Following are my list of contributions during the GSoC period.
In this post, I shall compare and contrast the different benchmarking tools available for Benchmarking GPU Devices running CUDA and OpenCL. The main aim is to be able to make an informed decision about the benchmarks to be selected for testing HPXCL.
The native CUDA events can be used to measure the performance of tasks in the stream.
The first task is to use HPXCL Cuda and execute the Mandelbrot set example. Mandelbrot set is one of the classic examples which can help utilize the data parallelism in calculating the points. The computation process is extrememly data parallel and my task was to execute this using multiple devices.
So, I had my second VC (Video-Conference) meeting with my mentor diehlpk. The major point discussed in this week’s meeting was on which should be the more focusing point over the course of the following two weeks till the official coding window begins.
So, as such I had previously compiled and built hpx, hpxcl for Windows with Visual Studio 2015 during the proposal phase of GSoC.
Today, in my first attempts to ssh into the Rostam Cluster hosted at LSU, was hit with a roadblock but in the process learnt quite a bit about the encryption system and how the public/private keys are used for authentication, though was not relevant to the solution of the problem.
This shall be the beginning of my blogging journey. I am very excited to be a part of the Ste||ar Group through the Google Summer of Code Program 2017 . For the foreseeable future, this blog shall only include development content pertaining to GSoC.