Pseudoinstructions and FLoating point for MIPS

[ghost]

The specification requires MIPS I assembly.

However, to confirm the ambiguity for floating point and pseudo instructions I want to clarify which instructions should be supported.

To conform with the clang outputs i believe that pseudoinstructions should be ignored such as mov etc. Since a real assembler wouldn't create such instructions. Also, for floating point i believe it should be assumed that they are supported as they are part of the original MIPS specification.

[ghost]

aparently move works from todays ABI lecture so...

[m8pple]

Sorry, completely missed this. I realised github notifications weren't getting through
the filter, so now have a backlog.

In terms of mov/move, it is a purely syntactic thing - there are multiple base
instructions that can support it. However, there are other things like li with
arbitrary 32-bit constants where it isn't so simple.

So your point is a good one. What would you like it to be?:

• What is the point of reference for assembly? Which program needs to accept it,
  and should be considered to define the output language for your compiler?
• For things like floating-point, at what level should it be handled?

[ghost]

I believe that since both QEMU supports FPU and the mips-linux-gnu-gcc (flag -mips=march32) compiler support floating point instructions and registers that it should be handled as is. If there is any ambiguity we should refer to the gnu compiler.