Hi,
I have a question about 3 kind of add that we have for FP16.
device __half __hadd ( const __half a, const __half b )
Performs half addition in round-to-nearest-even mode.
device __half __hadd_rn ( const __half a, const __half b )
Performs half addition in round-to-nearest-even mode.
device __half __hadd_sat ( const __half a, const __half b )
Performs half addition in round-to-nearest-even mode, with saturation to [0.0, 1.0].
I do not understand the meaning of following phrases correctly. What exactly is happening?
Prevents floating-point contractions of mul+add into fma.
for the second type of add are we doing multiplication operation also? I was thinking that fma is related to r=(a*b)+c but here we are just doing add.
Do we have any example or some documents for more information?
Best regards,