[STUMP] profiling StumpWM

[Eric Abrahamsen]

Before I start a new job, I thought I'd take a whack at one more
thing...

I still can't get rid of the idea that Stump is slow, both in reaction
to input and in its own operations. I know very little about profiling,
but I thought I'd take a whack at it and see if I could learn anything.
So far I haven't learned very much.

I'm hoping to get a few pointers here. Anything would be great: from
comments on a better setup for the test, to how to read the output.
Right now, so far as I can tell, most time is being spent in creating
vectors and integers, but I don't really know what to do with that
information.

What would be ideal, of course, is if some greybeard was willing to walk
me through what I'm seeing here, and hopefully point me in the direction
of some (hopefully not premature!) optimizations. It wouldn't have to
take much time, just the occasional email exchange.

This is SBCL specific -- I'm using SBCL 1.1.16 on arch linux, though
obviously we'd rather have speedups that affected all implementations.

Below I've pasted the code I used to for profiling (I thought using
"run-commands" might best simulate user operation), and at the bottom a
pastebin link to the output.

Thanks!
Eric

(require :sb-sprof)
(sb-sprof:profile-call-counts "STUMPWM")

(defun stump-test-inner ()
  (stumpwm:run-commands
   "hsplit"
   "move-window right"
   "remove-split"
   "gnewbg bubba"
   "gnext"
   "gkill"
))

(defun stump-test-outer (n)
  (dotimes (i n)
    (stump-test-inner)))

(defun start-test ()
  (sb-sprof:with-profiling
   (:max-samples 10000 :mode :alloc :report :flat)
   (stump-test-outer 1000)))


http://paste.lisp.org/display/141772