[hdEmbree] fix for random number generation (previously: hdEmbree-UsdLux-PR02) #3184
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
use of std::bind was copying the underlying random number generator before use, resulting in the exact same sequence getting reused for, ie, every AO calculation within a given tile-group
This was referenced Jul 31, 2024
|
Filed as internal issue #USD-9902 |
|
/AzurePipelines run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
2 tasks
pmolodo
changed the title
pmolodo
changed the title
|
Closing this, as it's standalone duplicate was merged: Closing this PR as it's duplicate was merged (#3211) |
pmolodo
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Warning
Closing this PR as it's duplicate was merged (#3211)
Important
This is a duplicate of this PR:
...except that PR was pulled out the UsdLux-HdEmbree PR "chain", to make for easier standalone integration, as it has no other PR dependencies, and fixes an outstanding bug.
I will close this PR if the other is merged first, and vice-versa.
Description of Change(s)
use of std::bind was copying the underlying random number generator before use, resulting in the exact same sequence getting reused for, ie, every AO calculation within a given tile-group
Fixes Issue(s)
Related PRs:
This PR is part of a chain of PRs that provide a reference implementation of UsdLux as part of hdEmbree.
Due to a github limitation, this PR must include all previous PRs in the chain. To see JUST the changes in this PR vs the previous in the chain, click this link
⛓️ All PRs in this chain:
📃 Documentation PR:
This chain of PRs (implementing UsdLux support in hdEmbree), was made separate from the PR documenting expected UsdLux behavior:
✳️ All UsdLux update related PRs:
To see ALL UsdLux update related PRs (documentation AND reference implementation) in one place, see:
⌛ Original PR:
These PRs are an update of this original PR:
But... why??
Why make these PRs in the first place?
The current specifications of the various UsdLux prims + attributes are imprecise or vague in many places, and as a result, actual implementations of them by various renderers have diverged, sometimes quite significantly. For instance, here is Intel's 4004 Moore Lane scene, with the same UsdLux lights defined, in 3 different renderers:
Karma:
Arnold:
Omniverse RTX:
For a full descpription of the problem, see here:
Why so many PRs?
This was my attempt to break up a rather large change into smaller, more easily reviewable changes, that can be merged in incrementally, to help ease the burden for code reviewers.
If you find this confusing, and would rather just one big PR (ie, just this one), or have them organized in some other way, please let me know!
Why are the documentation changes in their own PR?
In some ways, the documentation changes are the heart of this effort - we wish to specify more exactly what the various UsdLux prims and attributes represent. However, building consensus on this may take time - so we expect some dialogue on the exact language or formulas.
Because that may take time, the hdEmbree reference implementation changes are separate, and broken up, so we can hopefully start integrating portions of them even before final consensus has been reached on the final form of the specification.
Why are the documentation changes not broken up into smaller pieces, like the hdEmbree reference implementation changes?
Because I wasn't sure if that would be desirable or not! If people think that would be helpful, I can do so - perhaps breaking out by schema or individual attribute?