...diagnostic any arithmetic use of NaN is unacceptable in our environment...are the requirements in the Fortran standard to trap exceptions? What...earlier question of how relevant IEEE754 (or any similar standard...to a discussion of Fortran. IEEE754 is meant to guide...has anything to do with Fortran standards. Having said that, I...when you attempt to use NaN. I believe the correct syntax...
... (snip) If you want "max(x,NaN) = x", then float max(float x, float ...> intrinsic built upon flawless hardware. Testing for NaN is not an expensive operation, as was ...The question in my mind is what should the behavior of the IEEE754/854 compliant FPU (or vector coprocessor, e.g. SSE) be? Can it ...gets a lot more complicated if you want NaN to work out right. Note that C89 ...
..., even if y = NaN. That's nothing :-( Look at pow: pow(NaN,0) = 1 pow(+-inf,0) is unspecified ... to do with the brokenness of IEEE754. The latter is intended for.... If you want "max(x,NaN) = x", then float max(float x...built upon flawless hardware. Testing for NaN is not an expensive operation, ...what should the behavior of the IEEE754/854 compliant FPU (or vector...
... the question |> is whether |> |> max(x,NaN) |> |> should return x or ... a floating point value should return NaN if any |> operand is ..."not a number", of which IEEE754 generally selects the one...With that meaning, 'max(x,NaN) = NaN'. The meaning where 'max...statistical calculations need 'max(x,NaN) = NaN' in all calculations EXCEPT...this operation, and 'max(x,NaN) = x' forces extra code for...
... IEEE standard and the Fortran 2003 standard are relevant. Hmm, regarding (a) to... number and the other a NaN. Otherwise it is either x or ... that this part is not in IEEE754-1985.) And therefore, I assumed:...: HUGE(x), INF, NaN (i.e. MAXVAL implementation using IEEE's MAX); if one implements MAXVAL using "<" etc. directly, I come also to the result NaN for (a) and (b). I am...
...all operations involving one or more NaNs" was talking about operations defined as.... They state that comparisons (<,>,>=,<=) with NaNs in either operand return quiet NaNs...verse is section 5.17 of IEEE754-1985: "Comparisons are exact and... exceptions even when operands are NaN. For example, an operator that ... in the standard, the corresponding Fortran operations are listed as ".UG.",...
... huge(x), NaN, INF ] ) c) MAXVAL( [ NaN ] ) d) MAXVAL( [ ] ) Following IEEE's max(),...NaN as the return value for MAXVAL. This issue has come up on...three cases. Unfortunately proposed new the IEEE standard (754r) adds new operations MIN... of this definition, max(5,NaN) is 5. The same reasoning (...and c) to come back as NaN. -- wade ward "Nicht verzagen, Bruder ...
... huge(x), NaN, INF ] ) c) MAXVAL( [ NaN ] ) d) MAXVAL( [ ] ) Following IEEE's max(),... NaN as the return value for MAXVAL. This issue has come up on ...three cases. Unfortunately proposed new the IEEE standard (754r) adds new operations MIN...>> unexceptional. If both operands are NaN, see section 6.2. If ... of this definition, max(5,NaN) is 5. The same reasoning (...
...(x), NaN, INF ] ) c) MAXVAL( [ NaN ] ) d) MAXVAL( [ ] ) Following IEEE's ... and (c) should all deliver NaN as the return value for ...the answer ought to be NaN in all the first three...cases. Unfortunately proposed new the IEEE standard (754r) adds new operations...unexceptional. If both operands are NaN, see section 6.2. If...of this definition, max(5,NaN) is 5. The same reasoning...
...: HUGE(x), INF, NaN (i.e. MAXVAL implementation using IEEE's MAX... NaN for (a) and (b). Tobias Fortran 2003, section 1.9, references IEC...", as the normative reference for IEEE arithmetic. The same section also says that IEC 60559 was originally IEEE754-1985 and "this standard refers...throughout section 14, which defines the IEEE arithmetic and exception handling facilities of...
