[09/13] Use GCC builtins for floor functions if desired.
Commit Message
This patch is using the corresponding GCC builtin for floorf, floor,
floorl and floorf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins.h.
This is the case for s390 if build with at least --march=z196 --mzarch.
Otherwise the generic implementation is used. The code of the generic
implementation is not changed except changes in code style.
---
sysdeps/generic/math-use-builtins.h | 5 ++
sysdeps/ieee754/dbl-64/s_floor.c | 5 ++
sysdeps/ieee754/float128/float128_private.h | 3 +
sysdeps/ieee754/flt-32/s_floorf.c | 60 ++++++++-----
sysdeps/ieee754/ldbl-128/s_floorl.c | 97 ++++++++++++++-------
sysdeps/s390/fpu/math-use-builtins.h | 6 ++
6 files changed, 121 insertions(+), 55 deletions(-)
Comments
On 02/12/2019 11:46, Stefan Liebler wrote:
> This patch is using the corresponding GCC builtin for floorf, floor,
> floorl and floorf128 if the USE_FUNCTION_BUILTIN macros are defined to one
> in math-use-builtins.h.
>
> This is the case for s390 if build with at least --march=z196 --mzarch.
> Otherwise the generic implementation is used. The code of the generic
> implementation is not changed except changes in code style.
LGTM with some changes below.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
> ---
> sysdeps/generic/math-use-builtins.h | 5 ++
> sysdeps/ieee754/dbl-64/s_floor.c | 5 ++
> sysdeps/ieee754/float128/float128_private.h | 3 +
> sysdeps/ieee754/flt-32/s_floorf.c | 60 ++++++++-----
> sysdeps/ieee754/ldbl-128/s_floorl.c | 97 ++++++++++++++-------
> sysdeps/s390/fpu/math-use-builtins.h | 6 ++
> 6 files changed, 121 insertions(+), 55 deletions(-)
>
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index 64b4a4bb5b..e1c5df62e4 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -31,4 +31,9 @@
> #define USE_RINTL_BUILTIN 0
> #define USE_RINTF128_BUILTIN 0
>
> +#define USE_FLOOR_BUILTIN 0
> +#define USE_FLOORF_BUILTIN 0
> +#define USE_FLOORL_BUILTIN 0
> +#define USE_FLOORF128_BUILTIN 0
> +
> #endif /* math-use-builtins.h */
Ok.
> diff --git a/sysdeps/ieee754/dbl-64/s_floor.c b/sysdeps/ieee754/dbl-64/s_floor.c
> index ce0c42bdb6..0ed22795ed 100644
> --- a/sysdeps/ieee754/dbl-64/s_floor.c
> +++ b/sysdeps/ieee754/dbl-64/s_floor.c
> @@ -35,6 +35,7 @@
> #include <math_private.h>
> #include <stdint.h>
> #include <libm-alias-double.h>
> +#include <math-use-builtins.h>
>
> /*
> * floor(x)
> @@ -47,6 +48,9 @@
> double
> __floor (double x)
> {
> +#if USE_FLOOR_BUILTIN
> + return __builtin_floor (x);
> +#else
> int64_t i0;
> EXTRACT_WORDS64 (i0, x);
> int32_t j0 = ((i0 >> 52) & 0x7ff) - 0x3ff;
> @@ -74,6 +78,7 @@ __floor (double x)
> else if (j0 == 0x400)
> return x + x; /* inf or NaN */
> return x;
> +#endif /* USE_FLOOR_BUILTIN */
> }
> #ifndef __floor
> libm_alias_double (__floor, floor)
Ok.
> diff --git a/sysdeps/ieee754/float128/float128_private.h b/sysdeps/ieee754/float128/float128_private.h
> index f458e7b85f..8c8a74a12c 100644
> --- a/sysdeps/ieee754/float128/float128_private.h
> +++ b/sysdeps/ieee754/float128/float128_private.h
> @@ -144,6 +144,8 @@
> #define USE_NEARBYINTL_BUILTIN USE_NEARBYINTF128_BUILTIN
> #undef USE_RINTL_BUILTIN
> #define USE_RINTL_BUILTIN USE_RINTF128_BUILTIN
> +#undef USE_FLOORL_BUILTIN
> +#define USE_FLOORL_BUILTIN USE_FLOORF128_BUILTIN
>
> /* IEEE function renames. */
> #define __ieee754_acoshl __ieee754_acoshf128
> @@ -349,6 +351,7 @@
> #define __builtin_signbitl __builtin_signbit
> #define __builtin_nearbyintl __builtin_nearbyintf128
> #define __builtin_rintl __builtin_rintf128
> +#define __builtin_floorl __builtin_floorf128
>
> /* Get the constant suffix from bits/floatn-compat.h. */
> #define L(x) __f128 (x)
Ok.
