MAGEEC Machine Learner. More...

#include <mageec-ml.h>

Inheritance diagram for mageec::mageec_ml:

Public Member Functions
virtual	~mageec_ml ()

virtual int	init (std::string dbfilename)
	Initilizes the MAGEEC Machine Learner. More...

virtual int	init (std::string compiler_version, std::string compiler_target)
	Initilizes the MAGEEC Machine Learner. More...

void	new_file (void)
	Prepares Machine Learner for a new file. More...

void	end_file (void)
	Declares to Machine Learner processing has finished. More...

void	finish (void)
	Hook for Machine Learner to e.g. More...

void	provide_features (std::vector< mageec_feature * > features)
	Provide a set of program features to the machine learner to base decisions on. More...

std::vector< mageec_pass >	get_passes ()
	Returns a set of passes the machine learner has chosen the compiler to execute. More...

std::vector< mageec_pass * >	all_passes ()
	Returns complete list of passes known to machine learner for disabling upon initialisation. More...

virtual decision	make_decision (mageec_pass pass, std::vector< mageec_feature > features)
	Make a single function/pass decision based on provided feature vector. More...

void	process_results ()
	Informs the machine learner that it can process all new data to update internal decisions. More...

void	add_result (std::vector< mageec_feature * > features, std::vector< mageec_pass * > passes, int64_t metric, bool good)
	Adds a result to the machine learner database. More...

void	add_result (result res)
	Adds a result to the machine learner database. More...

void	raw_result (std::vector< mageec_feature * > features, std::vector< mageec_pass * > passes, std::vector< int64_t > metrics, bool good)
	Adds a set of result points to machine learner raw database. More...

Detailed Description

MAGEEC Machine Learner.

Definition at line 37 of file mageec-ml.h.

Constructor & Destructor Documentation

virtual mageec::mageec_ml::~mageec_ml ( )

inlinevirtual

Definition at line 42 of file mageec-ml.h.

Member Function Documentation

void mageec::mageec_ml::add_result	(	std::vector< mageec_feature * >	features,
		std::vector< mageec_pass * >	passes,
		int64_t	metric,
		bool	good
	)

Adds a result to the machine learner database.

(Depending on the algorithm used, this may be for training purposes only)

Parameters

features	List of features for program.
passes	List of passes executed during compilation.
metric	Performance metric for optimisation.
good	Whether the result of compilation was valid (for machine learners that understand bad combinations.)

void mageec_ml::add_result ( result res)

Adds a result to the machine learner database.

Parameters

res	Result structure containing program, pass and metric info

Definition at line 89 of file ml.cpp.

std::vector< mageec_pass * > mageec_ml::all_passes ( void )

Returns complete list of passes known to machine learner for disabling upon initialisation.

Returns: list of all passes.

Definition at line 67 of file ml.cpp.

void mageec_ml::end_file ( void )

Declares to Machine Learner processing has finished.

Definition at line 55 of file ml.cpp.

void mageec_ml::finish ( void )

Hook for Machine Learner to e.g.

disconnect for any resources.

Definition at line 60 of file ml.cpp.

std::vector<mageec_pass> mageec::mageec_ml::get_passes ( )

Returns a set of passes the machine learner has chosen the compiler to execute.

Returns: list of passes to execute.

int mageec_ml::init ( std::string dbfilename)

virtual

Initilizes the MAGEEC Machine Learner.

Parameters

dbfilename Path of database file to open.

Returns: 0 if MAGEEC successfully set up, 1 otherwise.

Reimplemented in mageec::file_ml.

Definition at line 34 of file ml.cpp.

int mageec_ml::init	(	std::string	compiler_version,
		std::string	compiler_target
	)

virtual

Initilizes the MAGEEC Machine Learner.

Parameters

compiler_version	Compiler and version, e.g. GCC-4.8.
compiler_target	Compiler target, e.g. arm-none-gnueabi.

Returns: 0 if MAGEEC successfully set up, 1 otherwise.

Reimplemented in mageec::file_ml.

Definition at line 44 of file ml.cpp.

decision mageec_ml::make_decision	(	mageec_pass *	pass,
		std::vector< mageec_feature * >	features
	)

virtual

Make a single function/pass decision based on provided feature vector.

Parameters

pass	Pass to make a decision about.
features	Feature vector of function to run pass on.

Returns: decision

Reimplemented in mageec::file_ml.

Definition at line 96 of file ml.cpp.

void mageec_ml::new_file ( void )

Prepares Machine Learner for a new file.

Definition at line 50 of file ml.cpp.

void mageec_ml::process_results ( )

Informs the machine learner that it can process all new data to update internal decisions.

Definition at line 155 of file ml.cpp.

void mageec::mageec_ml::provide_features ( std::vector< mageec_feature * > features)

Provide a set of program features to the machine learner to base decisions on.

Parameters

features List of features for program.

void mageec::mageec_ml::raw_result	(	std::vector< mageec_feature * >	features,
		std::vector< mageec_pass * >	passes,
		std::vector< int64_t >	metrics,
		bool	good
	)

Adds a set of result points to machine learner raw database.

Parameters

features	List of features for program.
passes	List of passes executed during compilation.
metrics	Performance metrics.
good	Whether the result of compilation was valid (for machine learners that understand bad combinations.)

The documentation for this class was generated from the following files:

include/mageec/mageec-ml.h
ml/ml.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation