bpxwunix does a spawn() of /bin/sh.  If you pass the env var
_BPX_SHAREAS=YES a local spawn is done.
Address syscall should be far more efficient than running a shell command
that gives something similar,
not to even mention parsing the output.
A quick experiment on our development system shows 'readdir pathname dir.
attr.'
is about 17X faster than bpxwunix 'ls -l pathname',,'dir.'
and ls is a builtin.  Trying a non-builtin on a small file, readfile is
giving me about 2200X faster than head.
Using _BPX_SHAREAS=YES, ls -l stays about the same and readfile is about
1600X faster than head.

Bill Schoen


>
> Date:    Thu, 12 Dec 2019 11:44:25 -0600
> From:    Kirk Wolf <[email protected]>
> Subject: Re: OMVS Tuning ?
>
> I have a related question, maybe someone who is more familiar with
> bpxwunix() knows - does it do a spawn() or a fork()/exec() under the
> covers?   If spawn(), is it possible to do a local spawn?   The doc that
I
> see is silent.
>
> I'm also not a REXX expert, but what about running your REXX ISPF dialog
> with syscalls('ON') and then using the REXX syscalls directly?
> Could you use the spawn()/spawnp() syscall with a "local spawn"?   Then,
> you will only be dubbing once (at syscalls('ON')), and you wouldn't be
> starting a new OMVS address spaces.
>
> I know that you can have an ISPF dialog program written in C/C++ that
uses
> the C library to do Unix stuff:  the first thing that does a UNIX syscall
> will dub the address space.   You can use (local) spawn to invoke other
> commands in child processes but in the same address space.    We've done
> this before and it works well.
>
> On the other hand, maybe your performance problem is something specific
to
> your dev system, and otherwise not really significant?   Try posting a
tiny
> REXX test case and maybe others on the list will see what kind of
> performance they get.
>
> Kirk Wolf
> Dovetailed Technologies
> https://urldefense.proofpoint.com/v2/url?
> u=http-3A__dovetail.com&d=DwIFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=rFsK2G-
>
kj0QZo0UhxA9XKA&m=lbGfjbkv3Q2gb_9r2CrFg7iehrvchvT2AdhVuuMZPzc&s=8rrtkhpUfcLuQibwRUqnAjcMV0ZsTCjkGoB1ticrOaA&e=

>
> On Thu, Dec 12, 2019 at 6:07 AM Lionel B Dyck <[email protected]> wrote:
>
> > Thanks for the update on STEPLIB - I didn't see it in 2.1 but my search
> > may have been using the wrong keywords.
> >
> > I did some testing doing bpxwunix from rexx.  1 test was issuing the
env
> > command and the other a cd followed by an ls -la.
> >
> > With libraries in STEPLIB and in ISPLLIB the timing on my system (which
is
> > a z/VM guest) was 10 and 4 seconds respectively under ISPF and 4 and 2
> > seconds in native TSO. This pointed to the ISPLLIB as a tasklib so I
> > removed everything from it and the timing under ISPF went from 10 down
to 7
> > seconds and 4 to 1 seconds.  Removing the STEPLIB got me to 2 and 0.5
> > seconds.
> >
> > Still not subsecond so I suspect there is something in the z/VM setup
> > which I'm not sure how to diagnose but am looking.
> >
> > Thanks
> >
> >
> > Lionel B. Dyck <sdg><
> > Website: https://urldefense.proofpoint.com/v2/url?
> u=http-3A__www.lbdsoftware.com&d=DwIFaQ&c=jf_iaSHvJObTbx-
> siA1ZOg&r=rFsK2G-
>
kj0QZo0UhxA9XKA&m=lbGfjbkv3Q2gb_9r2CrFg7iehrvchvT2AdhVuuMZPzc&s=7tuhSdThQv_b82yhdAYgJk0cPl2I0hp5M2-0XQ44oWU&e=

