Practical usage : flags needed

[colindaven]

Hi,

EDIT - you can ignore this, I was hitting the cgroups limits on the cluster servers, so all measurements are invalid (restricted to a single core). It would still be good to know if AVX512 helps a lot.

I've been trying to speedup minimap2 for long ONT reads and this project fits the bill.

I have seen massive differences in benchmarking on our heterogenous cluster though. Maybe you can give some hints on CPU flags I need to look out for to get decent performance ? That would be very helpful.

Our machines are all very different. Will I only get really good performance on AVX512 systems ?

Thanks,
Colin

[colindaven]

So, my second round of results are more consistent.
Is this 3-5 % speedup expected, or do you expect more on full 100+ GB datasets?

Bigger direct comparison minimap vs mm2-fast

• Results
• mm2-fast is 3% quicker than minimap2 with 16 cores on this smallish 450MB dataset
• mm2-fast is 5% quicker than minimap2 with 24 cores on this smallish 450MB dataset

# working
>>> 709/731
0.9699042407660738
>>> 616/649
0.9491525423728814