Vectorized the evaluation of tensor expression (using SSE, AVX, NEON, ...)

Added the ability to parallelize the evaluation of a tensor expression over multiple cpu cores.
Added the ability to offload the evaluation of a tensor expression to a GPU.

1 files changed, 37 insertions, 0 deletions

diff --git a/unsupported/test/cxx11_tensor_thread_pool.cpp b/unsupported/test/cxx11_tensor_thread_pool.cpp
new file mode 100644
index 000000000..c9de71da3
--- /dev/null
+++ b/unsupported/test/cxx11_tensor_thread_pool.cpp
@@ -0,0 +1,37 @@
+// This file is part of Eigen, a lightweight C++ template library
+// for linear algebra.
+//
+// Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.com>
+//
+// This Source Code Form is subject to the terms of the Mozilla
+// Public License v. 2.0. If a copy of the MPL was not distributed
+// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
+
+#define EIGEN_USE_THREADS
+
+
+#include "main.h"
+#include <Eigen/CXX11/Tensor>
+
+using Eigen::Tensor;
+
+void test_cxx11_tensor_thread_pool()
+{
+  Eigen::Tensor<float, 3> in1(Eigen::array<ptrdiff_t, 3>(2,3,7));
+  Eigen::Tensor<float, 3> in2(Eigen::array<ptrdiff_t, 3>(2,3,7));
+  Eigen::Tensor<float, 3> out(Eigen::array<ptrdiff_t, 3>(2,3,7));
+
+  in1.setRandom();
+  in2.setRandom();
+
+  Eigen::ThreadPoolDevice thread_pool_device(3);
+  out.device(thread_pool_device) = in1 + in2 * 3.14;
+
+  for (int i = 0; i < 2; ++i) {
+    for (int j = 0; j < 3; ++j) {
+      for (int k = 0; k < 7; ++k) {
+        VERIFY_IS_APPROX(out(Eigen::array<ptrdiff_t, 3>(i,j,k)), in1(Eigen::array<ptrdiff_t, 3>(i,j,k)) + in2(Eigen::array<ptrdiff_t, 3>(i,j,k)) * 3.14f);
+      }
+    }
+  }
+}