diff --git a/Eigen/src/Core/MathFunctions.h b/Eigen/src/Core/MathFunctions.h
index 000cafee7..e6c7dfa08 100644
--- a/Eigen/src/Core/MathFunctions.h
+++ b/Eigen/src/Core/MathFunctions.h
@@ -1034,7 +1034,7 @@ double tan(const double &x) { return ::tan(x); }
 
 template<typename T>
 EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE
-T abs(const T &x) {
+typename NumTraits<T>::Real abs(const T &x) {
   EIGEN_USING_STD_MATH(abs);
   return abs(x);
 }
diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h
index fead02916..63a2d9f52 100644
--- a/Eigen/src/Core/arch/NEON/PacketMath.h
+++ b/Eigen/src/Core/arch/NEON/PacketMath.h
@@ -177,7 +177,9 @@ template<> EIGEN_STRONG_INLINE Packet4i pdiv<Packet4i>(const Packet4i& /*a*/, co
   return pset1<Packet4i>(0);
 }
 
-#ifdef __ARM_FEATURE_FMA
+// Clang/ARM wrongly advertises __ARM_FEATURE_FMA even when it's not available,
+// then implements a slow software scalar fallback calling fmaf()!
+#if (defined __ARM_FEATURE_FMA) && !(EIGEN_COMP_CLANG && EIGEN_ARCH_ARM)
 // See bug 936.
 // FMA is available on VFPv4 i.e. when compiling with -mfpu=neon-vfpv4.
 // FMA is a true fused multiply-add i.e. only 1 rounding at the end, no intermediate rounding.
@@ -186,7 +188,25 @@ template<> EIGEN_STRONG_INLINE Packet4i pdiv<Packet4i>(const Packet4i& /*a*/, co
 // MLA: 10 GFlop/s ; FMA: 12 GFlops/s.
 template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return vfmaq_f32(c,a,b); }
 #else
-template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return vmlaq_f32(c,a,b); }
+template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) {
+#if EIGEN_COMP_CLANG && EIGEN_ARCH_ARM
+  // Clang/ARM will replace VMLA by VMUL+VADD at least for some values of -mcpu,
+  // at least -mcpu=cortex-a8 and -mcpu=cortex-a7. Since the former is the default on
+  // -march=armv7-a, that is a very common case.
+  // See e.g. this thread:
+  //     http://lists.llvm.org/pipermail/llvm-dev/2013-December/068806.html
+  Packet4f r = c;
+  asm volatile(
+    "vmla.f32 %q[r], %q[a], %q[b]"
+    : [r] "+w" (r)
+    : [a] "w" (a),
+      [b] "w" (b)
+    : );
+  return r;
+#else
+  return vmlaq_f32(c,a,b);
+#endif
+}
 #endif
 
 // No FMA instruction for int, so use MLA unconditionally.
diff --git a/Eigen/src/Core/functors/UnaryFunctors.h b/Eigen/src/Core/functors/UnaryFunctors.h
index 826d84f69..26c02e4a7 100644
--- a/Eigen/src/Core/functors/UnaryFunctors.h
+++ b/Eigen/src/Core/functors/UnaryFunctors.h
@@ -41,7 +41,7 @@ struct functor_traits<scalar_opposite_op<Scalar> >
 template<typename Scalar> struct scalar_abs_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_abs_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { using std::abs; return abs(a); }
+  EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return numext::abs(a); }
   template<typename Packet>
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const
   { return internal::pabs(a); }
diff --git a/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h b/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h
index e30ad5b6d..481dfa91a 100755
--- a/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h
+++ b/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h
@@ -589,23 +589,24 @@ EIGEN_AUTODIFF_DECLARE_GLOBAL_UNARY(log,
   return ReturnType(log(x.value()),x.derivatives() * (Scalar(1)/x.value()));)
 
