math: Remove the error handling wrapper from fmod and fmodf

The error handling is moved to sysdeps/ieee754 version with no SVID
support.  The compatibility symbol versions still use the wrapper
with SVID error handling around the new code.  There is no new symbol
version nor compatibility code on !LIBM_SVID_COMPAT targets
(e.g. riscv).

The ia64 is unchanged, since it still uses the arch specific
__libm_error_region on its implementation.  For both i686 and m68k,
which provive arch specific implementation, wrappers are added so
no new symbol are added (which would require to change the
implementations).

It shows an small improvement, the results for fmod:

  Architecture     | Input           | master   | patch
  -----------------|-----------------|----------|--------
  x86_64 (Ryzen 9) | subnormals      | 12.5049  | 9.40992
  x86_64 (Ryzen 9) | normal          | 296.939  | 296.738
  x86_64 (Ryzen 9) | close-exponents | 16.0244  | 13.119
  aarch64 (N1)     | subnormal       | 6.81778  | 4.33313
  aarch64 (N1)     | normal          | 155.620  | 152.915
  aarch64 (N1)     | close-exponents | 8.21306  | 5.76138
  armhf (N1)       | subnormal       | 15.1083  | 14.5746
  armhf (N1)       | normal          | 244.833  | 241.738
  armhf (N1)       | close-exponents | 21.8182  | 22.457

Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu.
Reviewed-by: Wilco Dijkstra  <Wilco.Dijkstra@arm.com>
This commit is contained in:
Adhemerval Zanella Netto 2023-03-20 13:01:18 -03:00 committed by Adhemerval Zanella
parent cf9cf33199
commit 16439f419b
38 changed files with 172 additions and 13 deletions

View File

@ -631,4 +631,8 @@ libm {
# No SVID compatible error handling.
hypotf; hypot;
}
GLIBC_2.38 {
# No SVID compatible error handling.
fmod; fmodf;
}
}

View File

@ -20,10 +20,10 @@
#include <math-svid-compat.h>
#include <libm-alias-double.h>
#if LIBM_SVID_COMPAT
#if LIBM_SVID_COMPAT && SHLIB_COMPAT (libm, GLIBC_2_0, GLIBC_2_38)
/* wrapper fmod */
double
__fmod (double x, double y)
__fmod_compat (double x, double y)
{
if (__builtin_expect (isinf (x) || y == 0.0, 0)
&& _LIB_VERSION != _IEEE_ && !isnan (y) && !isnan (x))
@ -32,5 +32,12 @@ __fmod (double x, double y)
return __ieee754_fmod (x, y);
}
libm_alias_double (__fmod, fmod)
compat_symbol (libm, __fmod_compat, fmod, GLIBC_2_0);
# ifdef NO_LONG_DOUBLE
weak_alias (__fmod_compat, fmodl)
# endif
# ifdef LONG_DOUBLE_COMPAT
LONG_DOUBLE_COMPAT_CHOOSE_libm_fmodl (
compat_symbol (libm, __fmod_compat, fmodl, FIRST_VERSION_libm_fmodl), );
# endif
#endif

View File

@ -20,10 +20,10 @@
#include <math-svid-compat.h>
#include <libm-alias-float.h>
#if LIBM_SVID_COMPAT
#if LIBM_SVID_COMPAT && SHLIB_COMPAT (libm, GLIBC_2_0, GLIBC_2_38)
/* wrapper fmodf */
float
__fmodf (float x, float y)
__fmod_compatf (float x, float y)
{
if (__builtin_expect (isinf (x) || y == 0.0f, 0)
&& _LIB_VERSION != _IEEE_ && !isnan (y) && !isnan (x))
@ -32,5 +32,5 @@ __fmodf (float x, float y)
return __ieee754_fmodf (x, y);
}
libm_alias_float (__fmod, fmod)
compat_symbol (libm, __fmod_compatf, fmodf, GLIBC_2_0);
#endif

View File

@ -0,0 +1,14 @@
/* i386 provides an optimized __ieee752_fmod. */
#include <math-svid-compat.h>
#ifdef SHARED
# undef SHLIB_COMPAT
# define SHLIB_COMPAT(a, b, c) 1
# undef LIBM_SVID_COMPAT
# define LIBM_SVID_COMPAT 1
# undef compat_symbol
# define compat_symbol(a, b, c, d)
#endif
#include <math/w_fmod_compat.c>
#ifdef SHARED
libm_alias_double (__fmod_compat, fmod)
#endif

