https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109359
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2023-03-31
Target Milestone|--- |12.3
Summary|Compile-time rounding of |[12/13 Regression]
|double literal to float is |Compile-time rounding of
|incorrect with |double literal to float is
|-frounding-math |incorrect with
| |-frounding-math
Keywords| |needs-bisection, wrong-code
Status|UNCONFIRMED |NEW
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
With -frounding-math:
xs:
.long -639366012
.long 1063214053
.long 536561674
.long 1071918432
without:
xs:
.long 989519663
.long 1059154689
it looks like we fail to convert the double constant to single precision
and then end up outputting the double precision constants ...
The C frontend works.
