Performance regressions

[fhars]

I have a modarately sized repository with currently 23 renamed, 1 deleted and 3 changed files. Doing magit-status (with the version from melpa today) pegs emacs at 100% CPU for several minutes, it will only very occasionally respond to key presses or mouse clicks (if at all) , and if I manage to scroll down in the status window, it shows me line numbers of several thousend.

Currently it has the 35th file highlighted, it is at the top of the file, and the status line displays "Top L5839". It took about eight minutes to position the cursor there, and now the CPU is at 0%. If I press an arrow key, it takes about 5 CPU seconds to move to the next or previous line.

In general, the current version feels horribly sluggish compared to version 1.1.1, which I used until earlier today.

[dabrahams]

I have experienced the same.

[tarsius]

This is likely due to d976c0d which fixes #1128. That of course assumes that this is a recent regression (which we cannot possibly know for @fhars but for @dabrahams I am going to assume you update more often than every few months).

I will look into speeding up diff parsing (so that the question "should we delay diff parsing" hopefully becomes an irrelevant one. And I will also write a more complete faq entry on how to speed up magit by correctly configuring git. I also just had an idea how Magit calling Git several times to perform a single action, could potentially be improved, which I want to try out. Non of this will happen overnight.

[fhars]

I can confirm that undoing d976c0d makes magit usable again (vi ~/.emacs.d/elpa/magit-20140126.459/magit.el FTW). Staging stuff is still kind of sluggish, but at least it works now.

[tarsius]

Closing because the regression has been fixed. Of course performance has to be further improved still. I added a not about adding a faq with performance tricks to #1096.