public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] [PR tree-optimization/107355] Handle NANs in abs range-op entry.
@ 2022-10-24 13:33 Aldy Hernandez
  2022-10-24 13:34 ` Aldy Hernandez
  0 siblings, 1 reply; 2+ messages in thread
From: Aldy Hernandez @ 2022-10-24 13:33 UTC (permalink / raw)
  To: GCC patches; +Cc: Andrew MacLeod, Aldy Hernandez

The problem here is that the threader is coming up with a path where
the only valid result is a NAN.  When the abs op1_range entry is
trying to add the negative posibility, it attempts to get the bounds
of the working range.  NANs don't have bounds so they need to be
special cased.

	PR tree-optimization/107355

gcc/ChangeLog:

	* range-op-float.cc (foperator_abs::op1_range): Handle NAN.

gcc/testsuite/ChangeLog:

	* gcc.dg/tree-ssa/pr107355.c: New test.
---
 gcc/range-op-float.cc                    |  9 +++++++++
 gcc/testsuite/gcc.dg/tree-ssa/pr107355.c | 13 +++++++++++++
 2 files changed, 22 insertions(+)
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr107355.c

diff --git a/gcc/range-op-float.cc b/gcc/range-op-float.cc
index 8777bc70d71..04208c88dd1 100644
--- a/gcc/range-op-float.cc
+++ b/gcc/range-op-float.cc
@@ -1269,6 +1269,15 @@ foperator_abs::op1_range (frange &r, tree type,
   positives.update_nan (/*sign=*/false);
   positives.intersect (lhs);
   r = positives;
+  // Add -NAN if relevant.
+  if (r.maybe_isnan ())
+    {
+      frange neg_nan;
+      neg_nan.set_nan (type, true);
+      r.union_ (neg_nan);
+    }
+  if (r.known_isnan ())
+    return true;
   // Then add the negative of each pair:
   // ABS(op1) = [5,20] would yield op1 => [-20,-5][5,20].
   r.union_ (frange (type,
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr107355.c b/gcc/testsuite/gcc.dg/tree-ssa/pr107355.c
new file mode 100644
index 00000000000..40796344bfb
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr107355.c
@@ -0,0 +1,13 @@
+// { dg-do compile }
+// { dg-options "-O2 -fno-guess-branch-probability -fsanitize=float-cast-overflow --param=max-jump-thread-duplication-stmts=240" }
+
+float f;
+
+void
+foo (double d)
+{
+  (char) f;
+  long l = __builtin_fabs (d);
+  (char) f;
+  (long) d;
+}
-- 
2.37.3


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] [PR tree-optimization/107355] Handle NANs in abs range-op entry.
  2022-10-24 13:33 [PATCH] [PR tree-optimization/107355] Handle NANs in abs range-op entry Aldy Hernandez
@ 2022-10-24 13:34 ` Aldy Hernandez
  0 siblings, 0 replies; 2+ messages in thread
From: Aldy Hernandez @ 2022-10-24 13:34 UTC (permalink / raw)
  To: GCC patches; +Cc: Andrew MacLeod

Tested on x86-64 Linux.

Pushed.

On Mon, Oct 24, 2022 at 3:33 PM Aldy Hernandez <aldyh@redhat.com> wrote:
>
> The problem here is that the threader is coming up with a path where
> the only valid result is a NAN.  When the abs op1_range entry is
> trying to add the negative posibility, it attempts to get the bounds
> of the working range.  NANs don't have bounds so they need to be
> special cased.
>
>         PR tree-optimization/107355
>
> gcc/ChangeLog:
>
>         * range-op-float.cc (foperator_abs::op1_range): Handle NAN.
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.dg/tree-ssa/pr107355.c: New test.
> ---
>  gcc/range-op-float.cc                    |  9 +++++++++
>  gcc/testsuite/gcc.dg/tree-ssa/pr107355.c | 13 +++++++++++++
>  2 files changed, 22 insertions(+)
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr107355.c
>
> diff --git a/gcc/range-op-float.cc b/gcc/range-op-float.cc
> index 8777bc70d71..04208c88dd1 100644
> --- a/gcc/range-op-float.cc
> +++ b/gcc/range-op-float.cc
> @@ -1269,6 +1269,15 @@ foperator_abs::op1_range (frange &r, tree type,
>    positives.update_nan (/*sign=*/false);
>    positives.intersect (lhs);
>    r = positives;
> +  // Add -NAN if relevant.
> +  if (r.maybe_isnan ())
> +    {
> +      frange neg_nan;
> +      neg_nan.set_nan (type, true);
> +      r.union_ (neg_nan);
> +    }
> +  if (r.known_isnan ())
> +    return true;
>    // Then add the negative of each pair:
>    // ABS(op1) = [5,20] would yield op1 => [-20,-5][5,20].
>    r.union_ (frange (type,
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr107355.c b/gcc/testsuite/gcc.dg/tree-ssa/pr107355.c
> new file mode 100644
> index 00000000000..40796344bfb
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr107355.c
> @@ -0,0 +1,13 @@
> +// { dg-do compile }
> +// { dg-options "-O2 -fno-guess-branch-probability -fsanitize=float-cast-overflow --param=max-jump-thread-duplication-stmts=240" }
> +
> +float f;
> +
> +void
> +foo (double d)
> +{
> +  (char) f;
> +  long l = __builtin_fabs (d);
> +  (char) f;
> +  (long) d;
> +}
> --
> 2.37.3
>


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2022-10-24 13:34 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-24 13:33 [PATCH] [PR tree-optimization/107355] Handle NANs in abs range-op entry Aldy Hernandez
2022-10-24 13:34 ` Aldy Hernandez

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).