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

Clang-format tests, examples, libraries, benchmarks, etc.

Browse Source
This commit is contained in:
Antonio Sánchez
2023-12-05 21:22:55 +00:00
committed by Rasmus Munk Larsen
parent 3252ecc7a4
commit 46e9cdb7fe
876 changed files with 33453 additions and 37795 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
unsupported/test/cxx11_tensor_reduction.cpp
View File

@@ -53,7 +53,7 @@ static void test_trivial_reductions() {
}
}
template <typename Scalar,int DataLayout>
template <typename Scalar, int DataLayout>
static void test_simple_reductions() {
Tensor<Scalar, 4, DataLayout> tensor(2, 3, 5, 7);
tensor.setRandom();
@@ -227,13 +227,13 @@ static void test_simple_reductions() {
Tensor<int, 1> ints(10);
std::iota(ints.data(), ints.data() + ints.dimension(0), 0);
TensorFixedSize<bool, Sizes<> > all_;
TensorFixedSize<bool, Sizes<>> all_;
all_ = ints.all();
VERIFY(!all_());
all_ = (ints >= ints.constant(0)).all();
VERIFY(all_());
TensorFixedSize<bool, Sizes<> > any;
TensorFixedSize<bool, Sizes<>> any;
any = (ints > ints.constant(10)).any();
VERIFY(!any());
any = (ints < ints.constant(1)).any();
@@ -241,7 +241,6 @@ static void test_simple_reductions() {
}
}
template <int DataLayout>
static void test_reductions_in_expr() {
Tensor<float, 4, DataLayout> tensor(2, 3, 5, 7);
@@ -267,7 +266,6 @@ static void test_reductions_in_expr() {
}
}
template <int DataLayout>
static void test_full_reductions() {
Tensor<float, 2, DataLayout> tensor(2, 3);
@@ -333,11 +331,9 @@ static void test_user_defined_reductions() {
template <int DataLayout>
static void test_tensor_maps() {
int inputs[2 * 3 * 5 * 7];
TensorMap<Tensor<int, 4, DataLayout> > tensor_map(inputs, 2, 3, 5, 7);
TensorMap<Tensor<const int, 4, DataLayout> > tensor_map_const(inputs, 2, 3, 5,
7);
const TensorMap<Tensor<const int, 4, DataLayout> > tensor_map_const_const(
inputs, 2, 3, 5, 7);
TensorMap<Tensor<int, 4, DataLayout>> tensor_map(inputs, 2, 3, 5, 7);
TensorMap<Tensor<const int, 4, DataLayout>> tensor_map_const(inputs, 2, 3, 5, 7);
const TensorMap<Tensor<const int, 4, DataLayout>> tensor_map_const_const(inputs, 2, 3, 5, 7);
tensor_map.setRandom();
array<ptrdiff_t, 2> reduction_axis;
@@ -346,8 +342,7 @@ static void test_tensor_maps() {
Tensor<int, 2, DataLayout> result = tensor_map.sum(reduction_axis);
Tensor<int, 2, DataLayout> result2 = tensor_map_const.sum(reduction_axis);
Tensor<int, 2, DataLayout> result3 =
tensor_map_const_const.sum(reduction_axis);
Tensor<int, 2, DataLayout> result3 = tensor_map_const_const.sum(reduction_axis);
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 5; ++j) {
@@ -370,7 +365,7 @@ static void test_static_dims() {
Tensor<float, 2, DataLayout> out(72, 97);
in.setRandom();
Eigen::IndexList<Eigen::type2index<1>, Eigen::type2index<3> > reduction_axis;
Eigen::IndexList<Eigen::type2index<1>, Eigen::type2index<3>> reduction_axis;
out = in.maximum(reduction_axis);
@@ -393,9 +388,9 @@ static void test_innermost_last_dims() {
Tensor<float, 2, DataLayout> out(97, 113);
in.setRandom();
// Reduce on the innermost dimensions.
// Reduce on the innermost dimensions.
// This triggers the use of packets for ColMajor.
Eigen::IndexList<Eigen::type2index<0>, Eigen::type2index<1> > reduction_axis;
Eigen::IndexList<Eigen::type2index<0>, Eigen::type2index<1>> reduction_axis;
out = in.maximum(reduction_axis);
@@ -418,7 +413,7 @@ static void test_innermost_first_dims() {
Tensor<float, 2, DataLayout> out(72, 53);
in.setRandom();
