This is a simple benchmark of calls to Emacs require and load functions. It can be used to keep track of where time is being spent during Emacs startup in order to optimize startup times.
The code was originally based on init-benchmarking.el by Steve Purcell with many additional modifications having gone into it since.
Ensure that you have added MELPA to your package archives. For instance by having the following in your Emacs configuration.
(require 'package)
(add-to-list 'package-archives
'("melpa" . "http://melpa.org/packages/") t)Then update the local package list using M-x package-refresh-contents RET.
After this your can install benchmark-init by running M-x package-install RET benchmark-init RET. After installation completes you need to add the
following near the beginning of your configuration file but after setting up
package.
(require 'benchmark-init)
;; To disable collection of benchmark data after init is done.
(add-hook 'after-init-hook 'benchmark-init/deactivate)Add the following as early as possible to your Emacs configuration, but after
setting up use-package.
(use-package benchmark-init
:ensure t
:config
;; To disable collection of benchmark data after init is done.
(add-hook 'after-init-hook 'benchmark-init/deactivate))If you are not using el-get to manage your Emacs packages you can skip this section.
Since benchmark-init must be activated as early as possible so that it can
measure calls to load and require it should be loaded before el-get starts
bringing in other packages. To achieve that, add something like the following
snippet as early as possible in your Emacs initialization script, before
calling el-get. Replace /path/to/el-get with the path to your el-get
directory.
(load "/path/to/el-get/benchmark-init/benchmark-init.el"
'no-error nil 'no-suffix)The first time you start Emacs after adding this nothing will be benchmarked since el-get will only install the package. Simply quit and restart Emacs and everything should be benchmarked from now on.
Run make inside the directory where you installed benchmark-init, this will
produce the benchmark-init-loaddefs.el file. Then place the following code
as early as possible in your Emacs initialization script. Replace
/path/to/benchmark-init with the path to the directory where you put
benchmark-init.
(add-to-list 'load-path "/path/to/benchmark-init/")
(require 'benchmark-init-loaddefs)
(benchmark-init/activate)Data collection will begin immediately after the call to
benchmark-init/activate.
There are two ways in which benchmark-init's results can be presented, as a table or in a tree. The table can be displayed by running:
- benchmark-init/show-durations-tabulated
Which will bring up the results in a tabulated list:
| Module | Type | ms [^] | total ms |
+------------------------------+---------+--------+----------+
| eldoc-eval | require | 204 | 204 |
| eldoc | require | 183 | 183 |
| semantic/db-find | require | 19 | 92 |
| ispell | require | 16 | 16 |
| grep | require | 6 | 6 |
| ~/.emacs.d/benchmark-init.el | load | 1 | 1 |
The ms column lists the amount of time spent loading the entry itself and total ms is the duration spent loading the entry and its dependencies. In the tree mode each entry will only display the time spent loading the entry itself, not including children.
Tree mode can be displayed by running:
- benchmark-init/show-durations-tree
╼►[benchmark-init/root nil 0ms]
├─[benchmark-init-modes require 8ms]
├─[eldoc-eval require 2ms]
│ ╰─[eldoc require 125ms]
├─[~/.emacs.d/el-get/benchmark-init/benchmark-init.el load 4ms]
╰─[auto-dictionary require 72ms]
╰─[flyspell require 9ms]
╰─[ispell require 24ms]
It is possible to control when benchmark-init should collect data by using the following two functions:
- benchmark-init/activate
- benchmark-init/deactivate