Re: perfm of Date\Time\Timestamp

On 11-Jul-2011 08:49 , Henrik Rützou wrote:

As I remember that date support in DB2 and thereby in RPG came along
in the early/mid 1990s but it did speed things down because at that
time processors where slow and not as to day where they are mainly
waiting for disk I/O.

As I recall the Date\Time support in the languages had some performance issues due to implementation choices. I recall also that [a couple releases] later some changes were made to improve against that original implementation [perhaps only in the RPG HLL?]; something to do with utilization of external-form [visual presentation formats] versus internal-form [integer data] which required conversion from the former to the latter before doing calculations, and then the reverse conversion [to external form] for writing to the database. The original implementation in the database as I understood, was chosen to fit nicer with both the interface to and the [original design and] implementation of, the HLLs. I believe part of the changes to improve performance may have been via exposing database record buffers to an HLL [perhaps only RPG updated to take advantage] with date data in internal-form versus external-form for read, and similarly enabling the record buffer from the HLL to contain internal-form versus external-form for write.

FWiW in a recent [but unrelated] search for a compiler PTF I noticed a new option to turn off date [, time, and timestamp] validation which could give even further improvement for certain processing [according to statements in the APAR-text\PTF-cover-letter] where doing so might be justified while also being a trusted effect:

APAR SE46814 gives VALIDATE(*NODATETIME); note limitation on *USA:
http://www-912.ibm.com/n_dir/nas4apar.NSF/c79815e083182fec862564c00079d117/31a8843c41bb1b1e8625784900421ef3?OpenDocument

Regards, Chuck