diff mbox

RFA: PR target/53120, constraint modifier "+" on operand tied by matching-constraint, "0".

Message ID 4FA0FF4E.5030205@redhat.com
State New
Headers show

Commit Message

Nick Clifton May 2, 2012, 9:33 a.m. UTC 
Hi DJ,

   As pointed out by Hans-Peter in his patch for PR 53120 for the cris 
backend, the m32c/bitops.md file contains a pattern that uses both the 
"+" modifier and the "0" constraint.  The patch below is a 
straightforward fix for this, replacing the "0" constraint with a 
match_dup.  Tested with no regressions on an m32c-elf toolchain.

OK to apply ?

Cheers
   Nick

gcc/ChangeLog
2012-05-02  Nick Clifton  <nickc@redhat.com>

	PR target/53120
	* config/m32c/bitops.md (bset_qi): Change operand 2 from having
	a "0" constraint to being a (match_dup 0).

Comments

DJ Delorie May 3, 2012, 3:40 p.m. UTC | #1 
Make sure a match_dup will still match the generated pattern later,
I've had problems with match_dup not matching two rtx that
rtx_equals() says are "the same" but not physically the same.
Nick Clifton May 8, 2012, 4:08 p.m. UTC | #2 
Hi DJ,
> Make sure a match_dup will still match the generated pattern later,
> I've had problems with match_dup not matching two rtx that
> rtx_equals() says are "the same" but not physically the same.

I have tried, but failed, to find a way to trigger the use of the 
bset_qi pattern. :-(  I tried rebuilding the toolchain and running the 
GCC testsuite, but neither of these worked.  Do you have a test case 
that triggers it ?

Cheers
   Nick
DJ Delorie May 8, 2012, 6:04 p.m. UTC | #3 
#define q ((char *)0x1234)

foo(int x)
{
  *q |= (1 << (char)x);
}

$ m32c-elf-gcc -S -O3 nick.c

        .global _foo
_foo:                                    
        mov.w   r1,a0                    ; 20                           movhi_op/3
        bset    4660[a0]                 ; 11                           bset_qi
        rts                              ; 23                           epilogue_rts
Nick Clifton May 9, 2012, 9:39 a.m. UTC | #4 
Hi DJ,

   Thanks for the test program.  With that I can confirm that the 
match_dup does work and that the bset instruction is still generated 
even with my patch applied.

   OK to apply ?

Cheers
   Nick
DJ Delorie May 9, 2012, 4:58 p.m. UTC | #5 
>    OK to apply ?

Ok.  Thanks!
diff mbox

Patch

Index: gcc/config/m32c/bitops.md
===================================================================
--- gcc/config/m32c/bitops.md	(revision 187036)
+++ gcc/config/m32c/bitops.md	(working copy)
@@ -43,7 +43,7 @@ 
    [(set (match_operand:QI 0 "memsym_operand" "+Si")
  	(ior:QI (subreg:QI (ashift:HI (const_int 1)
  				      (subreg:QI (match_operand:HI 1 "a_qi_operand" "Raa") 0)) 0)
-		(match_operand:QI 2 "memsym_operand" "0")))]
+		(match_dup 0)))]
    "TARGET_A16"
    "bset\t%0[%1]"
    [(set_attr "flags" "n")]