Re: java performance in batch of a packaged tool

[Vernon Hamberg <vhamberg@xxxxxxxxxxxxxxx>]

Hi Dieter

I agree strongly with your approach to use asynchronous calls. It is not 
something I've been able to take advantage of yet in our situation - 
this is even a useful approach outside of Java calls - I'm facing a 
situation of multiple calls to utilities in PASE - things like 
GhostScript and ImageMagick - and doing all that synchronously is adding 
quite a bit of time. In particular, this involves converting PNG files 
to GIF for use with the AFPRSC keyword (we are at 7.1 of the OS, which 
has issues that Host Print Transform does not know about AFPRSC at that 
level). And it includes converting the spooled files to raw PCL in 
*USERASCII spooled files.

I believe it should be possible and beneficial to make this into an 
asynchronous approach - the thing is, we probably can live with it for 
6-9 months, until we get version 7.2 of the OS on our Poser 8, then 
we'll be able to use AFPRSC fully, since Host Print Transform DOES 
support it at that release and above.

Regards
Vern

On 1/7/2017 10:48 AM, D*B wrote:
> <Vern>
> I personally know that all of these people did not just figure this 
> out last year.
> </Vern>
> Vern,
> thanks for your compliments about ArdGate!
>
> I was very happy to read, that well known real experts (I've named 
> some in my previous posting) came to the conclusion, that it's not 
> recommended to use the RPG built in JNI support. The problem are 
> those, who still recommend it and the actual documentation (ILE RPG 
> Programmers Guide V7R3). The Redbook is the one and only IBM official 
> publication and this version the first version telling the truth about 
> RPG JNI support and its limitations.
>
> Just read this thread and its recommendations, or actual discussions 
> at RPG cafe. I'm very sure, that changing the architecture from 
> synchronous call to asynchronous call with a prestarted JVM will solve 
> most performance problems of mixed RPG/Java applications.
>
> Dieter