This benchmark is a monte carlo simulation and game tree search Artificial Intelligence player for Go. For more details, see the SPEC CPU2017 documentation.
From the full report, we found that found that, other than memory accesses, the behavior of the benchmark varies little between workloads. The report was created with Kit 93.
We have created scripts that can randomly generate workloads from example Go positions. They can be downloaded here.
We created nine new workloads taken from the defunct No-Name Go Server's archive. Each of the workloads contains six go positions. Moves are culled from the end of games to ensure that the AI plays the game to its completion.