Sampling differences between Path and Branched integrators #42815

New Issue

Matt Heimlich · 2014-12-06T08:05:28+01:00

Matt Heimlich commented

2014-12-06 08:05:28 +01:00

System Information
Win7 64-bit

Blender Version
Broken: c8995ec

Short description of error
Branched tracing and normal tracing give different results for the same number of samples taken

Exact steps for others to reproduce the error
Open file, switch between integrators. The difference is visible after only a single sample.

branchedDifferences.blend

**System Information** Win7 64-bit **Blender Version** Broken: c8995ec **Short description of error** Branched tracing and normal tracing give different results for the same number of samples taken **Exact steps for others to reproduce the error** Open file, switch between integrators. The difference is visible after only a single sample. [branchedDifferences.blend](https://archive.blender.org/developer/F129740/branchedDifferences.blend)

Matt Heimlich commented

2014-12-06 08:05:28 +01:00

Changed status to: 'Open'

Matt Heimlich commented

2014-12-06 08:05:28 +01:00

Added subscriber: @MatthewHeimlich

Joel Godin commented

2014-12-06 14:08:11 +01:00

Added subscriber: @FloridaJo

Joel Godin commented

2014-12-06 14:08:11 +01:00

The devs can probably speak more to this, but
if they were the same I would think we wouldn't need to
include two separate algorithms for rendering.
I would expect they're going to give two different results.

The devs can probably speak more to this, but if they were the same I would think we wouldn't need to include two separate algorithms for rendering. I would expect they're going to give two different results.

Julian Eisel commented

2014-12-06 15:43:33 +01:00

Changed status from 'Open' to: 'Archived'

Julian Eisel closed this issue

2014-12-06 15:43:33 +01:00

Julian Eisel self-assigned this 2014-12-06 15:43:33 +01:00

Julian Eisel commented

2014-12-06 15:43:33 +01:00

Of course two different algorithms means two different results. So everything works as expexcted.

Thanks for the report @MatthewHeimlich, but (luckily) no bug here :)

Of course two different algorithms means two different results. So everything works as expexcted. Thanks for the report @MatthewHeimlich, but (luckily) no bug here :)

Matt Heimlich commented

2014-12-06 21:03:52 +01:00

I'm going to argue against that. In theory (as I understand it) the branched tracer taking one sample of all ray types should provide the same exact result as the regular path tracer taking a single sample. Changes should only occur (again, at least in theory) when more samples are given to individual ray types. As you can see, a single sample in the BPT does not equal a single sample in the standard tracer, and is in fact apparently much more efficient, which would imply that there are some improvements to be done to the sampling in the progressive sampler. I'm happy to be corrected by someone familiar with the actual cause though.

Matt Heimlich commented

2014-12-06 22:33:58 +01:00

I found where the disconnect is. The standard path tracer cannot sample all lights per sample, which causes the discontinuity in the results. So I guess this falls more into a feature request than a bug. Sorry for the confusion!

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Sampling differences between Path and Branched integrators #42815