[gcollecutt]

Description:
I downloaded OpenFOAM 1.4.1 and attempted to compile from scratch under Linux FC5 x86-64. The OpenFOAM-1.4.1/.OpenFOAM-1.4.1/bashrc script did not detect the 64 bit OS and as I only have the 64bit versions of the X11 libraries available, the compiliation failed. As a work around, I put "export WM_64=Y" in my $HOME/.bashrc file just prior calling the ". OpenFOAM-1.4.1/.OpenFOAM-1.4.1/bashrc" line.

Platform:
Linux 2.6.19-1.2288.2.4.fc5 (x86-64)

Version:
OpenFOAM 1.4.1

[olesen]

This is a feature, not a bug.
The default is 32bit. If you wish to compile and use 64bit (on a 64bit machine) you need to set WM_64 to something (as you did).

[henry]

This is not a bug, OpenFOAM is setup to build either 64bit or 32bit on a 64bit machine and 32bit is the default. To select a 64bit build you set WM_64.

Generally we run 32bit even on 64bit machines as on some machines it is faster and always takes less storage. On recent 64bit machines the 64bit version is faster and in future releases we will make the 64bit build the default on 64bit machines.

[gcollecutt]

When compiling double precision code, I have ALWAYS found 64 bit code much faster than 32 bit code. As double precision compilation is the default, may I endorse Henry's suggestion of making the 64 bit compilation the default on 64 bit OS.

Also a refence to setting the WM_64 macro in section 4 of the README file wouldn't go amiss.

[gcollecutt]

Sorry Guys - I just found the reference to WM_64 in the readme file - a bit further on than I had read to!

My Apologies.

[henry]

We have never found running 64bit much faster than 32bit but on some processors it is faster. Typically 64bit is faster than 32bit on AMD64/Opteron processors and on the latest 64bit Xeon processors but not on P4-EM64T processors. However, we expect all recent and future 64bit processors to provide better 64bit than 32bit performance so we will make the 64bit option the default for 64bit processors in the future.

[gcollecutt]

Henry, changing from 32 bit to 64 bit on a P4-EM64T gave me a 70% increase in speed when running LS-Dyna double precision. I have noted that, for single precision code, there is little improvement going from 32 bit to 64 bit. Perhaps my observed speed improvement was more a quick of the f77 compiler.

[henry]

My comments only relate to the performance of OpenFOAM on the machines we have at OpenCFD. We expected an increase in performance in using 64bit on our P4-EM64T machines but didn't get, but these are an early version of the P4-EM64T processor. We now have a Xeon 5160 machine on which we do get in increased performance when running 64bit but it is not anything like 70%.