> diff --git a/sysdeps/ieee754/flt-32/s_floorf.c b/sysdeps/ieee754/flt-32/s_floorf.c
> index b34d967f01..39bb570d10 100644
> --- a/sysdeps/ieee754/flt-32/s_floorf.c
> +++ b/sysdeps/ieee754/flt-32/s_floorf.c
> @@ -24,32 +24,48 @@
> #include <math.h>
> #include <math_private.h>
> #include <libm-alias-float.h>
> +#include <math-use-builtins.h>
>
> float
> -__floorf(float x)
> +__floorf (float x)
> {
> - int32_t i0,j0;
> - uint32_t i;
> - GET_FLOAT_WORD(i0,x);
> - j0 = ((i0>>23)&0xff)-0x7f;
> - if(j0<23) {
> - if(j0<0) {
> - /* return 0*sign(x) if |x|<1 */
> - if(i0>=0) {i0=0;}
> - else if((i0&0x7fffffff)!=0)
> - { i0=0xbf800000;}
> - } else {
> - i = (0x007fffff)>>j0;
> - if((i0&i)==0) return x; /* x is integral */
> - if(i0<0) i0 += (0x00800000)>>j0;
> - i0 &= (~i);
> - }
> - } else {
> - if(__builtin_expect(j0==0x80, 0)) return x+x; /* inf or NaN */
> - else return x; /* x is integral */
> +#if USE_FLOORF_BUILTIN
> + return __builtin_floorf (x);
> +#else
> + int32_t i0, j0;
> + uint32_t i;
> + GET_FLOAT_WORD (i0, x);
> + j0 = ((i0 >> 23) & 0xff) - 0x7f;
> + if (j0 < 23)
> + {
> + if (j0 < 0)
> + {
> + /* return 0 * sign (x) if |x| < 1 */
> + if (i0 >= 0)
> + i0 = 0;
> + else if ((i0 & 0x7fffffff) != 0)
> + i0 = 0xbf800000;
> }
> - SET_FLOAT_WORD(x,i0);
> - return x;
> + else
> + {
> + i = (0x007fffff) >> j0;
> + if ((i0 & i) == 0)
> + return x; /* x is integral */
> + if (i0 < 0)
> + i0 += (0x00800000) >> j0;
> + i0 &= (~i);
> + }
> + }
> + else
> + {
> + if (__glibc_unlikely (j0 == 0x80))
> + return x + x; /* inf or NaN */
> + else
> + return x; /* x is integral */
> + }
> + SET_FLOAT_WORD (x, i0);
> + return x;
> +#endif /* USE_FLOORF_BUILTIN */
> }
> #ifndef __floorf
> libm_alias_float (__floor, floor)
Ok, but fix the indentation in a separated patch.