View File

@ -0,0 +1,14 @@
/* i386 provides an optimized __ieee752_fmodf. */
#include <math-svid-compat.h>
#ifdef SHARED
# undef SHLIB_COMPAT
# define SHLIB_COMPAT(a, b, c) 1
# undef LIBM_SVID_COMPAT
# define LIBM_SVID_COMPAT 1
# undef compat_symbol
# define compat_symbol(a, b, c, d)
#endif
#include <math/w_fmodf_compat.c>
#ifdef SHARED
libm_alias_float (__fmod_compat, fmod)
#endif

View File

@ -16,7 +16,9 @@
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
#include <libm-alias-double.h>
#include <libm-alias-finite.h>
#include <math-svid-compat.h>
#include <math.h>
#include "math_config.h"
@ -55,7 +57,7 @@
} */
double
__ieee754_fmod (double x, double y)
__fmod (double x, double y)
{
uint64_t hx = asuint64 (x);
uint64_t hy = asuint64 (y);
@ -67,11 +69,16 @@ __ieee754_fmod (double x, double y)
/* Special cases:
- If x or y is a Nan, NaN is returned.
- If x is an inifinity, a NaN is returned.
- If x is an inifinity, a NaN is returned and EDOM is set.
- If y is zero, Nan is returned.
- If x is +0/-0, and y is not zero, +0/-0 is returned. */
if (__glibc_unlikely (hy == 0 || hx >= EXPONENT_MASK || hy > EXPONENT_MASK))
return (x * y) / (x * y);
if (__glibc_unlikely (hy == 0
|| hx >= EXPONENT_MASK || hy > EXPONENT_MASK))
{
if (is_nan (hx) || is_nan (hy))
return (x * y) / (x * y);
return __math_edom ((x * y) / (x * y));
}
if (__glibc_unlikely (hx <= hy))
{
@ -153,4 +160,11 @@ __ieee754_fmod (double x, double y)
return make_double (mx, ey, sx);
}
strong_alias (__fmod, __ieee754_fmod)
libm_alias_finite (__ieee754_fmod, __fmod)
#if LIBM_SVID_COMPAT
versioned_symbol (libm, __fmod, fmod, GLIBC_2_38);
libm_alias_double_other (__fmod, fmod)
#else
libm_alias_double (__fmod, fmod)
#endif

View File

@ -170,6 +170,9 @@ attribute_hidden double __math_invalid (double);
/* Error handling using output checking, only for errno setting. */
/* Check if the result generated a demain error. */
attribute_hidden double __math_edom (double x);
/* Check if the result overflowed to infinity. */
attribute_hidden double __math_check_oflow (double);
/* Check if the result underflowed to 0. */

View File

@ -33,6 +33,12 @@ with_errno (double y, int e)
#define with_errno(x, e) (x)
#endif
attribute_hidden double
__math_edom (double y)
{
return with_errno (y, EDOM);
}
/* NOINLINE reduces code size. */
NOINLINE static double
xflow (uint32_t sign, double y)

View File

@ -0,0 +1 @@
/* Not needed. */

View File

@ -17,6 +17,8 @@
<https://www.gnu.org/licenses/>. */
#include <libm-alias-finite.h>
#include <libm-alias-float.h>
#include <math-svid-compat.h>
#include <math.h>
#include "math_config.h"
@ -55,7 +57,7 @@
} */
float
__ieee754_fmodf (float x, float y)
__fmodf (float x, float y)
{
uint32_t hx = asuint (x);
uint32_t hy = asuint (y);
@ -70,8 +72,13 @@ __ieee754_fmodf (float x, float y)
- If x is an inifinity, a NaN is returned.
- If y is zero, Nan is returned.
- If x is +0/-0, and y is not zero, +0/-0 is returned. */
if (__glibc_unlikely (hy == 0 || hx >= EXPONENT_MASK || hy > EXPONENT_MASK))
return (x * y) / (x * y);
if (__glibc_unlikely (hy == 0
|| hx >= EXPONENT_MASK || hy > EXPONENT_MASK))
{
if (is_nan (hx) || is_nan (hy))
return (x * y) / (x * y);
return __math_edomf ((x * y) / (x * y));
}
if (__glibc_unlikely (hx <= hy))
{
@ -152,4 +159,11 @@ __ieee754_fmodf (float x, float y)
return make_float (mx, ey, sx);
}
strong_alias (__fmodf, __ieee754_fmodf)
#if LIBM_SVID_COMPAT
versioned_symbol (libm, __fmodf, fmodf, GLIBC_2_38);
libm_alias_float_other (__fmod, fmod)
#else
libm_alias_float (__fmod, fmod)
#endif
libm_alias_finite (__ieee754_fmodf, __fmodf)

