SPPM renderer (CPU-only)

[Chawl]

Aah ok, I will refer the hash from now on Tx.

[helloworld922]

sppm

Bi-directional

Is it just me, or does it appear like there's a noticeable band of "lighter" ground across the top of the images? I noticed it near the beginning and hoped it would go away with time, but a day later and it's still there. The Bidir render of the same scene didn't have the same anomaly.

I'm using the default settings for SPPM/Bidir.

[Pilchard123]

Nope, not you. There's definitely some lighter ground.

[guibou]

sppm.png

bidir.png

Is it just me, or does it appear like there's a noticeable band of "lighter" ground across the top of the images? I noticed it near the beginning and hoped it would go away with time, but a day later and it's still there. The Bidir render of the same scene didn't have the same anomaly.

I'm using the default settings for SPPM/Bidir.

I noticed this issue few times ago, but it is something really weird and difficult to trace. It appears that there is less light in place far away from the camera, but I don't see where we can have forgot a weighting factor which depends on the camera distance (and moreover, which increase with distance).

By the way, this scene is interesting, because we can see where SPPM is powerfull (the floor under the glass character) and where there is issue (the dark band, the blurred caustics and all those nasty fireflies).

For the blurred caustics, I suggest you start with a smaller initial radius, it may help.

For the fireflies, I'm really looking for a simple test scene which can show them quickly, If someone has one ?

[Chawl]

Here are fresh binaries of random_parameter_filtering_test branch for Windows x64...

Edit: Deleted the junk attachment.

[Dade]

Here are fresh binaries of random_parameter_filtering_test branch for Windows x64...

Oh, wait, I have just started, far from having something usable ... and it is totally unrelated with SPPM

[helloworld922]

For the fireflies, I'm really looking for a simple test scene which can show them quickly, If someone has one ?

Something like this?

Here's an archive of the blender scene and luxblend exported files (hopefully I got everything):

untitled.7z

(186.06 KiB) Downloaded 12 times

[guibou]

I pushed a new accel structure which may use less memory than every other structure and build faster during the eyepass. The performances should be better than hashgrid and kdtree and a bit behind hybrid.

More information in the commit log: http://src.luxrender.net/lux/rev/3e2c1631ea47

[Chawl]

I can't compile 3e2c1631ea47 on VS2008 SP1 with this error:

Code: Select all

1>Compiling...
1>parallelhashgrid.cpp
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(52) : error C2664: 'boost::interprocess::detail::atomic_cas32' : cannot convert parameter 1 from 'unsigned int *' to 'volatile boost::uint32_t *'
1>        Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(54) : warning C4146: unary minus operator applied to unsigned type, result still unsigned
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(59) : error C2664: 'boost::interprocess::detail::atomic_cas32' : cannot convert parameter 1 from 'unsigned int *' to 'volatile boost::uint32_t *'
1>        Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(60) : warning C4146: unary minus operator applied to unsigned type, result still unsigned
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(79) : warning C4146: unary minus operator applied to unsigned type, result still unsigned
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(84) : warning C4146: unary minus operator applied to unsigned type, result still unsigned
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(119) : warning C4146: unary minus operator applied to unsigned type, result still unsigned
1>..\..\..\lux\renderers\sppm\parallelhashgrid.cpp(127) : warning C4146: unary minus operator applied to unsigned type, result still unsigned
1>Build log was saved at "file://***masked***\windows\Projects\luxrender\BuildTemp\liblux\x64\Release\BuildLog.htm"
1>liblux - 2 error(s), 6 warning(s)
========== Build: 0 succeeded, 1 failed, 1 up-to-date, 0 skipped ==========

PS: Also windows repo has to be updated for solution to contain the new parallelhasgrid.cpp

[binarycortex]

For the fireflies, I'm really looking for a simple test scene which can show them quickly, If someone has one ?

They usually show up on the first or second pass for me.