> diff --git a/sysdeps/ieee754/ldbl-128/s_floorl.c b/sysdeps/ieee754/ldbl-128/s_floorl.c
> index f340a3fbca..8d20a79fec 100644
> --- a/sysdeps/ieee754/ldbl-128/s_floorl.c
> +++ b/sysdeps/ieee754/ldbl-128/s_floorl.c
> @@ -13,7 +13,9 @@
> * ====================================================
> */
>
> -#if defined(LIBM_SCCS) && !defined(lint)
> +#include <math-use-builtins.h>
> +
> +#if ! USE_FLOORL_BUILTIN && defined (LIBM_SCCS) && ! defined (lint)
> static char rcsid[] = "$NetBSD: $";
> #endif
>
> @@ -29,41 +31,70 @@ static char rcsid[] = "$NetBSD: $";
> #include <math_private.h>
> #include <libm-alias-ldouble.h>
>
> -_Float128 __floorl(_Float128 x)
> +_Float128
> +__floorl (_Float128 x)
> {
> - int64_t i0,i1,j0;
> - uint64_t i,j;
> - GET_LDOUBLE_WORDS64(i0,i1,x);
> - j0 = ((i0>>48)&0x7fff)-0x3fff;
> - if(j0<48) {
> - if(j0<0) {
> - /* return 0*sign(x) if |x|<1 */
> - if(i0>=0) {i0=i1=0;}
> - else if(((i0&0x7fffffffffffffffLL)|i1)!=0)
> - { i0=0xbfff000000000000ULL;i1=0;}
> - } else {
> - i = (0x0000ffffffffffffULL)>>j0;
> - if(((i0&i)|i1)==0) return x; /* x is integral */
> - if(i0<0) i0 += (0x0001000000000000LL)>>j0;
> - i0 &= (~i); i1=0;
> +#if USE_FLOORL_BUILTIN
> + return __builtin_floorl (x);
> +#else
> + int64_t i0, i1, j0;
> + uint64_t i, j;
> + GET_LDOUBLE_WORDS64 (i0, i1, x);
> + j0 = ((i0 >> 48) & 0x7fff) - 0x3fff;
> + if (j0 < 48)
> + {
> + if (j0 < 0)
> + {
> + /* return 0 * sign (x) if |x| < 1 */
> + if (i0 >= 0)
> + {
> + i0 = i1 = 0;
> + }
> + else if (((i0 & 0x7fffffffffffffffLL) | i1) != 0)
> + {
> + i0 = 0xbfff000000000000ULL;
> + i1 = 0;
> }
> - } else if (j0>111) {
> - if(j0==0x4000) return x+x; /* inf or NaN */
> - else return x; /* x is integral */
> - } else {
> - i = -1ULL>>(j0-48);
> - if((i1&i)==0) return x; /* x is integral */
> - if(i0<0) {
> - if(j0==48) i0+=1;
> - else {
> - j = i1+(1LL<<(112-j0));
> - if(j<i1) i0 +=1 ; /* got a carry */
> - i1=j;
> - }
> + }
> + else
> + {
> + i = (0x0000ffffffffffffULL) >> j0;
> + if (((i0 & i) | i1) == 0)
> + return x; /* x is integral */
> + if (i0 < 0)
> + i0 += (0x0001000000000000LL) >> j0;
> + i0 &= (~i);
> + i1 = 0;
> + }
> + }
> + else if (j0 > 111)
> + {
> + if (j0 == 0x4000)
> + return x + x; /* inf or NaN */
> + else
> + return x; /* x is integral */
> + }
> + else
> + {
> + i = -1ULL >> (j0 - 48);
> + if ((i1 & i) == 0)
> + return x; /* x is integral */
> + if (i0 < 0)
> + {
> + if (j0 == 48)
> + i0 += 1;
> + else
> + {
> + j = i1 + (1LL << (112 - j0));
> + if (j < i1)
> + i0 += 1 ; /* got a carry */
> + i1 = j;
> }
> - i1 &= (~i);
> }
> - SET_LDOUBLE_WORDS64(x,i0,i1);
> - return x;
> + i1 &= (~i);
> + }
> + SET_LDOUBLE_WORDS64 (x, i0, i1);
> + return x;
> +#endif /* USE_FLOORL_BUILTIN */
> }
> libm_alias_ldouble (__floor, floor)
Ok, but fix the indentation in a separated patch.
> diff --git a/sysdeps/s390/fpu/math-use-builtins.h b/sysdeps/s390/fpu/math-use-builtins.h
> index 15705fa74a..63cc9d7a23 100644
> --- a/sysdeps/s390/fpu/math-use-builtins.h
> +++ b/sysdeps/s390/fpu/math-use-builtins.h
> @@ -34,12 +34,18 @@
> # define USE_RINTF_BUILTIN 1
> # define USE_RINTL_BUILTIN 1
>
> +# define USE_FLOOR_BUILTIN 1
> +# define USE_FLOORF_BUILTIN 1
> +# define USE_FLOORL_BUILTIN 1
> +
> # if __GNUC_PREREQ (8, 1)
> # define USE_NEARBYINTF128_BUILTIN 1
> # define USE_RINTF128_BUILTIN 1
> +# define USE_FLOORF128_BUILTIN 1
> # else
> # define USE_NEARBYINTF128_BUILTIN 0
> # define USE_RINTF128_BUILTIN 0
> +# define USE_FLOORF128_BUILTIN 0
> # endif
>
> #else
>
Ok.
