[Bug target/124266] [16 Regression] Incorrect unwind info for vector math functions

[wilco at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124266

Wilco <wilco at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[16 Regression]             |[16 Regression] Incorrect
                   |gcc.target/aarch64/torture/ |unwind info for vector math
                   |pr111677.c fails at -O3     |functions
                   |-fomit-frame-pointer        |
                   |-funroll-loops -fpeel-loops |
                   |-ftracer -finline-functions |
                   |after r16-7666              |
           Assignee|unassigned at gcc dot gnu.org      |wdijkstr at arm dot com
          Component|rtl-optimization            |target

--- Comment #6 from Wilco <wilco at gcc dot gnu.org> ---
Simple reproducer:

void f0(void);

__attribute__ ((aarch64_vector_pcs)) 
void f1 (void)
{
  f0();
}

This should emit cfi entries for q24-q31, but this is omitted on the saves and
only 4 of the 8 are emitted for the restores. Present since GCC10.