"long double" Variable Precision Info.

"long double" Variable Precision Info.

Under icpc, I need "long double" variables to support ~ 34 - NOT just ~ 19 - significant decimal digits of floating-point precision. So far, I just get ~ 19. So, what compiler switches do I need
to set, please?
Under, ifort, I got ~ 34 digits very easily, and was very impressed: thank you!
Happy Hol[y]days!

3 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Well, I assume when you say "~19 digits" you're referring to 64 bits (8 bytes, IEEE 754 double precision?) and when you say "~34 digits" you're referring to 128 bits (16 bytes, IEEE 754 Quad precision). To the best of my recollection, 128 bit fp numbers are not supported in the C++ standard (nor are they supported by the hardware). I suspect the fortran compiler handles them with some sort of library, so you might look around to see if there is some C++ library that provides support for larger fp precision than the language standard provides.


I have not been attempting to answer, as I didn't understand how the question is not answered by the earlier thread on this topic.

Leave a Comment

Please sign in to add a comment. Not a member? Join today