@@ -31,4 +31,9 @@
#define USE_RINTL_BUILTIN 0
#define USE_RINTF128_BUILTIN 0
+#define USE_FLOOR_BUILTIN 0
+#define USE_FLOORF_BUILTIN 0
+#define USE_FLOORL_BUILTIN 0
+#define USE_FLOORF128_BUILTIN 0
+
#endif /* math-use-builtins.h */
@@ -35,6 +35,7 @@
#include <math_private.h>
#include <stdint.h>
#include <libm-alias-double.h>
+#include <math-use-builtins.h>
/*
* floor(x)
@@ -47,6 +48,9 @@
double
__floor (double x)
{
+#if USE_FLOOR_BUILTIN
+ return __builtin_floor (x);
+#else
int64_t i0;
EXTRACT_WORDS64 (i0, x);
int32_t j0 = ((i0 >> 52) & 0x7ff) - 0x3ff;
@@ -74,6 +78,7 @@ __floor (double x)
else if (j0 == 0x400)
return x + x; /* inf or NaN */
return x;
+#endif /* USE_FLOOR_BUILTIN */
}
#ifndef __floor
libm_alias_double (__floor, floor)
@@ -144,6 +144,8 @@
#define USE_NEARBYINTL_BUILTIN USE_NEARBYINTF128_BUILTIN
#undef USE_RINTL_BUILTIN
#define USE_RINTL_BUILTIN USE_RINTF128_BUILTIN
+#undef USE_FLOORL_BUILTIN
+#define USE_FLOORL_BUILTIN USE_FLOORF128_BUILTIN
/* IEEE function renames. */
#define __ieee754_acoshl __ieee754_acoshf128
@@ -349,6 +351,7 @@
#define __builtin_signbitl __builtin_signbit
#define __builtin_nearbyintl __builtin_nearbyintf128
#define __builtin_rintl __builtin_rintf128
+#define __builtin_floorl __builtin_floorf128
/* Get the constant suffix from bits/floatn-compat.h. */
#define L(x) __f128 (x)
@@ -24,32 +24,48 @@
#include <math.h>
#include <math_private.h>
#include <libm-alias-float.h>
+#include <math-use-builtins.h>
float
-__floorf(float x)
+__floorf (float x)
{
- int32_t i0,j0;
- uint32_t i;
- GET_FLOAT_WORD(i0,x);
- j0 = ((i0>>23)&0xff)-0x7f;
- if(j0<23) {
- if(j0<0) {
- /* return 0*sign(x) if |x|<1 */
- if(i0>=0) {i0=0;}
- else if((i0&0x7fffffff)!=0)
- { i0=0xbf800000;}
- } else {
- i = (0x007fffff)>>j0;
- if((i0&i)==0) return x; /* x is integral */
- if(i0<0) i0 += (0x00800000)>>j0;
- i0 &= (~i);
- }
- } else {
- if(__builtin_expect(j0==0x80, 0)) return x+x; /* inf or NaN */
- else return x; /* x is integral */
+#if USE_FLOORF_BUILTIN
+ return __builtin_floorf (x);
+#else
+ int32_t i0, j0;
+ uint32_t i;
+ GET_FLOAT_WORD (i0, x);
+ j0 = ((i0 >> 23) & 0xff) - 0x7f;
+ if (j0 < 23)
+ {
+ if (j0 < 0)
+ {
+ /* return 0 * sign (x) if |x| < 1 */
+ if (i0 >= 0)
+ i0 = 0;
+ else if ((i0 & 0x7fffffff) != 0)
+ i0 = 0xbf800000;
}
- SET_FLOAT_WORD(x,i0);
- return x;
+ else
+ {
+ i = (0x007fffff) >> j0;
+ if ((i0 & i) == 0)
+ return x; /* x is integral */
+ if (i0 < 0)
+ i0 += (0x00800000) >> j0;
+ i0 &= (~i);
+ }
+ }
+ else
+ {
+ if (__glibc_unlikely (j0 == 0x80))
+ return x + x; /* inf or NaN */
+ else
+ return x; /* x is integral */
+ }
+ SET_FLOAT_WORD (x, i0);
+ return x;
+#endif /* USE_FLOORF_BUILTIN */
}
#ifndef __floorf
libm_alias_float (__floor, floor)
