devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add support for asynchronous evaluation of tensor casting expressions.

Browse Source
This commit is contained in:
Rasmus Munk Larsen
2019-09-19 13:54:49 -07:00
parent 28b6786498
commit 1d5af0693c
2 changed files with 37 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
unsupported/test/cxx11_tensor_thread_pool.cpp
View File

@@ -40,19 +40,19 @@ void test_multithread_elementwise()
{
Tensor<float, 3> in1(200, 30, 70);
Tensor<float, 3> in2(200, 30, 70);
Tensor<float, 3> out(200, 30, 70);
Tensor<double, 3> out(200, 30, 70);
in1.setRandom();
in2.setRandom();
Eigen::ThreadPool tp(internal::random<int>(3, 11));
Eigen::ThreadPoolDevice thread_pool_device(&tp, internal::random<int>(3, 11));
out.device(thread_pool_device) = in1 + in2 * 3.14f;
out.device(thread_pool_device) = (in1 + in2 * 3.14f).cast<double>();
for (int i = 0; i < 200; ++i) {
for (int j = 0; j < 30; ++j) {
for (int k = 0; k < 70; ++k) {
VERIFY_IS_APPROX(out(i, j, k), in1(i, j, k) + in2(i, j, k) * 3.14f);
VERIFY_IS_APPROX(out(i, j, k), static_cast<double>(in1(i, j, k) + in2(i, j, k) * 3.14f));
}
}
}
@@ -62,7 +62,7 @@ void test_async_multithread_elementwise()
{
Tensor<float, 3> in1(200, 30, 70);
Tensor<float, 3> in2(200, 30, 70);
Tensor<float, 3> out(200, 30, 70);
Tensor<double, 3> out(200, 30, 70);
in1.setRandom();
in2.setRandom();
@@ -71,13 +71,13 @@ void test_async_multithread_elementwise()
Eigen::ThreadPoolDevice thread_pool_device(&tp, internal::random<int>(3, 11));
Eigen::Barrier b(1);
out.device(thread_pool_device, [&b]() { b.Notify(); }) = in1 + in2 * 3.14f;
out.device(thread_pool_device, [&b]() { b.Notify(); }) = (in1 + in2 * 3.14f).cast<double>();
b.Wait();
for (int i = 0; i < 200; ++i) {
for (int j = 0; j < 30; ++j) {
for (int k = 0; k < 70; ++k) {
VERIFY_IS_APPROX(out(i, j, k), in1(i, j, k) + in2(i, j, k) * 3.14f);
VERIFY_IS_APPROX(out(i, j, k), static_cast<double>(in1(i, j, k) + in2(i, j, k) * 3.14f));
}
}
}