rs6000: floating point cast oddities?

[degger@fhm.edu]

On 17 Dec, Dale Johannesen wrote:

>   (c89) The type of FP constants is in 3.1.3.1:  "An unsuffixed
> floating point constant has type double."
> Promotion rules are in 3.2.1.5; after some cases that don't apply,
> "if either operand has type double, the other operand is converted
> to double."

Thanks. That makes the behaviour explainable. Though I believe that
iff it is safe to calculate in single precision we should optimize
the case. If people agree on that I'd like to figure out how complicated
it is to implement.

--
Servus,
       Daniel