diff mbox series

combine: Fix up -fcompare-debug issue in the combiner [PR104544]

Message ID 20220216085334.GX2646553@tucnak
State New
Headers show
Series combine: Fix up -fcompare-debug issue in the combiner [PR104544] | expand

Commit Message

Jakub Jelinek Feb. 16, 2022, 8:53 a.m. UTC 
Hi!

On the following testcase on aarch64-linux, we behave differently
with -g and -g0.

The problem is that on:
(insn 10011 10010 10012 2 (set (reg:CC 66 cc)
        (compare:CC (reg:DI 105)
            (const_int 0 [0]))) "pr104544.c":18:3 407 {cmpdi}
     (expr_list:REG_DEAD (reg:DI 105)
        (nil)))
(insn 10012 10011 10013 2 (set (reg:SI 109)
        (eq:SI (reg:CC 66 cc)
            (const_int 0 [0]))) "pr104544.c":18:3 444 {aarch64_cstoresi}
     (expr_list:REG_DEAD (reg:CC 66 cc)
        (nil)))
(insn 10013 10012 10016 2 (set (reg:DI 110)
        (zero_extend:DI (reg:SI 109))) "pr104544.c":18:3 111 {*zero_extendsidi2_aarch64}
     (expr_list:REG_DEAD (reg:SI 109)
        (nil)))
(insn 10016 10013 10017 2 (parallel [
            (set (reg:CC 66 cc)
                (compare:CC (const_int 0 [0])
                    (reg:DI 110)))
            (set (reg:DI 111)
                (neg:DI (reg:DI 110)))
        ]) "pr104544.c":18:3 281 {negdi_carryout}
     (expr_list:REG_DEAD (reg:DI 110)
        (nil)))
...
(debug_insn 6 5 7 2 (var_location:SI y (debug_expr:SI D#5)) "pr104544.c":18:3 -1
     (nil))
(debug_insn 7 6 10033 2 (debug_marker) "pr104544.c":11:3 -1
     (nil))
(insn 10033 7 10034 2 (set (reg:DI 117 [ _14 ])
        (ior:DI (reg:DI 111)
            (reg:DI 112))) "pr104544.c":11:6 496 {iordi3}
     (expr_list:REG_DEAD (reg:DI 112)
        (expr_list:REG_DEAD (reg:DI 111)
            (nil))))
we successfully split 3 insns into two:

Trying 10011, 10013 -> 10016:
 10011: cc:CC=cmp(r105:DI,0)
      REG_DEAD r105:DI
 10013: r110:DI=cc:CC==0
      REG_DEAD cc:CC
 10016: {cc:CC=cmp(0,r110:DI);r111:DI=-r110:DI;}
      REG_DEAD r110:DI
Failed to match this instruction:
(parallel [
        (set (reg:CC 66 cc)
            (compare:CC (reg:DI 105)
                (const_int 0 [0])))
        (set (reg:DI 111)
            (neg:DI (eq:DI (reg:DI 105)
                    (const_int 0 [0]))))
    ])
Failed to match this instruction:
(parallel [
        (set (reg:CC 66 cc)
            (compare:CC (reg:DI 105)
                (const_int 0 [0])))
        (set (reg:DI 111)
            (neg:DI (eq:DI (reg:DI 105)
                    (const_int 0 [0]))))
    ])
Successfully matched this instruction:
(set (reg:DI 111)
    (neg:DI (eq:DI (reg:DI 105)
            (const_int 0 [0]))))
Successfully matched this instruction:
(set (reg:CC 66 cc)
    (compare:CC (reg:DI 105)
        (const_int 0 [0])))
Successfully matched this instruction:
(set (reg:DI 112)
    (neg:DI (eq:DI (reg:CC 66 cc)
            (const_int 0 [0]))))
allowing combination of insns 10011, 10013 and 10016
original costs 4 + 4 + 4 = 16
replacement costs 4 + 4 = 12
deferring deletion of insn with uid = 10011.

but the code that searches forward for insns to update their log
links (before the change there is a link from insn 10033 to insn 10016
for pseudo 111) only finds insn 10033 and updates the log link if
-g isn't enabled, otherwise it stops earlier because there are debug insns
in between.  So, with -g LOG_LINKS of 10033 isn't updated, points eventually
to NOTE_INSN_DELETED and so we do not attempt to combine 10033 with other
insns, while with -g0 we do.

The following patch fixes that by instead ignoring debug insns during the
searching.  We can still check BLOCK_FOR_INSN (insn) on those, because
if we notice DEBUG_INSN in a following basic block, necessarily there won't
be any further normal insns in the current block after it.

Bootstrapped/regtested on x86_64-linux and i686-linux, bootstrapped
on aarch64-linux, regtest on aarch64-linux still pending, ok for trunk
if it succeeds?

2022-02-16  Jakub Jelinek  <jakub@redhat.com>

	PR rtl-optimization/104544
	* combine.cc (try_combine): When looking for insn whose links
	should be updated from i3 to i2, don't stop on debug insns, instead
	skip over them.

	* gcc.dg/pr104544.c: New test.


	Jakub

Comments

Segher Boessenkool Feb. 16, 2022, 10:44 a.m. UTC | #1 
Hi!

