On Thu, 27 Feb 2020, Aleksandar Markovic wrote:
I totally disagree with your using the term "hardfloat feature enabled" in
this context, speaking about this particulat patch. This may be just
wishful thinking. The right wording would be "hardfloat feature hacked", or
"hardfloat feature fooled".
The patch itself has the wrong, intentionally misleading and confusing
title from the outset. It should be something like "target/ppc: Cheat
hardfloat feature into beleiving that inexact flag is always set"
May I point out that the patch is RFC, meaning it's not meant to be merged
only to test it and provide feedback. Also the limitations were stated in
the commit message. There's no other easy way that I know to test if
hardfloat would work with PPC than forcing inexact bit to have it run with
hardfloat most of the time. Once it's tested what regression this would
cause (other than the expected inexact bit) then we can see if there are
any other problem with hardfloat and PPC or only this bit. Then we can
either change it to only not clear inexact bit like it's done on other
archs or do something else as even not clearing sticky inexact bit would
break the non-sticky counterpart PPC has. Breakage would be limited to the
non-sticky version and discussion was about if even that's unacceptable.
Regards,
BALATON Zoltan
