[llvm] [AMDGPU] Change the representation of double literals in operands (PR #68740)

[Jay Foad via llvm-commits]

jayfoad wrote:

> we need to change the representation to store a full 64-bit value

I think this is clearly a Good Thing. There are just some details of the implementation that I don't understand.

https://github.com/llvm/llvm-project/pull/68740