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

Merged with upstream eigen

Browse Source
This commit is contained in:
Eugene Zhulenev
2018-08-08 16:57:58 -07:00
parent 1b0373ae10 131ed1191f
commit 1c8b9e10a7
22 changed files with 190 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
unsupported/Eigen/CXX11/src/Tensor/TensorBase.h
View File

@@ -538,8 +538,8 @@ class TensorBase<Derived, ReadOnlyAccessors>
// Fourier transforms
template <int FFTDataType, int FFTDirection, typename FFT> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const TensorFFTOp<const FFT, const Derived, FFTDataType, FFTDirection>
fft(const FFT& fft) const {
return TensorFFTOp<const FFT, const Derived, FFTDataType, FFTDirection>(derived(), fft);
fft(const FFT& dims) const {
return TensorFFTOp<const FFT, const Derived, FFTDataType, FFTDirection>(derived(), dims);
}
// Scan.
@@ -723,8 +723,8 @@ class TensorBase<Derived, ReadOnlyAccessors>
template <typename Broadcast> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const TensorBroadcastingOp<const Broadcast, const Derived>
broadcast(const Broadcast& broadcast) const {
return TensorBroadcastingOp<const Broadcast, const Derived>(derived(), broadcast);
broadcast(const Broadcast& bcast) const {
return TensorBroadcastingOp<const Broadcast, const Derived>(derived(), bcast);
}
template <typename Axis, typename OtherDerived> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
@@ -832,8 +832,8 @@ class TensorBase<Derived, ReadOnlyAccessors>
}
template <typename Shuffle> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const TensorShufflingOp<const Shuffle, const Derived>
shuffle(const Shuffle& shuffle) const {
return TensorShufflingOp<const Shuffle, const Derived>(derived(), shuffle);
shuffle(const Shuffle& shfl) const {
return TensorShufflingOp<const Shuffle, const Derived>(derived(), shfl);
}
template <typename Strides> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const TensorStridingOp<const Strides, const Derived>
@@ -1030,13 +1030,13 @@ class TensorBase : public TensorBase<Derived, ReadOnlyAccessors> {
template <typename Shuffle> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const TensorShufflingOp<const Shuffle, const Derived>
shuffle(const Shuffle& shuffle) const {
return TensorShufflingOp<const Shuffle, const Derived>(derived(), shuffle);
shuffle(const Shuffle& shfl) const {
return TensorShufflingOp<const Shuffle, const Derived>(derived(), shfl);
}
template <typename Shuffle> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
TensorShufflingOp<const Shuffle, Derived>
shuffle(const Shuffle& shuffle) {
return TensorShufflingOp<const Shuffle, Derived>(derived(), shuffle);
shuffle(const Shuffle& shfl) {
return TensorShufflingOp<const Shuffle, Derived>(derived(), shfl);
}
template <typename Strides> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
@@ -1052,8 +1052,8 @@ class TensorBase : public TensorBase<Derived, ReadOnlyAccessors> {
// Select the device on which to evaluate the expression.
template <typename DeviceType>
TensorDevice<Derived, DeviceType> device(const DeviceType& device) {
return TensorDevice<Derived, DeviceType>(device, derived());
TensorDevice<Derived, DeviceType> device(const DeviceType& dev) {
return TensorDevice<Derived, DeviceType>(dev, derived());
}
protected:

2
unsupported/Eigen/CXX11/src/Tensor/TensorBlock.h
View File

@@ -89,7 +89,7 @@ EIGEN_STRONG_INLINE void MergeResourceRequirements(
// policy if block shapes/sizes conflict).
*block_shape = resources[0].block_shape;
*block_total_size = resources[0].block_total_size;
for (int i = 1; i < resources.size(); ++i) {
for (std::vector<TensorOpResourceRequirements>::size_type i = 1; i < resources.size(); ++i) {
if (resources[i].block_shape == TensorBlockShapeType::kSkewedInnerDims &&
*block_shape != TensorBlockShapeType::kSkewedInnerDims) {
*block_shape = TensorBlockShapeType::kSkewedInnerDims;

