Research Buffet
Wednesday, September 14, 2011
3147 MEB
Refreshments 3:40 p.m.
Lecture 4:00 p.m.
Erik Brunvand
Associate Professor
Title: Hardware for Ray Tracing
Wednesday, September 14, 2011
3147 MEB
Refreshments 3:40 p.m.
Lecture 4:00 p.m.
Erik Brunvand
Associate Professor
Title: Hardware for Ray Tracing
Abstract
Commercial graphics processing units (GPUs) are tremendously powerful computational engines for graphics. They have also become general enough to be useable for more than just graphics. However, their basic organization is focused on wide SIMD parallelism, which is well suited to rendering graphics based on Z-buffering rasterization. What would a GPU architecture look like if it was designed to support an alternative rendering algorithm like Ray Tracing? Ray Tracing has some huge potential benefits in terms of rendering highly realistic graphics images, but as an algorithm is not necessarily well suited to wide SIMD parallel hardware. We are exploring alternative GPU organizations that better support algorithms like Ray Tracing though using MIMD or SPMD (Single Program Multiple Data) style parallelism.
Rajeev Balasubramonian
Associate Professor
Title: Problems and Opportunities in Memory System Design
Commercial graphics processing units (GPUs) are tremendously powerful computational engines for graphics. They have also become general enough to be useable for more than just graphics. However, their basic organization is focused on wide SIMD parallelism, which is well suited to rendering graphics based on Z-buffering rasterization. What would a GPU architecture look like if it was designed to support an alternative rendering algorithm like Ray Tracing? Ray Tracing has some huge potential benefits in terms of rendering highly realistic graphics images, but as an algorithm is not necessarily well suited to wide SIMD parallel hardware. We are exploring alternative GPU organizations that better support algorithms like Ray Tracing though using MIMD or SPMD (Single Program Multiple Data) style parallelism.
Rajeev Balasubramonian
Associate Professor
Title: Problems and Opportunities in Memory System Design
Abstract
The talk will provide a brief overview of memory system design, upcoming challenges, and potential solutions. Among the topics to be covered: low-power memory chips, high bandwidth memory channels, emerging non-volatile memory cells, and processing-in-memory.
The talk will provide a brief overview of memory system design, upcoming challenges, and potential solutions. Among the topics to be covered: low-power memory chips, high bandwidth memory channels, emerging non-volatile memory cells, and processing-in-memory.