This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patches] Re: x86_64 merger part 24 - string operations
> On Sun, Mar 25, 2001 at 01:48:24AM +0100, Jan Hubicka wrote:
> > [movstrsi]
> > ! "TARGET_64BIT"
>
> Should be !TARGET_64BIT, surely.
Actually the test should be removed, since my expander handles
SImode movstr in optimized way (SImode counter is cheaper)
> > (define_expand "clrstrsi"
> > [(use (match_operand:BLK 0 "memory_operand" ""))
> > (use (match_operand:SI 1 "nonmemory_operand" ""))
> > ! (use (match_operand 2 "const_int_operand" ""))]
> > ""
>
> !TARGET_64BIT?
Similary here - I handle both 64bit and 32bit modes.
>
> > ! if (TARGET_64BIT)
> > ! emit_insn (gen_adddi3 (out, out, const1_rtx));
> > ! else
> > ! emit_insn (gen_addsi3 (out, out, const1_rtx));
>
> Isn't this why you introduced adjust_counter?
Not exactly - I use adjust_counter just to decrease the count register,
here I do increase the pointer, so using the function looked quite confusing
(the call would also need to say -1)
>
> Otherwise ok.
>
> There seems to be a real need for patterns to be able to use Pmode.
> If you have the time to hack genemit and genrecog, that would be
> appreciated.
I was thinking about the same. I've run into problems with generecog atempting
to generate switch/case sequence on these, but I will try to do so once again.
Probably not before next week tought :(
Honza
>
>
> r~