This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Your change to emit_group_{load,store}
- From: Aldy Hernandez <aldyh at redhat dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: gcc at gcc dot gnu dot org, rth at redhat dot com
- Date: Tue, 26 Oct 2004 13:58:26 -0400
- Subject: Re: Your change to emit_group_{load,store}
- References: <10410261754.AA21817@vlsi1.ultra.nyu.edu>
On Tue, Oct 26, 2004 at 01:54:05PM -0400, Richard Kenner wrote:
> I don't understand the test for !SCALAR_INT_MODE_P. That include BLKmode,
> but you call emit_move_insn in the new code, which won't work for BLKmode.
> What did you mean here? Perhaps SCALAR_FLOAT_MODE_P?
>
> The test case is compiling the Ada RTS file a-calend.adb with -O2 on x864-64.
Hmmm... perhaps what we want is SCALAR_FLOAT_MODE_P or
(!SCALAR_INT_MODE_P && !BLKmode). rth had suggested I use SCALAR_INT_MODE_P;
Originally I was just checking that the class was MODE_FLOAT (which I
now see is the definition to SCALAR_FLOAT_MODE_P).
(Further hmmmm... I can't recall wether this affected complex floats
or not. If it does, it will have to be FLOAT_MODE_P. But I'll worry
about that when I'm twiddling with complex, which isn't implemented
in the back-end yet.)
rth, what do you think?
Aldy