Timothy,
To summarize:
REXX performance is terrible. It's used extensively on z/OS, a platform
where CPU consumption costs big bucks. In some environments like NetView
IBM gave away the REXX compiler for free. But like I mentioned the
compiler has serious flaws. Some IBM products like FileManager and
Workload Simulator provide REXX-lite subsets to work around the
performance problems. Interpreted Lua thrashes the REXX compiler in
every test I've run which suggests it's not optimal. How can a
interpreter beat a compiler? Does compiled REXX even have an optimizer?
Mainframe folks don't want to give up REXX. They're not interested in
new languages no matter how superior they may be. So a REXX JIT would be
manna from heaven. It will save money. REXX programmers would have a
language that they can use in production for intensive workloads.
However, it would take a rewrite of REXX and that's not going to happen.
On 28/10/2017 3:34 PM, Timothy Sipples wrote:
David Crayford wrote:
We want a JIT for classic REXX so it can run
at speeds similar to compiled languages.
Yes, but IBM already has the REXX Compiler and Library for Z:
https://www.ibm.com/us-en/marketplace/compiler-and-library-for-rexx-on-ibm-z
REXX is not just "similar" to compiled languages. It *is* a compiled
language. Or interpreted. Or some of both. You can choose, and you've been
able to choose since August 30, 1991 on MVS (since June 30, 1989, on VM).
Release 4 of the REXX Compiler and Library was released on August 29, 2003,
and is IBM marketed and fully supported.
Again, I'm asking the "Why?" here before someone else does. :-) I'm
intrigued, but what's the real problem/need (that actually exists) that
needs to be solved? If, for example, it's having the Object REXX
programming language available for scripting in TSO/ISPF....
Kirk Wolf wrote:
FWIW, the service that z/OSMF uses for starting TSO address spaces is:
https://www.ibm.com/support/knowledgecenter/en/SSLTBW_2.2.0/com.ibm.zos.v2r2.ieac100/cea-intro.htm
John McKown was kind enough to compile Open Object REXX for z/OS UNIX
(thanks, John!):
https://github.com/JohnArchieMckown/oorexx
This TSO service interface looks like a strong candidate for those who'd
like to extend OOREXX on z/OS to interact with TSO.
--------------------------------------------------------------------------------------------------------
Timothy Sipples
IT Architect Executive, Industry Solutions, IBM z Systems, AP/GCG/MEA
E-Mail: [email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