View File

@ -158,6 +158,7 @@ attribute_hidden float __math_uflowf (uint32_t);
attribute_hidden float __math_may_uflowf (uint32_t);
attribute_hidden float __math_divzerof (uint32_t);
attribute_hidden float __math_invalidf (float);
attribute_hidden float __math_edomf (float x);
/* Shared between expf, exp2f, exp10f, and powf. */
#define EXP2F_TABLE_BITS 5

View File

@ -31,6 +31,12 @@ with_errnof (float y, int e)
# define with_errnof(x, e) (x)
#endif
attribute_hidden float
__math_edomf (float y)
{
return with_errnof (y, EDOM);
}
/* NOINLINE prevents fenv semantics breaking optimizations. */
NOINLINE static float
xflowf (uint32_t sign, float y)

View File

@ -0,0 +1 @@
/* Not needed. */

View File

@ -0,0 +1,14 @@
/* m68k provides an optimized __ieee752_fmod. */
#include <math-svid-compat.h>
#ifdef SHARED
# undef SHLIB_COMPAT
# define SHLIB_COMPAT(a, b, c) 1
# undef LIBM_SVID_COMPAT
# define LIBM_SVID_COMPAT 1
# undef compat_symbol
# define compat_symbol(a, b, c, d)
#endif
#include <math/w_fmod_compat.c>
#ifdef SHARED
libm_alias_double (__fmod_compat, fmod)
#endif

View File

@ -0,0 +1,14 @@
/* m68k provides an optimized __ieee752_fmodf. */
#include <math-svid-compat.h>
#ifdef SHARED
# undef SHLIB_COMPAT
# define SHLIB_COMPAT(a, b, c) 1
# undef LIBM_SVID_COMPAT
# define LIBM_SVID_COMPAT 1
# undef compat_symbol
# define compat_symbol(a, b, c, d)
#endif
#include <math/w_fmodf_compat.c>
#ifdef SHARED
libm_alias_float (__fmod_compat, fmod)
#endif

View File

@ -1146,3 +1146,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -1203,6 +1203,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -533,6 +533,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 _LIB_VERSION D 0x4
GLIBC_2.4 __clog10 F
GLIBC_2.4 __clog10f F

View File

@ -533,6 +533,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 _LIB_VERSION D 0x4
GLIBC_2.4 __clog10 F
GLIBC_2.4 __clog10f F

View File

@ -844,4 +844,6 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 exp2l F

View File

@ -533,6 +533,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 _LIB_VERSION D 0x4
GLIBC_2.4 __clog10 F
GLIBC_2.4 __clog10f F

View File

@ -845,3 +845,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -845,3 +845,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -844,4 +844,6 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 exp2l F

View File

@ -1146,3 +1146,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -845,3 +845,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -890,6 +890,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -889,6 +889,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -883,6 +883,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -1318,3 +1318,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -1147,6 +1147,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -1147,6 +1147,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -844,4 +844,6 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 exp2l F

View File

@ -844,4 +844,6 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 exp2l F

View File

@ -1154,6 +1154,8 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F
GLIBC_2.4 __clog10l F
GLIBC_2.4 __finitel F
GLIBC_2.4 __fpclassifyl F

View File

@ -1146,3 +1146,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -1179,3 +1179,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F

View File

@ -1179,3 +1179,5 @@ GLIBC_2.35 fsqrt F
GLIBC_2.35 fsqrtl F
GLIBC_2.35 hypot F
GLIBC_2.35 hypotf F
GLIBC_2.38 fmod F
GLIBC_2.38 fmodf F