24.02.1
|
Go to the documentation of this file.
24 #ifndef ACL_SRC_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADSKETCHIMPL_H
25 #define ACL_SRC_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADSKETCHIMPL_H
37 namespace experimental
39 namespace dynamic_fusion
50 : _context{
context}, _comp_services{}, _component_graph{_context, &_comp_services}, _operator_group{}
74 return _component_graph;
79 return _component_graph;
84 return _operator_group;
89 return _operator_group;
134 #endif // ACL_SRC_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADSKETCHIMPL_H
Context * context()
Get workload context.
Implementation(Context *context)
Constructor.
ITensorInfo * create_virtual_tensor()
Create a virtual (see MemoryType) tensor info and save it.
const MemoryDescriptorMap & mem_map() const
Get memory descriptor registry.
const GpuOperatorGroup & operator_group() const
Get operator group.
A multi-input (tensors), multi-output (tensors) acyclic directed graph of gpu kernel components Its m...
ITensorInfo * create_auxiliary_tensor(const ITensorInfo &tensor_info)
Create an auxiliary (see MemoryType) tensor info and save it.
GpuWorkloadSourceCode generate_source_code() const
Generate GpuWorkloadSourceCode from the workload sketch.
GpuKernelComponentStream fuse(const MemoryDescriptorMap &mem_map) const
Perform component fusion and serialize the graph into a stream of component groups.
GpuWorkloadSourceCode write_workload_code()
Generate and assemble GpuWorkloadSourceCode from the stream.
ITensorInfo * create_virtual_tensor()
Create a virtual (see MemoryType) tensor info and save it.
Impl & implementation()
Get the internal implementation.
Implementation & operator=(const Implementation &impl)=delete
Prevent instances of this class from being copied.
ITensorInfo * get_tensor_info(ITensorInfo::Id id)
const GpuKernelComponentGraph & component_graph() const
Get component graph.
Provide context necessary for the creation and configuration of a workload e.g.
int32_t Id
An id that uniquely identifies an ITensorInfo within some domain (e.g.
const Context * context() const
Get workload context.
GpuOperatorGroup & operator_group()
Get operator group.
Hold the generated kernel source code and other information required to compile and run the workload.
ITensorInfo * get_tensor_info(ITensorInfo::Id id)
Get tensor info created by this context, from id.
Copyright (c) 2017-2024 Arm Limited.
Services that are used throughout the creation phase of workload code.
GpuKernelComponentGraph & component_graph()
Get component graph.
Store the tensor's metadata.
Internal implementation of GpuWorkloadSketch.
TensorInfo tensor_info
Associated tensor info.
ITensorInfo * create_auxiliary_tensor(const ITensorInfo &tensor_info)
Create an auxiliary (see MemoryType) tensor info and save it.
A linear sequence of operators to be fused in a workload For the time being, this class is only used ...