rmtheis / mummergpu Goto Github PK

Observing long reads of 9kb is 1.2x slower than an equivalent total basepair version of short reads on GPU dataset while 18kb long read dataset is 1.83x slower than its equivalent short read dataset. We figured out that V100 has much higher l1 and l2 caches than 8800. Is there any parameter that we could change to optimize mummergpu for longer reads on the powerful v100?

I kept getting the segmentations fault error when running with the sample command. Any idea why? Btw I used cuda 10.1 for the project