[Halld-online] Revised evio2et rates on gluon20

[Elliott Wolin]

Hi,

A typo in evio2et in the previous test caused the program to skip some 
of the work, revised rates are:

  - uncompressed events, single producer, 22 kHz at 99% (17%).
  - compressed events, single producer, 55 kHz at 99% (30%).

So the revised rates are about half the previous ones, I didn't try 
multiple producers.  In any case this should suffice for initial 
testing, and I'll speed things up internally as well.

I also created et2xml, which dumps events from an ET system to the 
screen in xml format.







On 06/28/2013 11:18 AM, Elliott Wolin wrote:
> Hi,
>
> I'm testing inserting evio events into an ET system on gluon20 (program
> is evio2et).  I get events from a file and reopen the file when I run
> out of events.
>
> It runs at 45 kHz on gluon20 reading uncompressed events from halldraid1
> and inserting them into an ET system on gluon20 using the traditional
> format (memcpy, no evio buffer interface) at 99% cpu (et system takes 10%).
>
> Using compressed events (about 1/3 the size) I get 100 kHz at 99% cpu (23%).
>
> This should suffice for all our initial testing.  If we need higher
> input rates we can run multiple copies of evio2et.  With all evio2et
> copies at 99% cpu, I get:
>
>    - uncompressed events two copies, 72 kHz (et system at 20%), three
> copies 85 kHz (25%)
>    - compressed events two copies 170 kHz (40%), three copies 190 kHz (50%)
>
> At three copies I think they are all fighting over the same input file.
>
> There are options to speed things up in evio2et, if needed.
>

-- 

				Sincerely,
					Elliott
  

================================================================================


  Those raised in a morally relative or neutral environment will hold
		    no truths to be self-evident.
				

Elliott Wolin
Staff Physicist, Jefferson Lab
12000 Jefferson Ave
Suite 8 MS 12A1
Newport News, VA 23606
757-269-7365

================================================================================