jhuber6 wrote: > Turned out I've forgotten to add SVE-guarding condition, so it always been > trying to build scalable types even on X86.
Well I guess it's a good sign that it didn't break even harder. In general I think this is a more promising direction, because it allows you to do function overloading based on whether or not it's scalable. Not everything will work with scalable vectors and that's okay, but it makes it way easier to provide a unified interface. But it would be nice to get some more opinions from people familiar with this stuff. There should probably be a more formal RFC on the LLVM discource once we've gotten some of the basic issues worked out. What is really important is that a lot of the 'basic' operations work out of the box, like initializers and masks. https://github.com/llvm/llvm-project/pull/183307 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
