Add floating-point operations, ceil and floor

[elpinal]

This PR adds 2 floating-point operations:

• ceil : float -> float
• floor : float -> float

[nyuichi]

@elpinal

Nice pull request 👍
Could you possibly add some more functions like the following too?

• truncate : float -> float
• round : float -> float
• pow : float -> float -> float (pow x y = x^y)
• exp : float -> float (exp x = e^x)
• sqrt : float -> float
• log : float -> float -> float
• log10 : float -> float

[elpinal]

@nyuichi SATySFi already has a primitive named round, with type float -> int. SATySFi's round is defined as OCaml's int_of_float, which in turn is equal to OCaml's truncate. What do you think of that?

[nyuichi]

@elpinal

Thanks for your feedback!

SATySFi already has a primitive named round, with type float -> int.

Aw maybe we need to use a new name other than round. One better name is float-round, I think. (cc: @gfngfn)

SATySFi's round is defined as OCaml's int_of_float, which in turn is equal to OCaml's truncate

What I want is the function called Float.trunc in OCaml's standard library, which is not equivalent to truncate in Pervasives.
I would name this float-truncate following the above convention.
To be honest, their actual names do not matter to me because I will just re-export them in satysfi-li as Float.round and Float.truncate.
So I don't have a strong opinion about this.

[gfngfn]

I’m so sorry for the late response… I merged only the implementation of ceil and floor for now.

Note: currently, tools/gencode/vminst.ml is automatically generated from src/frontend/bytecomp/vminstdef.yaml. Thus I updated the latter so that it generates the former one, which you had extended. We will probably put only the former under version control in the future (See: #226).

Thank you for the improvement!