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

Fix calls to device functions from host code

Browse Source 
(cherry picked from commit 972cf0c28a)
This commit is contained in:
Nathan Luehr
2021-05-11 22:47:49 +00:00
committed by Rasmus Munk Larsen
parent d1825cbb68
commit 82f13830e6
31 changed files with 182 additions and 210 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
View File

@@ -320,7 +320,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
typedef internal::TensorBlockNotImplemented TensorBlock;
//===--------------------------------------------------------------------===//
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorEvaluator(const XprType& op, const Device& device)
EIGEN_STRONG_INLINE TensorEvaluator(const XprType& op, const Device& device)
: m_inputImpl(op.inputExpression(), device), m_kernelImpl(op.kernelExpression(), device), m_kernelArg(op.kernelExpression()), m_kernel(NULL), m_local_kernel(false), m_device(device)
{
EIGEN_STATIC_ASSERT((static_cast<int>(TensorEvaluator<InputArgType, Device>::Layout) == static_cast<int>(TensorEvaluator<KernelArgType, Device>::Layout)), YOU_MADE_A_PROGRAMMING_MISTAKE);
@@ -384,12 +384,12 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return m_dimensions; }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(Scalar*) {
EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(Scalar*) {
m_inputImpl.evalSubExprsIfNeeded(NULL);
preloadKernel();
return true;
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void cleanup() {
EIGEN_STRONG_INLINE void cleanup() {
m_inputImpl.cleanup();
if (m_local_kernel) {
m_device.deallocate((void*)m_kernel);
@@ -797,7 +797,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
typedef internal::TensorBlockNotImplemented TensorBlock;
//===--------------------------------------------------------------------===//
EIGEN_DEVICE_FUNC TensorEvaluator(const XprType& op, const GpuDevice& device)
TensorEvaluator(const XprType& op, const GpuDevice& 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);