RE: [EXTERNAL] Re: Performance issues using GAWK 3.1.6 ->from Win 2008 to Win 2016

[Koleti, Haritha]

Arnold -> I see the below email from you .    Is this the one  you are asking 
to try ?

- W  -> I asked my server team to look at i/o buffer size.    They did not give 
me clear answer so did not respond on this.  But when I checked on task manager 
Cache is showing ~4.6 GB used 1.4 GB free.  and Page memory also 178 MB free.

Once I have complete details from my server team I will let you know i/o buffer 
size.

Thanks
Haritha




On mobile phones, forward message to Cyber Security.

Wolfgang Laun <wolfgang.laun@gmail.com> wrote:

> The durations 10 min and 90 min suggest to me that a lot of i/o is
> going on. I have experienced performance changes of a similar order of
> magnitude due to changes in the default i/o buffer size.
> -W

This is an interesting idea. Eli, what if you supply a binary built with the 
following patch?

Thanks,

Arnold
--------------------
diff --git a/pc/gawkmisc.pc b/pc/gawkmisc.pc index 4e58b0a8..4be131b8 100644
--- a/pc/gawkmisc.pc
+++ b/pc/gawkmisc.pc
@@ -165,6 +165,9 @@ optimal_bufsize(fd, stb)  int fd;  struct stat *stb;  {
+#if 1
+return 4096;
+#else
/* force all members to zero in case OS doesn't use all of them. */
memset(stb, '\0', sizeof(struct stat));

@@ -182,6 +185,7 @@ struct stat *stb;
    && 0 < stb->st_size && stb->st_size < DEFBLKSIZE) /* small file */
return stb->st_size;
return DEFBLKSIZE;
+#endif
}

/* ispath --- return true if path has directory components */

[https://www.pseg.com/images/global/email/PSEG_emailsignature_PSEGw-tag_version2.png]<http://www.pseg.com>
[http://facebook.com/pseg]<http://www.facebook.com/pseg>        [Twitter] 
<http://www.twitter.com/psegdelivers>         [LinkedIn] 
<http://www.linkedin.com/company/pseg>       
[https://www.pseg.com/images/global/WP_LOGOgrey.png] <http://energizepseg.com/>


PSEGSC
-----Original Message-----
From: arnold@skeeve.com <arnold@skeeve.com>
Sent: Thursday, June 17, 2021 4:26 AM
To: eliz@gnu.org; arnold@skeeve.com
Cc: wolfgang.laun@gmail.com; Pereira, Ricardo <Ricardo_D.Pereira@pseg.com>; 
mortoneccc@comcast.net; mcollado2011@gmail.com; Pirane, Marco 
<Marco.Pirane@pseg.com>; Koleti, Haritha <Haritha.Koleti@pseg.com>; 
bug-gawk@gnu.org
Subject: Re: [EXTERNAL] Re: Performance issues using GAWK 3.1.6 ->from Win 2008 
to Win 2016

***CAUTION******CAUTION******CAUTION***This e-mail is from an EXTERNAL address. 
 The actual sender is  (arnold@skeeve.com) which may be different from the 
display address in the From: field. Be cautious of clicking on links or opening 
attachments. Suspicious? Report it via the Report Phishing button.  On mobile 
phones, forward message to Cyber Security.

Hi.

Eli Zaretskii <eliz@gnu.org> wrote:

> Are you sure it isn't a Gawk problem?  Did you look at the gprof
> profile I posted, and if so, does everything there look as expected?

I did look at it, and it looked fine. I apologize for not responding to the 
group about that.  I have many other things going on. :-(

> Because it could be that these scripts, no matter how inefficient and
> badly written, expose some issue with Gawk, which somehow rears its
> ugly head on Windows 10.  Until we have eliminated that possibility, I
> don't see how we can decide this is off-topic here.

What's clear is:

1. The scripts are poorly written. They should be fixed no matter what.

2. Haritha and company could use some external help in doing that
  (consultant, Ed, whatever, but inappropriate to the list).

3. There is *some* difference between the environments, given the
  change in runtimes between the same gawk binary (3.1.6) on
  both systems.  Trying to isolate that is also inappropriate to
  the list.

>From everything I've seen so far, I see no indication of any problem in gawk 
>itself.

> I do agree that the continuing discussion of how to improve the
> scripts may have crossed the line of being on-topic here.  But that
> wasn't what I was asking about.  The only issue that still bothers me
> is the sudden performance regression when the scripts were used on
> another version of the same OS.  We don't have any explanation for
> that, and in the experiment I conducted (see my report yesterday,
> which surprisingly didn't get any responses) I couldn't reproduce the
> regression.  Does the lack of responses to that experiment mean we
> consider the regression to be some fluke in the OP's environment
> unworthy of our attention?  If so, we can indeed stop talking about
> the problem here.

>From my point of view, it looks like "some fluke in the OP's environment"
and thus need not be discussed here.

I'm NOT saying "let's leave the user out in the cold". I'm simply asking that 
discussion and further attempts to help be done elsewhere.

Thanks,

Arnold

The information contained in this e-mail, including any attachment(s), is 
intended solely for use by the named addressee(s). If you are not the intended 
recipient, or a person designated as responsible for delivering such messages 
to the intended recipient, you are not authorized to disclose, copy, distribute 
or retain this message, in whole or in part, without written authorization from 
PSEG. This e-mail may contain proprietary, confidential or privileged 
information. If you have received this message in error, please notify the 
sender immediately. This notice is included in all e-mail messages leaving 
PSEG. Thank you for your cooperation.