Gene Wu Email & Phone Number

Austin, Texas, United States

• Aurora Supercomputer (2016-2018) o Developed and analyzed a novel fabric partitioning scheme for the Dragonfly topology, which eliminated communication interference between applications and limited bandwidth loss using a pseudo-direct link & routing scheme o Characterized the Megafly topology, a new highly scalable topology, using at-scale simulations (over 100,000 nodes) with multiple traffic patterns to uncover potential performance bottlenecks • Accelerator to Accelerator Fabric (2019) o Explored trade-offs for a GPU-to-GPU accelerator fabric in simulation o Compared multiple topologies using machine learning communication patterns

Austin, Texas, United States

• Developed a GPGPU power and performance scalability modeling methodology using k-means clustering and neural networks• Used k-means clustering to form groups of kernels with similar scaling behaviors• Constructed classifiers using neural networks to predict kernel scaling behaviors from performance counter values

Hillsboro, Oregon, United States

• Worked on a methodology to integrate light-weight timing core and memory models into a virtual platform for early stage software debug• Helped develop light-weight memory models using data from detailed DRAM simulation

Hillsboro, Oregon, United States

• Developed a new RTL-level power and confidence modeling approach • Created an infrastructure to automatically generate power and confidence models for SoCs • Evaluated modeling approach on a full graphics processor design

Austin, Texas, United States

• Modified an existing architectural simulator to support the ARM ISA• Tested and debugged the simulator against internal processor test suites

Doctor Of Philosophy - Phd, Computer Engineering, 3.98/4.00

Master'S Degree, Computer Engineering, 3.73/4.00

Bachelor'S Degree, Computer Engineering, 3.89/4.00