[ptx] auto fine-tune on large images

[Rich]

douglas wilkins wrote:
> --- Bruno Postle <bruno at postle.net> wrote:
> 
>>On Thu 19-May-2005 at 16:49 +0300, Rich wrote:
>>
>>>there probably is no way to run engine on one machine and frontend
>>>on another ? :)
>>
>>Two ideas:
>>
>>- You can run everything on the faster machine and display it on
>>  your laptop via remote X, just `ssh -X monster.machine hugin`

that's exactly what i would have done - unfortunately it is a server 
without x :)

>>- If both machines share the same home partition via NFS, you can
>>  change just your enblend command to run on the faster machine:
>>  `ssh monster.machine enblend`

right now i'm running enblend manually on monster.machine ;) - this way 
i can schedule it to run at the night and i don't have to worry about 
availability of my workstation

> If you are running linux on monster.machine, these two suggestions make
> perfect sense (and make me wonder why I didn't think of it). You could use
> the same technique as suggestion 2 to run PTOptimizer as well.
> 
> If monster.machine is a windows machine, things are a little trickier. :-)
> 
> This doesn't help with auto-tuning though.

which was the slowest part, but during weekend i tried managing all 
images in single project, as pablo suggested, and limited enblends 
memory usage. this really helps, autotuning was as fast as with two 
images and enblend also was much faster.

an idea for post-0.5 releases : maybe a lower enblend memory limit 
should be enabled by default to protect others (default being 1gb is 
pretty high, although many have that amount...)

to increase intuitivity, it could be a textbox where only amount should 
be entered and then it would be passed to enblend as -m <value> (of 
course, this can be done right now by adding -m to arguments that are 
passed to enblend, but i believe that proposed approach would make hugin 
more user friendly :) )

> regards,
> Doug
-- 
  Rich