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 some typos found

Browse Source
This commit is contained in:
Kolja Brix
2021-09-23 15:22:00 +00:00
committed by Antonio Sánchez
parent 76bb29c0c2
commit afa616bc9e
19 changed files with 36 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -107,7 +107,7 @@ struct TensorEvaluator<const TensorBroadcastingOp<Broadcast, ArgType>, Device>
typedef typename XprType::CoeffReturnType CoeffReturnType;
typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType;
static const int PacketSize = PacketType<CoeffReturnType, Device>::size;
protected: // all the non-static fields must have the same access control, otherwise the TensorEvaluator wont be standard layout;
protected: // all the non-static fields must have the same access control, otherwise the TensorEvaluator won't be standard layout;
bool isCopy, nByOne, oneByN;
public:
typedef StorageMemory<CoeffReturnType, Device> Storage;

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

@@ -112,7 +112,7 @@ struct TTPanelSize {
// BC : determines if supporting bank conflict is required
static EIGEN_CONSTEXPR bool BC = true;
// DoubleBuffer: determines if double buffering technique should be used (This can be disabled by
// EIGEN_SYCL_DISABLE_DOUBLE_BUFFER macro when the device doesnot have sufficient local memory)
// EIGEN_SYCL_DISABLE_DOUBLE_BUFFER macro when the device does not have sufficient local memory)
static EIGEN_CONSTEXPR bool DoubleBuffer =
#ifdef EIGEN_SYCL_DISABLE_DOUBLE_BUFFER
false;
@@ -430,7 +430,7 @@ struct ThreadProperties {
Otherwise, the result of contraction will be written iin a temporary buffer. This is the case when Tall/Skinny
contraction is used. So in this case, a final reduction step is required to compute final output.
* \tparam contraction_tp: it is an enum value representing whether the local memroy/no local memory implementation of
* \tparam contraction_tp: it is an enum value representing whether the local memory/no local memory implementation of
the algorithm to be used
*
* \param scratch: local memory containing tiles of LHS and RHS tensors for each work-group
@@ -495,7 +495,7 @@ class TensorContractionKernel {
* the TiledMemory for both local and private memory, the MemHolder structs is used as a helper to abstract out
* different type of memory needed when local/no_local memory computation is called.
*
* \tparam contraction_type: it is an enum value representing whether the local memroy/no local memory implementation
* \tparam contraction_type: it is an enum value representing whether the local memory/no local memory implementation
of the algorithm to be used
* \tparam the private memory size
* \param ptr the tile memory pointer type

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

@@ -897,7 +897,7 @@ struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgT
} else {
// If we can't guarantee that all kernels in `k` slice will be
// executed sequentially in current thread, it's no longer safe to use
// thread local memory in followig slices along the k dimensions.
// thread local memory in following slices along the k dimensions.
eigen_assert(k > 0);
can_use_thread_local_packed_[n].store(false,
std::memory_order_relaxed);

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

@@ -715,7 +715,7 @@ class QueueInterface {
EIGEN_STRONG_INLINE int majorDeviceVersion() const { return 1; }
EIGEN_STRONG_INLINE unsigned long maxSyclThreadsPerMultiProcessor() const {
// OpenCL doesnot have such concept
// OpenCL does not have such a concept
return 2;
}
@@ -1035,7 +1035,7 @@ struct SyclDevice : public SyclDeviceBase {
return queue_stream()->maxWorkItemSizes();
}
EIGEN_STRONG_INLINE unsigned long maxSyclThreadsPerMultiProcessor() const {
// OpenCL doesnot have such concept
// OpenCL does not have such a concept
return queue_stream()->maxSyclThreadsPerMultiProcessor();
}
EIGEN_STRONG_INLINE size_t sharedMemPerBlock() const {

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

@@ -133,7 +133,7 @@ template <typename T> class UniformRandomGenerator {
m_state = PCG_XSH_RS_state(seed);
#ifdef EIGEN_USE_SYCL
// In SYCL it is not possible to build PCG_XSH_RS_state in one step.
// Therefor, we need two step to initializate the m_state.
// Therefore, we need two steps to initializate the m_state.
// IN SYCL, the constructor of the functor is s called on the CPU
// and we get the clock seed here from the CPU. However, This seed is
//the same for all the thread. As unlike CUDA, the thread.ID, BlockID, etc is not a global function.
@@ -246,7 +246,7 @@ template <typename T> class NormalRandomGenerator {
m_state = PCG_XSH_RS_state(seed);
#ifdef EIGEN_USE_SYCL
// In SYCL it is not possible to build PCG_XSH_RS_state in one step.
// Therefor, we need two steps to initializate the m_state.
// Therefore, we need two steps to initializate the m_state.
// IN SYCL, the constructor of the functor is s called on the CPU
// and we get the clock seed here from the CPU. However, This seed is
//the same for all the thread. As unlike CUDA, the thread.ID, BlockID, etc is not a global function.

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

@@ -25,7 +25,7 @@
* buffer is given as an input and all the threads within a work-group scan and
* reduces the boundaries between the blocks (generated from the previous
* kernel). and write the data on the temporary buffer. If the second kernel is
* required, the third and final kerenl (ScanAdjustmentKernelFunctor) will
* required, the third and final kernel (ScanAdjustmentKernelFunctor) will
* adjust the final result into the output buffer.
* The original algorithm for the parallel prefix sum can be found here:
*