HPCGPU's Blog

This page has no Portuguese version.
(Format of the dates: day.month.year)

10.03.2010

Good morning.

I've uploaded resumed results about all 3 implemented benchmarks. The results diverge considerably. While we almost didn't obtain a speedup for the CG Benchmark, the EP Benchmark showed a speedup of 20. All these results can be reproduced on different systems, since all our code is available online.

Best regards for all,

Laércio Lima Pilla

/*------------------------------------------------*/

13.01.2010

Hello.

The CUDA version of the CG benchmark is now available here. Differently from the other benchmarks edit in this project, this version can be compiled to run with CUDA but also sequentially (CPU only).

Best regards,

Laércio Lima Pilla

/*------------------------------------------------*/

09.11.2009

Hi again!

The new version of FT benchmark, with coalesced memory access, is now available here. It has been compiled and executed on Windows and Linux environments.

Laércio Lima Pilla

/*------------------------------------------------*/

09.11.2009

Hi everyone.

Now that I've settled down after the 21st International Symposium on Computer Architecture and High Performance Computing in São Paulo, Brazil, I'll present some news.

First, the Symposium was really good. We had some special keynotes this easier. You can see below some pictures of them.

keynote-speeches-2.jpg
Dan Reed, from Microsoft, presenting the talk "Multicore, HPC and Clouds: Raining Multicore Services"

keynote-speeches-3.jpg
Prof. Yale Patt, from The University of Texas at Austin, presenting his talk "The 1000-Core Microprocessor: Will We Be Ready For It?"

keynote-speeches-14.jpg
Prof. Wen-Mei Hwu, from The University of Illinois at Urbana-Champaign, presenting the talk "GPUs in High-Performance Computing -- Architectures, Software Stack, Education, and Applications"

We also had some industrial talks from Microsoft, Intel and AMD.

I've participated on the 4th Marathon of Parallel Programming, and guess what? My team won the marathon!

maratona-11.jpg
The winning team (from left to right): Claudio Schepke, me and Stefano Mór.

Next year, the symposium is going to be at Petrópolis, in the Brazilian state of Rio de Janeiro. Hope to see you there.

Laércio Lima Pilla

/*------------------------------------------------*/

22.09.2009

Hello.

The CUDA version of the EP benchmark is already available for download. It works on both Windows and Linux with CUDA 2.3 and a GPU with Compute Capability 1.2 or above.
A comparison between the CUDA version and an OpenMP version is available here.

Best regards,

Laércio Lima Pilla

/*------------------------------------------------*/
23.04.2009

Hi there!

I've been working a while with the NAS Parallel Benchmarks, and found some interesting results. For an example, the IS benchmark on NPB 2.3 gets segmentation faults when running the problems of classes B and C! I'll be probably focusing on IS and EP benchmarks, but I'm still considering CG and MG too.

You can check out how everything is going here and here.

/*------------------------------------------------*/
07.04.2009

Hello.

The evaluation of the performance of the configuration was postponed due to problems with the new CUDA driver in Ubuntu 8.10. When I find a solution for that, the evaluation will be done.

Meanwhile, I'm working on the NAS Parallel Benchmarks, which can be seen in the Section 5.

Laércio Lima Pilla

/*------------------------------------------------*/
17.03.2009

Hello everyone.

I forgot to post here that the guide for CUDA's installation on Windows Server 2008 is complete, and can be seen in here. Next posts will probably treat about parallel benchmarks and CUDA. Just wait and see.

Laércio Lima Pilla

/*------------------------------------------------*/
29.01.2009

Hy!

To all interested in the small guide for CUDA's installation on Windows Server 2008, you can see a preview here. As I progress installing Visual Studio 2008 and integrating it with CUDA, you will see the progress there.

Laércio Lima Pilla

/*------------------------------------------------*/
05.12.2008

Hello.

Today we had Mark Ihimoyan from Microsoft giving us a lecture about entitled High Productivity Computing with Window. It was a very interesting talk, where he presented numerous solutions that Microsoft have been developing.

Imagem195.jpg
Mark Ihimoyan and his presentation_

Today I also installed our GTX 280 in our new computer. After Mark's talk, I'm convinced to install the Windows Server 2008 in the machine, because the NVIDIA drivers to Windows Vista should be compatible with the Windows Server. This will probably make my life much easier.

Meanwhile, I installed CUDA in Linux, and to make others' lives easier, I just wrote The small guide for CUDA's installation in Linux. Enjoy.

Laércio Lima Pilla

/*------------------------------------------------*/
21.11.2008

Howdy!

I wrote to you all a little piece of code with CUDA! And then, I discovered how painful is to publish source code in a Wiki Page...

I hope it was worth it.

Until sooner than you expect,

Laércio Lima Pilla

/*------------------------------------------------*/
19.11.2008

Hello my friends. Stay a while and listen.

Santa Claus is coming sooner this year. In the beginning of December we will have some beautiful GeForce GTX 280 to work with!!!

Meanwhile, I've been working with Windows XP 32 and 64 bits. I was hoping to see which would have the best performance, or if they were equivalent. It happened that Windows XP 64 was slower when I was installing everything, and was a little buggy. Because of this, I will install Windows XP 32 bits in our new machine. I shall post its configuration soon.

I finished writing about the architecture of CUDA compatible GPUs. In the next days I will probably finish writing about gpus.

See you all soon!

Laércio Lima Pilla

/*------------------------------------------------*/
05.11.2008

Hello!

This is my first post in this blog.
My name is Laércio Lima Pilla and I'm an undergraduate student at the Federal University of Rio Grande do Sul (UFRGS). I'm the newest member in our Open Source and Interoperability Development Nucleous (NDOS-BR).

In a continuation of the HPCCCS Project, I'm working in High-Performance Computing with Graphic Processing Units. In the next days, I'll present some of the motivation for our research and what I've discovered so far.

Just keep watching :)

Laércio Lima Pilla

Last edited Mar 10, 2010 at 11:40 AM by Pilla, version 20

Comments

No comments yet.