[PATCH v3 2/4] iee754: provide gcc builtins based generic fma functions

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@xxxxxxxxxx>
---
 sysdeps/generic/math-use-builtins.h         | 5 +++++
 sysdeps/ieee754/dbl-64/s_fma.c              | 6 ++++++
 sysdeps/ieee754/dbl-64/s_fmaf.c             | 6 ++++++
 sysdeps/ieee754/float128/float128_private.h | 2 ++
 sysdeps/ieee754/ldbl-128/s_fma.c            | 5 +++++
 sysdeps/ieee754/ldbl-128/s_fmal.c           | 5 +++++
 sysdeps/s390/fpu/math-use-builtins.h        | 5 +++++
 7 files changed, 34 insertions(+)

diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
index fc724c824a17..cf25ed8a2138 100644
--- a/sysdeps/generic/math-use-builtins.h
+++ b/sysdeps/generic/math-use-builtins.h
@@ -63,4 +63,9 @@
 #define USE_SQRT_BUILTIN 0
 #define USE_SQRTF_BUILTIN 0
 
+#define USE_FMA_BUILTIN 0
+#define USE_FMAF_BUILTIN 0
+#define USE_FMAL_BUILTIN 0
+#define USE_FMAF128_BUILTIN 0
+
 #endif /* math-use-builtins.h */
diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
index 876df6e78bdc..9dc5b132b9ee 100644
--- a/sysdeps/ieee754/dbl-64/s_fma.c
+++ b/sysdeps/ieee754/dbl-64/s_fma.c
@@ -25,6 +25,7 @@
 #include <fenv_private.h>
 #include <libm-alias-double.h>
 #include <tininess.h>
+#include <math-use-builtins.h>
 
 /* This implementation uses rounding to odd to avoid problems with
    double rounding.  See a paper by Boldo and Melquiond:
@@ -33,6 +34,10 @@
 double
 __fma (double x, double y, double z)
 {
+#if USE_FMA_BUILTIN
+  return __builtin_fma (x, y, z);
+#else
+  /* Use generic implementation.  */
   union ieee754_double u, v, w;
   int adjust = 0;
   u.d = x;
@@ -292,6 +297,7 @@ __fma (double x, double y, double z)
       v.ieee.mantissa1 |= j;
       return v.d * 0x1p-108;
     }
+#endif /* ! USE_FMA_BUILTIN  */
 }
 #ifndef __fma
 libm_alias_double (__fma, fma)
diff --git a/sysdeps/ieee754/dbl-64/s_fmaf.c b/sysdeps/ieee754/dbl-64/s_fmaf.c
index 57329d0a87fe..93b8660d5242 100644
--- a/sysdeps/ieee754/dbl-64/s_fmaf.c
+++ b/sysdeps/ieee754/dbl-64/s_fmaf.c
@@ -23,6 +23,7 @@
 #include <math-barriers.h>
 #include <fenv_private.h>
 #include <libm-alias-float.h>
+#include <math-use-builtins.h>
 
 /* This implementation relies on double being more than twice as
    precise as float and uses rounding to odd in order to avoid problems
@@ -33,6 +34,10 @@
 float
 __fmaf (float x, float y, float z)
 {
+#if USE_FMAF_BUILTIN
+  return __builtin_fmaf (x, y, z);
+#else
+  /* Use generic implementation.  */
   fenv_t env;
 
   /* Multiplication is always exact.  */
@@ -60,6 +65,7 @@ __fmaf (float x, float y, float z)
 
   /* And finally truncation with round to nearest.  */
   return (float) u.d;
+#endif /* ! USE_FMAF_BUILTIN  */
 }
 #ifndef __fmaf
 libm_alias_float (__fma, fma)
diff --git a/sysdeps/ieee754/float128/float128_private.h b/sysdeps/ieee754/float128/float128_private.h
index f97463d9dc1b..ab6fc9f3c9cf 100644
--- a/sysdeps/ieee754/float128/float128_private.h
+++ b/sysdeps/ieee754/float128/float128_private.h
@@ -154,6 +154,8 @@
 #define USE_ROUNDL_BUILTIN USE_ROUNDF128_BUILTIN
 #undef USE_COPYSIGNL_BUILTIN
 #define USE_COPYSIGNL_BUILTIN USE_COPYSIGNF128_BUILTIN
+#undef USE_FMAL_BUILTIN
+#define USE_FMAL_BUILTIN USE_FMAF128_BUILTIN
 
