Closed Incarnation-p-lee closed 10 months ago
I think vf(w)redsum uses rounding mode. The others do not.
Yes, vfw from INT to FP may never lose precisions, thus rounding is not required up to a point.
Floating-point rounding mode intrinsics are added now. Closing this issue.
Hi Eop,
Just go thru all the float point FRM intrinisc API and locate some of them that may need double confirmation with you, for the FRM is honored or not.
CC @kito-cheng and @zhongjuzhe for awareness.
Pan