diff mbox series

PR tree-optimization/104420: Fix checks for constant folding X*0.0

Message ID 007401d81d42$67e6dde0$37b499a0$@nextmovesoftware.com
State New
Headers show
Series PR tree-optimization/104420: Fix checks for constant folding X*0.0 | expand

Commit Message

Roger Sayle Feb. 8, 2022, 11:20 p.m. UTC
This patch resolves PR tree-optimization/104420, which is a P1 regression
where, as observed by Jakub Jelinek, the conditions for constant folding
x*0.0 are incorrect (following my patch for PR tree-optimization/96392).
The multiplication x*0.0 may yield a negative zero result, -0.0, if X is
negative (not just if x may be negative zero).  Hence (without -ffast-math)
(int)x*0.0 can't be optimized to 0.0, but (unsigned)x*0.0 can be constant
folded.  This adds a bunch of test cases to confirm the desired behaviour,
and removes an incorrect test from gcc.dg/pr96392.c which checked for the
wrong behaviour.

This patch has been tested on x86_64-pc-linux-gnu with make bootstrap and
make -k check no new failures.  Ok for mainline?

2022-02-08  Roger Sayle  <roger@nextmovesoftware.com>

gcc/ChangeLog
	PR tree-optimization/104420
	* match.pd (mult @0 real_zerop): Tweak conditions for constant
	folding X*0.0 (or X*-0.0) to HONOR_SIGNED_ZEROS when appropriate.

gcc/testsuite/ChangeLog
	PR tree-optimization/104420
	* gcc.dg/pr104420-1.c: New test case.
	* gcc.dg/pr104420-2.c: New test case.
	* gcc.dg/pr104420-3.c: New test case.
	* gcc.dg/pr104420-4.c: New test case.
	* gcc.dg/pr96392.c: Remove incorrect test.

Thanks in advance (and sorry for the breakage/thinko).
Roger
--

Comments

Richard Biener Feb. 9, 2022, 8:32 a.m. UTC | #1
On Wed, Feb 9, 2022 at 12:20 AM Roger Sayle <roger@nextmovesoftware.com> wrote:
>
>
> This patch resolves PR tree-optimization/104420, which is a P1 regression
> where, as observed by Jakub Jelinek, the conditions for constant folding
> x*0.0 are incorrect (following my patch for PR tree-optimization/96392).
> The multiplication x*0.0 may yield a negative zero result, -0.0, if X is
> negative (not just if x may be negative zero).  Hence (without -ffast-math)
> (int)x*0.0 can't be optimized to 0.0, but (unsigned)x*0.0 can be constant
> folded.  This adds a bunch of test cases to confirm the desired behaviour,
> and removes an incorrect test from gcc.dg/pr96392.c which checked for the
> wrong behaviour.
>
> This patch has been tested on x86_64-pc-linux-gnu with make bootstrap and
> make -k check no new failures.  Ok for mainline?

OK.

Thanks,
Richard.

> 2022-02-08  Roger Sayle  <roger@nextmovesoftware.com>
>
> gcc/ChangeLog
>         PR tree-optimization/104420
>         * match.pd (mult @0 real_zerop): Tweak conditions for constant
>         folding X*0.0 (or X*-0.0) to HONOR_SIGNED_ZEROS when appropriate.
>
> gcc/testsuite/ChangeLog
>         PR tree-optimization/104420
>         * gcc.dg/pr104420-1.c: New test case.
>         * gcc.dg/pr104420-2.c: New test case.
>         * gcc.dg/pr104420-3.c: New test case.
>         * gcc.dg/pr104420-4.c: New test case.
>         * gcc.dg/pr96392.c: Remove incorrect test.
>
> Thanks in advance (and sorry for the breakage/thinko).
> Roger
> --
>
diff mbox series

Patch

diff --git a/gcc/match.pd b/gcc/match.pd
index 7bbb801..4fe5909 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -262,8 +262,7 @@  DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
  (mult @0 real_zerop@1)
  (if (!tree_expr_maybe_nan_p (@0)
       && (!HONOR_NANS (type) || !tree_expr_maybe_infinite_p (@0))
-      && !tree_expr_maybe_real_minus_zero_p (@0)
-      && !tree_expr_maybe_real_minus_zero_p (@1))
+      && (!HONOR_SIGNED_ZEROS (type) || tree_expr_nonnegative_p (@0)))
   @1))
 
 /* In IEEE floating point, x*1 is not equivalent to x for snans.
diff --git a/gcc/testsuite/gcc.dg/pr104420-1.c b/gcc/testsuite/gcc.dg/pr104420-1.c
new file mode 100644
index 0000000..48385fa
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr104420-1.c
@@ -0,0 +1,10 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-add-options ieee } */
+
+double f(int a)
+{
+  return a * 0.0;
+}
+
+/* { dg-final { scan-tree-dump " \\\* 0.0" "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/pr104420-2.c b/gcc/testsuite/gcc.dg/pr104420-2.c
new file mode 100644
index 0000000..49d0189
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr104420-2.c
@@ -0,0 +1,10 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-add-options ieee } */
+
+double f(int a)
+{
+  return a * -0.0;
+}
+
+/* { dg-final { scan-tree-dump " \\\* -0.0" "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/pr104420-3.c b/gcc/testsuite/gcc.dg/pr104420-3.c
new file mode 100644
index 0000000..962dfff
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr104420-3.c
@@ -0,0 +1,10 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-add-options ieee } */
+
+double f(unsigned int a)
+{
+  return a * 0.0;
+}
+
+/* { dg-final { scan-tree-dump "return 0.0" "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/pr104420-4.c b/gcc/testsuite/gcc.dg/pr104420-4.c
new file mode 100644
index 0000000..95ed0cc
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr104420-4.c
@@ -0,0 +1,10 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-add-options ieee } */
+
+double f(unsigned int a)
+{
+  return a * -0.0;
+}
+
+/* { dg-final { scan-tree-dump "return -0.0" "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/pr96392.c b/gcc/testsuite/gcc.dg/pr96392.c
index 662bacb..fb7de21 100644
--- a/gcc/testsuite/gcc.dg/pr96392.c
+++ b/gcc/testsuite/gcc.dg/pr96392.c
@@ -12,11 +12,6 @@  double sub0(int x)
   return x - 0.0;
 }
 
-double mult0(int x)
-{
-  return 0.0 * x;
-}
-
 double negate(int x)
 {
   return 0.0 - x;
@@ -29,5 +24,4 @@  double subtract(int x)
 
 /* { dg-final { scan-tree-dump-not " \\+ " "optimized" } } */
 /* { dg-final { scan-tree-dump-not " \\- " "optimized" } } */
-/* { dg-final { scan-tree-dump-not " \\* " "optimized" } } */