2
unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h
View File

@@ -274,8 +274,8 @@ struct TensorContractionEvaluatorBase
op.lhsExpression(), op.rhsExpression()), device),
m_rightImpl(choose(Cond<static_cast<int>(Layout) == static_cast<int>(ColMajor)>(),
op.rhsExpression(), op.lhsExpression()), device),
m_output_kernel(op.outputKernel()),
m_device(device),
m_output_kernel(op.outputKernel()),
m_result(NULL) {
EIGEN_STATIC_ASSERT((static_cast<int>(TensorEvaluator<LeftArgType, Device>::Layout) ==
static_cast<int>(TensorEvaluator<RightArgType, Device>::Layout)),

6
unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
View File

@@ -527,8 +527,8 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
Scalar* local = (Scalar*)m_device.allocate(kernel_sz);
typedef TensorEvalToOp<const KernelArgType> EvalTo;
EvalTo evalToTmp(local, m_kernelArg);
const bool PacketAccess = internal::IsVectorizable<Device, KernelArgType>::value;
internal::TensorExecutor<const EvalTo, Device, PacketAccess>::run(evalToTmp, m_device);
const bool Vectorize = internal::IsVectorizable<Device, KernelArgType>::value;
internal::TensorExecutor<const EvalTo, Device, Vectorize>::run(evalToTmp, m_device);
m_kernel = local;
m_local_kernel = true;
@@ -786,7 +786,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
};
EIGEN_DEVICE_FUNC TensorEvaluator(const XprType& op, const GpuDevice& device)
: m_inputImpl(op.inputExpression(), device), m_kernelArg(op.kernelExpression()), m_kernelImpl(op.kernelExpression(), device), m_indices(op.indices()), m_buf(NULL), m_kernel(NULL), m_local_kernel(false), m_device(device)
: m_inputImpl(op.inputExpression(), device), m_kernelImpl(op.kernelExpression(), device), m_kernelArg(op.kernelExpression()), m_indices(op.indices()), m_buf(NULL), m_kernel(NULL), m_local_kernel(false), m_device(device)
{
EIGEN_STATIC_ASSERT((static_cast<int>(TensorEvaluator<InputArgType, GpuDevice>::Layout) == static_cast<int>(TensorEvaluator<KernelArgType, GpuDevice>::Layout)), YOU_MADE_A_PROGRAMMING_MISTAKE);

23
unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h
View File

@@ -91,18 +91,31 @@ static EIGEN_STRONG_INLINE void wait_until_ready(SyncType* n) {
}
}
// An abstract interface to a device specific memory allocator.
class Allocator {
public:
virtual ~Allocator() {}
EIGEN_DEVICE_FUNC virtual void* allocate(size_t num_bytes) const = 0;
EIGEN_DEVICE_FUNC virtual void deallocate(void* buffer) const = 0;
};
// Build a thread pool device on top the an existing pool of threads.
struct ThreadPoolDevice {
// The ownership of the thread pool remains with the caller.
ThreadPoolDevice(ThreadPoolInterface* pool, int num_cores) : pool_(pool), num_threads_(num_cores) { }
ThreadPoolDevice(ThreadPoolInterface* pool, int num_cores, Allocator* allocator = nullptr)
: pool_(pool), num_threads_(num_cores), allocator_(allocator) { }
EIGEN_STRONG_INLINE void* allocate(size_t num_bytes) const {
return internal::aligned_malloc(num_bytes);
return allocator_ ? allocator_->allocate(num_bytes)
: internal::aligned_malloc(num_bytes);
}
EIGEN_STRONG_INLINE void deallocate(void* buffer) const {
internal::aligned_free(buffer);
if (allocator_) {
allocator_->deallocate(buffer);
} else {
internal::aligned_free(buffer);
}
}
EIGEN_STRONG_INLINE void* allocate_temp(size_t num_bytes) const {
@@ -275,9 +288,13 @@ struct ThreadPoolDevice {
// Thread pool accessor.
ThreadPoolInterface* getPool() const { return pool_; }
// Allocator accessor.
Allocator* allocator() const { return allocator_; }
private:
ThreadPoolInterface* pool_;
int num_threads_;
Allocator* allocator_;
};

4
unsupported/Eigen/CXX11/src/Tensor/TensorEvaluator.h
View File

@@ -126,7 +126,7 @@ struct TensorEvaluator
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void getResourceRequirements(
std::vector<internal::TensorOpResourceRequirements>* resources) const {}
std::vector<internal::TensorOpResourceRequirements>*) const {}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void block(TensorBlock* block) const {
assert(m_data != NULL);
@@ -255,7 +255,7 @@ struct TensorEvaluator<const Derived, Device>
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void getResourceRequirements(
std::vector<internal::TensorOpResourceRequirements>* resources) const {}
std::vector<internal::TensorOpResourceRequirements>*) const {}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void block(TensorBlock* block) const {
assert(m_data != NULL);

4
unsupported/Eigen/CXX11/src/Tensor/TensorForcedEval.h
View File

@@ -124,8 +124,8 @@ struct TensorEvaluator<const TensorForcedEvalOp<ArgType>, Device>
}
typedef TensorEvalToOp< const typename internal::remove_const<ArgType>::type > EvalTo;
EvalTo evalToTmp(m_buffer, m_op);
const bool PacketAccess = internal::IsVectorizable<Device, const ArgType>::value;
internal::TensorExecutor<const EvalTo, typename internal::remove_const<Device>::type, PacketAccess>::run(evalToTmp, m_device);
const bool Vectorize = internal::IsVectorizable<Device, const ArgType>::value;
internal::TensorExecutor<const EvalTo, typename internal::remove_const<Device>::type, Vectorize>::run(evalToTmp, m_device);
return true;
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void cleanup() {

3
unsupported/Eigen/CXX11/src/Tensor/TensorForwardDeclarations.h
View File

@@ -21,6 +21,7 @@ namespace Eigen {
template<typename T> struct MakePointer {
typedef T* Type;
typedef T& RefType;
typedef T ScalarType;
};
namespace internal{
@@ -97,7 +98,7 @@ template<typename XprType> class TensorForcedEvalOp;
template<typename ExpressionType, typename DeviceType> class TensorDevice;
template<typename Derived, typename Device> struct TensorEvaluator;
class NoOpOutputKernel;
struct NoOpOutputKernel;
struct DefaultDevice;
struct ThreadPoolDevice;

4
unsupported/Eigen/CXX11/src/Tensor/TensorShuffling.h
View File

@@ -61,8 +61,8 @@ class TensorShufflingOp : public TensorBase<TensorShufflingOp<Shuffle, XprType>
typedef typename Eigen::internal::traits<TensorShufflingOp>::StorageKind StorageKind;
typedef typename Eigen::internal::traits<TensorShufflingOp>::Index Index;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorShufflingOp(const XprType& expr, const Shuffle& shuffle)
: m_xpr(expr), m_shuffle(shuffle) {}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorShufflingOp(const XprType& expr, const Shuffle& shfl)
: m_xpr(expr), m_shuffle(shfl) {}
EIGEN_DEVICE_FUNC
const Shuffle& shufflePermutation() const { return m_shuffle; }

4
unsupported/Eigen/CXX11/src/Tensor/TensorTrace.h
View File

@@ -273,11 +273,11 @@ struct TensorEvaluator<const TensorTraceOp<Dims, ArgType>, Device>
Dimensions m_dimensions;
TensorEvaluator<ArgType, Device> m_impl;
// Initialize the size of the trace dimension
Index m_traceDim;
const Device& m_device;
array<bool, NumInputDims> m_reduced;
array<Index, NumReducedDims> m_reducedDims;
// Initialize the size of the trace dimension
Index m_traceDim;
array<Index, NumOutputDims> m_outputStrides;
array<Index, NumReducedDims> m_reducedStrides;
array<Index, NumOutputDims> m_preservedStrides;

