Project 2: Saksham Nagpal

[thecapablesnakekeeper]

Repo Link

Features implemented:

• CPU: scan, compact without + with scan
• GPU: Naive scan, Work-Efficient scan, Stream compaction
• Thrust scan
• Attempted optimizing the work-efficient scan by launching only as many threads as the number of resulting elements in each iteration

Feedback:

• It would have been good to make us work on making the GPU implementations more performant that the CPU ones.
• This project could also have been a good opportunity to show us how to use various profiling tools to identify the different kinds of bottlenecks that can occur, and tell us how to go from there.