Re: Do functions never get inlined by jvm?

[sorry for a premature send...] I had the official documentation (http://clojure.github.io/clojure/clojure.core-api.html) open in a browser and was looking for things with plain search function of my browser. I always found the docstrings to be very clear and to the point. But then, I did it the

Re: Do functions never get inlined by jvm?

I never understood why people complain about the documentation of clojure/core. From the very beginning, I have found the docstrings to be exactly what I needed; when I first began 4clojure, I had On 26 April 2013 01:30, u1204 wrote: > ...0? :-) > > Tim Daly > > -- > -- > You received this messag

Re: Do functions never get inlined by jvm?

...0? :-) Tim Daly -- -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this gr

Re: Do functions never get inlined by jvm?

I am not defending it at all cost. Just tired of earing you whining the same complaint in around 70% of your posts and in some cases, like this one, unrelated with the original issue. Like if "correct" documentation could replace self acquired knowledge about how things work. This is why civilisa

Re: Do functions never get inlined by jvm?

There seems to be some rule that given sufficient time and enough participants, all threads deteriorate into an argument about the current state of clojure documentation and a huge post from Tim Daly regarding literate programming in 3...2...1... On Thu, Apr 25, 2013 at 2:23 PM, Gary Trakhman wro

Re: Do functions never get inlined by jvm?

Hi guys, Maybe we'll see some change in the future, Atlas already seems to be interesting. I think there is also a huge potential with Codeq. >From what I gather Codeq only parse thing looking like (defsomething ) for now. But we can imagine that if we augment th

Re: Do functions never get inlined by jvm?

2013/4/25 Softaddicts > May I suggest you an upgrade ? > > http://www.ehow.com/how_6949396_record-78-vinyl-records-cd.html > Ah, a batch of fresh preaching from Mr. Defend Clojure/core At All Costs. Best Canadian export since Wayne Gretzky! -- MK http://github.com/michaelklishin http://twitt

Re: Do functions never get inlined by jvm?

Good vinyls are considered higher quality by audiophiles, because there are less stages in between the mastering and amplification. There is more potential of better performance. It can be considered a real-world case of inlining. On Thu, Apr 25, 2013 at 3:16 PM, Softaddicts wrote: > May I sug

Re: Do functions never get inlined by jvm?

May I suggest you an upgrade ? http://www.ehow.com/how_6949396_record-78-vinyl-records-cd.html > 2013/4/25 Softaddicts > > > Of course my answer comes in total antagonism with your usual position > > about the > > bad state of the existing documentation which is incomplete, wrong, ... > > and

Re: Do functions never get inlined by jvm?

2013/4/25 Softaddicts > Of course my answer comes in total antagonism with your usual position > about the > bad state of the existing documentation which is incomplete, wrong, ... > and so forth. > Your reaction does not suprise me, your behavior is quite predictable. > Like an old vinyl record

Re: Do functions never get inlined by jvm?

Well you looked quite outraged that it could not be found easily. I demonstrated that doc strings can be easily searched. Of course my answer comes in total antagonism with your usual position about the bad state of the existing documentation which is incomplete, wrong, ... and so forth. You

Re: Do functions never get inlined by jvm?

You could come up with definline yourself by thinking about what inlining is and wrapping things in macros, it seems to me the real problem definline solves is to also be able to use the output as a function, which is more about keeping convenience than performance gains. I think the people who wa

Re: Do functions never get inlined by jvm?

2013/4/25 Softaddicts > Inlining is a concept that existed for more than 40 years in many > programming > languages. It's not anything new. > The OP probably know what inlining is because, hm, the subject has that word. Then she is recommended to use something that only technically has documenta

Re: Do functions never get inlined by jvm?

You asked a simple question, you got a plain answer. Now if you are still grunting there's not much I can do about that. I do agree that the doc string could be a bit more descriptive. But what does it mean to be understandable by "normal" users ? I am still trying to size what is a "normal" Lisp

Re: Do functions never get inlined by jvm?

2013/4/25 Softaddicts > user=> (apropos "inline") > (definline) > Yeah, yeah. It all starts with (apropos "apropos"), right? I knew it. -- MK http://github.com/michaelklishin http://twitter.com/michaelklishin -- -- You received this message because you are subscribed to the Google Groups "

Re: Do functions never get inlined by jvm?

user=> (apropos "inline") (definline) user=> (doc inline) .. > 2013/4/25 David Nolen > > > (doc definline) > > > Macro > Experimental - like defmacro, except defines a named function whose > body is the expansion, calls to which may be expanded inline as if > it were a macro. Cannot b

Re: Do functions never get inlined by jvm?

