[daip] Aips on Intel and AMD
Nanni Mauro
nanni at ira.cnr.it
Fri May 30 04:15:28 EDT 2003
Hello.
In this days I have get up the current release of AIPS to upgrade
our computer in Bologna Radioastronomy Institute, about 60 PC in
nfs with linux RedHat 7.x, 8.0 and 9.0
I use the 2.95.3 gcc to compile, and I had a problem compiling in rh9.0
because the "ernum" modules was not found . This can be fixed with
a -ldbsd in LDOPTS.SH. Compilng on rh8.0 the -ldbsd is not
necessary, but the code do not run on rh7.x computer , and I decide to
compile -static that works on all the OS release.
Then I run y2k to test the some computers and I have strange results:
RH 8 Intel 1000 Mhz 32 AipsMark
RH 8 Intel 2000 Mhz 75 AipsMark
And this is OK
RH 9.0 AMD 2000 Mhz 31 AipsMark
RH 7.2 AMD 1000 Mhz 11 AipsMark
That are low value for these computers that works fast in others
applications.
Can you help me to understand this. Can the "static" compilation
have a so hight impact in the performance on the different architectures.
We try to compile on AMD but we don't have better results ....
Best regards
M.Nanni
More information about the Daip
mailing list