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
Currently we round the results of matrix multiplications, but the math is still done in floating point on the GPU. Per the spec the GTE did fixed point math. Would this be too much accuracy (on top of probably being slower)?
The text was updated successfully, but these errors were encountered:
Currently we round the results of matrix multiplications, but the math is still done in floating point on the GPU. Per the spec the GTE did fixed point math. Would this be too much accuracy (on top of probably being slower)?
The text was updated successfully, but these errors were encountered: