High Performance Computing

5 problem sets of parallel programming on CPU and GPU. University projects for High Performance Computing Systems (Fall 2016).

Description

Programmed with CUDA and C on GPU and CPU. Explored both homogeneous and heterogeneous parallel architects and introduced to performance measurement techniques, profiling, experimental evaluation of software interaction with underlying hardware and optimization.

Part 1:

• Operator Sobel
• PSNR
• Code optimization: Loop interchange, loop unrolling, loop fusion, function inling, loop invariant code motion, common subexpression elimination, strength reduction
• Performance analysis and profiling

Part 2:

• OpenMP
• Multi-threaded CPU programming
• K-means clustering
• Performance analysis and profiling

Part 3:

• Convolution filter
• Tiling
• Divergences
• PTXAS

Part 4:

• Convolution
• Histogram
• Mapping

Part 5:

• Contrast enhancement
• Histogram equalization

Prerequisites

• CUDA
• C
• Make (sudo apt install make)
• GCC (sudo apt install gcc)

Authors

• Panagiotis Nikitakis

Course website

ECE415 High Performance Computing Systems