2000-12-27 20:58:36 +01:00
|
|
|
#ifndef _FENV_H
|
1998-08-03 18:47:01 +02:00
|
|
|
#include <math/fenv.h>
|
Set/restore rounding mode only when needed
The most common use case of math functions is with default rounding
mode, i.e. rounding to nearest. Setting and restoring rounding mode
is an unnecessary overhead for this, so I've added support for a
context, which does the set/restore only if the FP status needs a
change. The code is written such that only x86 uses these. Other
architectures should be unaffected by it, but would definitely benefit
if the set/restore has as much overhead relative to the rest of the
code, as the x86 bits do.
Here's a summary of the performance improvement due to these
improvements; I've only mentioned functions that use the set/restore
and have benchmark inputs for x86_64:
Before:
cos(): ITERS:4.69335e+08: TOTAL:28884.6Mcy, MAX:4080.28cy, MIN:57.562cy, 16248.6 calls/Mcy
exp(): ITERS:4.47604e+08: TOTAL:28796.2Mcy, MAX:207.721cy, MIN:62.385cy, 15543.9 calls/Mcy
pow(): ITERS:1.63485e+08: TOTAL:28879.9Mcy, MAX:362.255cy, MIN:172.469cy, 5660.86 calls/Mcy
sin(): ITERS:3.89578e+08: TOTAL:28900Mcy, MAX:704.859cy, MIN:47.583cy, 13480.2 calls/Mcy
tan(): ITERS:7.0971e+07: TOTAL:28902.2Mcy, MAX:1357.79cy, MIN:388.58cy, 2455.55 calls/Mcy
After:
cos(): ITERS:6.0014e+08: TOTAL:28875.9Mcy, MAX:364.283cy, MIN:45.716cy, 20783.4 calls/Mcy
exp(): ITERS:5.48578e+08: TOTAL:28764.9Mcy, MAX:191.617cy, MIN:51.011cy, 19071.1 calls/Mcy
pow(): ITERS:1.70013e+08: TOTAL:28873.6Mcy, MAX:689.522cy, MIN:163.989cy, 5888.18 calls/Mcy
sin(): ITERS:4.64079e+08: TOTAL:28891.5Mcy, MAX:6959.3cy, MIN:36.189cy, 16062.8 calls/Mcy
tan(): ITERS:7.2354e+07: TOTAL:28898.9Mcy, MAX:1295.57cy, MIN:380.698cy, 2503.7 calls/Mcy
So the improvements are:
cos: 27.9089%
exp: 22.6919%
pow: 4.01564%
sin: 19.1585%
tan: 1.96086%
The downside of the change is that it will have an adverse performance
impact on non-default rounding modes, but I think the tradeoff is
justified.
2013-06-12 07:06:48 +02:00
|
|
|
#include <stdbool.h>
|
2000-12-27 20:58:36 +01:00
|
|
|
|
2012-11-03 00:21:36 +01:00
|
|
|
#ifndef _ISOMAC
|
2000-12-27 20:58:36 +01:00
|
|
|
/* Now define the internal interfaces. */
|
|
|
|
|
|
|
|
extern int __feclearexcept (int __excepts);
|
2008-04-12 02:51:34 +02:00
|
|
|
extern int __fegetexcept (void);
|
2000-12-27 20:58:36 +01:00
|
|
|
extern int __fegetexceptflag (fexcept_t *__flagp, int __excepts);
|
|
|
|
extern int __feraiseexcept (int __excepts);
|
2012-01-08 05:57:22 +01:00
|
|
|
extern int __fesetexceptflag (const fexcept_t *__flagp, int __excepts);
|
2000-12-27 20:58:36 +01:00
|
|
|
extern int __fegetenv (fenv_t *__envp);
|
2012-01-08 05:57:22 +01:00
|
|
|
extern int __fesetenv (const fenv_t *__envp);
|
|
|
|
extern int __feupdateenv (const fenv_t *__envp);
|
2000-12-27 20:58:36 +01:00
|
|
|
|
Update.
2002-09-09 Jakub Jelinek <jakub@redhat.com>
* include/math.h (__finite_internal, __finitef_internal,
__finitel_internal, __isinf_internal, __isnan_internal): Remove.
(isfinite): Remove.
