Ian Lance Taylor - Re: PATCH COMMITTED: More lower-subreg.c patches (original) (raw)

This is the mail archive of the gcc-patches@gcc.gnu.orgmailing list for the GCC project.

• From: Ian Lance Taylor
• To: DJ Delorie
• Cc: gcc-patches at gcc dot gnu dot org
• Date: 06 Feb 2007 18:17:02 -0800
• Subject: Re: PATCH COMMITTED: More lower-subreg.c patches
• References: <m3bqkazus8.fsf@localhost.localdomain> <200702070210.l172A04O032102@greed.delorie.com>

DJ Delorie dj@redhat.com writes:

This module is now trying to split my stack pointer in half. Is there a way to disable it for a given target, at least for now? The m32c has different sized registers, and some are subregable and others aren't, so you can copy the stack pointer to r1r3, but you can't copy half of sp to just r1 or just r3. I see it uses MODES_TIEABLE_P to see if a mode can be split, but IIRC having MODES_TIEABLE_P (no hw floating point on m32c, just integer registers) return anything but 1 causes other problems (it currently has some commented out code which return 0, but the only remaining code returns 1).

Perhaps lower-subreg.c could use HARD_REGNO_MODE_OK when one of the registers is a hard reg?

It already does use HARD_REGNO_MODE_OK, in can_decompose_p. Note that that patch was committed on Saturday.

In the place where I use MODES_TIEABLE_P, hard registers are already ignored, so changing that is unlikely to make any difference.

If you tell me how to configure gcc and give me a test case, I will look into the problem.

Ian

• Follow-Ups:
  • Re: PATCH COMMITTED: More lower-subreg.c patches
    * From: DJ Delorie
• References:
  • PATCH COMMITTED: More lower-subreg.c patches
    * From: Ian Lance Taylor
  • Re: PATCH COMMITTED: More lower-subreg.c patches
    * From: DJ Delorie