[GRASSLIST:530] Re: GRASS gism2 and speed

[Michael Barton]

With respect to speed of redraw, I got around to doing some informal tests
(watching screen redraws and timing them on my watch's stopwatch function).

Here are the results. I used the Spearfish 6 demo data, with region set to
match the 10m dem, and both monitors set to near my full screen size. I'm
using a iMac G5 with 1Gb RAM. Times are in seconds:

xmon and xdriver
10m dem = 2.37
10m dem + roads = 2.71
relief+10m dem + roads = 5.23
relief+10m dem + roads +streams= 5.50


tcltk canvas and png driver
10m dem (1st) =4.39
10m dem (2nd) =2.54
10m dem  (2nd) + roads (1st) = 3.22
10m dem  (2nd) + roads (2nd) = 2.59
relief (1st) +10m dem (1st) + roads  (2nd)= 7.15
relief (2nd) +10m dem (2nd) + roads (2nd) = 2.73
relief (2nd) +10m dem (2nd) + roads (2nd) + streams (1st) = 3.41
relief (2nd) +10m dem (2nd) + roads (2nd) + streams (2nd) = 2.73

*1st refers to the first time a layer is rendered; 2nd refers to all
subsequent renderings.


The first time a map is drawn, it is significantly slower in the tcltk
canvas. For a single layer map of the 10m dem file, the canvas/png-driver
combination the xmon draw time is only 54% of the canvas draw time-for the
FIRST time it is drawn. For subsequent redraws, both systems are about the
same.

If you add a vector layer on top of the raster file, the redraw time in the
xmon goes up a bit. With the canvas, it goes up more the first time (though
not as much as the first time the raster was drawn), then drops back to it's
one-map time for subsequent redraws.

If you drape a color raster on top of a shaded relief map (d.his) and add a
vector, again the xmon time goes up quite a bit. The canvas time also goes
up even more the first time it's drawn, but drops back to near the one-layer
map time for subsequent redraws. The real difference is adding another
vector layer (streams). The xmon draw time again goes up a bit. It is 5.5
sec. every time it is redrawn. The canvas draw time goes up a bit to 3.41
sec the first time it is rendered and again drops back to 2.73 sec. for
every subsequent re-rendering.

So, for simple maps, the xmon system is considerably faster for the first
time a map is drawn, and a little faster for any subsequent redraw. For more
complicated maps, the canvas can be slower or faster for the initial draw,
depending on how many maps are being rendered for the first time. However,
with each additional map, it becomes significantly faster than xmon because
only new maps or changed maps are re-rendered.

With respect to your problem with rendering, I don't see it, but am working
with the 18 March binary from Lorenzo.

Michael
__________________________________________
Michael Barton, Professor of Anthropology
School of Human Evolution & Social Change
Center for Social Dynamics and Complexity
Arizona State University

phone: 480-965-6213
fax: 480-965-7671
www: http://www.public.asu.edu/~cmbarton


> From: Renato Henriques <rhenriques at dct.uminho.pt>
> Date: Fri, 31 Mar 2006 19:38:02 +0100
> To: Michael Barton <michael.barton at asu.edu>
> Conversation: GRASS
> Subject: Re: GRASS
> 
> Hi Michael
> 
> GRASS 6.1.cvs-060325 for OSX (from Lorenzo Moretti)
> 
> I'm using your new GRASS interface almost in a daily basis. It's very nice
> to work with it. Although I noticed some problems in this latest release.
> You detailed the output with more graphical rich information. Somehow the
> maps display became very slow and I suppose that has something to do with
> the output window. Maybe you could make the detailed output optional or so.
> I tried to use the "Run (background)" and the "Run in Xterm" buttons but
> they seem to have no effect.
> The biggest bug (or possible bug) that affects my image output is related
> with the occultation of raster layers below vector ones. If I use, for
> instance, three raster layers and one or two active vector layers above
> them, the occultation of any raster layer will have no effect. They will all
> show up below the vector layers despite the state of the activation button
> on their left. If I turn the vector layers to inactive, all raster layers
> will work as expected, being possible to activate or inactivate them and get
> the expected result.
> I hope that I've made myself clear using my poor English.
> Cheers and keep on the good work.
> 
> 
> -- 
> Renato Henriques
> Department of Earth Sciences
> Campus de Gualtar
> University of Minho
> Braga Portugal
> Tel +351 253604031
> rhenriques at dct.uminho.pt
> 
> 
> 
> 
> 
> 
> 
> 
>