(__finite, __isinf, __isnan, __finitef, __isinff, __isnanf, __finitel,
__isinfl, __isnanl): Add hidden_proto.
(__fpclassify, __fpclassifyf, __fpclassifyl, __expm1l): Add
libm_hidden_proto.
* math/Makefile (libm-calls): Add s_isinf and s_isnan.
* stdio-common/printf_fp.c (__printf_fp): Remove INTUSE from
__is{inf,nan} calls.
* stdio-common/printf_size.c (printf_size): Likewise.
* sysdeps/generic/printf_fphex.c (__printf_fphex): Likewise.
* sysdeps/generic/s_ldexp.c (__ldexp): Likewise.
* sysdeps/generic/s_ldexpf.c (__ldexpf): Likewise.
* sysdeps/generic/s_ldexpl.c (__ldexpl): Likewise.
* sysdeps/generic/s_expm1l.c (__expm1l): Add libm_hidden_def.
* sysdeps/i386/fpu/s_finite.S (__finite_internal): Remove alias.
(__finite): Add hidden_def.
* sysdeps/i386/fpu/s_finitef.S (__finitef_internal): Remove alias.
(__finitef): Add hidden_def.
* sysdeps/i386/fpu/s_finitel.S (__finitel_internal): Remove alias.
(__finitel): Add hidden_def.
* sysdeps/i386/fpu/s_isinfl.c (__isinfl): Remove INTDEF. Add
hidden_def.
* sysdeps/i386/fpu/s_isnanl.c (__isnanl): Likewise.
* sysdeps/i386/fpu/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/i386/fpu/s_expm1l.S (__expm1l): Likewise.
* sysdeps/ieee754/dbl-64/s_finite.c (__finite): Remove INTDEF. Add
hidden_def.
* sysdeps/ieee754/dbl-64/s_isinf.c (__isinf): Likewise.
(__isinfl): Remove INTDEF.
* sysdeps/ieee754/dbl-64/s_isnan.c (__isnan): Remove INTDEF. Add
hidden_def.
(__isnanl): Remove INTDEF.
* sysdeps/ieee754/dbl-64/s_fpclassify.c (__fpclassify): Add
libm_hidden_def.
* sysdeps/ieee754/dbl-64/e_lgamma_r.c (sin_pi): Use __sin and __cos
instead of sin and cos.
* sysdeps/ieee754/flt-32/s_finitef.c (__finitef): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/flt-32/s_isinff.c (__isinff): Likewise.
* sysdeps/ieee754/flt-32/s_isnanf.c (__isnanf): Likewise.
* sysdeps/ieee754/flt-32/s_fpclassifyf.c (__fpclassifyf): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-128/s_finitel.c (__finitel): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/ldbl-128/s_isinfl.c (__isinfl): Likewise.
* sysdeps/ieee754/ldbl-128/s_isnanl.c (__isnanl): Likewise.
* sysdeps/ieee754/ldbl-128/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-128/s_expm1l.c (__expm1l): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-96/s_finitel.c (__finitel): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/ldbl-96/s_isinfl.c (__isinfl): Likewise.
* sysdeps/ieee754/ldbl-96/s_isnanl.c (__isnanl): Likewise.
* sysdeps/ieee754/ldbl-96/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/ia64/fpu/s_finite.S (__finite_internal, __finitef_internal,
__finitel_internal): Remove aliases.
(__finite, __finitef, __finitel): Add hidden_def.
* sysdeps/ia64/fpu/s_isnan.S (__isnan_internal, __isnanf_internal,
__isnanl_internal): Remove aliases.
(__isnan, __isnanf, __isnanl): Add hidden_def.
* sysdeps/ia64/fpu/s_isinf.S (__isinf_internal, __isinff_internal,
__isinfl_internal): Remove aliases.
(__isinf, __isinff, __isinfl): Add hidden_def.
* sysdeps/ia64/fpu/s_fpclassify.S (__fpclassify, __fpclassifyf,
__fpclassifyl): Add libm_hidden_def.
* sysdeps/ia64/fpu/s_expm1l.S (__expm1l): Likewise.
* sysdeps/m68k/s_isinfl.c (__isinfl): Remove INTDEF. Add hidden_def.
* sysdeps/m68k/fpu/s_isinf.c (INTDEFX): Remove.
(hidden_defx): Define and use.
* sysdeps/m68k/fpu/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/m68k/fpu/s_expm1l.c (__expm1l): Likewise.
