Thanks - Re: quad precision question

[Stu Anderson 425-865-3595]

My thanks to all who replied.  I now see what has happened.

In case you are interested, the question relates to legacy
Fortran 77 codes being compiled under doubling switches: e.g.,
-i8 -r8 -d16.  Obviously, new machine constants are needed in
such cases.  That is not a problem with modern languages such
as C and Fortran 90, but Fortran 77 lacks any environmental
inquiry.

"Automatic" machine-constant codes such as LAPACK's xLAMCH
don't always work.  The Bell Labs codes x1MACH don't support
most of these switches.  The replacements for x1MACH (Gay &
Grosse, TOMS 1999 25(1):123-126) often won't work either.

The switches are being used to migrate Classic Cray codes to
workstations.  Also, some huge sparse linear algebra codes
need 64-bit integers.  While we rarely need quad precision
today, that won't last.

And I used to think that the introduction of IEEE arithmetic
was going to simplify my life! *grin*

--Stu Anderson
_______________________________________________________________
stu.andersonaboeing.com -- Mathematics and Computing Technology
http://www.rt.cs.boeing.com/MEA/comp_math/sla/
http://www.halcyon.com/stuander/
Who speaks for Boeing?  Not me!