This is the mail archive of the gcc-patches@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]
Other format: [Raw text]

Re: partial def patch

Ok, I got round to reviewing this patch.

*************** df_read_modify_subreg_p (rtx x)
*** 2702,2708 ****
return false;
isize = GET_MODE_SIZE (GET_MODE (SUBREG_REG (x)));
osize = GET_MODE_SIZE (GET_MODE (x));
! return (isize > osize && isize > UNITS_PER_WORD);
}
--- 2702,2708 ----
return false;
isize = GET_MODE_SIZE (GET_MODE (SUBREG_REG (x)));
osize = GET_MODE_SIZE (GET_MODE (x));
! return isize > osize;
}

Leave out this hunk, and put a FIXME saying that it could use REGMODE_NATURAL_SIZE instead of UNITS_PER_WORD. A (subreg:QI (reg:SI A) B) clobbers the entire contents of the pseudo A.

  	 || GET_CODE (dst) == ZERO_EXTRACT
  	 || df_read_modify_subreg_p (dst))
      {
!       flags |= DF_REF_READ_WRITE;
!       if (GET_CODE (dst) != SUBREG
! 	  || GET_MODE_SIZE (GET_MODE (dst))
! 	     >= REGMODE_NATURAL_SIZE (GET_MODE (dst)))
! 	flags |= DF_REF_PARTIAL;

Just do "flags |= DF_REF_READ_WRITE | DF_REF_PARTIAL"; the check for REGMODE_NATURAL_SIZE, if anywhere, should go in df_read_modify_subreg_p.

Thanks, and sorry for the delay.

Paolo

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