[mythtv] Scheduler needs table keys?

[Stuart Auchterlonie]

Chris Pinkham wrote:
> * On Mon Jan 29, 2007 at 02:11:41PM +1000, Paul Andreassen wrote:
>> What other RECTABLE is there?
> 
> I think that's how the speculative scheduler works.
> 
>> CREATE INDEX sr ON record (search, recordid);
>> CREATE INDEX cvc ON channel (chanid, visible, callsign);
>> CREATE INDEX mt ON program (manualid, title);
>> CREATE INDEX sourceid ON cardinput (sourceid);
>> CREATE INDEX rd ON record (recordid, dupmethod);
>> CREATE INDEX tdp ON oldrecorded (title, duplicate, programid);
>> CREATE INDEX tdf ON oldrecorded (title, duplicate, findid);
>> CREATE INDEX tds ON oldrecorded (title, duplicate, subtitle);
>> CREATE INDEX tdd ON oldrecorded (title, duplicate, description(128));
>> CREATE INDEX csmgps ON program (chanid, starttime, manualid, generic, 
>>     programid, subtitle);
>> CREATE INDEX csmgpd ON program (chanid, starttime, manualid, generic, 
>>     programid, description(128));
>> CREATE INDEX trdf ON recorded (title, recgroup, duplicate, findid);
>> CREATE INDEX trdps ON recorded (title, recgroup, duplicate, programid, 
>>     subtitle);
>> CREATE INDEX trdpd ON recorded (title, recgroup, duplicate, programid, 
>>     description(128));
> 
> No real difference on any of these that I could tell.  I applied them one
> by one and did a "mythbackend --resched" after each.
> 

Well that's not really going to be a good test, since after the first
reschedule all of the referenced DB tables will have been read into
memory, so having an index isn't going to speed things up much.

Where it will help is when doing a reschedule after your machine has
been busy doing other things, so then just the indexes need reading
into memory.

Stuart

!DSPAM:10,45bf16bf279581340614851!