Differential D12610
Fixes T91632 by stopping the sample correlation between dimensions which was causing rendering artifacts on simple scenes. Authored by William Leeson (leesonw) on Sep 23 2021, 11:23 AM.
Details
Fix T91632: Stops the sample correlation between dimensions which was causing rendering artefacts on simple scenes. This is done by increasing the amount of jitter the Cranley Patterson Rotation is allowed to add. Also, it uses the y dimension of the of the sample table for 1D sampling which causes further decorrelation between dimensions. As an additional measure the x and y dimensions are swapped randomly to provide further decorrelation.
Diff Detail
Event TimelineComment Actions If you are posting Diffs here regularly [which I think you are/will], it might be good to look into ARC Also please check https://wiki.blender.org/wiki/Style_Guide/Commit_Messages Comment Actions I see I got the fix format wrong so I'll rectify that. I used arc to submit this diff. Am I doing any thing else wrong? Please let me know as I am pretty new to arc so it's highly possible :-/ Comment Actions Doing the Cranley-Patterson on the full pattern instead of the grid seems ok. When this is committed all reference renders will need to be updated along with it. That means committing those the svn repository first, and then right that finishes after pushing the git code, so that the buildbot build tests with the right reference renders.
Comment Actions I did a BLENDER_TEST_UPDATE=1 ctest -R cycles cpu and all the following files will change in the SVN repo (quite a few) 38 - cycles_bake_cpu (Failed) 39 - cycles_bsdf_cpu (Failed) 40 - cycles_denoise_cpu (Failed) 41 - cycles_displacement_cpu (Failed) 42 - cycles_hair_cpu (Failed) 43 - cycles_image_colorspace_cpu (Failed) 45 - cycles_image_mapping_cpu (Failed) 47 - cycles_integrator_cpu (Failed) 48 - cycles_light_cpu (Failed) 49 - cycles_mesh_cpu (Failed) 50 - cycles_motion_blur_cpu (Failed) 51 - cycles_openvdb_cpu (Failed) 52 - cycles_render_layer_cpu (Failed) 53 - cycles_reports_cpu (Failed) 54 - cycles_shader_cpu (Failed) 55 - cycles_shadow_catcher_cpu (Failed) 56 - cycles_sss_cpu (Failed) 57 - cycles_volume_cpu (Failed) 58 - cycles_osl_cpu (Failed) However the following tests fail on the master branch 38 - cycles_bake_cpu (Failed) 39 - cycles_bsdf_cpu (Failed) 40 - cycles_denoise_cpu (Failed) 41 - cycles_displacement_cpu (Failed) 42 - cycles_hair_cpu (Failed) 43 - cycles_image_colorspace_cpu (Failed) 45 - cycles_image_mapping_cpu (Failed) 47 - cycles_integrator_cpu (Failed) 48 - cycles_light_cpu (Failed) 49 - cycles_mesh_cpu (Failed) 50 - cycles_motion_blur_cpu (Failed) 51 - cycles_openvdb_cpu (Failed) 52 - cycles_render_layer_cpu (Failed) 53 - cycles_reports_cpu (Failed) 54 - cycles_shader_cpu (Failed) 55 - cycles_shadow_catcher_cpu (Failed) 56 - cycles_sss_cpu (Failed) 57 - cycles_volume_cpu (Failed) 58 - cycles_osl_cpu (Failed) resulting in almost the same set of files being added except the 2 bake ones at the top (even though they failed in both cases :-/). So the question is should I remove the files for the already broken tests from the set to be submitted or should they be submitted also? Comment Actions The test should pass in the master branch, they are passing on the buildbot. Maybe you have not updated the SVN repository recently? I suggest to svn revert all changes in the SVN repo and update it, then check if it builds with master. But yes, this will require committing updated versions of practically all reference images. | ||||||||||||||||