tc::CudaCache::CachedEntry Struct Reference

#include <compilation_cache.h>

Collaboration diagram for tc::CudaCache::CachedEntry:

Classes
struct	Key
struct	Values

Public Member Functions
	CachedEntry (const std::string &id, const std::string &kernelSpecializedName, const std::vector< int > &kernelParameters, const Grid &grid, const Block &block, const MappingOptions &mappingOptions, const std::vector< const DLTensor * > &inputs, const std::vector< const DLTensor * > &outputs, const std::string &cudaSource, const std::string &deviceStr)
	CachedEntry (const CudaCacheEntryProto &buf)
CudaCacheEntryProto	toProtobuf () const

Public Attributes
Key	key
Values	values

Detailed Description

A CudaCache holds multiple CachedEntry's. Each CachedEntry is split to two conceptual parts the key and the values. The values are: the specialized (wrt inputs) Cuda source code, the kernel's specialized name, the kernel parameters, the Cuda block and grid dimensions The key is: the kernel/op's unique id (string), the specialized input dimensions, the isl options when the kernel was optimized, the target architecture (string), tc's version (string),

Constructor & Destructor Documentation

tc::CudaCache::CachedEntry::CachedEntry	(	const std::string &	id,
		const std::string &	kernelSpecializedName,
		const std::vector< int > &	kernelParameters,
		const Grid &	grid,
		const Block &	block,
		const MappingOptions &	mappingOptions,
		const std::vector< const DLTensor * > &	inputs,
		const std::vector< const DLTensor * > &	outputs,
		const std::string &	cudaSource,
		const std::string &	deviceStr
	)

tc::CudaCache::CachedEntry::CachedEntry

(

const CudaCacheEntryProto &

buf

)

Member Function Documentation

CudaCacheEntryProto tc::CudaCache::CachedEntry::toProtobuf

(

)

const

Member Data Documentation

Key tc::CudaCache::CachedEntry::key

Values tc::CudaCache::CachedEntry::values

---

The documentation for this struct was generated from the following file:

• include/tc/core/compilation_cache.h