VDF Alliance FPGA Contest Round 1 Results
Congratulations to Eric Pearson for his winning entry with 28.6 ns/sq latency!!
Thank you to all those who participated and submitted entries!
This repository contains the results and designs submitted for round 1 of the FPGA competition. See FPGA Contest Wiki for more information about the contest.
Five teams submitted a total of 10 designs in round 1. All were based on the provided Ozturk baseline and optimized in various ways for the FPGA platform.
The round 2 baseline latency will be set at 28.6 ns/sq based on the winning entry from Eric.
For updates on Round 2 of the competition please see FPGA Contest Round 2
Results
The following designs were submitted and are available in this repository. The submission_form.txt file provides information from the contestants about their design.
Designs were evaluated for performance as follows:
- Run hardware emulation to test basic functionality
- Synthesis using the provided scripts
- Run for 2^30 iterations on AWS F1 and check for correctness and performance.
Unfortunately we had a number of designs that where not fully functional for varous reasons. Please see the round 2 wiki page linked above for further discussion on what happened and advice for round 2.
DNQ = Did not qualify
Team Name | Directory | Expected | Actual |
---|---|---|---|
Eric Pearson | eric_pearson-1 | 31.5 ns/sq | 31.5 ns/sq |
Eric Pearson | eric_pearson-2 | 28.6 ns/sq | 28.6 ns/sq |
FPGA Enthusiast | fpga_enthusiast-1 | 45.7 ns/sq | Clocking issues caused slowdown to 56 ns/sq |
FPGA Enthusiast | fpga_enthusiast-2 | 45.7 ns/sq | DNQ - did not route successfully |
FPGA Enthusiast | fpga_enthusiast-3 | 45.7 ns/sq | DNQ - did not meet timing |
Silicon Tailor | silicon_tailor-30 | 30 ns/sq | DNQ - clocking/power faults on F1 |
Silicon Tailor | silicon_tailor-291 | 29.1 ns/sq | DNQ - clocking/power faults on F1 |
Silicon Tailor | silicon_tailor-296 | 29.6 ns/sq | DNQ - clocking/power faults on F1 |
Geriatric Guys with Gates | geriatric_guys_with_gates | 44.96 ns/sq | 48 ns/sq - slower than expected due to clocking issues |
Andreas Brokalakis | andreas_brokalakis | 24 ns/sq | DNQ - simulation/synthesis errors |
Optimization ideas for round 2
- Combine concepts from the round 1 submissions
- Improve upon round 1 submissions
- Explore approaches and algorithms other than Ozturk, such as Chinese Remainder Theorem
- Explore options to do fewer reductions, such as squaring twice then reducing