diff options
| author |  Antonio Sanchez <cantonios@google.com> | 2021-01-22 11:10:54 -0800 |
|---|---|---|
| committer | Antonio Sanchez <cantonios@google.com> | 2021-01-22 11:10:54 -0800 |
| commit | f0e46ed5d41eeb450cbcbdb1ce3233d524ad3acd (patch) | |
| tree | b2e862ad5f8c0788db4f3c39c0732db64fb5e217 /Eigen/src/Core/functors | |
| parent | f19bcffee6b8018ca101ceb370e6e550a940289f (diff) | |
Fix pow and other cwise ops for half/bfloat16.
The new `generic_pow` implementation was failing for half/bfloat16 since
their construction from int/float is not `constexpr`. Modified
in `GenericPacketMathFunctions` to remove `constexpr`.
While adding tests for half/bfloat16, found other issues related to
implicit conversions.
Also needed to implement `numext::arg` for non-integer, non-complex,
non-float/double/long double types. These seem to be implicitly
converted to `std::complex<T>`, which then fails for half/bfloat16.
Diffstat (limited to 'Eigen/src/Core/functors')
| -rw-r--r-- | Eigen/src/Core/functors/UnaryFunctors.h | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/Eigen/src/Core/functors/UnaryFunctors.h b/Eigen/src/Core/functors/UnaryFunctors.h index 976ecba59..c98fa573c 100644 --- a/Eigen/src/Core/functors/UnaryFunctors.h +++ b/Eigen/src/Core/functors/UnaryFunctors.h @@ -403,7 +403,7 @@ struct functor_traits<scalar_log10_op<Scalar> > */ template<typename Scalar> struct scalar_log2_op { EIGEN_EMPTY_STRUCT_CTOR(scalar_log2_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return Scalar(EIGEN_LOG2E) * std::log(a); } + EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return Scalar(EIGEN_LOG2E) * numext::log(a); } template <typename Packet> EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog2(a); } }; |