2013/4/25 David Nolen > (doc definline) Macro Experimental - like defmacro, except defines a named function whose body is the expansion, calls to which may be expanded inline as if it were a macro. Cannot be used with variadic (&) args. If you think this is useful to regular users (who h

Re: Do functions never get inlined by jvm?

(doc definline) On Thu, Apr 25, 2013 at 1:17 PM, Michael Klishin < michael.s.klis...@gmail.com> wrote: > > 2013/4/25 David Nolen > >> + :inline metadata > > > Which is not documented anywhere and might as well not exist for regular > Clojure users. > -- > MK > > http://github.com/michaelklishin

Re: Do functions never get inlined by jvm?

2013/4/25 David Nolen > + :inline metadata Which is not documented anywhere and might as well not exist for regular Clojure users. -- MK http://github.com/michaelklishin http://twitter.com/michaelklishin -- -- You received this message because you are subscribed to the Google Groups "Cloju

Re: Do functions never get inlined by jvm?

It just doesn't accurately describe how primitive fns work when used non-higher order, nor is it correct for the tricks you can play with non-higher order fn usage + :inline metadata. JVM will aggressively inline these cases. On Thu, Apr 25, 2013 at 1:03 PM, Alice wrote: > Care to elaborate whi

Re: Do functions never get inlined by jvm?

Care to elaborate which part is out of date? On Apr 26, 1:48 am, David Nolen wrote: > Which is out of date. > > > > > > > > On Thu, Apr 25, 2013 at 12:47 PM, Alice wrote: > > Found this blog post written by fogus: > > > "To provide this level of flexibility Clojure establishes a level of > > ind

Re: Do functions never get inlined by jvm?

Which is out of date. On Thu, Apr 25, 2013 at 12:47 PM, Alice wrote: > Found this blog post written by fogus: > > "To provide this level of flexibility Clojure establishes a level of > indirection. Specifically, all function lookups through a Var occur, > at the lowest level, through an atomic

Re: Do functions never get inlined by jvm?

Found this blog post written by fogus: "To provide this level of flexibility Clojure establishes a level of indirection. Specifically, all function lookups through a Var occur, at the lowest level, through an atomic volatile. This happens every time that a function bound using the def/defn special

Re: Do functions never get inlined by jvm?

You have to be very careful with microbenchmarks like this. I recommend writing less trivial benchmarks. For example http://github.com/clojure/test.benchmark/blob/master/src/main/clojure/alioth/spectral_norm.clj This code demonstrates performance on par with plain Java. There are many other simila

Re: Do functions never get inlined by jvm?

If that's a problem, you could try https://github.com/hugoduncan/criterium On Thu, Apr 25, 2013 at 5:38 PM, Phil Hagelberg wrote: > Three repetitions is not nearly enough to get a feel for how hotspot > optimizes functions when it detects they're in a tight loop. I don't know > how javac works,

Re: Do functions never get inlined by jvm?

Three repetitions is not nearly enough to get a feel for how hotspot optimizes functions when it detects they're in a tight loop. I don't know how javac works, but Clojure doesn't optimize much for cases where hotspot can do a much better job over time. -Phil -- -- You received this message bec

Re: Do functions never get inlined by jvm?

jvm.tools.analyzer is a nice tool for exploration in this area. I don't personally know all the subtleties here, but after some playing I managed to emit an unboxing function. I could tell from the AST. https://gist.github.com/frenchy64/5459989 Thanks, Ambrose On Thu, Apr 25, 2013 at 9:44 PM,

Re: Do functions never get inlined by jvm?

Primitive hinted funtions seem to be not an exception. (defn my-inc ^long [^long l] (inc l)) (defn f1 [^long l] (inc l)) (defn f2 [^long l] (my-inc l)) (time (dotimes [n 1] (f1 1))) (time (dotimes [n 1] (f1 1))) (time (dotimes [n 1] (f1 1))) (time (dotimes [n 1] (

Re: Do functions never get inlined by jvm?

primitive hinted fns will get inlined. You can also play the same kinds of games that Clojure does with definterface+deftype and fns that declare :inline metadata. If you don't want to learn the subtleties of Clojure performance tuning then you can always write your performance critical bits in Ja

Do functions never get inlined by jvm?

I create many small methods in java without worrying about the performance since it's usually the target of inline optimization. For example, public class Foo { public static long inc(long l) { return ++l; } public static long f1() { long l = 0; for (int i=0; i < 10; i++