[go: up one dir, main page]

Skip to content

PostgreSQL extension that provides high-accuracy timing of SQL expressions.

License

Notifications You must be signed in to change notification settings

joelonsql/pg-timeit

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

82 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

⏱️🐘timeit

  1. About
  2. Dependencies
  3. Installation
  4. Usage
  5. API
    1. timeit.c
    2. timeit.t
    3. timeit.measure
  6. Types
    1. timeit.measure_type
  7. Internal functions
    1. timeit.round_to_sig_figs
    2. timeit.compute_regression_metrics
    3. timeit.pretty_time
    4. timeit.measure_time
    5. timeit.measure_cycles
    6. timeit.eval
  8. Examples

1. About

timeit is a PostgreSQL extension to measure the execution time of built-in internal C-functions with high resolution. On x86_64, it's also possible to measure clock cycles.

The number of necessary iterations to obtain a stable measurement is determined automatically, by starting with one iteration, and then doubling, until the latest measurements form a more or less straight line, using linear regression, controlled via the r_squared_threshold and sample_size parameters, or until reaching a timeout.

This approach allows quickly measuring all types of functions, from just a few clock cycles up to functions that take seconds, while at the same time ensuring a stable measurement is obtained.

2. Dependencies

None, except PostgreSQL.

3. Installation

Install the timeit extension with:

$ git clone https://github.com/joelonsql/pg-timeit.git
$ cd pg-timeit
$ make
$ sudo make install
$ make installcheck

4. Usage

Use with:

$ psql
# CREATE EXTENSION timeit;
CREATE EXTENSION;

5. API

timeit.c → bigint

Input Parameter Type Default
function_name text
input_values text[]
significant_figures integer 1
r_squared_threshold float8 0.99
sample_size integer 10
timeout interval 1 second
core_id integer -1

Returns measured clock cycles, rounded to significant figures.

timeit.t → text

Input Parameter Type Default
function_name text
input_values text[]
significant_figures integer 1
r_squared_threshold float8 0.99
sample_size integer 10
timeout interval 1 second
core_id integer -1

Returns measured execution time in human-readable format.

timeit.measure → TABLE

Input Parameter Type Default
function_name text
input_values text[]
r_squared_threshold float8 0.99
sample_size integer 10
timeout interval 1 second
measure_type timeit.measure_type 'time'
core_id integer -1
Output Column Type
x float8[]
y float8[]
r_squared float8
slope float8
intercept float8
iterations bigint

The timeit.measure function benchmarks the execution time or cycles of a specified function (function_name) using provided arguments (input_values).

It measures performance by iterating the function call, collecting a specified number of samples (sample_size).

The measure_type can be either 'time' or 'cycles', and it can be run on a specified CPU core (core_id).

During benchmarking, the number of iterations is doubled each time until all of these conditions are met:

  • The R-squared value meets or exceeds the threshold.
  • The slope is greater than zero, indicating that more iterations naturally lead to longer measured times or higher cycle counts.
  • The intercept is greater than zero, accounting for the inherent overhead in executing the function.

Alternatively, benchmarking stops if the timeout is reached and at least two measurements have been completed.

The timeit.measure function returns a table with the following columns: x, y, r_squared, slope, intercept, and iterations.

The x and y arrays contain the independent and dependent variable values used for regression analysis.

The r_squared value indicates how well the regression line fits the data.

The slope represents the execution time per iteration in microseconds if measure_type is 'time', or cycles if measure_type is 'cycles'.

The intercept is in the same unit as the slope and represents the overhead of executing the function.

The iterations value is the total number of iterations performed in the last measurement and corresponds to the last element of the x array.

6. Types

timeit.measure_type

timeit.measure_type is an ENUM with the following elements:

  • cycles: Measure clock cycles.
  • time: Measure time in microseconds.

7. Internal functions

timeit.round_to_sig_figs → numeric

Input Parameter Type
numeric_value numeric
significant_figures integer

Round numeric value to integer number of significant figures.

SELECT timeit.round_to_sig_figs(1234,2);
 round_to_sig_figs
-------------------
              1200
(1 row)

SELECT timeit.round_to_sig_figs(12.456,3);
 round_to_sig_figs
-------------------
              12.5
(1 row)

SELECT timeit.round_to_sig_figs(0.00012456,3);
 round_to_sig_figs
-------------------
          0.000125
(1 row)

There is also a bigint overload: timeit.round_to_sig_figs(bigint, integer) → bigint

timeit.compute_regression_metrics → TABLE

Input Parameter Type
x float8[]
y float8[]
Output Column Type
r_squared float8
slope float8
intercept float8

Calculates the regression metrics for benchmarking the execution time or cycles of functions, returning the coefficient of determination (R-squared), the slope (time or cycles), and the intercept (overhead).

Parameters

  • x: Independent variable values representing the number of iterations.
  • y: Dependent variable values representing the measured execution times or cycles.

Returns

A table with:

  • r_squared: Indicates how well the regression line fits the data (0 to 1).
  • slope: Represents the execution time or cycles per iteration.
  • intercept: Represents the fixed overhead time or cycles.

Requirements

  • Arrays must have at least two elements.
  • Arrays must be of the same length.

Exceptions

  • Raises an exception if arrays have fewer than two elements.
  • Raises an exception if arrays are of different lengths.

Example

SELECT * FROM timeit.compute_regression_metrics(
    ARRAY[1, 2, 4, 8, 16, 32, 64, 128, 256, 512],
    ARRAY[11.3, 21.5, 42.0, 81.2, 167.3, 334.4, 650.7, 1292.2, 2572.3, 5132.3]
);

     r_squared      |       slope        |     intercept
--------------------+--------------------+-------------------
 0.9999925121104006 | 10.018782621621654 | 5.598537808104993
(1 row)

timeit.pretty_time(numeric) → text

Returns measured execution time in human-readable output using time unit suffixes.

There is also an overload: timeit.pretty_time(numeric, significant_figures integer) → text

timeit.measure_time → bigint

Input Parameter Type
internal_function text
input_values text[]
iterations bigint
core_id integer

Measures execution time for internal_function with input_values over a specified number of iterations, on given CPU core_id.

If -1 is specified as core_id, the kernel will be responsible for CPU core scheduling.

timeit.measure_cycles

Input Parameter Type
internal_function text
input_values text[]
iterations bigint
core_id integer

Measures clock cycles for internal_function with input_values over a specified number of iterations, on given CPU core_id.

If -1 is specified as core_id, the kernel will be responsible for CPU core scheduling.

timeit.eval → text

Input Parameter Type
function_name text
input_values text[]

Performs a single execution of function_name with arguments passed via input_values.

Returns the result cast to text.

8. Examples

SELECT timeit.t('pg_sleep', ARRAY['0.01']);
   t
-------
 10 ms
(1 row)
SELECT timeit.t('now', ARRAY[]::text[]);
  t
------
 3 ns
(1 row)
SELECT timeit.t('clock_timestamp', ARRAY[]::text[]);
   t
-------
 20 ns
(1 row)
SELECT timeit.t('numeric_add', ARRAY['1.5','2.5']);
   t
-------
 50 ns
(1 row)

By default, a result with one significant figure is produced.

If we instead want two significant figures:

SELECT timeit.t('numeric_add', ARRAY['1.5','2.5'], 2);
   t
-------
 52 ns
(1 row)

About

PostgreSQL extension that provides high-accuracy timing of SQL expressions.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published