Tags » Parallel Computing

Simplifying efficiency - The actor model

When you work for the same company for some time, you might end up reviewing code you wrote years ago. The feeling is always shocking. Imprudent, brave, naive, it’s like watching the first season of “The Simpsons” again. 1,687 more words

Actor Model


雖然建基於 Map-Reduce (https://hadoop.apache.org/) 的分散式平行運算技術已經非常成熟, 而且受到廣泛應用, 但隨科技發展, 由流動裝置所產生網絡數據會繼續以幾何級數增長, 所以就慢慢發展出兩個層級的平行運算技術 – 以Map-Reduce作為主幹, 再配合 CUDA GPU平行運算技術 (http://www.nvidia.com/object/cuda_home_new.html) 來再進一步提升運算速度. CUDA 技術的好處是硬件(nVidia Display Card)相對於一部 hadoop cluster server 來得平, 但其運算能力卻遠超 cluster server上的x86/x64 CPU, 而且對於一些需要超低延遲(ultra low latency)同時又要大量運算的工, 例如 high frequency algo trading, CUDA 可以免卻 hadoop cluster node 之間的通訊延遲. 250 more words


NVIDIA to establish deep learning lab with China partners

NVIDIA is teaming up with China high performance computing firm Sugon and the Institute of Computing Technology, of the Chinese Academy of Sciences to jointly operate a deep learning laboratory. 40 more words


Setup TORQUE PBS and MPI on a Ubuntu Cluster

The goal in this blog is to document my experience in setting up mpi with torque such that one can submit multiple jobs as well as MPI jobs which is running on more than one worker node (i.e. 1,193 more words

Optimizing Django Selenium Test Speed - Test Faster!

Functional testing works wonderfully with Selenium in Django >1.8. Here are a number of simple improvements that can can lead to speedups of 4x or more. 704 more words


UPGRADE TORQUE PBS from 2.5.1 to 5.1.1 on a Ubuntu cluster

We have an Ubuntu cluster in our group. But for some reason, I cannot run a parallel job with multiple nodes with the torque PBS. Plus there seems to be other bugs on the cluster. 925 more words

More OpenCL/GL Demos

I came up with several other demos for OpenCL/GL. Here they are :D

In the future I plan to add some physics simulation demos…