 template<typename DerType>
-inline const Eigen::AutoDiffScalar<Eigen::CwiseUnaryOp<Eigen::internal::scalar_multiple_op<typename Eigen::internal::traits<DerType>::Scalar>, const DerType> >
-pow(const Eigen::AutoDiffScalar<DerType>& x, typename Eigen::internal::traits<DerType>::Scalar y)
+inline const Eigen::AutoDiffScalar<Eigen::CwiseUnaryOp<Eigen::internal::scalar_multiple_op<typename internal::traits<typename internal::remove_all<DerType>::type>::Scalar>, const typename internal::remove_all<DerType>::type> >
+pow(const Eigen::AutoDiffScalar<DerType>& x, typename internal::traits<typename internal::remove_all<DerType>::type>::Scalar y)
 {
   using namespace Eigen;
-  typedef typename Eigen::internal::traits<DerType>::Scalar Scalar;
-  return AutoDiffScalar<CwiseUnaryOp<Eigen::internal::scalar_multiple_op<Scalar>, const DerType> >(
+  typedef typename internal::remove_all<DerType>::type DerTypeCleaned;
+  typedef typename Eigen::internal::traits<DerTypeCleaned>::Scalar Scalar;
+  return AutoDiffScalar<CwiseUnaryOp<Eigen::internal::scalar_multiple_op<Scalar>, const DerTypeCleaned> >(
     std::pow(x.value(),y),
     x.derivatives() * (y * std::pow(x.value(),y-1)));
 }
 
 
 template<typename DerTypeA,typename DerTypeB>
-inline const AutoDiffScalar<Matrix<typename internal::traits<DerTypeA>::Scalar,Dynamic,1> >
+inline const AutoDiffScalar<Matrix<typename internal::traits<typename internal::remove_all<DerTypeA>::type>::Scalar,Dynamic,1> >
 atan2(const AutoDiffScalar<DerTypeA>& a, const AutoDiffScalar<DerTypeB>& b)
 {
   using std::atan2;
-  typedef typename internal::traits<DerTypeA>::Scalar Scalar;
+  typedef typename internal::traits<typename internal::remove_all<DerTypeA>::type>::Scalar Scalar;
   typedef AutoDiffScalar<Matrix<Scalar,Dynamic,1> > PlainADS;
   PlainADS ret;
   ret.value() = atan2(a.value(), b.value());
diff --git a/unsupported/test/autodiff.cpp b/unsupported/test/autodiff.cpp
index 1aa1b3d2d..374f86df9 100644
--- a/unsupported/test/autodiff.cpp
+++ b/unsupported/test/autodiff.cpp
@@ -16,7 +16,7 @@ EIGEN_DONT_INLINE Scalar foo(const Scalar& x, const Scalar& y)
   using namespace std;
 //   return x+std::sin(y);
   EIGEN_ASM_COMMENT("mybegin");
-  return static_cast<Scalar>(x*2 - pow(x,2) + 2*sqrt(y*y) - 4 * sin(x) + 2 * cos(y) - exp(-0.5*x*x));
+  return static_cast<Scalar>(x*2 - 1 + pow(1+x,2) + 2*sqrt(y*y+0) - 4 * sin(0+x) + 2 * cos(y+0) - exp(-0.5*x*x+0));
   //return x+2*y*x;//x*2 -std::pow(x,2);//(2*y/x);// - y*2;
   EIGEN_ASM_COMMENT("myend");
 }
diff --git a/unsupported/test/autodiff_scalar.cpp b/unsupported/test/autodiff_scalar.cpp
index ba4b5aec4..c631c734a 100644
--- a/unsupported/test/autodiff_scalar.cpp
+++ b/unsupported/test/autodiff_scalar.cpp
@@ -30,6 +30,10 @@ template<typename Scalar> void check_atan2()
   
   VERIFY_IS_APPROX(res.value(), x.value());
   VERIFY_IS_APPROX(res.derivatives(), x.derivatives());
+
+  res = atan2(r*s+0, r*c+0);
+  VERIFY_IS_APPROX(res.value(), x.value());
+  VERIFY_IS_APPROX(res.derivatives(), x.derivatives());
 }