For integer to floating conversions I've added frange_drop_infs
call into the handler. Supposedly I thought that integers converted
to floating point are never +-inf, but that is clearly not the case
as the testcases show. For _Float16 it can be +-inf very easily,
as the finite range is just [-65504.0f16,65504.0f16], for others
all one needs is a large enough _BitInt.
The following patch just drops that call. In the common cases,
+-inf will not appear in the range anyway, lb and ub will be usually
finite.
2026-08-04 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/126549
* range-op-float.cc (operator_cast::fold_range): Don't call
frange_drop_infs.
* gcc.dg/torture/pr126549.c: New test.
* gcc.dg/torture/bitint-107.c: New test.
Reviewed-by: Aldy Hernandez <aldy@quesejoda.com>
frange_nextafter (mode, ub, dconstinf);
}
r.set (type, lb, ub, nan_state (false));
- frange_drop_infs (r, type);
if (r.undefined_p ())
r.set_varying (type);
return true;
--- /dev/null
+/* PR tree-optimization/126549 */
+/* { dg-do run { target { float32 && bitint575 } } } */
+/* { dg-add-options float32 } */
+
+[[gnu::noipa]] int
+foo (unsigned _BitInt(133) a)
+{
+ unsigned _BitInt(133) u = a % 5444517870735015415413993718908291383300uwb;
+ _Float32 h = (_Float32) u;
+ return h > __FLT32_MAX__;
+}
+
+int
+main ()
+{
+ if (foo (5444517870735015415413993718908291383295uwb) != 1)
+ __builtin_abort ();
+}
--- /dev/null
+/* PR tree-optimization/126549 */
+/* { dg-do run { target float16 } } */
+/* { dg-add-options float16 } */
+
+[[gnu::noipa]] int
+foo (unsigned a)
+{
+ unsigned u = a % 100001U;
+ _Float16 h = (_Float16) u;
+ return h > 65504.0f16;
+}
+
+int
+main ()
+{
+ if (foo (70000U) != 1)
+ __builtin_abort ();
+}