4
unsupported/Eigen/CXX11/src/Tensor/TensorTraits.h
View File

@@ -59,6 +59,7 @@ struct traits<Tensor<Scalar_, NumIndices_, Options_, IndexType_> >
template <typename T> struct MakePointer {
typedef T* Type;
typedef T& RefType;
typedef T ScalarType;
};
typedef typename MakePointer<Scalar>::Type PointerType;
@@ -80,6 +81,7 @@ struct traits<TensorFixedSize<Scalar_, Dimensions, Options_, IndexType_> >
template <typename T> struct MakePointer {
typedef T* Type;
typedef T& RefType;
typedef T ScalarType;
};
typedef typename MakePointer<Scalar>::Type PointerType;
@@ -105,6 +107,8 @@ struct traits<TensorMap<PlainObjectType, Options_, MakePointer_> >
typedef MakePointer_<T> MakePointerT;
typedef typename MakePointerT::Type Type;
typedef typename MakePointerT::RefType RefType;
typedef typename MakePointerT::ScalarType ScalarType;
};
typedef typename MakePointer<Scalar>::Type PointerType;

5
unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h
View File

@@ -684,10 +684,15 @@ template<typename DerType> struct NumTraits<AutoDiffScalar<DerType> >
}
namespace std {
template <typename T>
class numeric_limits<Eigen::AutoDiffScalar<T> >
: public numeric_limits<typename T::Scalar> {};
template <typename T>
class numeric_limits<Eigen::AutoDiffScalar<T&> >
: public numeric_limits<typename T::Scalar> {};
} // namespace std
#endif // EIGEN_AUTODIFF_SCALAR_H

36
unsupported/Eigen/src/SpecialFunctions/SpecialFunctionsImpl.h
View File

@@ -193,6 +193,8 @@ struct lgamma_impl<float> {
#if !defined(EIGEN_GPU_COMPILE_PHASE) && (defined(_BSD_SOURCE) || defined(_SVID_SOURCE)) && !defined(__APPLE__)
int dummy;
return ::lgammaf_r(x, &dummy);
#elif defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
return cl::sycl::lgamma(x);
#else
return ::lgammaf(x);
#endif
@@ -206,6 +208,8 @@ struct lgamma_impl<double> {
#if !defined(EIGEN_GPU_COMPILE_PHASE) && (defined(_BSD_SOURCE) || defined(_SVID_SOURCE)) && !defined(__APPLE__)
int dummy;
return ::lgamma_r(x, &dummy);
#elif defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
return cl::sycl::lgamma(x);
#else
return ::lgamma(x);
#endif
@@ -423,13 +427,25 @@ struct erf_retval {
template <>
struct erf_impl<float> {
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE float run(float x) { return ::erff(x); }
static EIGEN_STRONG_INLINE float run(float x) {
#if defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
return cl::sycl::erf(x);
#else
return ::erff(x);
#endif
}
};
template <>
struct erf_impl<double> {
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE double run(double x) { return ::erf(x); }
static EIGEN_STRONG_INLINE double run(double x) {
#if defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
return cl::sycl::erf(x);
#else
return ::erf(x);
#endif
}
};
#endif // EIGEN_HAS_C99_MATH
@@ -456,13 +472,25 @@ struct erfc_retval {
template <>
struct erfc_impl<float> {
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE float run(const float x) { return ::erfcf(x); }
static EIGEN_STRONG_INLINE float run(const float x) {
#if defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
return cl::sycl::erfc(x);
#else
return ::erfcf(x);
#endif
}
};
template <>
struct erfc_impl<double> {
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE double run(const double x) { return ::erfc(x); }
static EIGEN_STRONG_INLINE double run(const double x) {
#if defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
return cl::sycl::erfc(x);
#else
return ::erfc(x);
#endif
}
};
#endif // EIGEN_HAS_C99_MATH

