This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fix ifcvt on memory operands
- From: Richard Henderson <rth at redhat dot com>
- To: Jan Hubicka <hubicka at ucw dot cz>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 7 Jan 2003 13:38:45 -0800
- Subject: Re: Fix ifcvt on memory operands
- References: <20021204193501.GE5173@kam.mff.cuni.cz>
On Wed, Dec 04, 2002 at 08:35:01PM +0100, Jan Hubicka wrote:
> + if (!nonoverlapping_memrefs_p (x, (rtx)data))
I don't think you should be using nonoverlapping_memrefs_p
directly. Use
anti_dependence (x, data) || output_dependence (x, data)
Or better yet change write_depenence_p to let you do both
at once.
> + memory_modified = 1;
= true;
> + rtx insn;
> + if (start == end)
> + return 0;
Whitespace before if.
> case MEM:
> ! if (RTX_UNCHANGING_P (x))
> ! return 0;
> ! if (modified_between_p (XEXP (x, 0), start, end))
> return 1;
> + for (insn = NEXT_INSN (start); insn != end; insn = NEXT_INSN (insn))
> + if (memory_modified_in_insn_p (x, insn))
> + return 1;
> + return 0;
Why are you checking the address?
You'll need to change the function comment that says that
alias info isn't used. ;-)
r~