Richi recently changed tree-ssa-dom.c::record_equality to use
tree_swap_operands_p to canonicalize the implied copy we record for
equality comparisons. This is a good thing.
However, there is a case where tree_swap_operands_p gives us operands in
an undesirable order for this routine. Specifically when both operands
are SSA_NAMEs, but just one has a single use (in the equality test). In
that case, we want the single use SSA_NAME on the LHS of the recorded copy.
That, in effect, prevents DOM from destroying the single use nature of
that SSA_NAME. And if we're ultimately able to remove the comparison,
the statements that fed the single use SSA_NAME can be removed as
they'll be dead.
Given this issue is so closely tied to DOM, Richi and I agreed to fix
this in DOM rather than in tree_swap_operands_p. Had we tried to
address this in tree_swap_operands_p, we'll end up regressing elsewhere,
which is clearly not good.
Anyway, bootstrapped and regression tested on x86_64-linux-gnu. Fixes
the recently XFAILed 65217 test. Installed on the trunk.
Jeff
commit a787980e8012128b408ca8b716028e71d1b21373
Author: root <root@localhost.localdomain>
Date: Mon Apr 27 21:58:52 2015 -0600
PR tree-optimization/65217
* tree-ssa-dom.c (record_equality): Given two SSA_NAMEs, if just one
of them has a single use, make sure it is the LHS of the implied
copy.
PR tree-optimization/65217
* gcc.target/i386/pr65217.c: Remove XFAIL.
diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index fff0015..1a82f17 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,10 @@
+2015-04-27 Jeff Law <l...@redhat.com>
+
+ PR tree-optimization/65217
+ * tree-ssa-dom.c (record_equality): Given two SSA_NAMEs, if just one
+ of them has a single use, make sure it is the LHS of the implied
+ copy.
+
2015-04-28 Alan Modra <amo...@gmail.com>
PR target/65810
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 67bdd69..0fc2384 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2015-04-27 Jeff Law <l...@redhat.com>
+
+ PR tree-optimization/65217
+ * gcc.target/i386/pr65217.c: Remove XFAIL.
+
2015-04-27 Andre Vehreschild <ve...@gmx.de>
PR fortran/60322
diff --git a/gcc/testsuite/gcc.target/i386/pr65217.c
b/gcc/testsuite/gcc.target/i386/pr65217.c
index 3f495b2..d5c9be5 100644
--- a/gcc/testsuite/gcc.target/i386/pr65217.c
+++ b/gcc/testsuite/gcc.target/i386/pr65217.c
@@ -1,7 +1,7 @@
/* { dg-do compile } */
/* { dg-options "-O" } */
-/* { dg-final { scan-assembler-not "negl" { xfail *-*-* } } } */
-/* { dg-final { scan-assembler-not "andl" { xfail *-*-* } } } */
+/* { dg-final { scan-assembler-not "negl" } } */
+/* { dg-final { scan-assembler-not "andl" } } */
int
test(int n)
diff --git a/gcc/tree-ssa-dom.c b/gcc/tree-ssa-dom.c
index a67b4e4..c7d427b 100644
--- a/gcc/tree-ssa-dom.c
+++ b/gcc/tree-ssa-dom.c
@@ -1762,6 +1762,20 @@ record_equality (tree x, tree y)
if (tree_swap_operands_p (x, y, false))
std::swap (x, y);
+ /* Most of the time tree_swap_operands_p does what we want. But there's
+ cases where we we know one operand is better for copy propagation than
+ the other. Given no other code cares about ordering of equality
+ comparison operators for that purpose, we just handle the special cases
+ here. */
+ if (TREE_CODE (x) == SSA_NAME && TREE_CODE (y) == SSA_NAME)
+ {
+ /* If one operand is a single use operand, then make it
+ X. This will preserve its single use properly and if this
+ conditional is eliminated, the computation of X can be
+ eliminated as well. */
+ if (has_single_use (y) && ! has_single_use (x))
+ std::swap (x, y);
+ }
if (TREE_CODE (x) == SSA_NAME)
prev_x = SSA_NAME_VALUE (x);
if (TREE_CODE (y) == SSA_NAME)
