https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118533
Peter Bergner <bergner at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |law at gcc dot gnu.org,
| |rsandifo at gcc dot gnu.org,
| |segher at gcc dot gnu.org,
| |vmakarov at gcc dot gnu.org
--- Comment #1 from Peter Bergner <bergner at gcc dot gnu.org> ---
(In reply to Surya Kumari Jangala from comment #0)
> The array allocated_hardreg_p[] specifies if a hard register has already
> been allocated to an allocno. This array is updated whenever a register is
> assigned. This array should also be updated if a hard reg is being used (due
> to inline asm specifying a register for a local variable).
So the process is, when we first attempt to assign a non-volatile/callee-saved
register to a pseudo, then we must increase its allocation cost to account for
the extra prologue/epilogue insns that are required to save/restore that hard
register. If however, some other pseudo has already assigned that hard
register, then any further assignments of that hard register don't need to be
charged that save/restore cost, since we've already accounted for it.
I agree that any hard register usage in an inline asm should count as that
"first" usage, so updating allocated_hardreg_p[] does seem appropriate. That
said, there can be other hard register usage other than in inline asms. A
target's machine pattern my just decide (for better or worse) to explicitly
reference a hard register in its patterns. Therefore, we should look for any
hard register usage in normal rtl insns and add those to allocated_hardreg_p[]
too.
