Benchmarking a distributed version of clang

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Benchmarking a distributed version of clang

Mike Miller
Me and my partner(Joshua) are nearing the point of being able to test our distributed extensions to clang. However, we need a good program to benchmark compilation! On #llvm, some people suggested we compile clang itself.

I'm not well-versed in Makefiles, and I'm not sure if this would be easy to do, given how we've structured our compiler, so I was wondering if someone here could shed some light on how we might do some benchmarking.

The way we've structured our distribute option is that the user invokes "clang -cc1 -distribute <other args> myFile.c -o myOutputFile.o". Regardless of what args the user provides, we always output an object file, and do no linking whatsoever(we use an EmitObjAction). We also only support 1 input file at a time. (We obey all other args as you normally would). Given these limitations, would it be easy to get clang building in this manner? If not, is there another large program anyone can recommend which we could benchmark compilation with?

(As an aside, we realize that long-term we need a a true drop-in replacement for clang, but given our impending deadline, we need to get some results with our code as is.)

Thanks!
Mike

_______________________________________________
cfe-dev mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev