Skip to content

fma: Correct Underflow logic

Eclipse Webmaster requested to merge github/fork/gregdavill/fma-underflow into develop

Created by: gregdavill

This change produces the correct behavior when executing MUL,FMADD type instruction. It flags underflow if result with unbounded exponent would lie between +/-b^(emin). Even if final rounded result is exactly [+/-]01.000000. We use the Use the sticky bits to validate this is the case for both MUL and FMADD.

Issue: #94 (closed)

Merge request reports

Loading