https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91050
Segher Boessenkool <segher at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2019-07-03
Ever confirmed|0 |1
--- Comment #4 from Segher Boessenkool <segher at gcc dot gnu.org> ---
(In reply to Alan Modra from comment #3)
> We don't always emit .machine because that is a waste of time if the
> .machine cpu matches the original cpu passed on the command line to gas.
> .machine reloads the gas opcode table, which may take a significant time.
That is a legitimate concern, of course, but how bad is that? More and
more code uses .machine freely, in inline asm directly, or with function
attributes for example.
It would be a lot simpler for GCC to always use .machine.
Maybe gas should delay initialising these tables until it sees the first
actual insn?
Some of my configurations *always* use .machine, and I never noticed
any slowdown.
I'll benchmark something.
