Page MenuHome

flickery artifacts with mirrored reflection of bumpmapped surface *needs proper info*
Closed, ArchivedPublicTO DO

Description

Both balls are mirrored reflections. Observe the surface as it moves away from the reflectors.

Event Timeline

the flickering persists even when removing the bumpmapping

I'm sorry dude, but you have to follow our bug reporting guidelines.

Read the tips here:
http://projects.blender.org/tracker/?func=add&group_id=9&atid=498

Most importantly we need you test our official binaries (from blender.org) and provide us with the simplest example .blend possible to recreate the issue.

I originally tried to post both the video and the blend, but then the bugreport never showed up. Then I posted it with just the video.zip and intended to add the blend later. I added the blend to Attachments and clicked on "Save Changes". It took a while to submit and then I got returned to the bugtracker index. I never checked if it was actually uploaded, and it turns out it didn't. I've now tried a few times again and it just doesn't work, both when when using the "save changes" under comment and the "save changes" at the top...
I have uploaded it to http://www.pasteall.org/blend/5056 so you may add it yourself if it works for you
As far as the version goes, It is not the official release because this was intended to be tested along sparky's new bumpmapping.
However it apparently doesn't relate to bumpmapping and the artifacts are also in the official release.
For the record, I'm on Win7x64

OK; this error is two-fold:
- The mirror doesn't project the texture space correctly on the sphere. There's a proposal from a developer being reviewed to solve this
- For procedural textures we don't have an adaptive filtering at all, which means that small renders of procedurals always will give noise, mirrored or not.

Using image textures should give much better results already.

We'll try to work on this for our 2.6 version (shader system improvements). For now I can only quote a ILM director I once spoke in LA about this issue "we just render everhthing two times larger". :)

I am aware that procedurals do not antialias properly. The used texture is a 4k by 4k image that was rendered to texture from a procedural and does not exhibit any aliasing.

Ah, i was mislead by the looks of the texture, it looked procedural. For that case there's a proposal to tackle it. That's on todo and hopefully gets solved in days. I will use your test file to check and post an update here.

This was never fixed was it?

Blend file is not available anymore, please upload here.

The file has been attached. The bug is the result of derivatives not being recalculated during reflective/refractive bounces.
So it affects anything that relies on the derivatives such as texture filtering and bump mapping.

This file just won't upload. Dunno why. If anyone else can attach it that'd be cool.
Until then I've made it available from here --> http://jbit.net/~sparky/temp/bumptest2.blend

Now what's the status?

No activity. Discussed with Brecht and it's indeed a known TODO issue, so closing then.

Sergey Sharybin (sergey) changed the task status from Unknown Status to Unknown Status.Nov 30 2011, 9:57 PM