Deprecated List

Member docc::target::et::et_scheduling_passes (sdfg::builder::StructuredSDFGBuilder &builder, sdfg::analysis::AnalysisManager &analysis_manager, const TargetOptions &options)

Legacy. Use the handler on the DoccTarget instead, which will be auto-discoverable

Member sdfg::builder::StructuredSDFGBuilder::move ()

the unique ptr required SDFGs were removed from the registry during modification to make sense. This builder does not change the pointer. But this will release any references the builder has to the SDFG to end any modification

Member sdfg::codegen::CodeGenerator::as_source (const std::filesystem::path &header_path, const std::filesystem::path &source_path)=0

this requires generate() being run manually prior and will output directly into filepaths for header and source at the same time. Which is quite restrictive. New parallel builder require [emit_header] and [emit_main_source] instead

Member sdfg::codegen::LibraryNodeDispatcherRegistry::register_library_node_dispatcher (std::string library_node_code, LibraryNodeDispatcherFn fn)

to decouple how this registry indexes from its users

Member sdfg::codegen::MapDispatcher

Use the new nyme [SchedTypeMapDispatcher]. The old name suggests it is perhaps a base class for other dispatchers, which it is not

Member sdfg::codegen::ReduceDispatcher

Use the new name [SchedTypeReduceDispatcher]. The old name suggests it is perhaps a base class for other dispatchers, which it is not

Member sdfg::cuda::register_cuda_plugin ()

use the variant with explicit context

Class sdfg::cuda::ScheduleType_CUDA

This class is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Member sdfg::data_flow::DataFlowGraph::in_edge (const data_flow::AccessNode &node) const

AccessNodes can have multiple input edges, because the node is not an operation, it is almost irrelevant to the operation represented by the edges Will throw if there are more than 1 edge on the AccessNode

Member sdfg::data_flow::LibraryNode::implementation_type ()

use setter allow for future type-changes of ImplementationType that can still work with the existing API

Member sdfg::gpu::gpu_block_size (const structured_control_flow::ScheduleType &schedule)

This function is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Member sdfg::gpu::gpu_block_size (structured_control_flow::ScheduleType &schedule, const symbolic::Expression block_size)

This function is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Member sdfg::gpu::gpu_dimension (const structured_control_flow::ScheduleType &schedule)

This function is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Member sdfg::gpu::gpu_dimension (structured_control_flow::ScheduleType &schedule, const GPUDimension &dimension)

This function is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Class sdfg::gpu::ScheduleType_GPU_Base< Derived >

This class is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Member sdfg::math::tensor::find_usable_input_access_node (data_flow::DataFlowGraph &dataflow, data_flow::LibraryNode &node, const std::string &input_conn)

not compatible with new expansion scheme

Member sdfg::parallelization::register_parallelization_plugin ()

use the variant with explicit context

Class sdfg::passes::scheduler::SchedulerRegistry

Global state and superfluous, do not use

Member sdfg::plugins::Context::Context (serializer::LibraryNodeSerializerRegistry &library_node_serializer_registry, codegen::NodeDispatcherRegistry &node_dispatcher_registry, codegen::MapDispatcherRegistry &map_dispatcher_registry, codegen::ReduceDispatcherRegistry &reduce_dispatcher_registry, codegen::LibraryNodeDispatcherRegistry &library_node_dispatcher_registry, passes::scheduler::SchedulerRegistry &scheduler_registry, tiles::TileTargetRegistry &tile_target_registry)

Member sdfg::plugins::Context::library_node_dispatcher_registry

Member sdfg::plugins::Context::library_node_serializer_registry

Member sdfg::plugins::Context::map_dispatcher_registry

Member sdfg::plugins::Context::node_dispatcher_registry

Member sdfg::plugins::Context::reduce_dispatcher_registry

Member sdfg::plugins::Context::scheduler_registry

Member sdfg::rocm::register_rocm_plugin ()

use the variant with explicit context

Class sdfg::rocm::ScheduleType_ROCM

This class is deprecated and will be removed in future versions. Use the new GPU schedule type classes instead.

Member sdfg::tiles::register_tiles_plugin ()

use the variant with explicit context

Member sdfg::transformations::MultiLevelTiling::middle_loop ()

Kept for backwards compatibility

Member sdfg::types::Tensor::offset () const

use TensorLayout

Member sdfg::types::Tensor::shape () const

use TensorLayout

Member sdfg::types::Tensor::strides () const

use TensorLayout

Member sdfg::types::Tensor::Tensor (const Scalar &element_type, const symbolic::MultiExpression &shape)

use TensorLayout

Member sdfg::types::Tensor::Tensor (const Scalar &element_type, const symbolic::MultiExpression &shape, const symbolic::MultiExpression &strides, const symbolic::Expression &offset=symbolic::zero())

use TensorLayout

Member sdfg::types::Tensor::Tensor (StorageType storage_type, size_t alignment, const std::string &initializer, const Scalar &element_type, const symbolic::MultiExpression &shape, const symbolic::MultiExpression &strides, const symbolic::Expression &offset=symbolic::zero())

use TensorLayout