diff mbox series

[v5,01/12] math: Simplify hypotf implementation

Message ID 20211207190353.3282666-2-adhemerval.zanella@linaro.org
State New
Headers show
Series Improve hypot | expand

Commit Message

Adhemerval Zanella Dec. 7, 2021, 7:03 p.m. UTC 
Use a more optimized comparison for check for NaN and infinite and
add an inlined issignaling implementation for float.  With gcc it
results in 2 FP comparisons.

The file Copyright is also changed to use  GPL, the implementation was
completely changed by 7c10fd3515f to use double precision instead of
scaling and this change removes all the GET_FLOAT_WORD usage.

Checked on x86_64-linux-gnu.
---
 sysdeps/ieee754/flt-32/e_hypotf.c    | 64 +++++++++++++---------------
 sysdeps/ieee754/flt-32/math_config.h |  9 ++++
 2 files changed, 38 insertions(+), 35 deletions(-)
diff mbox series

Patch

diff --git a/sysdeps/ieee754/flt-32/e_hypotf.c b/sysdeps/ieee754/flt-32/e_hypotf.c
index e770947dc1..436e3c2da8 100644
--- a/sysdeps/ieee754/flt-32/e_hypotf.c
+++ b/sysdeps/ieee754/flt-32/e_hypotf.c
@@ -1,46 +1,40 @@ 
-/* e_hypotf.c -- float version of e_hypot.c.
- */
+/* Euclidean distance function.  Float/Binary32 version.
+   Copyright (C) 2012-2021 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
 
-/*
- * ====================================================
- * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
- *
- * Developed at SunPro, a Sun Microsystems, Inc. business.
- * Permission to use, copy, modify, and distribute this
- * software is freely granted, provided that this notice
- * is preserved.
- * ====================================================
- */
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
 
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#include <libm-alias-finite.h>
 #include <math.h>
+#include "math_config.h"
+#include <math-narrow-eval.h>
 #include <math_private.h>
-#include <libm-alias-finite.h>
 
 float
-__ieee754_hypotf(float x, float y)
+__ieee754_hypotf (float x, float y)
 {
-	double d_x, d_y;
-	int32_t ha, hb;
-
-	GET_FLOAT_WORD(ha,x);
-	ha &= 0x7fffffff;
-	GET_FLOAT_WORD(hb,y);
-	hb &= 0x7fffffff;
-	if (ha == 0x7f800000 && !issignaling (y))
-	  return fabsf(x);
-	else if (hb == 0x7f800000 && !issignaling (x))
-	  return fabsf(y);
-	else if (ha > 0x7f800000 || hb > 0x7f800000)
-	  return fabsf(x) * fabsf(y);
-	else if (ha == 0)
-	  return fabsf(y);
-	else if (hb == 0)
-	  return fabsf(x);
-
-	d_x = (double) x;
-	d_y = (double) y;
+   if (!isfinite(x) || !isfinite(y))
+     {
+       if ((isinf (x) || isinf (y))
+	   && !issignalingf_inline (x) && !issignalingf_inline (y))
+	 return INFINITY;
+       return x + y;
+     }
 
-	return (float) sqrt(d_x * d_x + d_y * d_y);
+  return math_narrow_eval ((float) sqrt ((double) x * (double) x
+					 + (double) y * (double) y));
 }
 #ifndef __ieee754_hypotf
 libm_alias_finite (__ieee754_hypotf, __hypotf)
diff --git a/sysdeps/ieee754/flt-32/math_config.h b/sysdeps/ieee754/flt-32/math_config.h
index 513454a297..daa8a82f99 100644
--- a/sysdeps/ieee754/flt-32/math_config.h
+++ b/sysdeps/ieee754/flt-32/math_config.h
@@ -101,6 +101,15 @@  asdouble (uint64_t i)
   return u.f;
 }
 
+static inline int
+issignalingf_inline (float x)
+{
+  uint32_t ix = asuint (x);
+  if (HIGH_ORDER_BIT_IS_SET_FOR_SNAN)
+    return (ix & 0x7fc00000) == 0x7fc00000;
+  return 2 * (ix ^ 0x00400000) > 2 * 0x7fc00000UL;
+}
+
 #define NOINLINE __attribute__ ((noinline))
 
 attribute_hidden float __math_oflowf (uint32_t);