Rexx2Nrx release 4.0 
Author Message
 Rexx2Nrx release 4.0

Hello there,

   back from vacations, and back in business ...

   I am now currently preparing release 4.00 of Rexx2Nrx (the classic
Rexx to NetRexx converter). As I do now have a new SONY notebook (1.6
MHZ, 255 MB RAM) the whole project progresses. As you know, it's  a
'free time' project, anyhow ... I expect 4.00 to be ready end of
september (hopefully this year).

The major current task is to put the documentation and samples
together, and run the Quality assurance tests on both IBM CMS compiled
Rexx, IBM PC Object Rexx, and NetRexx under Windows, and Java ... I
still do NOT have LINUX installed, so any parties willing to run the
LINUX testcases are very welcom, of course.

Also, if you feel 'your testcase' is interesting, or you tried the
software and found bugs, please forward the source by mail. Thanks in
advance

   One small note (to BIG BLUE), and Mike Cowlishaw in particular:

- when I use 'Object REXX' (release 2.1.1) for bigger programs  (like
my REXX parser and REXX2NRX translator) on my new PC (mentioned above)
then Object REXX performs approximately 90% slower than the equivalent
NetRexx (Java) program!!

- hence, NetRexx performance is 10 times quicker than Object Rexx with
bigger
programs ....   CONGRATULATIONS, Mike, you did a very good job! !!!

- when somebody would like the 'international REXX source' code to
verify this
testcase, simply send a mail ...

best regards, Th. Schneider.



Sun, 06 Feb 2005 02:38:56 GMT  
 Rexx2Nrx release 4.0
Thomas:

Quote:
> - when I use 'Object REXX' (release 2.1.1) for bigger programs  (like
> my REXX parser and REXX2NRX translator) on my new PC (mentioned above)
> then Object REXX performs approximately 90% slower than the equivalent
> NetRexx (Java) program!!

> - hence, NetRexx performance is 10 times quicker than Object Rexx with
> bigger programs ....

Hmm, I do not doubt your findings, just your conclusions. Speculation: you
are creating a large output string, where NetRexx/Java may use the
StringBuffer class under the hoods, and Rexx/Object Rexx uses Strings
instead.

Now, if you do something like "outputString=outputString newChunkofText"
over and over again, this might hit the system rather hard, if run under
Rexx/Object Rexx. To mimickry the same (Java-StringBuffer) behaviour in
Rexx/Object Rexx you could use stemy ("stem-arrays") or under Object Rexx
the Array class. (Would be nice if the Array class would have the
counterpart of the String' class "makearray", something like "makestring"
for dealing more efficiently with such situations. )

Or with other words: I bet that Java/NetRexx would not be able to beat out
Rexx/Object Rexx in that arena, if using stem-arrays.

Regards,

---rony



Sun, 06 Feb 2005 10:59:39 GMT  
 Rexx2Nrx release 4.0

Quote:
> > - hence, NetRexx performance is 10 times quicker than Object Rexx with
> > bigger programs ....

> Hmm, I do not doubt your findings, just your conclusions.

Heh heh . well, Rony, don't entirely forget that NetRexx is
compiled (to bytecodes, which are then compiled to native code by the JIT).
And, Thomas, if you use stems a lot, there has been a big improvement
in Object Rexx in that area very recently.

But I look foward to you gentlemen agreeing a small(ish) benchmark with
which we can compare both OO approaches -- a sort of 21st-century
RexxCPS?

:-)

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Mike Cowlishaw, IBM Fellow



Sun, 06 Feb 2005 22:39:08 GMT  
 Rexx2Nrx release 4.0
Mike Cowlishaw schrieb:

Quote:
> But I look foward to you gentlemen agreeing a small(ish) benchmark with which
> we can compare both OO approaches -- a sort of 21st-century RexxCPS?

> :-)

Wow, *that* would be a challenge, wouldn't it ?

(When are you coming to Vienna the next time, we need to sit down and have a
beer or two thinking this over? Or maybe on the next International Rexx
symposium next April/May?)

---rony



Mon, 07 Feb 2005 02:13:44 GMT  
 Rexx2Nrx release 4.0
Hello Rony, Lieber Herr Professor Flatscher!

I'll take your bet. I do suggeset the winner gets 1 bear ...

My bet is:

 If you are reading a 1.5 MB file, my 'Rexx2Nrx.RexxFile' routines are
10 times quicker than Object Rexx stream/linein/lineout version 2.1.1.

  (needless to say that NO STEMS are involved, simple strings only!!!)

