aboutsummaryrefslogtreecommitdiff
path: root/zola-static/content/blog/emacs-garbage-collection
diff options
context:
space:
mode:
authorJack Jamison <jackqjamison@gmail.com>2025-12-26 18:48:58 -0500
committerJack Jamison <jackqjamison@gmail.com>2025-12-26 18:48:58 -0500
commitfad3843b2ba6176a4c0834b66e27584fcf7aed17 (patch)
tree5e53bf449392605d78912b620b22273a950c1516 /zola-static/content/blog/emacs-garbage-collection
parentbe3a5f44a4eba7ca68839e1456553d5cc7399395 (diff)
move articles to directories
Diffstat (limited to 'zola-static/content/blog/emacs-garbage-collection')
-rw-r--r--zola-static/content/blog/emacs-garbage-collection/index.md43
1 files changed, 43 insertions, 0 deletions
diff --git a/zola-static/content/blog/emacs-garbage-collection/index.md b/zola-static/content/blog/emacs-garbage-collection/index.md
new file mode 100644
index 0000000..26cf5d9
--- /dev/null
+++ b/zola-static/content/blog/emacs-garbage-collection/index.md
@@ -0,0 +1,43 @@
++++
+title = "A Solution to Emacs Garbage Collection Stutters"
+date = 2025-05-12
+updated = 2025-05-14
++++
+
+The more you load Emacs up with packages, the more it starts to chug. Throw on a theme and an lsp client, and you might start to notice frequent stuttering.
+
+This problem was very frustrating for me - I want to be able to hold `C-n` and linearly move down the document, micro-stutter *free*. I spent weeks searching for the problem, and considered quitting Emacs altogether.
+
+# Garbage Collection
+It turns out that the problem wasn't that packages were executing slowly, it was the garbage that they were building up.
+
+Emacs' garbage collector waits until a byte threshold (the `gc-cons-threshold` variable) has been reached before executing. Its default value feels relatively low for a modern computer - only `800KB`. This results in frequent garbage collections, each causing a small stutter. Luckily, as with most things in Emacs, you can tune this behaviour.
+> If you want to see how frequently the GC is triggered, try setting `garbage-collection-messages` to non nil (you might be surprised).
+
+Opinions on the ideal solution differ.
+- Some people recommend **increasing** the threshold, leading to longer times without triggering the GC but longer stutters.
+- Others recommend **lowering** the threshold, causing more frequent GC triggers but smaller stutters.
+- The minibuffer (especially with packages like `vertico`) can use a lot of memory, so some [recommend](https://bling.github.io/blog/2016/01/18/why-are-you-changing-gc-cons-threshold/) increasing the threshold only when using it.
+
+# The Solution
+In my opinion, the best solution is to garbage collect **when idle**. When editing, you're often pausing briefly - and if there is little garbage, Emacs can collect it very quickly.
+I also *basically* disable the threshold entirely when in the minibuffer. Here's the code from my `init.el`.
+
+```lisp
+(defun my-minibuffer-setup-hook ()
+ (setq gc-cons-threshold most-positive-fixnum))
+
+(defun my-minibuffer-exit-hook ()
+ (setq gc-cons-threshold 800000000))
+
+(setq gc-cons-threshold most-positive-fixnum)
+
+(run-with-idle-timer 1.2 t 'garbage-collect)
+```
+
+I have been using this for a few months and it's worked perfectly. I've tried my best to notice the garbage collections, I haven't been able to.
+
+> There is a package for this, called [the Garbage Collector Magic Hack](https://github.com/emacsmirror/gcmh). Personally I prefer a lower idle time than its default, and I like keeping it in my own config.
+
+# Update: IGC Branch
+After posting this article, [some users on reddit](https://www.reddit.com/r/emacs/comments/1km1by3/comment/ms750w3/) informed me that there is an [Emacs branch](https://git.savannah.gnu.org/cgit/emacs.git/tree/README-IGC?h=feature/igc#n1) in development which greatly improves the garbage collector efficiency. People are saying that it has completely eliminated their stutters. I recommend trying it out.