* sysdeps/m68k/s_isnanl.c (__isnanl): Add hidden_def.
* sysdeps/powerpc/fpu/s_isnan.c (__isnan, __isnanf, __isnanl):
Remove INTDEF.
(__isnan, __isnanf): Add hidden_def.
* sysdeps/x86_64/fpu/s_finitel.S (__finitel_internal): Remove alias.
(__finitel): Add libm_hidden_def.
* sysdeps/x86_64/fpu/s_expm1l.S (__expm1l): Likewise.
* include/fenv.h (feraiseexcept, fesetenv): Add libm_hidden_proto.
* sysdeps/alpha/fpu/fesetenv.c (fesetenv): Add libm_hidden_ver.
* sysdeps/alpha/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/arm/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/arm/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/generic/fesetenv.c (fesetenv): Likewise.
* sysdeps/generic/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/i386/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/i386/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/m68k/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/m68k/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/mips/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/mips/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/powerpc/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/powerpc/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/sparc/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/sparc/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/hppa/fpu/fesetenv.c (fesetenv): Add libm_hidden_def.
* sysdeps/hppa/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/ia64/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/ia64/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/sh/sh4/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/sh/sh4/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/s390/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/s390/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/x86_64/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/x86_64/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
2002-09-10 03:40:26 +02:00
|
|
|
libm_hidden_proto (feraiseexcept)
|
2010-02-10 07:34:17 +01:00
|
|
|
libm_hidden_proto (fegetenv)
|
Update.
2002-09-09 Jakub Jelinek <jakub@redhat.com>
* include/math.h (__finite_internal, __finitef_internal,
__finitel_internal, __isinf_internal, __isnan_internal): Remove.
(isfinite): Remove.
(__finite, __isinf, __isnan, __finitef, __isinff, __isnanf, __finitel,
__isinfl, __isnanl): Add hidden_proto.
(__fpclassify, __fpclassifyf, __fpclassifyl, __expm1l): Add
libm_hidden_proto.
* math/Makefile (libm-calls): Add s_isinf and s_isnan.
* stdio-common/printf_fp.c (__printf_fp): Remove INTUSE from
__is{inf,nan} calls.
* stdio-common/printf_size.c (printf_size): Likewise.
* sysdeps/generic/printf_fphex.c (__printf_fphex): Likewise.
* sysdeps/generic/s_ldexp.c (__ldexp): Likewise.
* sysdeps/generic/s_ldexpf.c (__ldexpf): Likewise.
* sysdeps/generic/s_ldexpl.c (__ldexpl): Likewise.
* sysdeps/generic/s_expm1l.c (__expm1l): Add libm_hidden_def.
* sysdeps/i386/fpu/s_finite.S (__finite_internal): Remove alias.
(__finite): Add hidden_def.
* sysdeps/i386/fpu/s_finitef.S (__finitef_internal): Remove alias.
(__finitef): Add hidden_def.
* sysdeps/i386/fpu/s_finitel.S (__finitel_internal): Remove alias.
(__finitel): Add hidden_def.
* sysdeps/i386/fpu/s_isinfl.c (__isinfl): Remove INTDEF. Add
hidden_def.
* sysdeps/i386/fpu/s_isnanl.c (__isnanl): Likewise.
* sysdeps/i386/fpu/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/i386/fpu/s_expm1l.S (__expm1l): Likewise.
* sysdeps/ieee754/dbl-64/s_finite.c (__finite): Remove INTDEF. Add
hidden_def.
* sysdeps/ieee754/dbl-64/s_isinf.c (__isinf): Likewise.
(__isinfl): Remove INTDEF.
* sysdeps/ieee754/dbl-64/s_isnan.c (__isnan): Remove INTDEF. Add
hidden_def.
(__isnanl): Remove INTDEF.
* sysdeps/ieee754/dbl-64/s_fpclassify.c (__fpclassify): Add
libm_hidden_def.
* sysdeps/ieee754/dbl-64/e_lgamma_r.c (sin_pi): Use __sin and __cos
instead of sin and cos.
* sysdeps/ieee754/flt-32/s_finitef.c (__finitef): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/flt-32/s_isinff.c (__isinff): Likewise.
* sysdeps/ieee754/flt-32/s_isnanf.c (__isnanf): Likewise.
* sysdeps/ieee754/flt-32/s_fpclassifyf.c (__fpclassifyf): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-128/s_finitel.c (__finitel): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/ldbl-128/s_isinfl.c (__isinfl): Likewise.
