On Jun 23, 2005, at 6:23 AM, F2k3 wrote:

> USE, INTRINSIC ::  IEEE_FEATURES, ONLY: IEEE_ROUND

I assume you mean IEEE_ROUNDING. There is no IEEE_ROUND that I see 
offhand (though I don't have this stuff all internalized and have to 
look it up, so it is always possible that I missed something, but this 
looks like you mean IEE_ROUNDING).

> USE, INTRINSIC :: IEEE_ARITHMETIC
> PRINT *, IEEE_SUPPORT_DENORMAL()       ! <--(A)
> END
>
> Q. Assuming that there is hardware support for
> denormal numbers for all reals

Point 1. Hardware support has absolutely nothing to do with anything in 
the standard. The standard's use of the term "processor" does not mean 
the hardware processor. It means the whole system, specifically 
including the compiler. It is more accurate to read the standard's 
"processor" as "compiler"; not quite completely correct, but that's a 
closer approximation to the meaning than "hardware". A compiler can 
support all the IEEE stuff without any hardware floating point at all 
(slow, but it is allowed). Conversely, a compiler can refuse to support 
any of the IEEE features even if the hardware supports them all. This 
really is quite an important distinction.

So I will translate your statement as "assuming the compiler supports 
denormals for all reals."

>  and IEEE_FEATURES
> module defines the named constant, IEEE_DENORMAL.

That doesn't matter since you don't access it. The fact that the module 
defines that constant means that the compiler is capable of supporting 
IEEE denormals. But it is possible that the support might make code 
inefficient on some processors (in fact, yes, there are processors 
where it does). Therefore, the processor doesn't have to use its 
support of that if you don't demand it. USEing IEEE_DENORMAL would be 
how you demand it.

>  What should be the output for the print statement in (A)?

Processor-dependent (but either true or false). That's exactly the 
point of the function. You haven't demanded support, but you can 
inquire whether you are getting it anyway. You have basically left it 
up to the compiler to choose. Odds are that the result will probably be 
true if the support is "cheap" (in terms of efficiency), but false if 
the support is expensive.

-- 
Richard Maine                |  Good judgment comes from experience;
[log in to unmask]       |  experience comes from bad judgment.
                             |        -- Mark Twain