public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r12-4764] middle-end/57245 - honor -frounding-math in real truncation
@ 2021-10-28 9:29 Richard Biener
0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2021-10-28 9:29 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:a84b9d5373c7e67fd0ab2a412c22162cdf969c91
commit r12-4764-ga84b9d5373c7e67fd0ab2a412c22162cdf969c91
Author: Richard Biener <rguenther@suse.de>
Date: Wed Oct 27 14:27:40 2021 +0200
middle-end/57245 - honor -frounding-math in real truncation
The following honors -frounding-math when converting a FP constant
to another FP type.
2021-10-27 Richard Biener <rguenther@suse.de>
PR middle-end/57245
* fold-const.c (fold_convert_const_real_from_real): Honor
-frounding-math if the conversion is not exact.
* simplify-rtx.c (simplify_const_unary_operation): Do not
simplify FLOAT_TRUNCATE with sign dependent rounding.
* gcc.dg/torture/fp-double-convert-float-1.c: New testcase.
Diff:
---
gcc/fold-const.c | 6 ++++
gcc/simplify-rtx.c | 5 +++
.../gcc.dg/torture/fp-double-convert-float-1.c | 41 ++++++++++++++++++++++
3 files changed, 52 insertions(+)
diff --git a/gcc/fold-const.c b/gcc/fold-const.c
index ff23f12f33c..18950aeb760 100644
--- a/gcc/fold-const.c
+++ b/gcc/fold-const.c
@@ -2139,6 +2139,12 @@ fold_convert_const_real_from_real (tree type, const_tree arg1)
&& REAL_VALUE_ISSIGNALING_NAN (TREE_REAL_CST (arg1)))
return NULL_TREE;
+ /* With flag_rounding_math we should respect the current rounding mode
+ unless the conversion is exact. */
+ if (HONOR_SIGN_DEPENDENT_ROUNDING (arg1)
+ && !exact_real_truncate (TYPE_MODE (type), &TREE_REAL_CST (arg1)))
+ return NULL_TREE;
+
real_convert (&value, TYPE_MODE (type), &TREE_REAL_CST (arg1));
t = build_real (type, value);
diff --git a/gcc/simplify-rtx.c b/gcc/simplify-rtx.c
index bbbd6b74942..f38b6d7d31c 100644
--- a/gcc/simplify-rtx.c
+++ b/gcc/simplify-rtx.c
@@ -2068,6 +2068,11 @@ simplify_const_unary_operation (enum rtx_code code, machine_mode mode,
and the operand is a signaling NaN. */
if (HONOR_SNANS (mode) && REAL_VALUE_ISSIGNALING_NAN (d))
return NULL_RTX;
+ /* Or if flag_rounding_math is on and the truncation is not
+ exact. */
+ if (HONOR_SIGN_DEPENDENT_ROUNDING (mode)
+ && !exact_real_truncate (mode, &d))
+ return NULL_RTX;
d = real_value_truncate (mode, d);
break;
case FLOAT_EXTEND:
diff --git a/gcc/testsuite/gcc.dg/torture/fp-double-convert-float-1.c b/gcc/testsuite/gcc.dg/torture/fp-double-convert-float-1.c
new file mode 100644
index 00000000000..ec23274ea98
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/fp-double-convert-float-1.c
@@ -0,0 +1,41 @@
+/* PR57245 */
+/* { dg-do run } */
+/* { dg-require-effective-target fenv } */
+/* { dg-additional-options "-frounding-math" } */
+
+#include <fenv.h>
+#include <stdlib.h>
+
+int
+main ()
+{
+#if __DBL_MANT_DIG__ == 53 && __FLT_MANT_DIG__ == 24
+#ifdef FE_UPWARD
+ fesetround (FE_UPWARD);
+ float f = 1.3;
+ if (f != 0x1.4ccccep+0f)
+ __builtin_abort ();
+#endif
+#ifdef FE_TONEAREST
+ fesetround (FE_TONEAREST);
+ /* Use different actual values so the bogus CSE we perform does not
+ break things. */
+ f = 1.33;
+ if (f != 0x1.547ae2p+0f)
+ abort ();
+#endif
+#ifdef FE_DOWNWARD
+ fesetround (FE_DOWNWARD);
+ f = 1.333;
+ if (f != 0x1.553f7cp+0f)
+ abort ();
+#endif
+#ifdef FE_TOWARDZERO
+ fesetround (FE_TOWARDZERO);
+ f = 1.3333;
+ if (f != 0x1.555326p+0f)
+ abort ();
+#endif
+#endif
+ return 0;
+}
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2021-10-28 9:29 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-10-28 9:29 [gcc r12-4764] middle-end/57245 - honor -frounding-math in real truncation Richard Biener
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).