Fixed CMakeLists.txt

This commit is contained in:
gothictomato
2022-07-17 19:23:53 -04:00
parent e5df861d66
commit c7a2f3f95a
2 changed files with 3 additions and 1 deletions

View File

@@ -9,3 +9,5 @@ Plenty of optimizations left to be done, as well as algorithmic changes, particu
For now the algorithm only operates on a single local workgroup, but it'll be expanded to maintain and run on several independent counters. After that, I plan on integrating work re-balancing: periodically, the work groups will sync with eachother, and in case of a work group having proven its range unsat, the next largest work group will be split between the two.
The CPU-bound version of this algorithm is currently about 2x as fast, when I set the local workgroup size to 256. However, the CPU-bound version also takes advantage of early exit scenarios that this algorithm can, but doesn't take advantage of (yet).
I'm currently not sure what the best approach is, optimization-wise, out of the ideas I currently have,