[Info-vax] READ and WRITE vs. SEARCH/OUTPUT

[Paul Sture]

On Mon, 23 Jan 2012 20:08:10 -0500, Arne Vajhøj wrote:

> On 1/23/2012 6:58 AM, Paul Sture wrote:
>> On Sat, 21 Jan 2012 16:29:00 +0000, Phillip Helbig---undress to reply
>> wrote:
>>> Some tests indicate that SEARCH is much faster.  Can one depend on
>>> this always being the case?
>>
>> 1. If you try the comparison on a VAX, you are probably looking at
>> SEARCH being many times faster than DCL.
>>
>> 2. Empirical testing I did back in VAX days suggested that given
>> sufficient memory SEARCH uses larger buffers than, say, a COBOL
>> executable.
> 
> Wouldn't that depend on how the COBOL program was written?
> 

In COBOL the buffer sizes are defined using the RESERVE clause, somewhere 
up in the Data Division.  I know of no way of altering that at run time 
to suit available resources.  Set it too high and you might run out of 
nonpaged pool, too low and normal COBOL reads won't go as fast as SEARCH.

Given plenty of memory though, the performance gain could be quite 
impressive whipping up a large RMS indexed file though. It gave me a 
faster way to count records in an indexed file than for example CONVERT 
or ANAL/RMS/STATISTICS.

-- 
Paul Sture