search

cucapra / pollen

generating hardware accelerators for pangenomic graph queries
MIT License
24 stars 1 forks source link

use separate turnt outputs to reduce clobbering. closes #34 #43

Closed anshumanmohan closed 1 year ago

anshumanmohan commented 1 year ago

A simple PR that helps tidy up our testing a fair bit! As #34 advises, I use turnt's output.extension feature to make bespoke .expect files for each algorithm. A much less stressful testing experience!

sampsyo commented 1 year ago

Excellent! With this, we're getting closer to being able to just do turnt -j test/*.gfa and letting everything run in parallel. There is still the dependency that requires us to do a --save run first on the oracle before doing a "check" run on the saved results, but resolving that will take some careful thinking on the Turnt side. Maybe something very roughly in the ballpark of https://github.com/cucapra/turnt/issues/26.

anshumanmohan commented 1 year ago

Ah I didn't fully realize that this was the goal! I have written a little more in https://github.com/cucapra/turnt/issues/26#issuecomment-1505738139 but, unless I'm mistaken, my comment there suggests that turnt -j test/*.gfa is not going to be possible in our case, right?

sampsyo commented 1 year ago

Not for now, at least! I think we should be able to get there eventually, but the problem right now is that we need the "oracles" to go first and the "tests" to go second. Maybe Turnt can help us enforce that itself someday.