require new mismatch ratio of <0.8 in order to get 1577775 - 1578027 30 (thermohpilus). Therefore. Consider removing the entire dyad enumeration and just relying on mutant checking in the genome. This means that we save a fair bit on memory with regards to the number of dyads we send to the crispr kernel. Rather we just end up working with the genome and each cuda thread writing crispr arrays to its buffer based on mutant checks. I would expect this to actually improve the performance because here there's no need to do the dyad checks. #47