> >
> > "Worry more about your character than your reputation.  Character is
what
> > you are, reputation merely what others think you are." - John Wooden
> >
> > -----Original Message-----
> > From: IBM Mainframe Discussion List <[email protected]> On
Behalf
> > Of Kirk Wolf
> > Sent: Wednesday, December 11, 2019 7:20 PM
> > To: [email protected]
> > Subject: Re: OMVS Tuning ?
> >
> > Lionel,
> >
> > STEPLIB=none has been around for a long time, and I see it documented,
for
> > example, in the V1R13 z/OS UNIX Planning book.
> >
> > Can you offer any more details as to where the performance is bad or
how
> > you notice?   I would assume that what you are describing is something
> > relative to the time required to do a fork/exec or spawn.   Which?
local
> > spawn?
> >
> > On Wed, Dec 11, 2019 at 2:04 PM Lionel B Dyck <[email protected]> wrote:
> >
> > > Found one major issue - the systems where the issue is are z/OS 2.1
> > > and the STEPLIB=none variable looks to have been new in 2.2.
Reducing
> > > the STEPLIB and ISPLLIB solved things.
> > >
> > >
> > > Lionel B. Dyck <sdg><
> > > Website: https://urldefense.proofpoint.com/v2/url?
> u=http-3A__www.lbdsoftware.com&d=DwIFaQ&c=jf_iaSHvJObTbx-
> siA1ZOg&r=rFsK2G-
>
kj0QZo0UhxA9XKA&m=lbGfjbkv3Q2gb_9r2CrFg7iehrvchvT2AdhVuuMZPzc&s=7tuhSdThQv_b82yhdAYgJk0cPl2I0hp5M2-0XQ44oWU&e=

> > >
> > > "Worry more about your character than your reputation.  Character is
> > > what you are, reputation merely what others think you are." - John
> > > Wooden
> > >
> > > -----Original Message-----
> > > From: IBM Mainframe Discussion List <[email protected]> On
> > > Behalf Of Lionel B Dyck
> > > Sent: Wednesday, December 11, 2019 12:58 PM
> > > To: [email protected]
> > > Subject: OMVS Tuning ?
> > >
> > > I am working on a project where the omvs performance on some systems
> > > is slower than one would like -definitely not sub-second.
> > >
> > >
> > >
> > > I'm looking for tuning suggestions.
> > >
> > >
> > >
> > > When under ISPF the elapsed time is more than double the elapsed time
> > > when in native TSO.
> > >
> > >
> > >
> > > Note that the OMVS environment includes STEPLIB=none but my guess is
> > > that the ISPLLIB may be the culprit when under ISPF - I just can't
> > > prove that (yet).
> > >
> > >
> > >
> > > Thanks for any guidance.
> > >
> > >
> > >
> > >
> > >
> > > Lionel B. Dyck <sdg><
> > > Website:  <https://urldefense.proofpoint.com/v2/url?
> u=http-3A__www.lbdsoftware.com_&d=DwIFaQ&c=jf_iaSHvJObTbx-
> siA1ZOg&r=rFsK2G-
>
kj0QZo0UhxA9XKA&m=lbGfjbkv3Q2gb_9r2CrFg7iehrvchvT2AdhVuuMZPzc&s=ltyuZpKNwSpNExe-7RTeZ8HQj5a4b642JlbYuNkLQxc&e=

> > https://urldefense.proofpoint.com/v2/url?
> u=http-3A__www.lbdsoftware.com&d=DwIFaQ&c=jf_iaSHvJObTbx-
> siA1ZOg&r=rFsK2G-
>
kj0QZo0UhxA9XKA&m=lbGfjbkv3Q2gb_9r2CrFg7iehrvchvT2AdhVuuMZPzc&s=7tuhSdThQv_b82yhdAYgJk0cPl2I0hp5M2-0XQ44oWU&e=

> > >
> > > "Worry more about your character than your reputation.  Character is
> > > what you are, reputation merely what others think you are." - John
> > > Wooden
> > >

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to