Batching and prefetching #45
Description
The paper writes that the C++ version of AWRY processes batches of 4 queries at a time, and prefetches the memory for the next rank query as soon as possible, and indeed the code seems to do this:
batch processing: https://github.com/TravisWheelerLab/AvxWindowFmIndex/blob/92b849fddaea62a3b6bf47c7ff6db3ac3d896a10/src/AwFmParallelSearch.c#L273
prefetching: https://github.com/TravisWheelerLab/AvxWindowFmIndex/blob/92b849fddaea62a3b6bf47c7ff6db3ac3d896a10/src/AwFmSearch.c#L73
Looking at the Rust implementation, I don't find hits for either batch or prefetch.
Am I correct that these aren't implemented? (I guess that could be the reason for the 50% performance difference with the C++ implementation?)