[LLVMdev] Forwarded to llvm list ...

[John Criswell]

Mark ONeill wrote:
> This was originally sent to Vikram Adve -- I have forwarded it to the list
> as per his suggestion ...
> 
> 
> Regards
> 
> Mark O'Neill
> 
> ---------- Forwarded message ----------
> Date: Thu, 8 Dec 2005 11:51:40 +0000 (GMT)
> From: Mark ONeill <mao at raven.linux.ox.ac.uk>
> To: vadve at cs.uiuc.edu
> Cc: mflesner at cs.uiuc.edu
> Subject: LLVM help ...
> 
> 
> Dear Vikram
> 
> I spoke to you quite a long time ago about using LLVM to shrink wrap P3 a
> POSIX based clustering environment I have written to support coarse
> grained parallel computations. At the time, LLVM (and P3!) were a little
> immature, and trying to create a VM version of my system would have been
> somewhat premature. However, I appear to be getting on a lot better with
> the 1.6 distribution of LLVM and have compiled portions of the system with
> LLVM -- and they run!
> 
> To do the job properly I need a little help. I notice your compiler
> produces a shell script invoking the virtual machine thus:
> 
> ..
> 
> exec $lli
> 
> ..
> 
> If the application to be invoked is called <app> how hard would it be to
> change that line in the shell script to:

This would be a very easy change.  I believe the shell script is 
generated by the gccld program (llvm/tools/gccld).  It should be pretty 
simple to change it to do what you describe.

If you want to make the change yourself, test it, and then supply a 
patch to the list, that would be great.  We'd be happy to incorporate 
it.  Just make sure that standard bourne shell (as found on Solaris, for 
example) supports the -a option you describe.

Otherwise, please feel free to file a bug report requesting the feature 
at http://llvm.cs.uiuc.edu/bugs and one of us will look into it.

-- John T.

> 
> ..
> 
> exec -a <app> $lli
> 
> ..
> 
> e.g. If I do llvm-gcc -o <app> <app>.c ????
> 
> 
> I use named tracking of processes so if the "binbary" name as seen by ps
> etc actually matches <app> this would be very useful. I could add as a
> post processing step, but the logical place seems to be in LLVM ...
> 
> Sorry to bother you, but if I can get this fixed, I can handle the rest of
> my integration ...
> 
> 
> Regards
> 
> Mark O'Neill
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev


-- 
John T. Criswell
Research Programmer
University of Illinois at Urbana-Champaign
"It's today!" said Piglet. "My favorite day," said Pooh.