GPUs and Random Number Generation

>> Sunday, November 20, 2011

I had a great time at SC11 and I learned a lot about the current state of high-performance computing. I enjoyed meeting people who work in the industry, which included academics, developers, salesmen, and a few readers of this blog (Hello John and Jamie).

One mysterious man came to my booth and asked what I knew about generating random numbers with OpenCL. I told him about implementations of the Mersenne Twister, but he wasn't impressed. He told me about a new technical paper that explains how to generate random numbers on GPUs by combining integer counters and block ciphers. In reverential tones, he said that counter-based random number generators (CBRNGs) produce numbers with greater statistical randomness than the MT and with much greater speed.

I promised to look into the matter, and I have. The paper can be downloaded here, and though I did my best, I don't have the cryptography background needed to fully grasp the theory. However, the paper was a finalist in SC11's Best Paper competition, so it's been subjected to scrutiny by folks a lot smarter than me.

In addition to writing the paper, Dr. Salmon has made his code freely available, including an implementation in OpenCL. The full code release is here and the documentation is here. I don't need to generate random numbers at the moment, but if ever I do, this is the first place I'll go.

Read more...

OpenCL 1.2 Specification Released

>> Tuesday, November 15, 2011

The Khronos Group has released the OpenCL 1.2 specification (main site here). As stated on the site, the new functionalities include "seamless sharing of media and surfaces with DirectX® 9 and 11, enhanced image support, custom devices and kernels, device partitioning and separate compilation and linking of objects."

Upon hearing the news, I felt like something of an imbecile. After all, I'm here at the Supercomputing conference, purportedly an OpenCL expert, and I had no idea that a new spec was available. Ah well.

Read more...

Supercomputing 2011

>> Friday, November 11, 2011

The time has come. I'm just about to start driving up to Seattle for the Supercomputing 2011 conference. In addition to the laptop, monitor, and cables, I have posters, poster stands, article printouts, business cards, markers, duct tape, and at least a million copies of OpenCL in Action. I'll be selling the books at a discount in Booth 4613 on Level 6 of the Washington State Convention Center.

I'll also be demonstrating an application that combines OpenCL and OpenGL to portray rigid bodies accelerating, decelerating, and colliding in space. If you have any questions about OpenCL-OpenGL interoperability, feel free to drop by.

As if that's not enough, I'll be giving a presentation on simulating physics with OpenCL and OpenGL. Specifically, my presentation will be 2:30-3:00 on Thursday in WSCC 613/614.

Read more...

Nuisances

>> Thursday, November 3, 2011

To promote the book and demonstrate dynamic simulation with OpenCL and OpenGL, I've reserved a tiny booth at the Supercomputing 2011 conference in Seattle. Since making the reservation, I've been deluged with e-mails from contractors offering services such as advertising, plumbing, shipping, and lead retrieval. I've ignored them thus far, assuming that the high cost of reserving the booth would cover the services I need.

But as a saleswoman explained to me, this assumption was mistaken. You see, SC11 does not provide any electricity to the booths. So unless you're demonstrating an innovation in abacus design, you need to pay a contractor to rig an outlet for you. As you might imagine, this is not cheap. Now that I've paid for electricity, I'm concerned about the necessities I haven't paid for. Will I have light? Oxygen? Gravity?

In other news, the military announced the results of its solicitation calling for finite-difference time-domain (FDTD) simulation software. They want code capable of running on CPUs and GPUs, and because I'd worked with FDTD in grad school, I submitted a proposal explaining how OpenCL could do the job. I included plenty of kernel code for FDTD, but I didn't win and I wasn't expecting to.

However, I was very interested to find out which small business won the contract. After all, any company working on OpenCL-based electromagnetic simulation is a company I want to know about. But the military didn't give the award to anybody. Bummer.

Read more...

  © Blogger template Werd by Ourblogtemplates.com 2009

Back to TOP