Hello Adarsh, At the time I was doing this integration, I used Pipes because it's C++ based, which is easier for porting CUDA C code I assume.
It actually is quite straightforward I think. What I have done can be summarized into: 1) a simple abstract class that implements the common protocol for Mapper/Reducer such that each CUDA mapper/reducer does not have to deal with the Pipes communication. They just launch CUDA kernel and collect results. 2) a set of simple shell scripts for uploading the mapper/reducer/CUDA kernel to the Hadoop DFS and kickstarts the runs. 3) modify the CUDA program into a mapper inheriting from the wrapper class so that it takes input from Pipes and dump results also to Pipes. If you want, I can try to find the code for you but I think you get the idea. I didn't do a whole lot because it's mainly a class project. We just want to demonstrate its feasibility and power saving. Thank you for your interest in our work! Let me know if you need further assistance. Peng > Thanks Chen for your guidance, > > To translate CUDA in Hadoop environment, following are the options : > > -- Hadoop Streaming: Standard I/O ( Recommended by you ) > -- Hadoop Pipes: C++ library, Socket connection, ( I'm looking for ) > -- JNI, JNI-based CUDA wrapper (JCUDA) ( Done and able to run a Jcuda > Program in Hadoop Cluster ) > > We use Hadoop Pipes for our proposed technique > . MapReduce applications/CUDA kernel ? written in C++ > > So i.e why I am trying a C Cuda code to run through Hadoop Pipes, is it > difficult or not possible. > I am looking for a demo Cuda program that is up and running in Hadoop > Cluster that clarifies my basic concepts so that I program accordingly > in future. > > > Looking forward for some more guidance. > > Thanks once again for your wishes . > > > With best Regards, > > Adarsh Sharma > > > > He Chen wrote: >> Hi, Adarsh Sharma >> >> For C code >> >> My friend employ hadoop streaming to run CUDA C code. You can send email >> to >> him. [email protected]. >> >> Best wishes! >> >> Chen >> >> >> On Thu, Mar 3, 2011 at 11:18 PM, Adarsh Sharma >> <[email protected]>wrote: >> >> >>> Dear all, >>> >>> I followed a fantastic tutorial and able to run the Wordcont C++ >>> Program in >>> Hadoop Cluster. >>> >>> >>> http://cs.smith.edu/dftwiki/index.php/Hadoop_Tutorial_2.2_--_Running_C%2B%2B_Programs_on_Hadoop >>> >>> But know I want to run a Cuda Program in the Hadoop Cluster but results >>> in >>> errors. >>> Is anyone has done it before and guide me how to do this. >>> >>> I attached the both files. Please find the attachment. >>> >>> >>> Thanks & best Regards, >>> >>> Adarsh Sharma >>> >>> >> >> > >
