z5
z5 copied to clipboard
Published
20 hours ago •
constantinpape
constantinpape
Use xsimd vector API
With release 1.4.1, support for compiling with xsimd was added and this is automatically used for xtensor containers.
However, I/O and compression is not performed on xtensor containers, but on flat buffers implemented as std::vector. We could use the xsimd vector API to potentially speed this up via vector instructions.
@weilewei If you want to look more into performance, checking out xsimd might be a good idea.
@constantinpape Thanks! Sorry for my late response last two weeks as I was busy with wrapping up my summer project and traveling back home. I will look into this later when time permits. Also, my mentor @halehawk will look into this as well.
