[daip] AIPS on the x86_64.
Eric Greisen
egreisen at nrao.edu
Wed Nov 3 12:05:15 EST 2004
Omri Schwarz writes:
> I work in Linux support for the Center For Space Research, and I have an
> AIPS related hardware purchase to decide on, so I thought I'd ask whether
> anyone has begun running AIPS on Opteron machines and whether the AIPS
> code base makes enough use of 64 bit floating point operations to justify
> spending the extra money for a 64 bit processor. I'd be much obliged
> if you had any information.
>
AIPS does very little DOUBLE PRECISION computation since 32-bit floats
are adequate for almost all scientific data. In fact, we have had
problems with our dynamic memory algorithm on 64-bit machines since
some of them allocate dynamic memory at addresses requiring almost all
of the 64 bits while placing the compiled code at addresses near 0.
We have a solution to that problem which we will implement at some
point soon.
I do not know what sorts of machines AIPS is run on - in fact I do not
even know what the Italian machine was on which I debugged the dynamic
memory problem.
Eric Greisen
More information about the Daip
mailing list