Marcel Martin

I’m not sure where, but @ksahlin has some measurements that show that the multi-context seeds branch is slower than `main`. I did some profiling to find out why that is...

I tested all commits in this branch to see which ones contribute to the slowdown (using the sim3 datasetand drosophila-200 reads, single-end, mapping only). This table shows the commits at...

Here are also runtimes for disabling multi-context seeds (by just commenting out the appropriate `else` in `find_nams` or `find_nams_rescue`: * 13.1s when disabling mcs only in `find_nams` * 13.0s when...

Oh no, sorry to hear! Yeah, the assertions are ignored when compiling in Release mode. And I think the swap is very cheap; that should be something like two machine...

One of the remaining to-do items here is to enable multi-context seeds only when the read length is below a certain threshold. I just pushed commit 47f92575467569ecf75e13866f81f291b947a85a, which makes it...

I’m finding it difficult to work on some aspects of strobealign at the moment because many changes would actually be in conflict with the multi-context seeds PR. Also, as we...

> Would the difference between green and pink mostly be from new seeding parameters? I was wondering that as well, so I ran an evaluation for this PR but without...

Everything in this PR has now been split out into other, smaller PRs. Since we’ve been discussing speed of the mcs implementation here and that is still ongoing, I’ll leave...

I have also tried to vary the `MAX_LINEAR_SEARCH` parameter (currently set to 4). Interestingly, it makes no measurable difference in total runtime whether I set it to 0 or 10000...

Good point trying on CHM13, will do so.