// Reduce on the innermost dimensions.
// Reduce on the innermost dimensions.
// This triggers the use of packets for RowMajor.
Eigen::IndexList<Eigen::type2index<2>, Eigen::type2index<3>> reduction_axis;
@@ -443,7 +438,7 @@ static void test_reduce_middle_dims() {
Tensor<float, 2, DataLayout> out(72, 53);
in.setRandom();
// Reduce on the innermost dimensions.
// Reduce on the innermost dimensions.
// This triggers the use of packets for RowMajor.
Eigen::IndexList<Eigen::type2index<1>, Eigen::type2index<2>> reduction_axis;
@@ -466,7 +461,8 @@ template <typename ScalarType, int num_elements, int max_mean>
void test_sum_accuracy() {
Tensor<double, 1> double_tensor(num_elements);
Tensor<ScalarType, 1> tensor(num_elements);
for (double prescribed_mean = 0; prescribed_mean <= max_mean; prescribed_mean = numext::maxi(1.0, prescribed_mean*3.99)) {
for (double prescribed_mean = 0; prescribed_mean <= max_mean;
prescribed_mean = numext::maxi(1.0, prescribed_mean * 3.99)) {
// FIXME: NormalRandomGenerator doesn't work in bfloat and half.
double_tensor.setRandom<Eigen::internal::NormalRandomGenerator<double>>();
double_tensor += double_tensor.constant(prescribed_mean);
@@ -485,7 +481,8 @@ void test_sum_accuracy() {
// Test against probabilistic forward error bound. In reality, the error is much smaller
// when we use tree summation.
double err = Eigen::numext::abs(static_cast<double>(sum()) - expected_sum);
double tol = numext::sqrt(static_cast<double>(num_elements)) * static_cast<double>(NumTraits<ScalarType>::epsilon()) * abs_sum;
double tol = numext::sqrt(static_cast<double>(num_elements)) *
static_cast<double>(NumTraits<ScalarType>::epsilon()) * abs_sum;
VERIFY_LE(err, tol);
}
}
@@ -493,10 +490,10 @@ void test_sum_accuracy() {
EIGEN_DECLARE_TEST(cxx11_tensor_reduction) {
CALL_SUBTEST(test_trivial_reductions<ColMajor>());
CALL_SUBTEST(test_trivial_reductions<RowMajor>());
CALL_SUBTEST(( test_simple_reductions<float,ColMajor>() ));
CALL_SUBTEST(( test_simple_reductions<float,RowMajor>() ));
CALL_SUBTEST(( test_simple_reductions<Eigen::half,ColMajor>() ));
CALL_SUBTEST(( test_simple_reductions<Eigen::bfloat16,ColMajor>() ));
CALL_SUBTEST((test_simple_reductions<float, ColMajor>()));
CALL_SUBTEST((test_simple_reductions<float, RowMajor>()));
CALL_SUBTEST((test_simple_reductions<Eigen::half, ColMajor>()));
CALL_SUBTEST((test_simple_reductions<Eigen::bfloat16, ColMajor>()));
CALL_SUBTEST(test_reductions_in_expr<ColMajor>());
CALL_SUBTEST(test_reductions_in_expr<RowMajor>());
CALL_SUBTEST(test_full_reductions<ColMajor>());
@@ -513,11 +510,11 @@ EIGEN_DECLARE_TEST(cxx11_tensor_reduction) {
CALL_SUBTEST(test_innermost_first_dims<RowMajor>());
CALL_SUBTEST(test_reduce_middle_dims<ColMajor>());
CALL_SUBTEST(test_reduce_middle_dims<RowMajor>());
CALL_SUBTEST((test_sum_accuracy<float,10*1024*1024,8*1024>()));
CALL_SUBTEST((test_sum_accuracy<Eigen::bfloat16,10*1024*1024,8*1024>()));
CALL_SUBTEST((test_sum_accuracy<float, 10 * 1024 * 1024, 8 * 1024>()));
CALL_SUBTEST((test_sum_accuracy<Eigen::bfloat16, 10 * 1024 * 1024, 8 * 1024>()));
// The range of half is limited to 65519 when using round-to-even,
// so we are severely limited in the size and mean of the tensors
// we can reduce without overflow.
CALL_SUBTEST((test_sum_accuracy<Eigen::half,4*1024,16>()));
CALL_SUBTEST((test_sum_accuracy<Eigen::half,10*1024*1024,0>()));
CALL_SUBTEST((test_sum_accuracy<Eigen::half, 4 * 1024, 16>()));
CALL_SUBTEST((test_sum_accuracy<Eigen::half, 10 * 1024 * 1024, 0>()));
}