[yt-users] Incoming: Threaded volume rendering, and changes to off-axis projections

Matthew Turk matthewturk at gmail.com
Fri Jun 15 15:58:27 PDT 2012


Hi all,

This landed today in the development tip of yt.

If you update and run into any weird problems, please stop by IRC or
email the list and we can work through them with you.

-Matt

On Wed, Jun 13, 2012 at 11:23 AM, Matthew Turk <matthewturk at gmail.com> wrote:
> Hi all,
>
> SInce last December, Sam and I have been working on and off on
> threading the volume rendering system and providing a major step
> forward in its speed and quality.
>
> We've issued a pull request to finish off these efforts and move them
> into the main yt core.  This includes a threaded volume renderer,
> transfer functions with a degree of opacity, and the ability to (in
> the future) much more explicitly control the behavior of the volume
> renderer on the cell-level.  All of the code is now separated into
> very small component pieces, and it is more maintainable.  The
> threading works in conjunction with both MPI parallelism and MPI
> parallelism for tiling, which means in theory the yt volume renderer
> now operates at three levels of parallelism.  (Which is kind of neat,
> I think.)
>
> However, once this is accepted, the minimum requirements to build the
> development branch of yt will be upgraded to include Cython 0.16.
> This is because Cython 0.16 includes OpenMP support.  The installation
> script has been updated to include Cython 0.16, so if you re-run that
> (or install Cython 0.16 manually with "pip install -U Cython") then
> you should be set.  Otherwise you may run into problems during your
> next "yt update".  Additionally, _amr_utils has been renamed to lib,
> which I don't think will cause issues but there is an outside chance
> it will.
>
> Finally, the off-axis-projections have changed in behavior.  They no
> longer interpolate within a cell.  The end result is that the
> projections are truer to the data, but this comes at the expense of a
> bit of smoothness.  They are now also ~20x faster.
>
> This will be a core component of the yt 2.4 release, scheduled for a
> few weeks from now.  We encourage you to test this branch and new
> behavior.  Also, please reply if you have any questions or comments
> about any of this!
>
> Reference info:
>
> http://lists.spacepope.org/pipermail/yt-dev-spacepope.org/2012-June/002039.html
> https://bitbucket.org/yt_analysis/yt/pull-request/168/volume-rendering-refactor
>
> example "samplers":
>
> https://bitbucket.org/MatthewTurk/yt-refactor/src/ec374f56d5b6/yt/utilities/lib/grid_traversal.pyx#cl-310
> https://bitbucket.org/MatthewTurk/yt-refactor/src/ec374f56d5b6/yt/utilities/lib/grid_traversal.pyx#cl-344
>
> -Matt



More information about the yt-users mailing list