search

pulp-platform / ara

The PULP Ara is a 64-bit Vector Unit, compatible with the RISC-V Vector Extension Version 1.0, working as a coprocessor to CORE-V's CVA6 core
Other
357 stars 126 forks source link

half-precision floating point example code #332

Closed shkim-emily closed 1 month ago

shkim-emily commented 2 months ago

RVV 1.0 supports floating-point operations only for single precision and double precision. However, Ara2 hardware also supports half-precision floating-point and bfloat operations. Could you provide example code for testing a 16-bit floating-point matmul on Ara2?

mp-17 commented 1 month ago

Hey @shkim-emily,

Here we have an example of 16-bit (half) float matmul.