You can possibly get rid of the var dereference by dereferencing at
compile-time. Of course this means you lose the benefits of vars, too.
It's just super-easy when needed :-).
(defn a [] 1)
(defn b [] (dotimes [_ 10] (a)))
(time (b))
"Elapsed time: 1097.082556 msecs"
(time (b))
"Elap
I would not say "shame on you" -- it is reasonably common to find Clojure
code from experienced Clojure developers that puts type hints in places
where they do not have the desired effect. It isn't an error, so there are
no error messages. Finding occurrences of such things while I have been
test
Shame on me! Now the difference is around 10%. Much better!
Case closed, I suppose :)
On Sunday, April 6, 2014 11:24:05 PM UTC+4, Andy Fingerhut wrote:
>
> Sorry I am not taking the time to try out the change for you and see
> whether it makes the desired difference in performance happen, but I
Sorry I am not taking the time to try out the change for you and see
whether it makes the desired difference in performance happen, but I do
know that the ^double type hint on return values should be on the argument
vector, not on the var. So instead of your abs-diff, try this, and
similarly for a
For some time I had a suspicion that in Clojure we have a fundamental
problem with efficient math and today I've tried to test it.
Let's say we have a pretty simple function:
(ns testapp.core
(:require
[criterium.core :as cr]
[primitive-math :as p]
[no.disassemble :as nd]))
(set! *wa
