This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: ssa bootstrap problem on x86 (cmpstrsi_1 pattern)


>>>>> "Richard" == Richard Henderson <rth@cygnus.com> writes:

    Richard> The SEQUENCE construct is to be read as a unit.
    Richard> Therefore it was set in only one unit, therefore only set
    Richard> once.

I understand.  But, that's still confusing.  For example, when given
the insn where something is SET, you can't stop when you find the SET
-- you have to keep looking to see if there's a CLOBBER.  I guess I
don't see a better alternative solution for SUBREGs, and perhaps this
is analagous, but I'm unconvinced.

And, as far as I can tell, we don't have a solution for the kind of
constructs that appear in i386.md; we can use match_scratch there, but
are we willing to forbid this kind of usage in all machine
descriptions?

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]