[Bug target/67603] [aarch64] constant folded of floating-point expression at compile-time value does not match run-time value

[david.abdurachmanov at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67603

--- Comment #1 from David Abdurachmanov  
---
Created attachment 36345
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36345=edit
Annotated assembly

Shows that one value is computed at run-time, while another one is loaded from
literal pool.

[Bug target/67603] [aarch64] constant folded of floating-point expression at compile-time value does not match run-time value

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67603

Andrew Pinski  changed:

   What|Removed |Added

 Status|UNCONFIRMED |RESOLVED
 Resolution|--- |INVALID

--- Comment #2 from Andrew Pinski  ---
This is invalid.  You need -ffp-contract=off to get the same behavior for the
constant folding as for the instructions emitted.

Basically the same behavior would happen on PowerPC and on x86_64 (with -mfma).