Melvin's Grid and HPC Blog

Yesterday, I was playing around UPC implementation from Berkeley. UPC (Unified Parallel C) is a parallel programming language based on the concept of partitioned shared memory. It is an extension of C for large scale parallel machines, including those with a common global address space (SMP and NUMA) and those with distributed memory (eg. clusters). There are a few opensource UPC implementations available out there:

* GASNet UPC compiler from UC Berkeley
* v1.1 MuPC compiler from Michigan Tech
* Intrepid GCC UPC compiler


Compiling and installing Berkeley UPC on a new Solaris 10 machine is pretty straightforward, except I have to comment out the trunc() function in the emulate.cxx file, since it is already defined in Solaris 10. As with all programming languages, I start with the Hello World example.[Read More]

Recently, I sense an increase in interest in using GPGPUs for computing power in HPC cluster. I guess many people already knows about this for a while, but over here in my part of the world, there is still little awareness in this type of technology. I recently got to know of CUDA, Nvidia's toolkit for programming your apps to use their GPUs, and you can do it in C. There is also a presentation by NVidia in last year Sun HPC Consortium in Reno. I certainly like to hear more from everyone if they any recommendation on GPGPUs and what are they using it for.