elliottt
change pretty-printing of floats because of LLVM IR restrictions
As was pointed out by @tomsmeding here: https://github.com/GaloisInc/llvm-pretty/issues/135#issuecomment-2227477305 we cannot use 32-bit (8 characters) "float" constants in LLVM IR.
Instead, we must embed the 32-bit float as 64-bit double constants with the same meaning.
The particular case that triggered the problem for me is fixed with this PR, thanks! But this is only really a spot test, because this only accidentally popped up in the test suite of a larger codebase.
Thanks for the very quick PR!
In a discussion with the group, we feel like we may just want to always print the hexadecimal version, possibly providing a flag to add a best-effort decimal expansion as a comment nearby where applicable. Setting this as a draft until I have more time to handle these considerations.