8
unsupported/test/cxx11_tensor_concatenation.cpp
View File

@@ -50,7 +50,13 @@ static void test_static_dimension_failure()
.reshape(Tensor<int, 3>::Dimensions(2, 3, 1))
.concatenate(right, 0);
Tensor<int, 2, DataLayout> alternative = left
.concatenate(right.reshape(Tensor<int, 2>::Dimensions{{{2, 3}}}), 0);
// Clang compiler break with {{{}}} with an ambigous error on copy constructor
// the variadic DSize constructor added for #ifndef EIGEN_EMULATE_CXX11_META_H.
// Solution:
// either the code should change to
// Tensor<int, 2>::Dimensions{{2, 3}}
// or Tensor<int, 2>::Dimensions{Tensor<int, 2>::Dimensions{{2, 3}}}
.concatenate(right.reshape(Tensor<int, 2>::Dimensions{{2, 3}}), 0);
}
template<int DataLayout>

45
unsupported/test/cxx11_tensor_thread_pool.cpp
View File

@@ -16,6 +16,25 @@
using Eigen::Tensor;
class TestAllocator : public Allocator {
public:
~TestAllocator() override {}
EIGEN_DEVICE_FUNC void* allocate(size_t num_bytes) const override {
const_cast<TestAllocator*>(this)->alloc_count_++;
return internal::aligned_malloc(num_bytes);
}
EIGEN_DEVICE_FUNC void deallocate(void* buffer) const override {
const_cast<TestAllocator*>(this)->dealloc_count_++;
internal::aligned_free(buffer);
}
int alloc_count() const { return alloc_count_; }
int dealloc_count() const { return dealloc_count_; }
private:
int alloc_count_ = 0;
int dealloc_count_ = 0;
};
void test_multithread_elementwise()
{
@@ -374,14 +393,14 @@ void test_multithread_random()
}
template<int DataLayout>
void test_multithread_shuffle()
void test_multithread_shuffle(Allocator* allocator)
{
Tensor<float, 4, DataLayout> tensor(17,5,7,11);
tensor.setRandom();
const int num_threads = internal::random<int>(2, 11);
ThreadPool threads(num_threads);
Eigen::ThreadPoolDevice device(&threads, num_threads);
Eigen::ThreadPoolDevice device(&threads, num_threads, allocator);
Tensor<float, 4, DataLayout> shuffle(7,5,11,17);
array<ptrdiff_t, 4> shuffles = {{2,1,3,0}};
@@ -398,6 +417,21 @@ void test_multithread_shuffle()
}
}
void test_threadpool_allocate(TestAllocator* allocator)
{
const int num_threads = internal::random<int>(2, 11);
const int num_allocs = internal::random<int>(2, 11);
ThreadPool threads(num_threads);
Eigen::ThreadPoolDevice device(&threads, num_threads, allocator);
for (int a = 0; a < num_allocs; ++a) {
void* ptr = device.allocate(512);
device.deallocate(ptr);
}
VERIFY(allocator != nullptr);
VERIFY_IS_EQUAL(allocator->alloc_count(), num_allocs);
VERIFY_IS_EQUAL(allocator->dealloc_count(), num_allocs);
}
EIGEN_DECLARE_TEST(cxx11_tensor_thread_pool)
{
@@ -424,6 +458,9 @@ EIGEN_DECLARE_TEST(cxx11_tensor_thread_pool)
CALL_SUBTEST_6(test_memcpy());
CALL_SUBTEST_6(test_multithread_random());
CALL_SUBTEST_6(test_multithread_shuffle<ColMajor>());
CALL_SUBTEST_6(test_multithread_shuffle<RowMajor>());
TestAllocator test_allocator;
CALL_SUBTEST_6(test_multithread_shuffle<ColMajor>(nullptr));
CALL_SUBTEST_6(test_multithread_shuffle<RowMajor>(&test_allocator));
CALL_SUBTEST_6(test_threadpool_allocate(&test_allocator));
}

