CostModel.cpp File Reference

#include <mutable/catalog/CostModel.hpp>
#include "storage/ColumnStore.hpp"
#include "storage/store_manip.hpp"
#include "util/GridSearch.hpp"
#include "util/stream.hpp"
#include <mutable/catalog/TrainedCostFunction.hpp>
#include <mutable/mutable.hpp>
#include <type_traits>

Include dependency graph for CostModel.cpp:

Go to the source code of this file.

Macros
#define	DEFINE(TYPE)

Functions
uint64_t	get_column_offset_in_bytes (const DataLayout &layout, std::size_t idx)
	Returns the offset in bytes of the idx-th column in the DataLayout layout which is considered to represent a PAX-layout.
void	save_csv (const std::string &csv_path, const Eigen::MatrixXd &matrix, const std::string &header="")
	Save matrix in a csv file.
Timer::duration	time_select_query_execution (Database &DB, const std::string &input)
	Executes the given query and returns the median of query execution times for all SELECT queries.
template<typename T >
std::pair< Eigen::MatrixXd, Eigen::VectorXd >	generate_training_suite_filter ()
	Generates data for training filter models.
template<typename T >
std::pair< Eigen::MatrixXd, Eigen::VectorXd >	generate_training_suite_group_by ()
	Generates data for training group-by-models.
template<typename T >
std::pair< Eigen::MatrixXd, Eigen::VectorXd >	generate_training_suite_join ()
	Generates data for training join-models.
	DEFINE (int8_t)
	DEFINE (int16_t)
	DEFINE (int32_t)
	DEFINE (int64_t)
	DEFINE (float)
	DEFINE (double)

Variables
static constexpr std::size_t	NUM_DISTINCT_VALUES_IN_FILTER_EXPERIMENT = 100
static constexpr unsigned	DEFAULT_FILTER_POLYNOMIAL_DEGREE = 9
constexpr unsigned	NUM_REPETITIONS = 5
	The number of times a benchmark should be repeated to reduce noise in the data.

Macro Definition Documentation

DEFINE

#define DEFINE

(

TYPE

)

Value:

template CostModel CostModelFactory::generate_filter_cost_model<TYPE>(unsigned degree, const char *csv_folder_path); \
template CostModel CostModelFactory::generate_group_by_cost_model<TYPE>(const char *csv_folder_path); \
template CostModel CostModelFactory::generate_join_cost_model<TYPE>(const char *csv_folder_path)

Definition at line 598 of file CostModel.cpp.

Function Documentation

DEFINE() [1/6]

DEFINE

(

double

)

DEFINE() [2/6]

DEFINE

(

float

)

DEFINE() [3/6]

DEFINE

(

int16_t

)

DEFINE() [4/6]

DEFINE

(

int32_t

)

DEFINE() [5/6]

DEFINE

(

int64_t

)

DEFINE() [6/6]

DEFINE

(

int8_t

)

generate_training_suite_filter()

template<typename T >

std::pair< Eigen::MatrixXd, Eigen::VectorXd > generate_training_suite_filter

(

)

Generates data for training filter models.

Returns a pair of matrices for the features and the target.

Definition at line 93 of file CostModel.cpp.

References m::Catalog::add_database(), m::Database::add_table(), m::Catalog::create_store(), m::gs::LinearSpace< T >::delta(), m::Catalog::drop_database(), m::fill_uniform(), m::generate_primary_keys(), m::Catalog::Get(), get_column_offset_in_bytes(), m::Type::Get_Integer(), m::gs::LinearSpace< T >::hi(), m::gs::LinearSpace< T >::lo(), M_insist, m::storage::PAXLayoutFactory::NTuples, NUM_DISTINCT_VALUES_IN_FILTER_EXPERIMENT, m::gs::GridSearch< Spaces >::num_points(), m::Catalog::pool(), m::Table::push_back(), m::gs::LinearSpace< T >::sequence(), m::set_all_not_null(), m::T(), time_select_query_execution(), and m::Catalog::unset_database_in_use().

generate_training_suite_group_by()

template<typename T >

std::pair< Eigen::MatrixXd, Eigen::VectorXd > generate_training_suite_group_by

(

)

Generates data for training group-by-models.

