Date:         Mon, 19 Nov 2001 15:03:02 -0500
Reply-To:     Larry Bertolini <bertolini.1@OSU.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Larry Bertolini <bertolini.1@OSU.EDU>
Organization: Ohio State University
Subject:      Re: Second cousin, once removed, of Speed of SAS vs. COBOL vs.
Content-Type: text/plain; charset=us-ascii

Puddin' Man wrote:

> The curious, curious fella that I am, of course, hadda
> do exactly that.
>
> For your /* pair 1 */ I got  673 EXCP's.
>
> For your /* pair 2 */ I got   1291 EXCP's.
>
> What EXCP count would you expect if SAS
> transferred only one 175 b. record from
> DASD to memory? 1? 3?? 5???
>
> Direct access to a flat-file record with SAS?
> This po' boy remains unconvinced.
>
>   Puddin'

OK, on the unlikely chance that anyone cares,
here are my EXCP counts for my tests:

INFILE/INPUT method:     #records skipped  EXCPs   CPU time of SAS step
--------------------     ----------------  -----   --------------------
RECFM=N; INPUT @1750001    10,000            1     0.22 sec
FIRSTOBS=10001             10,000           67     0.25 sec
RECFM=N; INPUT @17500001  100,000            1     0.22 sec
FIRSTOBS=100001           100,000          633     0.42 sec

Note that I am counting EXCPs only to the file referenced by
the INFILE; I'm not counting EXCPs to other datasets.
(Data gathered from SMF Type 14 records.)

The EXCP counts for the "FIRSTOBS=n" tests are relatively
consistent with my expectation (LRECL=175,BLKSIZE=27825,
159 records per block).  SAS appears to start reading from
the beginning of the file, and discarding records before the
FIRSTOBS record.

The EXCP counts for the "RECFM=N" tests suggest (to me,
anyway) that SAS is going directly to the desired record.