search

I2PC / xmippCore

Core functions for Xmipp and sharing with other packages (like Scipion)
GNU General Public License v3.0
4 stars 2 forks source link

Add half precision floating point MRC file support #169

Closed oierlauzi closed 1 year ago

oierlauzi commented 1 year ago

Usage of half precision floating point numbers is becoming widespread in the context of CryoEM. Currently, Xmipp does not support this type of storage.

oierlauzi commented 1 year ago

In general, it is not useful to work with FP16 in CPUs, as it is supported only in rare cases. The approach would involve converting FP16 numbers to FP32 when reading and vice versa.