GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-12 03:39:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

Misc improvements and optimizations

Browse Source
This commit is contained in:
Benoit Steiner 2015-07-01 13:59:11 -07:00
parent 4ed213f97b
commit 1e911b276c
3 changed files with 8 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -510,7 +510,8 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
Scalar* local = (Scalar*)m_device.allocate(kernel_sz); Scalar* local = (Scalar*)m_device.allocate(kernel_sz);
typedef TensorEvalToOp<const KernelArgType> EvalTo; typedef TensorEvalToOp<const KernelArgType> EvalTo;
EvalTo evalToTmp(local, m_kernelArg); EvalTo evalToTmp(local, m_kernelArg);
internal::TensorExecutor<const EvalTo, Device>::run(evalToTmp, m_device); const bool PacketAccess = internal::IsVectorizable<Device, KernelArgType>::value;
internal::TensorExecutor<const EvalTo, Device, PacketAccess>::run(evalToTmp, m_device);
m_kernel = local; m_kernel = local;
m_local_kernel = true; m_local_kernel = true;
@ -815,7 +816,8 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
Scalar* local = (Scalar*)m_device.allocate(kernel_sz); Scalar* local = (Scalar*)m_device.allocate(kernel_sz);
typedef TensorEvalToOp<const KernelArgType> EvalTo; typedef TensorEvalToOp<const KernelArgType> EvalTo;
EvalTo evalToTmp(local, m_kernelArg); EvalTo evalToTmp(local, m_kernelArg);
internal::TensorExecutor<const EvalTo, GpuDevice>::run(evalToTmp, m_device); const bool PacketAccess = internal::IsVectorizable<GpuDevice, KernelArgType>::value;
internal::TensorExecutor<const EvalTo, GpuDevice, PacketAccess>::run(evalToTmp, m_device);
m_kernel = local; m_kernel = local;
m_local_kernel = true; m_local_kernel = true;

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

@ -113,9 +113,9 @@ struct TensorEvaluator<const TensorEvalToOp<ArgType>, Device>
EIGEN_DEVICE_FUNC const Dimensions& dimensions() const { return m_impl.dimensions(); } EIGEN_DEVICE_FUNC const Dimensions& dimensions() const { return m_impl.dimensions(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(CoeffReturnType*) { EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(CoeffReturnType* scalar) {
m_impl.evalSubExprsIfNeeded(NULL); eigen_assert(scalar == NULL);
return true; return m_impl.evalSubExprsIfNeeded(m_buffer);
} }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void evalScalar(Index i) { EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void evalScalar(Index i) {

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

@ -116,7 +116,7 @@ struct TensorEvaluator<const TensorForcedEvalOp<ArgType>, Device>
} }
typedef TensorEvalToOp<const ArgType> EvalTo; typedef TensorEvalToOp<const ArgType> EvalTo;
EvalTo evalToTmp(m_buffer, m_op); EvalTo evalToTmp(m_buffer, m_op);
static const bool PacketAccess = internal::IsVectorizable<Device, ArgType>::value; const bool PacketAccess = internal::IsVectorizable<Device, ArgType>::value;
internal::TensorExecutor<const EvalTo, Device, PacketAccess>::run(evalToTmp, m_device); internal::TensorExecutor<const EvalTo, Device, PacketAccess>::run(evalToTmp, m_device);
m_impl.cleanup(); m_impl.cleanup();
return true; return true;