 The bet is won by 2 test-cases:

                   C:cmd>java SOFTUNPK.class Rexx2Nrx.sof
                   C:cmd>rexx SOFTUNPK.ORX Rexx2Nrx.sof

 where 'SOFTUNPK.ORX' is the tokenized version of the original REXX
source code.

 where 'SOFTUNPK' delivers Provable, reliable, Portable, Platform
independent software distribution ... (just some advertising ...)

 and 'Rexx2Nrx.sof' is the Source of my 'classic Rexx2 to NetRexx
converter' (see www.Rexx2Nrx.com for details) (again some advertising)

Have a nice weekend!

Danke fuer das Bier im Voraus, Thomas Schneider.
Thanks for my beer in advance , --  " --



Wed, 09 Feb 2005 00:14:17 GMT  
 Rexx2Nrx release 4.0
Thomas:

Quote:
> I'll take your bet. I do suggeset the winner gets 1 bear ...

O.K., taken. Why don't we sit together, once I am back in Vienna and skim
over the code (NetRexx and [Object] Rexx together. (My assumption is that
you are building a lot of strings in a way like "newString=newString ||
someChunk". If this is the case, it would explain why [Object] Rexx is
slower than NetRexx/Java, because in NetRexx/Java mostlikely StringBuffers
are used for that, which would not impose the burden of creating new
String objects everytime you do a concatenation; if so, then one could
improve the performance of the [Object] Rexx code with stems or arrays.)

Regards,

---rony



Wed, 09 Feb 2005 00:36:38 GMT  
 Rexx2Nrx release 4.0
I am currently still busy with a lot of business work at my office,
and completing Rexx2Nrx release 4.00.

I suggest we setup plans for the benchmark tests for October/November
THIS year, and do NOT wait until the next REXX LA symposium next year.

best regards,
Thomas Schneider.



Sun, 20 Feb 2005 17:11:51 GMT  
 Rexx2Nrx release 4.0
Hello there,

   I am currentty collecting a couple of Quality assurance & benchmark
tests to
compare the 2 available Object Oriented 'Rexx' approaches:

   (1) IBM Object Rexx vs. 2.1.1
   (2) Rexx2Nrx + NetRexx + Java

   In the benchmark tests, always the 'Object Rexx' TOKENIZED version
will be
compared with the compiled Java version.

   I did already find major differences in run-time in 3 areas:

    1.) File I/O in general, but especially with bigger sequential
files (say around 1 MB).  Here the differences are 1:10 (NetRexx+Java
vs ORexx).

    2.) Heavy usage of 'string concatenations' in bigger programs.
Here the differences tend to be 1:50 thru 1:100 (NetRexx+Java vs.
ORexx).

    3.) heavy loops with SIGNAL (Rexx2Nrx only). Here Object Rexx
win's definitely (don't have factors yet). No surprise, as I am
currenlty using recursions to resolve SIGNAL's in Rexx2Nrx. A good
test to prove this will be welcome.

    If you do have good examples which should be included in the
benchmark pls. forward them to me and Jan Engehausen. But pls. keep in
mind that Rexx2Nrx currently does ONLY have the 'classic Rexx'
run-time routines implemented
(and INTERPRET ist still not available in Rexx2Nrx. I am working on
this).

  Also, good examples for testing the functionality are welcome.

  All parties interested to perform tests by their own: pls. send me a
mail, so that I can forward an 'EVALUATION licence (no costs)' to you.
As a re-pay, I would like to get a short survey of the results,
whatever they are.

best regards,
Thomas Schneider.



Tue, 08 Mar 2005 23:34:07 GMT  
 Rexx2Nrx release 4.0
The current beta version of Rexx2Nrx release 4.00 has been uplaoded to
my home-page 'www.Rexx2Nrx.com'.

Pls. note, that the online docu still needs to be updated. I am in the
progress to accomplish that, but it will take some time (I do hope to
get the final version 4.00 including docs out end of October 2002).

If you have any trouble to download or run the software, please send
me a mail.
Also, if you do have questions, pls. send a mail. I am eager to fill
the
'FAQ's' section on my home-page.

best regards, Thomas Schneider.



Thu, 24 Mar 2005 21:21:47 GMT  
 
 [ 13 post ] 

 Relevant Pages 

1. Rexx2Nrx release 4.01

2. Rexx2Nrx release 4.00 beta 22

3. Rexx2Nrx vs 3.02 pre-release 7

4. Rexx2Nrx release 3.01 run-time package - documentation update

5. Rexx2Nrx (the classic Rexx to NetRexx converter) release notes / issues

6. Rexx2Nrx 4.0 final release -- and it's now a FREE Software!

7. Dolphin 4.0 release

8. BackFlash 4.0 - Major new version release

9. ANN: SocketWrench Secure Edition 4.0 Released

10. Win32Forth version 4.0 Released

11. lina 4.0 released : an ISO Forth for Linux.

12. Introducing Rexx2Nrx to the world of REXX users.

 

 
Powered by phpBB® Forum Software