Hi Seth, if I understand correctly, from_pileup_mate_aware is run for each column in a pileup. This means it is grouping, sorting, grouping, etc for each position, often with the same sets of reads for each consecutive column.
Do I misread? If not then this will be an extreme bottleneck as it's O(n^2) or worse.
We can mitigate by checking if the end of the left-most read is less than the start of the right-most read. Then at least the cost could be minimized to only the percent of reads that overlap.
Hi Seth, if I understand correctly, from_pileup_mate_aware is run for each column in a pileup. This means it is grouping, sorting, grouping, etc for each position, often with the same sets of reads for each consecutive column.
Do I misread? If not then this will be an extreme bottleneck as it's O(n^2) or worse.
We can mitigate by checking if the end of the left-most read is less than the start of the right-most read. Then at least the cost could be minimized to only the percent of reads that overlap.