Share context/JS bindings

[Jim Laskey (Oracle)]

Tony,

We will keep this general issue open as https://bugs.openjdk.java.net/browse/JDK-8160103 and revisit later on.

— Jim



> On Jun 21, 2016, at 11:41 PM, Tony Zakula <tonyzakula at gmail.com> wrote:
> 
> Thank you for looking into this.  Much appreciated.  We are working on a
> large Nashorn project we hope to share soon.
> On Jun 21, 2016 8:33 AM, "Hannes Wallnöfer" <hannes.wallnoefer at oracle.com>
> wrote:
> 
>> Unfortunately, there is a problem with invalidation of callsites that
>> Sundar brought up in the review thread for the issue I filed.
>> 
>> I’ve thought about it, but there isn’t any easy way to do this without
>> giving up significant benefits on the other side. I’m closing the issue as
>> „won’t fix“, but I’ll continue thinking about a solution for quicker
>> creation of global objects.
>> 
>> Hannes
>> 
>> 
>>> Am 16.06.2016 um 21:11 schrieb Tony Zakula <tonyzakula at gmail.com>:
>>> 
>>> Thank you Hannes.  That would be great.  We have tried using clear
>> before too.
>>> 
>>> @Axel - Thank you for the stats.  Interesting to see that.  One thing we
>> did was use strict mode and use Immediate Functions.  This would keep
>> variables going to global scope I think.
>>> 
>>> Thanks,
>>> 
>>> Tony
>>> 
>>> On Wed, Jun 15, 2016 at 8:02 AM, Hannes Wallnöfer <
>> hannes.wallnoefer at oracle.com> wrote:
>>> I prototyped the proposed change to the clear() method. It seems to work
>> well and makes a lot of sense to me, so I filed a bug for it:
>>> 
>>> https://bugs.openjdk.java.net/browse/JDK-8159589
>>> 
>>> I haven’t discussed this with the other team members yet, so let’s see
>> what they think.
>>> 
>>> Hannes
>>> 
>>> 
>>>> Am 15.06.2016 um 11:24 schrieb Axel Dörfler <axeld at pinc-software.de>:
>>>> 
>>>> Am 14.06.2016 um 22:09 schrieb Tony Zakula:
>>>>> Interesting.  We use on global context and create a new engine context
>>>>> with each run.  The initial compile of several scripts is over 100ms,
>>>>> but then we see that drop to under 10ms.  The context creation seems
>> to
>>>>> have very little overhead.  We have not timed that specifically
>> though,
>>>>> just total process time.
>>>> 
>>>> In my special use case which was a very short JavaScript that was used
>> as Comparator, actual execution time was very short, so the impact of the
>> optimizations were pretty obvious.
>>>> Sorting about 15000 entries took way over 10 minutes before any
>> optimization. IIRC after using a single global engine this dropped to about
>> 2 minutes. After also reusing the JS context, it went down to 10 seconds.
>>>> 
>>>> Bye,
>>>>  Axel.
>>>> 
>>> 
>>> 
>> 
>>