New updates

> On Sat, 2015-06-27 at 10:32 -0300, Juan Vuletich wrote:
>
>> Hi Folks,
>>
>>
>> Our new community member Peter van Rooijen (welcome Peter!) reported an
>>  intolerable slowness in the Transcript on the Mac. It turned out to be
>>
>
> Hi Peter!
>
>
>> that direct Display access (without the double buffering Morphic uses)
>> is extremely slower on a Mac than on Linux or Windows (about 1000 times
>>  slower!). The fix was not trivial, but the results are good. What I
>> did is to have the display update deferred all the time, not just during
>>  Morphic update. This requires forking a new process that pushes
>> Display
>> to the actual screen if requested, but no more than 20 fps. It seems to
>> work ok on all 3 major platforms, with elimination of flicker, modest
>> speedup on Linux and Windows and very dramatic speedup on Mac.
>>
>
> One minor display update difference I notice as a result of this change
> is that when loading a large number of packages back-to-back, I used to see
> progress popups sprayed to the screen (i.e. one per package) that
> continued until all the packages were loaded.  Now I only see a few and
> they really don't get updated giving the appearance of being frozen, but
> the packages are still loading as they should.  It's not really an issue
> in this case (i.e. either behavior is a bit janky but not annoying enough
> to get me to do anything about right now) but something to keep an eye on
> with long-running processes (spanning several seconds or more) in other
> parts of the UI.

> Hi Phil, Peter,
>
> (below)
>
> On Sat, June 27, 2015 8:19 pm, Phil (list) wrote:
> > On Sat, 2015-06-27 at 10:32 -0300, Juan Vuletich wrote:
> >
> >> Hi Folks,
> >>
> >>
> >> Our new community member Peter van Rooijen (welcome Peter!) reported an
> >>  intolerable slowness in the Transcript on the Mac. It turned out to be
> >>
> >
> > Hi Peter!
> >
> >
> >> that direct Display access (without the double buffering Morphic uses)
> >> is extremely slower on a Mac than on Linux or Windows (about 1000 times
> >>  slower!). The fix was not trivial, but the results are good. What I
> >> did is to have the display update deferred all the time, not just during
> >>  Morphic update. This requires forking a new process that pushes
> >> Display
> >> to the actual screen if requested, but no more than 20 fps. It seems to
> >> work ok on all 3 major platforms, with elimination of flicker, modest
> >> speedup on Linux and Windows and very dramatic speedup on Mac.
> >>
> >
> > One minor display update difference I notice as a result of this change
> > is that when loading a large number of packages back-to-back, I used to see
> > progress popups sprayed to the screen (i.e. one per package) that
> > continued until all the packages were loaded.  Now I only see a few and
> > they really don't get updated giving the appearance of being frozen, but
> > the packages are still loading as they should.  It's not really an issue
> > in this case (i.e. either behavior is a bit janky but not annoying enough
> > to get me to do anything about right now) but something to keep an eye on
> > with long-running processes (spanning several seconds or more) in other
> > parts of the UI.
>
> Please try changing the Delay in #screenUpdater from 50 to 20, and do
> 'DisplayScreen installScreenUpdater'. They try loading the packages. Does
> this improve it? Specially if it does not, but in any case, what OS and VM
> are you using?
>

> Peter, does this higher frame rate (from 20fps to 50fps) make the
> Transcript much slower in your setup?
>
> Thanks,
> Juan Vuletich
>
> _______________________________________________
> Cuis mailing list
> [hidden email]
> http://jvuletich.org/mailman/listinfo/cuis_jvuletich.org

>>> ...
>>> One minor display update difference I notice as a result of this
>>> change is that when loading a large number of packages back-to-back, I
>>> used to see progress popups sprayed to the screen (i.e. one per
>>> package) that continued until all the packages were loaded.  Now I
>>> only see a few and they really don't get updated giving the appearance
>>> of being frozen, but the packages are still loading as they should.
>>> It's not really an issue
>>> in this case (i.e. either behavior is a bit janky but not annoying
>>> enough to get me to do anything about right now) but something to keep
>>> an eye on with long-running processes (spanning several seconds or
>>> more) in other parts of the UI.
>>
>> Please try changing the Delay in #screenUpdater from 50 to 20, and do
>> 'DisplayScreen installScreenUpdater'. They try loading the packages.
>> Does
>> this improve it? Specially if it does not, but in any case, what OS and
>> VM
>> are you using?
>>
>
> That resolves it initially (for about the first 5-10 seconds) but then
> the updates stop showing again.  If I recall correctly, doesn't Morphic
> back off on screen updates by default after a period without
> keyboard/mouse activity?  The interplay of your latest changes and that
> might be the issue.  I'm running on Linux with the Cog 15.14.3308 build.
> I've tested the new build with some long-running animated morphs and
> they seem to be fine so for this issue, I wouldn't worry about it too much
> unless you're concerned that it's a symptom of a larger problem.
>

> Hi Phil,
>
> >>> ...
> >
> > That resolves it initially (for about the first 5-10 seconds) but then
> > the updates stop showing again.  If I recall correctly, doesn't Morphic
> > back off on screen updates by default after a period without
> > keyboard/mouse activity?  The interplay of your latest changes and that
> > might be the issue.  I'm running on Linux with the Cog 15.14.3308 build.
> > I've tested the new build with some long-running animated morphs and
> > they seem to be fine so for this issue, I wouldn't worry about it too much
> > unless you're concerned that it's a symptom of a larger problem.
> >
>
> Well, I want to understand. And the best time is right now, when all this
> is fresh in mi mind. My Linux box is at the office, so I'll test on it on
> Monday.
>

> Thanks,
> Juan Vuletich
>
> On 6/28/2015 5:58 PM, Phil (list) wrote:
> > ...
> > Fair enough.  FYI, I downloaded the latest Cog (3390) and observed the
> > same behavior.  I am loading the packages via CodePackageFile
> > #installPackageStream: on a reasonably large set of packages (all core
> > packages plus 4 dozen or so more)
> > Thanks,
> > Phil
>