This is the mail archive of the libc-hacker@sourceware.org mailing list for the glibc project.
Note that libc-hacker is a closed list. You may look at the archives of this list, but subscription and posting are not open.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
Hi! Any reason why fsqrt{,s} use is conditional even in 64-bit glibc? sysdeps/powerpc/fpu/e_sqrt{,f}.c guards the insn use with: /* If the CPU is 64-bit we can use the optional FP instructions we. */ if ((GLRO (dl_hwcap) & PPC_FEATURE_64) != 0) and otherwise falls back to __slow_ieee754_sqrt{,f}. But in 64-bit code, I'd say we can always assume the CPU is 64-bit and PPC_FEATURE_64 is set and get rid of an unneeded runtime memory load, masking and more importantly quite large fallback routine. 2006-01-20 Jakub Jelinek <jakub@redhat.com> * sysdeps/powerpc/powerpc64/fpu/e_sqrt.c: New file. * sysdeps/powerpc/powerpc64/fpu/e_sqrtf.c: New file. * sysdeps/powerpc/fpu/bits/mathinline.h (__CPU_HAS_FSQRT): Define. (__ieee754_sqrt, __ieee754_sqrtf): Use it. --- libc/sysdeps/powerpc/powerpc64/fpu/e_sqrt.c.jj 2006-01-20 16:36:49.000000000 +0100 +++ libc/sysdeps/powerpc/powerpc64/fpu/e_sqrt.c 2006-01-20 16:39:05.000000000 +0100 @@ -0,0 +1,29 @@ +/* Double-precision floating point square root. + Copyright (C) 1997, 2002, 2003, 2004, 2006 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + 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, write to the Free + Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA + 02111-1307 USA. */ + +#include <math.h> +#include <math_private.h> + +double +__ieee754_sqrt (double x) +{ + double z; + __asm __volatile ("fsqrt %0,%1" : "=f" (z) : "f" (x)); + return z; +} --- libc/sysdeps/powerpc/powerpc64/fpu/e_sqrtf.c.jj 2006-01-20 16:36:52.000000000 +0100 +++ libc/sysdeps/powerpc/powerpc64/fpu/e_sqrtf.c 2006-01-20 16:37:55.000000000 +0100 @@ -0,0 +1,29 @@ +/* Single-precision floating point square root. + Copyright (C) 1997, 2003, 2004, 2006 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + 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, write to the Free + Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA + 02111-1307 USA. */ + +#include <math.h> +#include <math_private.h> + +float +__ieee754_sqrtf (float x) +{ + double z; + __asm ("fsqrts %0,%1" : "=f" (z) : "f" (x)); + return z; +} --- libc/sysdeps/powerpc/fpu/bits/mathinline.h.jj 2004-10-01 12:05:04.000000000 +0200 +++ libc/sysdeps/powerpc/fpu/bits/mathinline.h 2006-01-20 16:45:16.000000000 +0100 @@ -1,5 +1,5 @@ /* Inline math functions for powerpc. - Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2004 + Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2004, 2006 Free Software Foundation, Inc. This file is part of the GNU C Library. @@ -123,12 +123,18 @@ __NTH (fdimf (float __x, float __y)) #endif /* !__NO_MATH_INLINES && __OPTIMIZE__ */ /* This code is used internally in the GNU libc. */ -# ifdef __LIBC_INTERNAL_MATH_INLINES +#ifdef __LIBC_INTERNAL_MATH_INLINES #include <sysdep.h> #include <ldsodefs.h> #include <dl-procinfo.h> +# if __WORDSIZE == 64 +# define __CPU_HAS_FSQRT 1 +# else +# define __CPU_HAS_FSQRT ((GLRO(dl_hwcap) & PPC_FEATURE_64) != 0) +# endif + extern double __slow_ieee754_sqrt (double); __MATH_INLINE double __NTH (__ieee754_sqrt (double __x)) @@ -136,7 +142,7 @@ __NTH (__ieee754_sqrt (double __x)) double __z; /* If the CPU is 64-bit we can use the optional FP instructions we. */ - if ((GLRO(dl_hwcap) & PPC_FEATURE_64) != 0) + if (__CPU_HAS_FSQRT) { /* Volatile is required to prevent the compiler from moving the fsqrt instruction above the branch. */ @@ -158,7 +164,7 @@ __NTH (__ieee754_sqrtf (float __x)) float __z; /* If the CPU is 64-bit we can use the optional FP instructions we. */ - if ((GLRO(dl_hwcap) & PPC_FEATURE_64) != 0) + if (__CPU_HAS_FSQRT) { /* Volatile is required to prevent the compiler from moving the fsqrts instruction above the branch. */ @@ -172,5 +178,5 @@ __NTH (__ieee754_sqrtf (float __x)) return __z; } -# endif /* __LIBC_INTERNAL_MATH_INLINES */ +#endif /* __LIBC_INTERNAL_MATH_INLINES */ #endif /* __GNUC__ && !_SOFT_FLOAT */ Jakub
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |