Lossy doesn't make sense as a criterion with floats. Float to float addition can be lossy. Whether there is a conversion is architecture dependent. I believe

Lossy doesn't make sense as a criterion with floats. Float to float addition can be lossy. Whether there is a conversion is architecture dependent. I believe x86 has an opcode for direct multiplication.