On 24.02.2017 10:04, Remi Forax wrote:
----- Mail original -----
De: "Jochen Theodorou" <blackd...@gmx.org>
À: dev@groovy.apache.org
Envoyé: Mercredi 22 Février 2017 22:49:01
Objet: Re: Help with pitching Groovy and Grails
On 21.02.2017 22:51, Raviteja Lokineni wrote:
[...]
well, of course I should
also mention that for a example a method call using reflection use to be
like factor 30 slower compared with a direct method call using Java. But
that is no longer true for a long time.
yes, 30x was a long long time ago, i think before 1.4.2.
The last time the reflection code was touched for speed update was during the
Java 7 timeframe.
but the problem that a reflective call is a inlining barrier is still
true, or not? Otherwise my expected performance is something like 6-10x
slower
And then in Groovy we use runtime code generation to avoid Reflection and
optionally you could
also use the indy port, which is based on invokedynamic (the same one
that is used for method calls for lambdas) which can be pretty fast. And
if that is not enough, you can still use the static compiler to be
around the same level as Java.
Technically, invokedynamic is used from creating the lambda not for calling it,
hehe, right.
because after creation it's more like a plain reference on an object that
implement a functional interface, so it can be called like a plain old
interface. There has been a prototype to use invokedynamic to call the lambda
but performance was identical to an interface call so it was not integrated
because invokedynamic makes the startup a little slower (like reflection does).
I see.
In the jdk 9, invokedynamic is used when you do a concatenation with + because
invokedynamic is faster that using a chain of plain old Java calls to
StringBuilder.append().
that's the new and improved String handling, yes, I did hear about
that... But I guess it is a much better example than the one I had
bye Jochen
