[llvm] [AMDGPU] Do not use 64-bit literals with VOP3* encodings (PR #156602)

[Jay Foad via llvm-commits]

jayfoad wrote:

> Then in general we probably will need signed and unsigned integer operands to fully use it.

Exactly! If/when we know the signedness of all operands, then we won't have to make conservative decisions based on `!isInt<32>(Imm) || !isUInt<32>(Imm)`, and we can revert this patch.

https://github.com/llvm/llvm-project/pull/156602