Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
28559 Discussions

Errors defined for ASIN function?

ereisch
New Contributor II
572 Views

Are there any runtime errors defined for the ASIN function, other than the obvious domain error of ABS(A) > 1.0?  I'm getting an underflow error when I perform ASIN(4.15e-5), which should be entirely valid as the result is only 2.38e-3 (well within the bounds of REAL*4).  Input and output data types are both REAL*4.

0 Kudos
3 Replies
ereisch
New Contributor II
569 Views

Correction: Result is 4.15e-5, not 2.38e-3.  Still well within the capabilities of single precision though.

0 Kudos
mecej4
Honored Contributor III
546 Views

Please report the problem with sufficient detail that will allow it to be reproduced, as recommended.

Here is a counter-example:

program xasin
implicit none
real :: x = 4.15e-5
!
print *,asin(x)
end program

Compiled using Ifort 19.1U3 on Windows, this program prints out:

  4.1499999E-05
0 Kudos
ereisch
New Contributor II
538 Views
If you build the program to not mask underflow exceptions (which is not the default) and prevent flush-to-zero, it should throw an exception. I can't remember what those flags are offhand (the latter is -no-ftz IIRC).
0 Kudos
Reply