On Wed, Feb 16, 2022 at 09:53:34AM +0100, Jakub Jelinek wrote:
> On the following testcase on aarch64-linux, we behave differently
> with -g and -g0.

[ huge snip ]

> The following patch fixes that by instead ignoring debug insns during the
> searching.  We can still check BLOCK_FOR_INSN (insn) on those, because
> if we notice DEBUG_INSN in a following basic block, necessarily there won't
> be any further normal insns in the current block after it.

> --- gcc/combine.cc.jj	2022-02-11 13:51:56.294928090 +0100
> +++ gcc/combine.cc	2022-02-15 14:15:41.663012950 +0100
> @@ -4223,10 +4223,12 @@ try_combine (rtx_insn *i3, rtx_insn *i2,
>  	  for (rtx_insn *insn = NEXT_INSN (i3);
>  	       !done
>  	       && insn
> -	       && NONDEBUG_INSN_P (insn)
> +	       && INSN_P (insn)
>  	       && BLOCK_FOR_INSN (insn) == this_basic_block;
>  	       insn = NEXT_INSN (insn))
>  	    {
> +	      if (DEBUG_INSN_P (insn))
> +		continue;
>  	      struct insn_link *link;
>  	      FOR_EACH_LOG_LINK (link, insn)
>  		if (link->insn == i3 && link->regno == regno)

About half of the similar loops in combine.c are still broken this way,
from a quick sampling :-(

Okay for trunk and all backports you may want.  Thanks!


Segher
Jakub Jelinek Feb. 16, 2022, 10:55 a.m. UTC | #2 
On Wed, Feb 16, 2022 at 04:44:58AM -0600, Segher Boessenkool wrote:
> > --- gcc/combine.cc.jj	2022-02-11 13:51:56.294928090 +0100
> > +++ gcc/combine.cc	2022-02-15 14:15:41.663012950 +0100
> > @@ -4223,10 +4223,12 @@ try_combine (rtx_insn *i3, rtx_insn *i2,
> >  	  for (rtx_insn *insn = NEXT_INSN (i3);
> >  	       !done
> >  	       && insn
> > -	       && NONDEBUG_INSN_P (insn)
> > +	       && INSN_P (insn)
> >  	       && BLOCK_FOR_INSN (insn) == this_basic_block;
> >  	       insn = NEXT_INSN (insn))
> >  	    {
> > +	      if (DEBUG_INSN_P (insn))
> > +		continue;
> >  	      struct insn_link *link;
> >  	      FOR_EACH_LOG_LINK (link, insn)
> >  		if (link->insn == i3 && link->regno == regno)
> 
> About half of the similar loops in combine.c are still broken this way,
> from a quick sampling :-(

Looking for just NONDEBUG_INSN_P, I don't see any other than this.

> Okay for trunk and all backports you may want.  Thanks!

Thanks.

	Jakub
Segher Boessenkool Feb. 16, 2022, 3:11 p.m. UTC | #3 
On Wed, Feb 16, 2022 at 11:55:23AM +0100, Jakub Jelinek wrote:
> On Wed, Feb 16, 2022 at 04:44:58AM -0600, Segher Boessenkool wrote:
> > About half of the similar loops in combine.c are still broken this way,
> > from a quick sampling :-(
> 
> Looking for just NONDEBUG_INSN_P, I don't see any other than this.

Ah yes, I was confused by !NONDEBUG_INSN.  Too many inversions make my
head spin (NONDEBUG_INSN really means RTX_INSN && !DEBUG_INSN).

So everything looks fine here now.  Thanks for double checking!


Segher
diff mbox series

Patch

--- gcc/combine.cc.jj	2022-02-11 13:51:56.294928090 +0100
+++ gcc/combine.cc	2022-02-15 14:15:41.663012950 +0100
@@ -4223,10 +4223,12 @@  try_combine (rtx_insn *i3, rtx_insn *i2,
 	  for (rtx_insn *insn = NEXT_INSN (i3);
 	       !done
 	       && insn
-	       && NONDEBUG_INSN_P (insn)
+	       && INSN_P (insn)
 	       && BLOCK_FOR_INSN (insn) == this_basic_block;
 	       insn = NEXT_INSN (insn))
 	    {
+	      if (DEBUG_INSN_P (insn))
+		continue;
 	      struct insn_link *link;
 	      FOR_EACH_LOG_LINK (link, insn)
 		if (link->insn == i3 && link->regno == regno)
--- gcc/testsuite/gcc.dg/pr104544.c.jj	2022-02-15 14:17:50.154221461 +0100
+++ gcc/testsuite/gcc.dg/pr104544.c	2022-02-15 14:17:34.441440536 +0100
@@ -0,0 +1,19 @@ 
+/* PR rtl-optimization/104544 */
+/* { dg-do compile { target int128 } } */
+/* { dg-options "-O2 -fcompare-debug" } */
+
+int m, n;
+__int128 q;
+
+void
+bar (unsigned __int128 x, int y)
+{
+  if (x)
+    q += y;
+}
+
+void
+foo (void)
+{
+  bar (!!q - 1, (m += m ? m : 1) < n);
+}