuer one to experiment
with than gcc.
Cheers,
Ralph.
