dbinit.py

来自「linux下基于c++的处理器仿真平台。具有处理器流水线」· Python 代码 · 共 417 行 · 第 1/2 页

        #       element.
        #       For a distribution:
        #       -1: sum (for calculating standard deviation)
        #       -2: sum of squares (for calculating standard deviation)
        #       -3: total number of samples taken (for calculating
        #           standard deviation)
        #       -4: minimum value
        #       -5: maximum value
        #       -6: underflow
        #       -7: overflow
        #   'y' is used by a VECTORDIST and the VECTOR2D to describe the second
        #       dimension.
        #   'run' is the run that the data was generated from.  Details up in
        #       the run table
        #   'tick' is a timestamp generated by the simulator.
        #   'bin' is the name of the bin that the data was generated in, if
        #       any.
        #   'data' is the actual stat value.
        #
        # INDEXES:
        #   'stat' is indexed so that a user can find all of the data for a
        #       particular stat. It is not unique, because that specific stat
        #       can be found in many runs, bins, and samples, in addition to
        #       having entries for the mulidimensional cases. 
        #   'run' is indexed to allow a user to remove all of the data for a
        #       particular execution run.  It can also be used to allow the
        #       user to print out all of the data for a given run.
        #
        self.query('''
        CREATE TABLE data(
            dt_stat	SMALLINT UNSIGNED	NOT NULL,
            dt_x	SMALLINT		NOT NULL,
            dt_y	SMALLINT		NOT NULL,
            dt_run	SMALLINT UNSIGNED	NOT NULL,
            dt_tick	BIGINT UNSIGNED		NOT NULL,
            dt_bin	SMALLINT UNSIGNED	NOT NULL,
            dt_data	DOUBLE			NOT NULL, 
            INDEX (dt_stat),
            INDEX (dt_run),
            UNIQUE (dt_stat,dt_x,dt_y,dt_run,dt_tick,dt_bin)
        ) TYPE=InnoDB;''')

        #
        # Names and descriptions for multi-dimensional stats (vectors, etc.)
        # are stored here instead of having their own entry in the statistics
        # table. This allows all parts of a single stat to easily share a
        # single id. 
        #
        # COLUMNS:
        #   'stat' is the unique stat identifier from the stat table.
        #   'x' is the first dimension for multi-dimensional stats
        #       corresponding to the data table above.
        #   'y' is the second dimension for multi-dimensional stats
        #       corresponding to the data table above.
        #   'name' is the specific subname for the unique stat,x,y combination.
        #   'descr' is the specific description for the uniqe stat,x,y
        #        combination.
        #
        # INDEXES:
        #   'stat' is indexed so you can get the subdata for a specific stat.
        #
        self.query('''
        CREATE TABLE subdata(
            sd_stat	SMALLINT UNSIGNED	NOT NULL,
            sd_x	SMALLINT		NOT NULL,
            sd_y	SMALLINT		NOT NULL,
            sd_name	VARCHAR(255)		NOT NULL,
            sd_descr	TEXT,
            UNIQUE (sd_stat,sd_x,sd_y)
        ) TYPE=InnoDB''')


        #
        # The formula table is maintained separately from the data table
        # because formula data, unlike other stat data cannot be represented
        # there.
        #
        # COLUMNS:
        #   'stat' refers to the stat field generated in the stat table.
        #   'formula' is the actual string representation of the formula
        #       itself.
        #
        # INDEXES:
        #   'stat' is indexed so that you can just look up a formula.
        #
        self.query('''
        CREATE TABLE formulas(
            fm_stat	SMALLINT UNSIGNED	NOT NULL,
            fm_formula	BLOB			NOT NULL,
            PRIMARY KEY(fm_stat)
        ) TYPE=InnoDB''')

        #
        # Each stat used in each formula is kept in this table.  This way, if
        # you want to print out a particular formula, you can simply find out
        # which stats you need by looking in this table.  Additionally, when
        # you remove a stat from the stats table and data table, you remove
        # any references to the formula in this table.  When a formula is no
        # longer referred to, you remove its entry.
        #
        # COLUMNS:
        #   'stat' is the stat id from the stat table above.
        #   'child' is the stat id of a stat that is used for this formula.
        #       There may be many children for any given 'stat' (formula)
        #
        # INDEXES:
        #   'stat' is indexed so you can look up all of the children for a
        #       particular stat.
        #   'child' is indexed so that you can remove an entry when a stat is
        #       removed.
        #
        self.query('''
        CREATE TABLE formula_ref(
            fr_stat	SMALLINT UNSIGNED	NOT NULL,
            fr_run	SMALLINT UNSIGNED	NOT NULL,
            UNIQUE (fr_stat,fr_run),
            INDEX (fr_stat),
            INDEX (fr_run)
        ) TYPE=InnoDB''')

        # COLUMNS:
        #   'event' is the unique event id from the event_desc table
        #   'run' is simulation run id that this event took place in
        #   'tick' is the tick when the event happened
        #
        # INDEXES:
        #   'event' is indexed so you can look up all occurences of a
        #       specific event
        #   'run' is indexed so you can find all events in a run
        #   'tick' is indexed because we want the unique thing anyway
        #   'event,run,tick' is unique combination
        self.query('''
        CREATE TABLE events(
            ev_event	SMALLINT UNSIGNED	NOT NULL,
            ev_run	SMALLINT UNSIGNED	NOT NULL,
            ev_tick	BIGINT   UNSIGNED       NOT NULL,
            INDEX(ev_event),
            INDEX(ev_run),
            INDEX(ev_tick),
            UNIQUE(ev_event,ev_run,ev_tick)
        ) TYPE=InnoDB''')

        # COLUMNS:
        #   'id' is the unique description id
        #   'name' is the name of the event that occurred
        #
        # INDEXES:
        #   'id' is indexed because it is the primary key and is what you use
        #       to look up the descriptions
        #   'name' is indexed so one can find the event based on name
        #
        self.query('''
        CREATE TABLE event_names(
            en_id	SMALLINT UNSIGNED	NOT NULL AUTO_INCREMENT,
            en_name	VARCHAR(255)		NOT NULL,
            PRIMARY KEY (en_id),
            UNIQUE (en_name)
        ) TYPE=InnoDB''')

    def clean(self):
        self.query('''
        DELETE data
        FROM data
        LEFT JOIN runs ON dt_run=rn_id
        WHERE rn_id IS NULL''')

        self.query('''
        DELETE formula_ref
        FROM formula_ref
        LEFT JOIN runs ON fr_run=rn_id
        WHERE rn_id IS NULL''')

        self.query('''
        DELETE formulas
        FROM formulas
        LEFT JOIN formula_ref ON fm_stat=fr_stat
        WHERE fr_stat IS NULL''')

        self.query('''
        DELETE stats
        FROM stats
        LEFT JOIN data ON st_id=dt_stat
        WHERE dt_stat IS NULL''')

        self.query('''
        DELETE subdata
        FROM subdata
        LEFT JOIN data ON sd_stat=dt_stat
        WHERE dt_stat IS NULL''')

        self.query('''
        DELETE bins
        FROM bins
        LEFT JOIN data ON bn_id=dt_bin
        WHERE dt_bin IS NULL''')

        self.query('''
        DELETE events
        FROM events
        LEFT JOIN runs ON ev_run=rn_id
        WHERE rn_id IS NULL''')

        self.query('''
        DELETE event_names
        FROM event_names
        LEFT JOIN events ON en_id=ev_event
        WHERE ev_event IS NULL''')