Hacker News new | ask | show | jobs
by mmoskal 2082 days ago
It has a 32-bit FPU, with single cycle add and single cycle multiply; it also supports 16 bit float storage format. It lets you run small models, but note that you can't fit big ones in flash anyways.

There are also some DSP instructions, but it's only 2x16 bit mulitply and 4x8 bit saturating addition, so not so much use.