Skip to content

gabriel: Fix the puzzle benchmark

Peter Housel requested to merge phousel/cl-bench:puzzle-fix into master

According to the comments from Henry Baker and the papers referenced from his page, the puzzle benchmark is supposed to arrive at a solution in 2005 trials (as measured by kount). Due to off-by-one and logic errors converting DO to DOTIMES, among other problems, the code arrived at an incorrect solution in many fewer trials than are actually necessary.

Merge request reports