What are the opinions here about merging dragonegg into FSF gcc? It is in the odd position of straddling two projects so perhaps it could reside in both the LLVM and FSF gcc projects with regularly remerging. Certainly it would be an interesting addition to FSF gcc. For instance, without even attempting to target gfortran, in the last year, the llvm code performance has improved about 15% for the Polyhedron 2005 benchmarks...
http://lists.cs.uiuc.edu/pipermail/llvmdev/2010-April/030800.html For the darwin target, this would be a particularly big win since we would gain access to the llvm LTO. Thanks in advance for any comments. Jack
