[llvm] [AArch64][GlobalISel] Improve lowering of vector fp16 fptrunc and fpext (PR #163398)

[David Green via llvm-commits]

davemgreen wrote:

This looks like it should be two different patches, one for fptrunc and one for fpext. The fpext is the easier of the two IIRC, which should hopefully be a "lower" like we do for other extends. Trunc is more difficult but I don't think we can split it into two different trunc nodes and "fix" them later - we might need to introduce a new node type for the round to odd.

https://github.com/llvm/llvm-project/pull/163398