Returns a pair of matrices for the features and the target.

Definition at line 190 of file CostModel.cpp.

References m::Catalog::add_database(), m::Database::add_table(), m::Catalog::create_store(), m::fill_uniform(), m::generate_primary_keys(), m::Catalog::Get(), get_column_offset_in_bytes(), m::Type::Get_Integer(), m::gs::LinearSpace< T >::hi(), M_insist, m::storage::PAXLayoutFactory::NTuples, m::gs::GridSearch< Spaces >::num_points(), m::Catalog::pool(), m::Table::push_back(), m::gs::LinearSpace< T >::sequence(), m::set_all_not_null(), m::T(), and time_select_query_execution().

generate_training_suite_join()

template<typename T >

std::pair< Eigen::MatrixXd, Eigen::VectorXd > generate_training_suite_join

(

)

Generates data for training join-models.

Results are written in filepath.

Definition at line 322 of file CostModel.cpp.

References m::Catalog::add_database(), m::Database::add_table(), m::and, m::Catalog::create_store(), m::fill_uniform(), m::generate_primary_keys(), m::Catalog::Get(), get_column_offset_in_bytes(), m::Type::Get_Integer(), M_insist, m::storage::PAXLayoutFactory::NTuples, m::gs::GridSearch< Spaces >::num_points(), m::Catalog::pool(), m::Table::push_back(), m::gs::LinearSpace< T >::sequence(), m::set_all_not_null(), m::T(), and time_select_query_execution().

get_column_offset_in_bytes()

uint64_t get_column_offset_in_bytes	(	const DataLayout &	layout,
		std::size_t	idx
	)

Returns the offset in bytes of the idx-th column in the DataLayout layout which is considered to represent a PAX-layout.

Definition at line 28 of file CostModel.cpp.

References m::storage::DataLayout::child(), and M_insist.

Referenced by generate_training_suite_filter(), generate_training_suite_group_by(), and generate_training_suite_join().

save_csv()

void save_csv	(	const std::string &	csv_path,
		const Eigen::MatrixXd &	matrix,
		const std::string &	header = ""
	)

Save matrix in a csv file.

Definition at line 37 of file CostModel.cpp.

Referenced by m::CostModelFactory::generate_filter_cost_model(), m::CostModelFactory::generate_group_by_cost_model(), and m::CostModelFactory::generate_join_cost_model().

time_select_query_execution()

Timer::duration time_select_query_execution	(	Database &	DB,
		const std::string &	input
	)

Executes the given query and returns the median of query execution times for all SELECT queries.

Definition at line 60 of file CostModel.cpp.

References m::execute_query(), m::Catalog::Get(), NUM_REPETITIONS, and m::statement_from_string().

Referenced by generate_training_suite_filter(), generate_training_suite_group_by(), and generate_training_suite_join().

Variable Documentation

DEFAULT_FILTER_POLYNOMIAL_DEGREE

constexpr unsigned DEFAULT_FILTER_POLYNOMIAL_DEGREE = 9

staticconstexpr

Definition at line 17 of file CostModel.cpp.

Referenced by m::CostModelFactory::get_cost_function().

NUM_DISTINCT_VALUES_IN_FILTER_EXPERIMENT

constexpr std::size_t NUM_DISTINCT_VALUES_IN_FILTER_EXPERIMENT = 100

staticconstexpr

Definition at line 16 of file CostModel.cpp.

Referenced by generate_training_suite_filter().

NUM_REPETITIONS

constexpr unsigned NUM_REPETITIONS = 5

constexpr

The number of times a benchmark should be repeated to reduce noise in the data.

Definition at line 19 of file CostModel.cpp.

Referenced by time_select_query_execution().