I'm using emacs-eclim for programming Java. It gives me a lot of the goodness from Eclipse, including true auto completion and highlighting of warnings and errors in the buffer. However, I've found that sometimes it's close to unusuable because it consumes too much CPU and hangs for several seconds (~8 in my last case) before doing what I told it to.
Today, I thought I'd get to the bottom of this, and I ventured on my first adventure with the Emacs native profiler and it was surprisingly easy to use!
First, I started the profiler with
M-x profiler-start, then I did
what was the probblem in eclim, namely opening a Java file with lots
of errors. Once it was done hogging my CPU (100% for 5-8 seconds,
thank you very much), I asked the profiler to give me the function
calls that were generating CPU load by issuing
- command-execute 3122 84% - call-interactively 3122 84% - apply 3122 84% - ad-Advice-call-interactively 3122 84% - #<subr call-interactively> 3122 84% - Buffer-menu-this-window 2734 73% - switch-to-buffer 2734 73% - apply 2734 73% - ad-Advice-switch-to-buffer 2734 73% - eclim-problems-highlight 2734 73% - remove-if-not 2701 73% - apply 2701 73% - cl-remove 2701 73% - apply 2697 72% - cl-delete 2693 72% - #<compiled 0x1586729> 2693 72% - file-truename 2641 71% - file-truename 2420 65% - file-truename 2295 62% [..]
As you can see, it's the
eclim-problems-highlight method which is
the culprit and its call to
file-truename. You can also see that
there's some kind of looping going on from the
file-truename is constantly being called.
With that out of the way, I checkout the source code for emacs-eclim
and localed the function
eclim-problems-highlight. A quick
annotate later, I've found who made the change when and, most
importantly, why the developer made it. I also asked Emacs to tell me
file-truename does with
C-h f file-truename. So, it's code
that's there to resolve any symlinks present in the path to the source
code file we're looking at.
I now tried to remove the call to
file-truename to see if it would
improve performance. And what a difference it made! CPU usage went
down to normaly levels, seldomly rising above 5% and Emacs managed to
open Java files instantly without any lag or hang. Big success as far
as I'm concerned :-)
I've made a fix to emacs-eclim, but I'm not sure if this is something the maintainers of emacs-eclim wants merged. Nevertheless, this was a great lesson for me on how to investigate performance issue in Emacs and I can now use Emacs/eclim for projects than before. Yeah!