From 4896f049205013528f1f0df7f7cbb52e8c5664d1 Mon Sep 17 00:00:00 2001 From: Richard Henderson Date: Fri, 1 Aug 2014 12:13:50 -1000 Subject: Force eval for fma implementations --- sysdeps/ieee754/ldbl-96/s_fma.c | 11 ++++++----- sysdeps/ieee754/ldbl-96/s_fmal.c | 11 ++++++----- 2 files changed, 12 insertions(+), 10 deletions(-) (limited to 'sysdeps/ieee754/ldbl-96') diff --git a/sysdeps/ieee754/ldbl-96/s_fma.c b/sysdeps/ieee754/ldbl-96/s_fma.c index 354be4eae9..a27102eff1 100644 --- a/sysdeps/ieee754/ldbl-96/s_fma.c +++ b/sysdeps/ieee754/ldbl-96/s_fma.c @@ -64,16 +64,17 @@ __fma (double x, double y, double z) t1 = m1 - t1; t2 = z - t2; long double a2 = t1 + t2; + /* Ensure the arithmetic is not scheduled after feclearexcept call. */ + math_force_eval (m2); + math_force_eval (a2); feclearexcept (FE_INEXACT); - /* If the result is an exact zero, ensure it has the correct - sign. */ + /* If the result is an exact zero, ensure it has the correct sign. */ if (a1 == 0 && m2 == 0) { feupdateenv (&env); - /* Ensure that round-to-nearest value of z + m1 is not - reused. */ - asm volatile ("" : "=m" (z) : "m" (z)); + /* Ensure that round-to-nearest value of z + m1 is not reused. */ + z = math_opt_barrier (z); return z + m1; } diff --git a/sysdeps/ieee754/ldbl-96/s_fmal.c b/sysdeps/ieee754/ldbl-96/s_fmal.c index 4983eda9c6..0f299da7c1 100644 --- a/sysdeps/ieee754/ldbl-96/s_fmal.c +++ b/sysdeps/ieee754/ldbl-96/s_fmal.c @@ -201,16 +201,17 @@ __fmal (long double x, long double y, long double z) t1 = m1 - t1; t2 = z - t2; long double a2 = t1 + t2; + /* Ensure the arithmetic is not scheduled after feclearexcept call. */ + math_force_eval (m2); + math_force_eval (a2); feclearexcept (FE_INEXACT); - /* If the result is an exact zero, ensure it has the correct - sign. */ + /* If the result is an exact zero, ensure it has the correct sign. */ if (a1 == 0 && m2 == 0) { feupdateenv (&env); - /* Ensure that round-to-nearest value of z + m1 is not - reused. */ - asm volatile ("" : "=m" (z) : "m" (z)); + /* Ensure that round-to-nearest value of z + m1 is not reused. */ + z = math_opt_barrier (z); return z + m1; } -- cgit 1.4.1