If you can get this pushed into the LLVM tree as an optimization, then we can use it, or it needs to be licensed under a LLVM compatible license. We can't use any GPL'd code in our compiler stack for various reasons.
Originally posted by: viewon01 Hi,
I just would like to advertise a very interesting project based on OpenCL :
Maybe the AMD team can integrate it natively (if not done !!). Everybody can benefit from it !!!
I haven't read the paper but was hoping you could give a quick summary as to how this approach is different from many other auto-vectorization ideas?