Compile gromacs on Cygwin
May.27, 2010 in
Modeling & Simulation, 分子模拟
Just some notes:
1. manually download and install fftw3. Do not use the cgywin package, or the path will not be recognized by Gromacs.
2. Needed Cygwin packages: lapack, gcc, g77, make and nasm. nasm is required for both fftw and Gromacs in order to compile the assembly codes.
June 10th, 2010 at 6:24 pm
Dear Ting,
I am currently using GROMACS for simulations and I was wondering whether GROMACS on CYGWIN is less capable than GROMACS run on LINUX? i.e. concerning its speed and its ability to simulate larger macromolecules.
I would be very grateful for your help.
Thank you
June 11th, 2010 at 2:33 pm
Hi nbbb,
First, for serial compilation I don’t think there is any difference in functionality than between the binary on Cygwin and LINUX. However, I have seen some posts on internet that some of the MPI libraries do not support Cygwin, e.g. openmpi. In addition, to run a mpi program on Cygwin is a little weird to me. I never tried this.
Talking about the performance I think Cygwin is not a good platform. I prefer MinGW/MSYS or better the native LINUX virtual machine if you’d like to stick to Windows. The reason is Cgywin does not produce ‘native’ WIN32 but POSIX program. The programs compiled on Cygwin need cygwin.dll as an intermediate layer. Therefore, some performance sacrifice must be paid. (see also: http://forums.techarena.in/software-development/1196236.htm )
As my suggestion, you could run your testing on the virtual LINUX machine (e.g. oracle Virtual Box) and then make your final production run on a real LINUX server for the best performance. Or you could compile Gromacs with MinGW or VC++ to make native WIN32 binary and run it on Windows. I didn’t compare the speed of between the Windows version and the LINUX version. But I did compile Gromacs with VC++ successfully. (There is a Windows HPC VS solution of Gromacs on the official website as far as I remember.)
Further discussion is alway welcome!
Ting