* sysdeps/ieee754/ldbl-128/s_isnanl.c (__isnanl): Likewise.
* sysdeps/ieee754/ldbl-128/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-128/s_expm1l.c (__expm1l): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-96/s_finitel.c (__finitel): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/ldbl-96/s_isinfl.c (__isinfl): Likewise.
* sysdeps/ieee754/ldbl-96/s_isnanl.c (__isnanl): Likewise.
* sysdeps/ieee754/ldbl-96/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/ia64/fpu/s_finite.S (__finite_internal, __finitef_internal,
__finitel_internal): Remove aliases.
(__finite, __finitef, __finitel): Add hidden_def.
* sysdeps/ia64/fpu/s_isnan.S (__isnan_internal, __isnanf_internal,
__isnanl_internal): Remove aliases.
(__isnan, __isnanf, __isnanl): Add hidden_def.
* sysdeps/ia64/fpu/s_isinf.S (__isinf_internal, __isinff_internal,
__isinfl_internal): Remove aliases.
(__isinf, __isinff, __isinfl): Add hidden_def.
* sysdeps/ia64/fpu/s_fpclassify.S (__fpclassify, __fpclassifyf,
__fpclassifyl): Add libm_hidden_def.
* sysdeps/ia64/fpu/s_expm1l.S (__expm1l): Likewise.
* sysdeps/m68k/s_isinfl.c (__isinfl): Remove INTDEF. Add hidden_def.
* sysdeps/m68k/fpu/s_isinf.c (INTDEFX): Remove.
(hidden_defx): Define and use.
* sysdeps/m68k/fpu/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/m68k/fpu/s_expm1l.c (__expm1l): Likewise.
* sysdeps/m68k/s_isnanl.c (__isnanl): Add hidden_def.
* sysdeps/powerpc/fpu/s_isnan.c (__isnan, __isnanf, __isnanl):
Remove INTDEF.
(__isnan, __isnanf): Add hidden_def.
* sysdeps/x86_64/fpu/s_finitel.S (__finitel_internal): Remove alias.
(__finitel): Add libm_hidden_def.
* sysdeps/x86_64/fpu/s_expm1l.S (__expm1l): Likewise.
* include/fenv.h (feraiseexcept, fesetenv): Add libm_hidden_proto.
* sysdeps/alpha/fpu/fesetenv.c (fesetenv): Add libm_hidden_ver.
* sysdeps/alpha/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/arm/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/arm/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/generic/fesetenv.c (fesetenv): Likewise.
* sysdeps/generic/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/i386/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/i386/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/m68k/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/m68k/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/mips/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/mips/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/powerpc/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/powerpc/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/sparc/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/sparc/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/hppa/fpu/fesetenv.c (fesetenv): Add libm_hidden_def.
* sysdeps/hppa/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/ia64/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/ia64/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/sh/sh4/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/sh/sh4/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/s390/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/s390/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/x86_64/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/x86_64/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
2002-09-10 03:40:26 +02:00
|
|
|
libm_hidden_proto (fesetenv)
|
2005-07-08 20:54:49 +02:00
|
|
|
libm_hidden_proto (fesetround)
|
|
|
|
libm_hidden_proto (feholdexcept)
|
2010-10-11 15:27:05 +02:00
|
|
|
libm_hidden_proto (feupdateenv)
|
|
|
|
libm_hidden_proto (fetestexcept)
|
2012-11-03 20:48:53 +01:00
|
|
|
libm_hidden_proto (feclearexcept)
|
2012-11-03 00:21:36 +01:00
|
|
|
#endif
|
Update.
2002-09-09 Jakub Jelinek <jakub@redhat.com>
* include/math.h (__finite_internal, __finitef_internal,
__finitel_internal, __isinf_internal, __isnan_internal): Remove.
(isfinite): Remove.
(__finite, __isinf, __isnan, __finitef, __isinff, __isnanf, __finitel,
__isinfl, __isnanl): Add hidden_proto.
(__fpclassify, __fpclassifyf, __fpclassifyl, __expm1l): Add
libm_hidden_proto.
* math/Makefile (libm-calls): Add s_isinf and s_isnan.
* stdio-common/printf_fp.c (__printf_fp): Remove INTUSE from
__is{inf,nan} calls.
* stdio-common/printf_size.c (printf_size): Likewise.
