CpuTensorOps.Transfer.ixx File Reference

CPU tensor transfer operations partition. More...

#include <cstring>
#include <algorithm>
#include <memory>
#include <span>
#include <type_traits>
#include <stdexcept>
import Serialization.Tensor;
import Compute.IExecutionContext;
import Compute.ExecutionContext;
import Compute.CpuTensorDataTypeTraits;
import Dnn.TensorDataTypeTraits;
import Compute.CpuMemoryResource;
import Compute.DeviceType;
import Dnn.TensorDataType;
import Dnn.TensorDataTypeMap;
import Dnn.Tensor;

Classes
struct	Mila::Dnn::Compute::Cpu::TransferOps
	CPU specialization of TensorOps for transfer operations. More...

Namespaces
namespace	Mila
	Mila main API namespace.
namespace	Mila::Dnn
namespace	Mila::Dnn::Compute
namespace	Mila::Dnn::Compute::Cpu
namespace	Mila::Dnn::Compute::Cpu::Detail

Functions
template<TensorDataType TDataType>
void	Mila::Dnn::Compute::Cpu::Detail::copyHostToHostImpl (const void *src_data, void *dst_data, size_t count)
	Fast raw-memory copy for host-accessible tensors of same abstract type.
template<TensorDataType TSrcDataType, TensorDataType TDstDataType>
void	Mila::Dnn::Compute::Cpu::Detail::copyHostToHostWithConversionImpl (const void *src_data, void *dst_data, size_t count)
	Element-wise host copy with conversion between abstract data types.

Detailed Description

CPU tensor transfer operations partition.

Provides CPU-specific implementations of tensor transfer operations for host-accessible memory. All operations execute synchronously with no device synchronization overhead.

ExecutionContext handling:

• Accepts ExecutionContext parameter for API consistency with device implementations
• Parameter is unused for CPU operations (all operations are synchronous)
• No stream management needed on CPU