-----Original Message-----
From: David A. Case [mailto:case.scripps.edu]
Sent: Tuesday, March 28, 2006 10:49
To: Ross Walker
Subject: Re: test failures for QM/mm for SGI
On Mon, Mar 27, 2006, Ross Walker wrote:
>
> The only thing I can think is that the compiler is taking liberties with
> floating point precision.
>
> FFLAGS= -n32 -mips4 -r -OPT:roundoff=3:IEEE_arithmetic=3 -TENV:X=1 -O0
> $(LOCALFLAGS) $(AMBERBUILDFLAGS)
> FOPTFLAGS= -n32 -mips4 -r -OPT:roundoff=3:IEEE_arithmetic=3 -TENV:X=1
-O3
> $(LOCALFLAGS) $(AMBERBUILDFLAGS)
>
Removing the OPT and TENV flags seems to have done the trick with regard
to
not getting SCF convergence in the "built-in" codes. I've updated
configure.
Divcon is still broken on SGI_MIPS, at least with my version of the
compilers.
Can anyone else test this to see if it is a problem local to my machine?
(Xiongwu: have you run "make test.sander.DIVCON" and "make
test.sander.QMMM",
etc?)
...dac
Received on Wed Apr 05 2006 - 23:49:33 PDT
