Testing ideas by in-sample optimizing is not a good idea. First, it is slower, second, the results are misleading.
You can use multiple cores only for tasks that can be parallelized, such as the cycles of a Walk Forward optimization. A single cycle with a single asset cannot be further parallelized.