diff options
| author |  Benoit Steiner <benoit.steiner.goog@gmail.com> | 2016-12-05 02:31:11 +0000 |
|---|---|---|
| committer | Benoit Steiner <benoit.steiner.goog@gmail.com> | 2016-12-05 02:31:11 +0000 |
| commit | 462c28e77a57b1f5c296c90b15c21f999d193606 (patch) | |
| tree | 8a16c052d79a56126a2a5599bda4e8111976d726 /unsupported | |
| parent | 4465d20403921f9acd705ba3955057d729fd04b7 (diff) | |
| parent | 09ee7f0c80fb586835f4fa77f9129080ea63406e (diff) | |
Merged in srvasude/eigen (pull request PR-265)
Add Expm1 support to Eigen.
Diffstat (limited to 'unsupported')
4 files changed, 14 insertions, 0 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorBase.h b/unsupported/Eigen/CXX11/src/Tensor/TensorBase.h index 7a45a5cf4..fbe340820 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorBase.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorBase.h @@ -186,6 +186,12 @@ class TensorBase<Derived, ReadOnlyAccessors> } EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const TensorCwiseUnaryOp<internal::scalar_expm1_op<Scalar>, const Derived> + expm1() const { + return unaryExpr(internal::scalar_expm1_op<Scalar>()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const TensorCwiseUnaryOp<internal::scalar_log_op<Scalar>, const Derived> log() const { return unaryExpr(internal::scalar_log_op<Scalar>()); diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h b/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h index 25ce471f9..b5ef31d55 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h @@ -75,6 +75,7 @@ struct PacketType<half, GpuDevice> { HasSqrt = 1, HasRsqrt = 1, HasExp = 1, + HasExpm1 = 0, HasLog = 1, HasLog1p = 0, HasLog10 = 0, diff --git a/unsupported/test/cxx11_tensor_builtins_sycl.cpp b/unsupported/test/cxx11_tensor_builtins_sycl.cpp index d8c2898ca..e230b626f 100644 --- a/unsupported/test/cxx11_tensor_builtins_sycl.cpp +++ b/unsupported/test/cxx11_tensor_builtins_sycl.cpp @@ -91,6 +91,7 @@ template <typename T> T inverse(T x) { return 1 / x; } TEST_UNARY_BUILTINS_FOR_SCALAR(inverse, SCALAR, OPERATOR) \ TEST_UNARY_BUILTINS_FOR_SCALAR(tanh, SCALAR, OPERATOR) \ TEST_UNARY_BUILTINS_FOR_SCALAR(exp, SCALAR, OPERATOR) \ + TEST_UNARY_BUILTINS_FOR_SCALAR(expm1, SCALAR, OPERATOR) \ TEST_UNARY_BUILTINS_FOR_SCALAR(log, SCALAR, OPERATOR) \ TEST_UNARY_BUILTINS_FOR_SCALAR(abs, SCALAR, OPERATOR) \ TEST_UNARY_BUILTINS_FOR_SCALAR(ceil, SCALAR, OPERATOR) \ diff --git a/unsupported/test/cxx11_tensor_of_float16_cuda.cu b/unsupported/test/cxx11_tensor_of_float16_cuda.cu index 2f86980a2..908a5e5a9 100644 --- a/unsupported/test/cxx11_tensor_of_float16_cuda.cu +++ b/unsupported/test/cxx11_tensor_of_float16_cuda.cu @@ -200,6 +200,8 @@ void test_cuda_trancendental() { Eigen::TensorMap<Eigen::Tensor<Eigen::half, 1>, Eigen::Aligned> gpu_res2_float(d_res2_float, num_elem); Eigen::TensorMap<Eigen::Tensor<Eigen::half, 1>, Eigen::Aligned> gpu_res3_half(d_res3_half, num_elem); Eigen::TensorMap<Eigen::Tensor<Eigen::half, 1>, Eigen::Aligned> gpu_res3_float(d_res3_float, num_elem); + Eigen::TensorMap<Eigen::Tensor<Eigen::half, 1>, Eigen::Aligned> gpu_res4_half(d_res3_half, num_elem); + Eigen::TensorMap<Eigen::Tensor<Eigen::half, 1>, Eigen::Aligned> gpu_res4_float(d_res3_float, num_elem); gpu_float1.device(gpu_device) = gpu_float1.random() - gpu_float1.constant(0.5f); gpu_float2.device(gpu_device) = gpu_float2.random() + gpu_float1.constant(0.5f); @@ -207,6 +209,7 @@ void test_cuda_trancendental() { gpu_res1_float.device(gpu_device) = gpu_float1.exp().cast<Eigen::half>(); gpu_res2_float.device(gpu_device) = gpu_float2.log().cast<Eigen::half>(); gpu_res3_float.device(gpu_device) = gpu_float3.log1p().cast<Eigen::half>(); + gpu_res4_float.device(gpu_device) = gpu_float3.expm1().cast<Eigen::half>(); gpu_res1_half.device(gpu_device) = gpu_float1.cast<Eigen::half>(); gpu_res1_half.device(gpu_device) = gpu_res1_half.exp(); @@ -217,6 +220,9 @@ void test_cuda_trancendental() { gpu_res3_half.device(gpu_device) = gpu_float3.cast<Eigen::half>(); gpu_res3_half.device(gpu_device) = gpu_res3_half.log1p(); + gpu_res3_half.device(gpu_device) = gpu_float3.cast<Eigen::half>(); + gpu_res3_half.device(gpu_device) = gpu_res3_half.expm1(); + Tensor<float, 1> input1(num_elem); Tensor<Eigen::half, 1> half_prec1(num_elem); Tensor<Eigen::half, 1> full_prec1(num_elem); |