- This event has passed.
Graduate Defense: Kalyan Bhetwal
March 10 @ 1:00 pm - 2:00 pm MST
Event Navigation
Thesis Information
Title: High-Performance Domain-Specific Library For Hydrologic Data Processing
Program: Master of Science in Computer Science
Advisor: Dr. Catherine Olschanowsky, Computer Science
Committee Members: Dr. James Buffenbarger, Computer Science, and Dr. Alejandro Flores, Geosciences
Abstract
Hydrologists must process many gigabytes of data for hydrologic simulations, which takes significant time and resources even using powerful supercomputers. In my thesis, I demonstrate that using static compilation techniques significantly reduces the processing time and resources for these computations. I developed a Domain Specific Library (DSL) for hydrologists that minimizes the data processing time and memory needed for continental-scale hydrologic simulation. The DSL is a subset of Python and compiles to Sparse Polyhedral Framework – Intermediate Representation (SPF-IR), which allows opportunities for optimizations like read reduction fusion which are not available in Python. We fused the file I/O to perform computation on small chunks of data (stream computation) in order to reduce the memory footprint. The C code we generated from SPF-IR shows an average speed-up of 2.58 times over the existing implementations and can totally eliminate the temporary storage required. DSL users can still enjoy the ease of use of Python but get the performance gains of C code.