16
unsupported/test/cxx11_tensor_trace.cpp
View File

@@ -37,7 +37,7 @@ static void test_all_dimensions_trace() {
VERIFY_IS_EQUAL(result1(), sum);
Tensor<float, 5, DataLayout> tensor2(7, 7, 7, 7, 7);
array<ptrdiff_t, 5> dims({{2, 1, 0, 3, 4}});
array<ptrdiff_t, 5> dims = { { 2, 1, 0, 3, 4 } };
Tensor<float, 0, DataLayout> result2 = tensor2.trace(dims);
VERIFY_IS_EQUAL(result2.rank(), 0);
sum = 0.0f;
@@ -52,7 +52,7 @@ template <int DataLayout>
static void test_simple_trace() {
Tensor<float, 3, DataLayout> tensor1(3, 5, 3);
tensor1.setRandom();
array<ptrdiff_t, 2> dims1({{0, 2}});
array<ptrdiff_t, 2> dims1 = { { 0, 2 } };
Tensor<float, 1, DataLayout> result1 = tensor1.trace(dims1);
VERIFY_IS_EQUAL(result1.rank(), 1);
VERIFY_IS_EQUAL(result1.dimension(0), 5);
@@ -67,7 +67,7 @@ static void test_simple_trace() {
Tensor<float, 4, DataLayout> tensor2(5, 5, 7, 7);
tensor2.setRandom();
array<ptrdiff_t, 2> dims2({{2, 3}});
array<ptrdiff_t, 2> dims2 = { { 2, 3 } };
Tensor<float, 2, DataLayout> result2 = tensor2.trace(dims2);
VERIFY_IS_EQUAL(result2.rank(), 2);
VERIFY_IS_EQUAL(result2.dimension(0), 5);
@@ -82,7 +82,7 @@ static void test_simple_trace() {
}
}
array<ptrdiff_t, 2> dims3({{1, 0}});
array<ptrdiff_t, 2> dims3 = { { 1, 0 } };
Tensor<float, 2, DataLayout> result3 = tensor2.trace(dims3);
VERIFY_IS_EQUAL(result3.rank(), 2);
VERIFY_IS_EQUAL(result3.dimension(0), 7);
@@ -99,7 +99,7 @@ static void test_simple_trace() {
Tensor<float, 5, DataLayout> tensor3(3, 7, 3, 7, 3);
tensor3.setRandom();
array<ptrdiff_t, 3> dims4({{0, 2, 4}});
array<ptrdiff_t, 3> dims4 = { { 0, 2, 4 } };
Tensor<float, 2, DataLayout> result4 = tensor3.trace(dims4);
VERIFY_IS_EQUAL(result4.rank(), 2);
VERIFY_IS_EQUAL(result4.dimension(0), 7);
@@ -116,7 +116,7 @@ static void test_simple_trace() {
Tensor<float, 5, DataLayout> tensor4(3, 7, 4, 7, 5);
tensor4.setRandom();
array<ptrdiff_t, 2> dims5({{1, 3}});
array<ptrdiff_t, 2> dims5 = { { 1, 3 } };
Tensor<float, 3, DataLayout> result5 = tensor4.trace(dims5);
VERIFY_IS_EQUAL(result5.rank(), 3);
VERIFY_IS_EQUAL(result5.dimension(0), 3);
@@ -140,7 +140,7 @@ template<int DataLayout>
static void test_trace_in_expr() {
Tensor<float, 4, DataLayout> tensor(2, 3, 5, 3);
tensor.setRandom();
array<ptrdiff_t, 2> dims({{1, 3}});
array<ptrdiff_t, 2> dims = { { 1, 3 } };
Tensor<float, 2, DataLayout> result(2, 5);
result = result.constant(1.0f) - tensor.trace(dims);
VERIFY_IS_EQUAL(result.rank(), 2);
@@ -168,4 +168,4 @@ EIGEN_DECLARE_TEST(cxx11_tensor_trace) {
CALL_SUBTEST(test_simple_trace<RowMajor>());
CALL_SUBTEST(test_trace_in_expr<ColMajor>());
CALL_SUBTEST(test_trace_in_expr<RowMajor>());
}
}