more tidying of x86 feature tests
Tue Mar 6 16:30:00 GMT 2007
> As suggested somewhere in the thread, perform the mask in
> override options, and let the feature test reduce to a mere
> comparison vs zero.
does it really make CC1 smaller/faster? I would expect that in many
places we test things like TARGET_something || TARGET_something_else
that can be simplified into single & operation in the former scheme.
But probably the amount of unoptimized & operations dominate, I am just
THe individual flags could probably use some more comments, I will try
to put together something.
More information about the Gcc-patches