Re: USEOSTIME and performance

David Williams wrote: > > In article <34EAD1EA.218B@mcmail.com>, Paul Watson <watson@mcmail.com> > writes > >I have been asked to check the performance differences between USEOSTIME > >on and off. I expected the that there would be a slightly better > >performance (4-5% according to manual) with USEOSTIME = 0/No. This > >isn't the case on the tests done, in fact it is exactly the reverse. > >With USEOSTIME set to 1/Yes it is consistently better by 4-5%. > > > >The test script is a simple insert into a temp table, no time > >information used. > > > Try using something that needs time. Done that > > 1. USEOSTIME=NO > Online handles it internally. Hence some work is done every second. > Proably an alarm 1 and gettimeofday? Agreed > > 2. USEOSTIME=YES > Online calls the OS every time the time is required. Hence some > work is done every time the time is needed. > Agreed > I expect 1 has some internal overhead even if the time is not used and > 2 has less overhed if the time is needed more than once in a given > second. Agreed > Try it inserting rows that need the time. Done that - it still quicker if you do the opposite to the manual. > > > >Anybody care to comment, or spot something I've missed > > > >BTW Online DSA 7.12/AIX 3.2.something > > -- Paul Watson # WF Software Ltd # It said run using Windows 95 Tel: +44 1436 674629 # or better in the manual Fax: +44 1436 678693 # so I used Unix www.wfsoftware.co.uk #