Google Hash Code Final
- 1 minOut of about 6500 teams my team, Our Heaps Don’t Lie, reached 24th place in the 2019 world final held at the Google Office in Dublin.
If you’re not familiar with Hash Code it’s an annual programming competition held by Google where teams of up to four programmers compete in solving an optimization problem. That is a problem that cannot be solved exactly due to its size and therefore has to be solved approximately. After an online qualification round the top 40 teams are invited to compete in the five-hour long final.
The problems of Hash Code are interesting because they are modeled after real-life large-scale problems encountered at Google. This year the final problem was optimizing the compilation of a very large codebase (Google famously has a single monolithic code repository) on distributed servers. It was the contestants’ job to optimize the speed of compilation, prioritizing important binaries while not exceeding deadlines.
The problem came with five datasets with different characteristics and it was often necessary to modify the algorithm for each dataset. Therefore a good strategy usually is to split the team after an initial discussion and let each member specialize on a dataset.
Our solutions to the 2019 qualifier and final are available on Github.
Erik Gärtner
Ph.D. in Computer Vision and Machine Learning