You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Implemented prefixed and non-prefixed bfloat16 GER instructions
without accumulation and with positive or negative multiply and accumulate
---
There's a discrepancy between this implementation and mambo/the
hardware where implementing it with float32_mul then float32_muladd
results in incorrect signal in 0/infinite results, but implementing it
using the multiplication then muladd using FloatParts64 results in a
different result after rounding in an underflow, I've not been able to
solve this
0 commit comments