Marcel

> I recomputed the table above with the latest master on my old laptop. > > ``` > -------------------------------------------------------------------- > Benchmark Time CPU Iterations > -------------------------------------------------------------------- > dnax_revcomp 1022730 ns...

I wouldn't add this as a `#ifdef`. Why do we have the alphabet concept? My patch is just good for a certain use case and without carefully crafting your code...

Re-reading this thread, kloetzl proposed a "better" complement implementation for dna4 / rna4. He stated that > Adding the following function to dna4 makes complementing it about 16% faster. https://github.com/seqan/seqan3/issues/1970#issue-651799129...

I think for dna5 some potential arithmetic versions could be: ```c++ uint8_t dna5_complement_subtract(uint8_t rank) { return min(3 - rank, 4); // using underflow } uint8_t dna5_complement_xor_is_dna5(uint8_t rank) { bool is_n...

@rrahn I added https://github.com/seqan/seqan/issues/355 as a design consideration. **cc:** @smehringer, @h-2

Hi @tloka, this is currently not possible. Thank you for this use case. When re-discussing the design, I had the feeling that it would make sense to explicitly allow writing...

(Agree! We still need some way to modify a header that wasn't constructed before.)

Thank you for this idea. This would not be the "eierlegende Wollmilchsau" (all-in-one solution), but might be a great idea to make some of the problems with simd work in...

You don't need the second version at all. (at least possible since cmake 3.4, might also work with prior versions, but we require that as minimum for different reasons)

I once proposed in a strategy meeting to remove those variables, because this way of using CMake is out-dated.