Choosing a compiler - GCC-4.x.x, the GNU Compiler Collection

In order to compile or develop an HPC application, a solid compiler package is needed. From my experience, the GCC-4.x.x series of compilers offer the best balance of cost (FREE!!!), performance (competitive with commercial compilers from INTEL, IBM, PGI, PathScale, NAG, etc.), developer access (communication with developer community, rapid bug fixes, etc.), features, and standard compliance. In addition, the GCC-4.x.x compiler collection is the default compiler package on Linux, Unix, and Mac OS X operating systems. Commercial compilers from vendors such as those listed above offer their own unique advantages in certain areas; however, for my own personal (and professional) use the GCC-4.x.x is my compiler package of choice.

If the default compiler is sufficient (stability, features, performance, and relatively bug-free for your uses), then by all means use the default compiler on your system. However, if the system GCC installation is not sufficient for your needs (i.e. not a recent enough version , etc.), then your will need to either install a set of pre-built binaries or download the GCC source code of a more recent version and compile/bootstrap it yourself. In the near future, I plan to offer downloadable pre-built GCC binaries for a few select platforms (Mac OS X, RedHat Linux, Ubuntu Linux). In addition, future posts will descibe the process of downloading and compiling GCC-4.x.x along with its pre-requisite packages (GMP and MPFR). Once this is done, I will be discussing MPI package options and do a walk through of compiling and setting up OpenMPI. Until next time, may your clusters always crunch with minimal downtime! 

Intro to ClusterNut Notes

Hello and welcome to ClusterNut Notes! I am the ClusterNut and I will be writing on topics related to High Performance Computing (HPC), parallel computing clusters, MPI programming hints, and just about anything else related to scientific computing that crosses my mind. I am looking forward to sharing my knowledge and experience in the HPC field with you. Though I have quite a bit of experience in scientific computing and running clusters, I am by no means a guru and look forward to hearing your insight and suggestions as well.