@@ -13,7 +13,9 @@
* ====================================================
*/
-#if defined(LIBM_SCCS) && !defined(lint)
+#include <math-use-builtins.h>
+
+#if ! USE_FLOORL_BUILTIN && defined (LIBM_SCCS) && ! defined (lint)
static char rcsid[] = "$NetBSD: $";
#endif
@@ -29,41 +31,70 @@ static char rcsid[] = "$NetBSD: $";
#include <math_private.h>
#include <libm-alias-ldouble.h>
-_Float128 __floorl(_Float128 x)
+_Float128
+__floorl (_Float128 x)
{
- int64_t i0,i1,j0;
- uint64_t i,j;
- GET_LDOUBLE_WORDS64(i0,i1,x);
- j0 = ((i0>>48)&0x7fff)-0x3fff;
- if(j0<48) {
- if(j0<0) {
- /* return 0*sign(x) if |x|<1 */
- if(i0>=0) {i0=i1=0;}
- else if(((i0&0x7fffffffffffffffLL)|i1)!=0)
- { i0=0xbfff000000000000ULL;i1=0;}
- } else {
- i = (0x0000ffffffffffffULL)>>j0;
- if(((i0&i)|i1)==0) return x; /* x is integral */
- if(i0<0) i0 += (0x0001000000000000LL)>>j0;
- i0 &= (~i); i1=0;
+#if USE_FLOORL_BUILTIN
+ return __builtin_floorl (x);
+#else
+ int64_t i0, i1, j0;
+ uint64_t i, j;
+ GET_LDOUBLE_WORDS64 (i0, i1, x);
+ j0 = ((i0 >> 48) & 0x7fff) - 0x3fff;
+ if (j0 < 48)
+ {
+ if (j0 < 0)
+ {
+ /* return 0 * sign (x) if |x| < 1 */
+ if (i0 >= 0)
+ {
+ i0 = i1 = 0;
+ }
+ else if (((i0 & 0x7fffffffffffffffLL) | i1) != 0)
+ {
+ i0 = 0xbfff000000000000ULL;
+ i1 = 0;
}
- } else if (j0>111) {
- if(j0==0x4000) return x+x; /* inf or NaN */
- else return x; /* x is integral */
- } else {
- i = -1ULL>>(j0-48);
- if((i1&i)==0) return x; /* x is integral */
- if(i0<0) {
- if(j0==48) i0+=1;
- else {
- j = i1+(1LL<<(112-j0));
- if(j<i1) i0 +=1 ; /* got a carry */
- i1=j;
- }
+ }
+ else
+ {
+ i = (0x0000ffffffffffffULL) >> j0;
+ if (((i0 & i) | i1) == 0)
+ return x; /* x is integral */
+ if (i0 < 0)
+ i0 += (0x0001000000000000LL) >> j0;
+ i0 &= (~i);
+ i1 = 0;
+ }
+ }
+ else if (j0 > 111)
+ {
+ if (j0 == 0x4000)
+ return x + x; /* inf or NaN */
+ else
+ return x; /* x is integral */
+ }
+ else
+ {
+ i = -1ULL >> (j0 - 48);
+ if ((i1 & i) == 0)
+ return x; /* x is integral */
+ if (i0 < 0)
+ {
+ if (j0 == 48)
+ i0 += 1;
+ else
+ {
+ j = i1 + (1LL << (112 - j0));
+ if (j < i1)
+ i0 += 1 ; /* got a carry */
+ i1 = j;
}
- i1 &= (~i);
}
- SET_LDOUBLE_WORDS64(x,i0,i1);
- return x;
+ i1 &= (~i);
+ }
+ SET_LDOUBLE_WORDS64 (x, i0, i1);
+ return x;
+#endif /* USE_FLOORL_BUILTIN */
}
libm_alias_ldouble (__floor, floor)
@@ -34,12 +34,18 @@
# define USE_RINTF_BUILTIN 1
# define USE_RINTL_BUILTIN 1
+# define USE_FLOOR_BUILTIN 1
+# define USE_FLOORF_BUILTIN 1
+# define USE_FLOORL_BUILTIN 1
+
# if __GNUC_PREREQ (8, 1)
# define USE_NEARBYINTF128_BUILTIN 1
# define USE_RINTF128_BUILTIN 1
+# define USE_FLOORF128_BUILTIN 1
# else
# define USE_NEARBYINTF128_BUILTIN 0
# define USE_RINTF128_BUILTIN 0
+# define USE_FLOORF128_BUILTIN 0
# endif
#else