I graduated fromRio Americano HighSchool in 1997 and then moved on to receive myBSinComputer Science with honorsfrom theUniversity of California, SanDiego (UCSD) in 2001. During my time at UCSD, I worked at theSan Diego Supercomputer Center in theVisualization Group. I came to theStanford UniversityComputer Science department in the fallof 2001 and began pursuing myMS. I recieved my MSin 2003 and I have since transitioned to thePhD program. I'vecompleted myComps andQuals, soI'm now in thesis mode. My time at Stanford has been spent in theStanford Graphics Lab underthe advisement ofPat Hanrahan. My research for 2006 and 2007 has been supportedby anIntel PhD Fellowship.

Research Projects

Sequoia

Sequoia is a programming language designed to facilitate thedevelopment of memory hierarchy aware parallel programs that remainportable across modern machines with different memory hierarchyconfigurations. Sequoia abstractly exposes hierarchical memory in theprogramming model and provides language mechanisms to describecommunication vertically through the machine and to localizecomputation to particular memory locations within the machine. Wehave implemented a complete programming system, including a compilerand runtime systems for Cell processor-based blade systems anddistributed memory clusters, and have demonstrated efficientperformance running Sequoia programs on both of these platforms.

Folding@Home - GPGPU

I've been lucky enough to help the Folding@Home folks get the firstGPU based client out. This has been a long time coming and I'm gladthey had the patience to keep at it and keep working with Brook andGPGPU programming. I especially want to thank Vishal for the hardwork adapting algorithms to get the code running well. I also want tothank ATI for all the support getting things up and running. Thecurrent estimate is that a R580 is 20-40X a single core 2.8GHzPentium4. At these rates, we can now do in close to a month what usedto take years of simulation. See current statshere.

BrookGPU

BrookGPU is a compiler and runtime implementation of the Brook streamprogramming language which provides an easy, C-like programmingenvironment for today's GPU. As the programmability and performanceof modern GPUs continues to increase, many researchers are looking tographics hardware to solve problems previously performed on generalpurpose CPUs. In many cases, performing general purpose computation ongraphics hardware can provide a significant advantage overimplementations on traditional CPUs.

GPUBench

GPUBench is a benchmark suite designed to analyze the performance ofprogrammable graphics processors in areas of particular importance togeneral purpose computation. More so than being a measuring stick forGPU performance, GPUBench is intended to be a tool for developers ofGPU-accelerated applications, since a more complete understanding ofmachine capabilities will dictate better software designdecisions. Information included in GPUBench reports is useful inaccurately estimating the performance of shader programs as well asdetermining if a particular GPU is a viable target platform for aparticular computation.

Chromium

Chromium is a system for interactive rendering on clusters ofworkstations. It is a completely extensible architecture, so thatparallel rendering algorithms can be implemented on clusters withease. Various parallel rendering techniques such as sort-first andsort-last may be implemented with Chromium. Furthermore, Chromiumallows filtering and manipulation of OpenGL command streams fornon-invasive rendering algorithms. Chromium runs on Microsoft Windowsand several types of Unix such as Linux and IRIX.

Graphics Clusters

I've spent a great deal of time researching, developing, and buildinggraphics clusters. This link provides up to date information on ourlatest graphics cluster,SPIRE. There is also informationavailable about our previous generation cluster,Chromium. I've also givenmanytalkson building graphics clusters.

Raptor

Raptor is a parallel volume renderer written for use with Chromium.Raptor provides an example of a sort-last parallel applicationutilizing Chromium's distributed event model, CRUT, and compositingsystems, such as the binaryswap SPU.

NPACI Vistools

While at SDSC, I worked on the NPACI VisTools. NPACI's VisTools are acollection of software libraries and applications that allow users tocreate high resolution visualizations of large complex scientificdata. Current applications let users interactively navigate largevolumes on their desktop and do full color perspective volumerendering. The toolkits are designed to work with data sets that aretoo large to fit in memory using out-of-core techniques.