Hi Renyue,

I've just tested a similar script on the same changeset that you are on and had some amount of success. With yt-2.5.1, having a completely arbitrary subvolume to build the AMRKDTree was not yet fully functional, and has since been finished up. For example, it used to be that if your le/re edges didn't line up well with grids/cells you could run into problems.

It would be very useful to see your entire script and any output it creates. Could you paste each using http://paste.yt-project.org/ ?

If the code is just taking a long time (and not erroring out), it may just be that it is taking a long time to construct the ghost zones (requires a lot of disk i/o). One thing you could test is to use no_ghost=True in both lines of your script. This will degrade the quality of the rendering but would be useful information to know what the root cause of the problem is.

If, on the other hand, the code is erroring out after the volume=AMRKDTree(...) call, I've seen similar behavior on yt-2.5.1, which was fixed at some point in later development. I believe if you can update to yt-2.6.1 most if not all of the problems will go away.

I hope this helps,

Sam

On Sat, Mar 29, 2014 at 9:47 AM, Renyue Cen <cen@astro.princeton.edu> wrote:

Hi Nathan (and John),

with "yt instinfo" I got
8dce70a9c689

I was not very clear in my previous email. Here is a somewhat clearer description:

If I just use: volume = AMRKDTree(pf,fields=["Density"],no_ghost=False)
then code does not complain about this AMRKDTree call anymore, but the subsequent call

cam = pf.h.camera(c, L, W, (N,N), tf, volume=volume, no_ghost=False, north_vector=Ldm,steady_north=True)

appears to take forever to run.

On the other hand, if I use (I made sure the c and WW are sensible numbers):
volume = AMRKDTree(pf, le=c-0.5*WW,re=c+0.5*WW, fields=["Density"], no_ghost=False)

then I don't seem to pass this statement, because a print line following it does not get executed.

Thanks very much,
Renyue

Hi Renyue,

Which version of yt are you running? If you installed using the install script, the output of 'yt instinfo' run on the command line should be suffieicebt to figure out exactly which version you are using.

You may have to directly import AMRKDTree from yt.visualization.volume_rendering.api, although I believe it is still imported into yt.mods in the current stable release of yt.

Nathan

On Saturday, March 29, 2014, Renyue Cen <cen@astro.princeton.edu> wrote:
Hi folks,

Some time ago John Wise kindly gave me a yt script that uses AMRKDTree to speed up "camera" 3-d
rendering, it worked very well and fast (takes like a few minutes to do a rendering of a galaxy with 2048x2048 pixel).
It looks like

volume = AMRKDTree(pf, le=le, re=re, fields=["Density"], no_ghost=False)

then used in
cam = pf.h.camera(c, L, W, (N,N), tf, volume=volume, no_ghost=False, north_vector=Ldm,steady_north=True)

It appears that yt has evolved and the "volume ..." statement no longer works.
Is there a solution to that in the newer version?

Alternatively, is there another speedy way to do 3-d rendering?

Thanks very much,
Renyue

_______________________________________________
yt-users mailing list
yt-users@lists.spacepope.org
http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

_______________________________________________
yt-users mailing list
yt-users@lists.spacepope.org
http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org