Date: Fri, 7 Apr 2000 08:54:40 -0700
Reply-To: Puddin' Man <pudding_man@ALTAVISTA.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Puddin' Man <pudding_man@ALTAVISTA.COM>
Subject: Re: Fwd: Re: MVS -- U1319, No MKLES
Content-Type: text/plain
On Fri, 07 April 2000, "Paul Dorfman" wrote:
>
> Puddin',
>
> From your test, it appears that it is a V8 bug. Apparently, the transition
> from V6 to V8 has rendered MINSTG inoperative - exactly where MINSTG would
It (MINSTG) looks inoperative with respect to the data step. Period.
> be even more valuable since V8 consumes much more memory than V6 *at idle*.
Quite so. Surprisingly more memory. If it was only 10-20% more than V6
it might be expected. What I'm seeing is somewhat scary.
> Practically speaking, given the same amount of real storage OS/390 gives me
> now (70M), many of my smoothly running programs will bomb with a loud crack.
Ouch! And 70 mb of physical memory is not "small change". I'd guess
many users do well to get half that much.
> I am going to report it to SI and insist that our system people keep V6
> alive until at least V8.1 or V8.2, or until the bug has been fixed.
Well, I hope its a bug, and not a "design feature".
Is anybody in a position to run simple memory-usage tests on,
say, Intel SAS V6 vs V8? Unix??
> Thanks for your input.
Glad to help. Sorry I don't have better news.
Zalut,
Puddin'
> Kind regards,
> ======================
> Paul M. Dorfman
> Jacksonville, Fl
> ======================
> >From: Puddin' Man <pudding_man@ALTAVISTA.COM>
> >Reply-To: Puddin' Man <pudding_man@ALTAVISTA.COM>
> >To: SAS-L@LISTSERV.UGA.EDU
> >Subject: Fwd: Re: MVS -- U1319, No MKLES
> >Date: Fri, 7 Apr 2000 06:47:19 -0700
> >
> >On Thu, 06 April 2000, "Paul M. Dorfman" wrote:
> >
> > >
> > > Puddin',
> > >
> > > Right. But run a second step - after the first - allocating say 100,000
> > > elements, and I bet you will see a lot of difference.
> > >
> > > Cheers,
> > > Paul
> >
> >Take a peek, see what you think:
> >
> > *------------------------------------------------------------------*
> > *- SAS Version 8.00 Enhanced TS M0 -*
> > *------------------------------------------------------------------*
> > NOTE: SAS system options specified are:
> > SORT=4 FULLSTATS MINSTG
> >
> > NOTE: The initialization phase used 0.08 CPU seconds and 6068K.
> > 1 data _null_; array a(2000000) _temporary_; run;
> >
> > NOTE: The DATA statement used the following resources:
> > Task memory - 4276K (74K data, 4202K program)
> > Total memory - 22559K (17824K data, 4735K program)
> >
> > 2 data _null_; array a(100000) _temporary_; run;
> >
> > NOTE: The DATA statement used the following resources:
> > Task memory - 4192K (74K data, 4118K program)
> > Total memory - 22475K (17824K data, 4651K program)
> >
> > 3 data _null_; x=1 ; run;
> >
> > NOTE: The DATA statement used the following resources:
> > Task memory - 4188K (70K data, 4118K program)
> > Total memory - 22475K (17824K data, 4651K program)
> >
> >It appears that the data step grabs "17824K data" and holds
> >on to it regardless of the needs of subsequent data steps. Eh?
> >
> >Here's a quick/dirty test with PROC's:
> >
> >5 data a;
> >6 do i=1 to 250 by 4;
> >7 do j=1 to 250 by 4;
> >8 do k=1 to 250 by 4;
> >9 x=normal(0);
> >10 output;
> >11 end;
> >12 end;
> >13 end;
> >
> >NOTE: The data set WORK.A has 250047 observations and 4 variables.
> >NOTE: The DATA statement used the following resources:
> > Task memory - 4272K (70K data, 4202K program)
> > Total memory - 22583K (17824K data, 4759K program)
> >
> > 14 proc summary nway;
> > 15 class i j k;
> > 16 var x;
> > 17 output out=b;
> >
> > NOTE: Processing on disk occurred during summarization.
> >Peak disk usage was approximately 46 Mbytes. Adjusting SUMSIZE may
> >improve
> > performance.
> > NOTE: There were 250047 observations read from the dataset WORK.A.
> > NOTE: The data set WORK.B has 1250235 observations and 7 variables.
> > NOTE: The PROCEDURE SUMMARY used the following resources:
> > Task memory - -16805K (-20660K data, 3855K program)
> > Total memory - 27997K (23328K data, 4669K program)
> >
> > 18 proc print data=b(obs=20);
> > 3 The SAS Syst
> >
> > 19 ENDSAS;
> > NOTE: There were 20 observations read from the dataset WORK.B.
> > NOTE: The PROCEDURE PRINT printed page 1.
> > NOTE: The PROCEDURE PRINT used the following resources:
> > Task memory - 4333K (61K data, 4272K program)
> > Total memory - 22856K (17824K data, 5032K program)
> >
> > NOTE: The SAS session used 10.35 CPU seconds and 28360K.
> >
> >Up to "23328K data" then back down to "17824K data". It
> >doesn't seem to wanna "let loose" the data space
> >used by the data step.
> >
> >But it seems to have released some memory used by
> >PROC SUMMARY. Thats better than nothing ...
> >
> > Prosit,
> > Puddin'
> >
> >"Gimme back my wig
> > and, honey, let you head go bald!"
> > - Hound Dog Taylor, maybe 1972
> >
> >
> >
> >*******************************************************
> >* Arrest records (for the infamous Storeyville *
> >* district of New Orleans) at the turn of the *
> >* century list underworld characters with bizarre *
> >* aliases like Drop o' Sack, Cinderella, Pudding Man, *
> >* Willie the Pleaser, Diamond Dick, Lead Pencil, ... *
> >*******************************************************
> >*** Puddin' Man *** Pudding_Man@altavista.com *****
> >*******************************************************
> >;
> >____________________________________________________________________
> >
> >For the largest MP3 index on the Web, go to http://mp3.altavista.com
> >
> >____________________________________________________________________
> >
> >
> >*****************************************************
> >*** Puddin' Man *** pudding_man@altavista.com ***
> >*****************************************************;
> >
> >____________________________________________________________________
> >
> >For the largest MP3 index on the Web, go to http://mp3.altavista.com
> >
> >____________________________________________________________________
>
> ______________________________________________________
> Get Your Private, Free Email at http://www.hotmail.com
*****************************************************
*** Puddin' Man *** pudding_man@altavista.com ***
*****************************************************;
____________________________________________________________________
For the largest MP3 index on the Web, go to http://mp3.altavista.com
____________________________________________________________________
| 
