Re: using timers in the GUI

[Torsten]

On 21.03.2013 07:43, John W. Eaton wrote:
> On 03/20/2013 04:41 PM, Torsten wrote:
> 
>> Yes, the patches work fine and should also fix bug #38537.
>> Octave now consumes approximately 12% of cpu time on my system with only
>> a few variables in the workspace (without the patch 3..4%). So limiting
>> to one update per prompt would be helpful.
> 
> I checked in a series of changes so that the workspace list is updated
> just before the prompt is printed and the history list is updated just
> after a line of user input is accepted.  Now I don't see any CPU usage
> when Octave is idle.  Prior to that, I saw about 13%, similar to what
> you were seeing, because we were calling these update functions
> repeatedly even when there was nothing to do.  Even with my earlier
> patch that eliminated the timers but put the callbacks in the
> readline input event loop the CPU usage was about the same because the
> functions were still being called many times while readline was
> waiting for input.
> 
>   http://hg.savannah.gnu.org/hgweb/octave/rev/bf8397caeff1
>   http://hg.savannah.gnu.org/hgweb/octave/rev/072b3e6504c4
>   http://hg.savannah.gnu.org/hgweb/octave/rev/610617eb84d1
>   http://hg.savannah.gnu.org/hgweb/octave/rev/d4b6ad43bc87
>   http://hg.savannah.gnu.org/hgweb/octave/rev/d8babb249f65

Nice, not more than 0.5% cpu time when octave is idle now.

Torsten