* sysdeps/generic/printf_fphex.c (__printf_fphex): Likewise.
* sysdeps/generic/s_ldexp.c (__ldexp): Likewise.
* sysdeps/generic/s_ldexpf.c (__ldexpf): Likewise.
* sysdeps/generic/s_ldexpl.c (__ldexpl): Likewise.
* sysdeps/generic/s_expm1l.c (__expm1l): Add libm_hidden_def.
* sysdeps/i386/fpu/s_finite.S (__finite_internal): Remove alias.
(__finite): Add hidden_def.
* sysdeps/i386/fpu/s_finitef.S (__finitef_internal): Remove alias.
(__finitef): Add hidden_def.
* sysdeps/i386/fpu/s_finitel.S (__finitel_internal): Remove alias.
(__finitel): Add hidden_def.
* sysdeps/i386/fpu/s_isinfl.c (__isinfl): Remove INTDEF. Add
hidden_def.
* sysdeps/i386/fpu/s_isnanl.c (__isnanl): Likewise.
* sysdeps/i386/fpu/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/i386/fpu/s_expm1l.S (__expm1l): Likewise.
* sysdeps/ieee754/dbl-64/s_finite.c (__finite): Remove INTDEF. Add
hidden_def.
* sysdeps/ieee754/dbl-64/s_isinf.c (__isinf): Likewise.
(__isinfl): Remove INTDEF.
* sysdeps/ieee754/dbl-64/s_isnan.c (__isnan): Remove INTDEF. Add
hidden_def.
(__isnanl): Remove INTDEF.
* sysdeps/ieee754/dbl-64/s_fpclassify.c (__fpclassify): Add
libm_hidden_def.
* sysdeps/ieee754/dbl-64/e_lgamma_r.c (sin_pi): Use __sin and __cos
instead of sin and cos.
* sysdeps/ieee754/flt-32/s_finitef.c (__finitef): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/flt-32/s_isinff.c (__isinff): Likewise.
* sysdeps/ieee754/flt-32/s_isnanf.c (__isnanf): Likewise.
* sysdeps/ieee754/flt-32/s_fpclassifyf.c (__fpclassifyf): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-128/s_finitel.c (__finitel): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/ldbl-128/s_isinfl.c (__isinfl): Likewise.
* sysdeps/ieee754/ldbl-128/s_isnanl.c (__isnanl): Likewise.
* sysdeps/ieee754/ldbl-128/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-128/s_expm1l.c (__expm1l): Add
libm_hidden_def.
* sysdeps/ieee754/ldbl-96/s_finitel.c (__finitel): Remove INTDEF.
Add hidden_def.
* sysdeps/ieee754/ldbl-96/s_isinfl.c (__isinfl): Likewise.
* sysdeps/ieee754/ldbl-96/s_isnanl.c (__isnanl): Likewise.
* sysdeps/ieee754/ldbl-96/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/ia64/fpu/s_finite.S (__finite_internal, __finitef_internal,
__finitel_internal): Remove aliases.
(__finite, __finitef, __finitel): Add hidden_def.
* sysdeps/ia64/fpu/s_isnan.S (__isnan_internal, __isnanf_internal,
__isnanl_internal): Remove aliases.
(__isnan, __isnanf, __isnanl): Add hidden_def.
* sysdeps/ia64/fpu/s_isinf.S (__isinf_internal, __isinff_internal,
__isinfl_internal): Remove aliases.
(__isinf, __isinff, __isinfl): Add hidden_def.
* sysdeps/ia64/fpu/s_fpclassify.S (__fpclassify, __fpclassifyf,
__fpclassifyl): Add libm_hidden_def.
* sysdeps/ia64/fpu/s_expm1l.S (__expm1l): Likewise.
* sysdeps/m68k/s_isinfl.c (__isinfl): Remove INTDEF. Add hidden_def.
* sysdeps/m68k/fpu/s_isinf.c (INTDEFX): Remove.
(hidden_defx): Define and use.
* sysdeps/m68k/fpu/s_fpclassifyl.c (__fpclassifyl): Add
libm_hidden_def.
* sysdeps/m68k/fpu/s_expm1l.c (__expm1l): Likewise.
* sysdeps/m68k/s_isnanl.c (__isnanl): Add hidden_def.
