Current Research
I'm on leave at present, but currently exploring research avenues applying modern machine learning techniques to applications across computer architecture, from holistic caching policies (particularly in large multicore systems) to more scaleable techniques for microarchitectural performance evaluation.
Past Projects
Characterization of Irregular GPU Kernels
My master's thesis project investigated the performance behavior of irregular graph-based GPGPU benchmarks via cycle-level simulation and analyzed the impact of microarchitectural design tradeoffs.
[ Publications: Thesis'15, IISWC'14 ]
Parallel Traveling Salesman (TSP)
This project parallelized a random-restart iterative hill-climbing TSP solver for CUDA GPUs.
[ Publications: GPGPU'15, PDPTA'11 ]
GPU Lossless Data Compression
This project implemented a lossless compressor and decompressor for double-precision floating point data in CUDA, capable of compressing at 75 Gb/s.
[ Publication: GPGPU'11 ]
Software
GFC
GFC is a GPU-based compressor/decompressor written in CUDA C for binary IEEE 754 double-precision floating-point data.
Code: GFC
TSP_GPU
TSPGPU is a CUDA heuristic solver for the symmetric Traveling Salesman Problem based on iterative hill climbing with 2-opt local search.
Code: TSPGPU