Topic: Quad(ruple precision; REAL*16) support in General

Wolfgang Hoeppner

Posts: 8 Sibiu / Romania

Back to Top

27 Jul 2013 4:13 #12696

Will there be, in one of the next releases of FTN95, a support of quadruple precision arithmetic, at least for fundamental arithmetics (+ - * /), but preferably also for SQRT, SIN, COS, &c.?

Best wishes

Wolfgang.

PaulLaidler

Posts: 7975 Salford, UK

Back to Top

27 Jul 2013 7:35 #12697

At the moment there are no plans to support REAL*16 arithmetic. One approach might be to use the GMP (GNU) library. It would be helpful to know if anyone has experience of accessing this library from Fortran.

PaulLaidler

Posts: 7975 Salford, UK

Back to Top

28 Jul 2013 7:56 #12698

As a quick test I have added the following PRINT statement to a working 64 bit ClearWin+/gFortran program...

print*,sqrt(2.0_16)

and the output is

1.41421356237309504880168872420969798E+0000

so it looks like you could work with 64 bit gFortran and ClearWin+.

p.s. It think that this value is correct up to 32 decimal places.

DanRRight

Posts: 2877 South Pole, Antarctica

Back to Top

29 Jul 2013 5:02 #12709

What order of exponents real*16 allows?

davidb

Posts: 555 UK

Back to Top

29 Jul 2013 7:35 #12710

Approximately 1.0E-4932 to 1.0E4932.