* sysdeps/powerpc/fpu/s_isnan.c (__isnan, __isnanf, __isnanl):
Remove INTDEF.
(__isnan, __isnanf): Add hidden_def.
* sysdeps/x86_64/fpu/s_finitel.S (__finitel_internal): Remove alias.
(__finitel): Add libm_hidden_def.
* sysdeps/x86_64/fpu/s_expm1l.S (__expm1l): Likewise.
* include/fenv.h (feraiseexcept, fesetenv): Add libm_hidden_proto.
* sysdeps/alpha/fpu/fesetenv.c (fesetenv): Add libm_hidden_ver.
* sysdeps/alpha/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/arm/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/arm/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/generic/fesetenv.c (fesetenv): Likewise.
* sysdeps/generic/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/i386/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/i386/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/m68k/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/m68k/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/mips/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/mips/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/powerpc/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/powerpc/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/sparc/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/sparc/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/hppa/fpu/fesetenv.c (fesetenv): Add libm_hidden_def.
* sysdeps/hppa/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/ia64/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/ia64/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/sh/sh4/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/sh/sh4/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/s390/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/s390/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
* sysdeps/x86_64/fpu/fesetenv.c (fesetenv): Likewise.
* sysdeps/x86_64/fpu/fraiseexcpt.c (feraiseexcept): Likewise.
2002-09-10 03:40:26 +02:00
|
|
|
|
Set/restore rounding mode only when needed
The most common use case of math functions is with default rounding
mode, i.e. rounding to nearest. Setting and restoring rounding mode
is an unnecessary overhead for this, so I've added support for a
context, which does the set/restore only if the FP status needs a
change. The code is written such that only x86 uses these. Other
architectures should be unaffected by it, but would definitely benefit
if the set/restore has as much overhead relative to the rest of the
code, as the x86 bits do.
Here's a summary of the performance improvement due to these
improvements; I've only mentioned functions that use the set/restore
and have benchmark inputs for x86_64:
Before:
cos(): ITERS:4.69335e+08: TOTAL:28884.6Mcy, MAX:4080.28cy, MIN:57.562cy, 16248.6 calls/Mcy
exp(): ITERS:4.47604e+08: TOTAL:28796.2Mcy, MAX:207.721cy, MIN:62.385cy, 15543.9 calls/Mcy
pow(): ITERS:1.63485e+08: TOTAL:28879.9Mcy, MAX:362.255cy, MIN:172.469cy, 5660.86 calls/Mcy
sin(): ITERS:3.89578e+08: TOTAL:28900Mcy, MAX:704.859cy, MIN:47.583cy, 13480.2 calls/Mcy
tan(): ITERS:7.0971e+07: TOTAL:28902.2Mcy, MAX:1357.79cy, MIN:388.58cy, 2455.55 calls/Mcy
After:
cos(): ITERS:6.0014e+08: TOTAL:28875.9Mcy, MAX:364.283cy, MIN:45.716cy, 20783.4 calls/Mcy
exp(): ITERS:5.48578e+08: TOTAL:28764.9Mcy, MAX:191.617cy, MIN:51.011cy, 19071.1 calls/Mcy
pow(): ITERS:1.70013e+08: TOTAL:28873.6Mcy, MAX:689.522cy, MIN:163.989cy, 5888.18 calls/Mcy
sin(): ITERS:4.64079e+08: TOTAL:28891.5Mcy, MAX:6959.3cy, MIN:36.189cy, 16062.8 calls/Mcy
tan(): ITERS:7.2354e+07: TOTAL:28898.9Mcy, MAX:1295.57cy, MIN:380.698cy, 2503.7 calls/Mcy
So the improvements are:
cos: 27.9089%
exp: 22.6919%
pow: 4.01564%
sin: 19.1585%
tan: 1.96086%
The downside of the change is that it will have an adverse performance
impact on non-default rounding modes, but I think the tradeoff is
justified.
2013-06-12 07:06:48 +02:00
|
|
|
/* Rounding mode context. This allows functions to set/restore rounding mode
|
|
|
|
only when the desired rounding mode is different from the current rounding
|
|
|
|
mode. */
|
|
|
|
struct rm_ctx
|
|
|
|
{
|
|
|
|
fenv_t env;
|
|
|
|
bool updated_status;
|
|
|
|
};
|
|
|
|
|
2000-12-27 20:58:36 +01:00
|
|
|
#endif
|