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

Allow move-only done callback in TensorAsyncDevice

Browse Source
This commit is contained in:
Eugene Zhulenev
2019-09-03 17:20:56 -07:00
parent a8d264fa9c
commit 47fefa235f
6 changed files with 70 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
unsupported/test/cxx11_tensor_executor.cpp
View File

@@ -578,11 +578,15 @@ static void test_async_execute_unary_expr(Device d)
src.setRandom();
const auto expr = src.square();
using Assign = TensorAssignOp<decltype(dst), const decltype(expr)>;
using Executor = internal::TensorAsyncExecutor<const Assign, Device,
Vectorizable, Tileable>;
Eigen::Barrier done(1);
Executor::runAsync(Assign(dst, expr), d, [&done]() { done.Notify(); });
auto on_done = [&done]() { done.Notify(); };
using Assign = TensorAssignOp<decltype(dst), const decltype(expr)>;
using DoneCallback = decltype(on_done);
using Executor = internal::TensorAsyncExecutor<const Assign, Device, DoneCallback,
Vectorizable, Tileable>;
Executor::runAsync(Assign(dst, expr), d, on_done);
done.Wait();
for (Index i = 0; i < dst.dimensions().TotalSize(); ++i) {
@@ -610,12 +614,15 @@ static void test_async_execute_binary_expr(Device d)
const auto expr = lhs + rhs;
Eigen::Barrier done(1);
auto on_done = [&done]() { done.Notify(); };
using Assign = TensorAssignOp<decltype(dst), const decltype(expr)>;
using Executor = internal::TensorAsyncExecutor<const Assign, Device,
using DoneCallback = decltype(on_done);
using Executor = internal::TensorAsyncExecutor<const Assign, Device, DoneCallback,
Vectorizable, Tileable>;
Eigen::Barrier done(1);
Executor::runAsync(Assign(dst, expr), d, [&done]() { done.Notify(); });
Executor::runAsync(Assign(dst, expr), d, on_done);
done.Wait();
for (Index i = 0; i < dst.dimensions().TotalSize(); ++i) {

47
unsupported/test/cxx11_tensor_thread_pool.cpp
View File

@@ -683,34 +683,39 @@ EIGEN_DECLARE_TEST(cxx11_tensor_thread_pool)
CALL_SUBTEST_3(test_multithread_contraction_agrees_with_singlethread<RowMajor>());
CALL_SUBTEST_3(test_multithread_contraction_with_output_kernel<ColMajor>());
CALL_SUBTEST_3(test_multithread_contraction_with_output_kernel<RowMajor>());
CALL_SUBTEST_3(test_async_multithread_contraction_agrees_with_singlethread<ColMajor>());
CALL_SUBTEST_3(test_async_multithread_contraction_agrees_with_singlethread<RowMajor>());
CALL_SUBTEST_4(test_async_multithread_contraction_agrees_with_singlethread<ColMajor>());
CALL_SUBTEST_4(test_async_multithread_contraction_agrees_with_singlethread<RowMajor>());
// Test EvalShardedByInnerDimContext parallelization strategy.
CALL_SUBTEST_4(test_sharded_by_inner_dim_contraction<ColMajor>());
CALL_SUBTEST_4(test_sharded_by_inner_dim_contraction<RowMajor>());
CALL_SUBTEST_4(test_sharded_by_inner_dim_contraction_with_output_kernel<ColMajor>());
CALL_SUBTEST_4(test_sharded_by_inner_dim_contraction_with_output_kernel<RowMajor>());
CALL_SUBTEST_4(test_async_sharded_by_inner_dim_contraction<ColMajor>());
CALL_SUBTEST_4(test_async_sharded_by_inner_dim_contraction<RowMajor>());
CALL_SUBTEST_4(test_async_sharded_by_inner_dim_contraction_with_output_kernel<ColMajor>());
CALL_SUBTEST_4(test_async_sharded_by_inner_dim_contraction_with_output_kernel<RowMajor>());
CALL_SUBTEST_5(test_sharded_by_inner_dim_contraction<ColMajor>());
CALL_SUBTEST_5(test_sharded_by_inner_dim_contraction<RowMajor>());
CALL_SUBTEST_5(test_sharded_by_inner_dim_contraction_with_output_kernel<ColMajor>());
CALL_SUBTEST_5(test_sharded_by_inner_dim_contraction_with_output_kernel<RowMajor>());
CALL_SUBTEST_6(test_async_sharded_by_inner_dim_contraction<ColMajor>());
CALL_SUBTEST_6(test_async_sharded_by_inner_dim_contraction<RowMajor>());
CALL_SUBTEST_6(test_async_sharded_by_inner_dim_contraction_with_output_kernel<ColMajor>());
CALL_SUBTEST_6(test_async_sharded_by_inner_dim_contraction_with_output_kernel<RowMajor>());
// Exercise various cases that have been problematic in the past.
CALL_SUBTEST_5(test_contraction_corner_cases<ColMajor>());
CALL_SUBTEST_5(test_contraction_corner_cases<RowMajor>());
CALL_SUBTEST_7(test_contraction_corner_cases<ColMajor>());
CALL_SUBTEST_7(test_contraction_corner_cases<RowMajor>());
CALL_SUBTEST_6(test_full_contraction<ColMajor>());
CALL_SUBTEST_6(test_full_contraction<RowMajor>());
CALL_SUBTEST_8(test_full_contraction<ColMajor>());
CALL_SUBTEST_8(test_full_contraction<RowMajor>());
CALL_SUBTEST_7(test_multithreaded_reductions<ColMajor>());
CALL_SUBTEST_7(test_multithreaded_reductions<RowMajor>());
CALL_SUBTEST_9(test_multithreaded_reductions<ColMajor>());
CALL_SUBTEST_9(test_multithreaded_reductions<RowMajor>());
CALL_SUBTEST_7(test_memcpy());
CALL_SUBTEST_7(test_multithread_random());
CALL_SUBTEST_10(test_memcpy());
CALL_SUBTEST_10(test_multithread_random());
TestAllocator test_allocator;
CALL_SUBTEST_7(test_multithread_shuffle<ColMajor>(NULL));
CALL_SUBTEST_7(test_multithread_shuffle<RowMajor>(&test_allocator));
CALL_SUBTEST_7(test_threadpool_allocate(&test_allocator));
CALL_SUBTEST_11(test_multithread_shuffle<ColMajor>(NULL));
CALL_SUBTEST_11(test_multithread_shuffle<RowMajor>(&test_allocator));
CALL_SUBTEST_11(test_threadpool_allocate(&test_allocator));
// Force CMake to split this test.
// EIGEN_SUFFIXES;1;2;3;4;5;6;7;8;9;10;11
}