How to explicitly define float point registers as 'float'. about snowman

[yegord]

Right.
Types are inferred by TypeAnalyzer from the operations performed on the objects.
Currently IR does not have any floating-point operations; this should be fixed eventually, see also #28.

Topic-wise, I am not sure that it is a good idea to infer the type of a value based on its memory location.

[nihilus]

There are in the MIPS ISA special assembly opcodes for converting data types of float point registers.

[yegord]

Yep. But there are no floating-point operations in IR. You are welcome to add them. :)
Once you have added them and defined the semantics of floating-point instructions, you will get floating-point types.

[yegord]

Duplicate of #28

[nihilus]

Id say #28 is a sub-issue of this one. Since SSE2 is not about float pointing arithmetics but SIMD and is very x86 specific. Better close #28 and leave this more generic issue open.

[yegord]

#28 was submitted earlier.