This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH to hoist loads/stores out of loops
>>>>> "Richard" == Richard Henderson <rth@dot.cygnus.com> writes:
Richard> It seems like replace_loop_mem ought to be able to come
Richard> up with a number for our one register, rather than having
Richard> to do a general recalculation, but we need context info
Richard> from the containing rtx.
Richard> Perhaps for_each_rtx ought to also pass in the containing
Richard> rtx?
I thought about this, and chickened out in eagerness to get something
workable submitted so that people like yourself could look at it. You
also pointed out the ugly REG_USERVAR_P bit, which is (IMO) the
ugliest part of my patch.
I will think all this over, and come up with something a bit better in
the next day or two. Just to make sure I'm not mischaracterizing your
comments, you haven't suggested that the version I suggested was
*wrong* (in the sense that it generates incorrect code), only that it
is *sub-optimal* (in the sense that the same basic infrastructure
could be used to generate even better code), right?
--
Mark Mitchell mark@markmitchell.com
Mark Mitchell Consulting http://www.markmitchell.com