 /* IEEE function renames.  */
 #define __ieee754_acoshl __ieee754_acoshf128
diff --git a/sysdeps/ieee754/ldbl-128/s_fma.c b/sysdeps/ieee754/ldbl-128/s_fma.c
index 13f8ed3b87eb..0feff0a71cd1 100644
--- a/sysdeps/ieee754/ldbl-128/s_fma.c
+++ b/sysdeps/ieee754/ldbl-128/s_fma.c
@@ -21,6 +21,7 @@
 #include <fenv.h>
 #include <ieee754.h>
 #include <libm-alias-double.h>
+#include <math-use-builtins.h>
 
 /* This implementation relies on long double being more than twice as
    precise as double and uses rounding to odd in order to avoid problems
@@ -31,6 +32,9 @@
 double
 __fma (double x, double y, double z)
 {
+#if USE_FMA_BUILTIN
+  return __builtin_fma (x, y, z);
+#else
   fenv_t env;
   /* Multiplication is always exact.  */
   long double temp = (long double) x * (long double) y;
@@ -50,6 +54,7 @@ __fma (double x, double y, double z)
   feupdateenv (&env);
   /* And finally truncation with round to nearest.  */
   return (double) u.d;
+#endif /* ! USE_FMA_BUILTIN  */
 }
 #ifndef __fma
 libm_alias_double (__fma, fma)
diff --git a/sysdeps/ieee754/ldbl-128/s_fmal.c b/sysdeps/ieee754/ldbl-128/s_fmal.c
index 7475015bcec6..a610499e47c7 100644
--- a/sysdeps/ieee754/ldbl-128/s_fmal.c
+++ b/sysdeps/ieee754/ldbl-128/s_fmal.c
@@ -25,6 +25,7 @@
 #include <math_private.h>
 #include <libm-alias-ldouble.h>
 #include <tininess.h>
+#include <math-use-builtins.h>
 
 /* This implementation uses rounding to odd to avoid problems with
    double rounding.  See a paper by Boldo and Melquiond:
@@ -33,6 +34,9 @@
 _Float128
 __fmal (_Float128 x, _Float128 y, _Float128 z)
 {
+#if USE_FMAL_BUILTIN
+  return __builtin_fmal (x, y, z);
+#else
   union ieee854_long_double u, v, w;
   int adjust = 0;
   u.d = x;
@@ -296,5 +300,6 @@ __fmal (_Float128 x, _Float128 y, _Float128 z)
       v.ieee.mantissa3 |= j;
       return v.d * L(0x1p-228);
     }
+#endif /* ! USE_FMAL_BUILTIN  */
 }
 libm_alias_ldouble (__fma, fma)
diff --git a/sysdeps/s390/fpu/math-use-builtins.h b/sysdeps/s390/fpu/math-use-builtins.h
index eb87b6b0b16b..3a80f86b2e29 100644
--- a/sysdeps/s390/fpu/math-use-builtins.h
+++ b/sysdeps/s390/fpu/math-use-builtins.h
@@ -111,4 +111,9 @@
 #define USE_SQRT_BUILTIN 0
 #define USE_SQRTF_BUILTIN 0
 
+#define USE_FMA_BUILTIN 0
+#define USE_FMAF_BUILTIN 0
+#define USE_FMAL_BUILTIN 0
+#define USE_FMAF128_BUILTIN 0
+
 #endif /* math-use-builtins.h */
-- 
2.20.1


_______________________________________________
linux-snps-arc mailing list
linux-snps-arc@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/linux-snps-arc



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux