Fix fsqrt build in libm
Posted: Wed Feb 20, 2019 9:49 am
FYI:
sysdeps/powerpc/fpu/e_sqrtf.c:
sysdeps/powerpc/fpu/math_private.h:
Source: sourceware.org
sysdeps/powerpc/fpu/e_sqrt.c:Adhemerval Zanella wrote: powerpc: Fix fsqrt build in libm [BZ#16576]
Some powerpc64 processors (e5500 core for instance) does not provide the
fsqrt instruction, however current check to use in math_private.h is
__WORDSIZE and _ARCH_PWR4 (ISA 2.02). This is patch change it to use
the compiler flag _ARCH_PPCSQ (which is the same condition GCC uses to
decide whether to generate fsqrt instruction).
It fixes BZ#16576.
Code: Select all
--- a/sysdeps/powerpc/fpu/e_sqrt.c
+++ b/sysdeps/powerpc/fpu/e_sqrt.c
@@ -24,6 +24,7 @@
#include <sysdep.h>
#include <ldsodefs.h>
+#ifndef _ARCH_PPCSQ
static const double almost_half = 0.5000000000000001; /* 0.5 + 2^-53 */
static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 };
static const ieee_float_shape_type a_inf = {.word = 0x7f800000 };
@@ -152,6 +153,7 @@ __slow_ieee754_sqrt (double x)
}
return f_wash (x);
}
+#endif /* _ARCH_PPCSQ */
#undef __ieee754_sqrt
double
@@ -159,16 +161,11 @@ __ieee754_sqrt (double x)
{
double z;
- /* If the CPU is 64-bit we can use the optional FP instructions. */
- if (__CPU_HAS_FSQRT)
- {
- /* Volatile is required to prevent the compiler from moving the
- fsqrt instruction above the branch. */
- __asm __volatile (" fsqrt %0,%1\n"
- :"=f" (z):"f" (x));
- }
- else
- z = __slow_ieee754_sqrt (x);
+#ifdef _ARCH_PPCSQ
+ asm ("fsqrt %0,%1\n" :"=f" (z):"f" (x));
+#else
+ z = __slow_ieee754_sqrt (x);
+#endif
return z;
}
Code: Select all
--- a/sysdeps/powerpc/fpu/e_sqrtf.c
+++ b/sysdeps/powerpc/fpu/e_sqrtf.c
@@ -24,6 +24,7 @@
#include <sysdep.h>
#include <ldsodefs.h>
+#ifndef _ARCH_PPCSQ
static const float almost_half = 0.50000006; /* 0.5 + 2^-24 */
static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 };
static const ieee_float_shape_type a_inf = {.word = 0x7f800000 };
@@ -128,6 +129,7 @@ __slow_ieee754_sqrtf (float x)
}
return f_washf (x);
}
+#endif /* _ARCH_PPCSQ */
#undef __ieee754_sqrtf
float
@@ -135,16 +137,11 @@ __ieee754_sqrtf (float x)
{
double z;
- /* If the CPU is 64-bit we can use the optional FP instructions. */
- if (__CPU_HAS_FSQRT)
- {
- /* Volatile is required to prevent the compiler from moving the
- fsqrt instruction above the branch. */
- __asm __volatile (" fsqrts %0,%1\n"
- :"=f" (z):"f" (x));
- }
- else
- z = __slow_ieee754_sqrtf (x);
+#ifdef _ARCH_PPCSQ
+ asm ("fsqrts %0,%1\n" :"=f" (z):"f" (x));
+#else
+ z = __slow_ieee754_sqrtf (x);
+#endif
return z;
}
Code: Select all
--- a/sysdeps/powerpc/fpu/math_private.h
+++ b/sysdeps/powerpc/fpu/math_private.h
@@ -25,26 +25,17 @@
#include <fenv_private.h>
#include_next <math_private.h>
-# if __WORDSIZE == 64 || defined _ARCH_PWR4
-# define __CPU_HAS_FSQRT 1
-# else
-# define __CPU_HAS_FSQRT ((GLRO(dl_hwcap) & PPC_FEATURE_64) != 0)
-# endif
-
extern double __slow_ieee754_sqrt (double);
extern __always_inline double
__ieee754_sqrt (double __x)
{
double __z;
- if (__CPU_HAS_FSQRT)
- {
- /* Volatile is required to prevent the compiler from moving the
- fsqrt instruction above the branch. */
- __asm __volatile ("fsqrt %0,%1" : "=f" (__z) : "f" (__x));
- }
- else
- __z = __slow_ieee754_sqrt(__x);
+#ifdef _ARCH_PPCSQ
+ asm ("fsqrt %0,%1" : "=f" (__z) : "f" (__x));
+#else
+ __z = __slow_ieee754_sqrt(__x);
+#endif
return __z;
}
@@ -55,14 +46,11 @@ __ieee754_sqrtf (float __x)
{
float __z;
- if (__CPU_HAS_FSQRT)
- {
- /* Volatile is required to prevent the compiler from moving the
- fsqrts instruction above the branch. */
- __asm __volatile ("fsqrts %0,%1" : "=f" (__z) : "f" (__x));
- }
- else
- __z = __slow_ieee754_sqrtf(__x);
+#ifdef _ARCH_PPCSQ
+ asm ("fsqrts %0,%1" : "=f" (__z) : "f" (__x));
+#else
+ __z = __slow_ieee754_sqrtf(__x);
+#endif
return __z;
}