Bill Mark signup schedule
Tuesday/Wednesday, April 17 and 18, 2007
Bill Mark, UT-Austin
Insights into the future of multicore computer systems provided by today's real-time graphics systems.
Fill in your name, location next to a time slot. Create/merge time slots if necessary. Click on edit to fill in entries.
April 17, 2007
Flight Arrives 12:25pm
Lunch @ 1pm : ?
2:00 pm : Students (Multiscalar)
3:30 pm : Cookies before seminar
4:00 pm : Seminar 1210 CS
5:15 pm :
Dinner : Guri, David (most likely)
April 18, 2007
(Breakfast?)
9:30 am : Michael Gleicher
10:15 am : Karu (?)
11:00 am: Marc de Kruijf (Carl's office)
11:45 : Lunch (Mark Hill)
1:00 : Graphics students (1347 CS)
2:30 : Multifacet students
3:30 : PHARM and Jim's students
5:00 : Karu/Michael
Leave for airport: 5:30 PM
Insights into the future of multicore computer systems provided by today's real-time graphics systems.
It is clear that the primary mechanism for increasing the performance of commodity microprocessors over the next few years will be the addition of more cores. This impending transition to explicitly programmed parallelism on a single chip has raised a wide variety of questions about parallel programming models, parallel architectures, and overall system design. But graphics processors (GPUs) have already successfully made the transisition to parallelism. Modern graphics cores have many cores and over 100 floating point units. Surprisingly, they are relatively easy to program for their intended application (Z-buffer rendering). In this talk, I'll discuss how this ease of programming is delivered and how these ideas can be generalized to apply to multicore CPUs. I'll also discuss the changes that parallelism brings to the relationship between computer architects and other participants in the design of computer systems.
Bill Mark's research is primarily in the area of systems for real-time computer graphics. This area includes graphics algorithms, graphics architectures, and programming tools and interfaces for these systems. More broadly Bill is interested in programming models and hardware architectures for "general purpose" single-chip parallel computers (i.e. multicore machines). From 2001-2002 Bill worked at NVIDIA as the technical leader of the team that co-designed (with Microsoft) the Cg/HLSL language for programmable graphics hardware. Bill joined UT Austin as an assistant professor in 2003 where he has been working to design more flexible real-time graphics systems, including enhanced Z-buffer systems and real-time ray tracing systems. Bill served as papers co-chair for the 2003 Graphics Hardware Conference, and spent six months in 2006 as a visiting researcher in Intel's Applications Research Lab.