Well I've done a /usr/local/bin/scan-build xcodebuild -project   
against the iPhone VM. The LLVM static analysis points out a few  
interesting things I've not really
look too hard at.

One of which is that in many places in the interpret we can end up  
with using initialized variables when the success code is false.  
However in looking at
the problem areas, if the success flag is set to false (usually not  
possible) then continuing execution is impossible anyway.

Obviously someone *could* cleanup VMMaker to ensure LLVM static  
analysis pronounces it as clean.

Now as for the compiler  getting LLVM to compile the VM is chore, and  
we know that GCC 4.2 produces a slower interpreter than 4.0. This  
after *much* work to see if we
could fool the 4.2 compiler to product code at least as good as 4.0


On 21-Mar-09, at 10:58 AM, askoh wrote:

>
> I noticed that Mono and Python are doing some work with LLVM.
> http://www.mono-project.com/StudentProjects#LLVM_Back-End
> http://codespeak.net/pypy/dist/pypy/doc/translation.html#the-llvm-back-end
> http://llvm.org/
> What are they gaining from LLVM? Can Squeak benefit from LLVM?
>
> Thanks,
> Aik-Siong Koh
> -- 
> View this message in context: http://www.nabble.com/Is-LLVM-useful-to-Squeak--tp22638664p22638664.html
> Sent from the Squeak - Dev mailing list archive at Nabble.com.
>
>

--
= 
= 
= 
========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com>
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
= 
= 
= 
========================================================================