I sometimes need quad precision in a few calculations where speed is not
the primary concern. I have found a quad precision module at The Allen Miller
Fortran software site and have given it a try. It seems to work but I wondered if anyone else has tried this or knows of an alternative.
Hi Charles. I assume you mean ALAN Miller. He references double-double and quad-double from a Berkeley site. I have used these for several years and found them very reliable. Which particular package, from Alan Miller’s site are you using? He seems to have several depending on which compiler you are using!
I have been playing with the version called quad_df.f90 (as opposed to quad.f90) the difference being that the latter seems to assume the 80 bit
8087 register structure while the former does not. I compiled with the
compiler option -O0 and it seems to work fine. His test program compiles
and runs and the results seem reasonable. Sadly, so far I have not figured out a clever way to check the results by printing out the result of a known
computation as a long decimal number. I can think of a number of ways of doing this by hand be looking that the binary representation of the two
double precision pieces and converting to decimal…but I am too lazy. Thus
your experience is encouraging. Any tips will be greatly appreciated.