nv2a/glsl: Determinant scaling to avoid floating-point overflow #2510
+19
−8
Conversation
This file contains hidden or 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
…lows If output screen x,y coordinates from vertex shader are large, e.g. 1e22, multiplying them will overflow the maximum exponent of single-precision floating point numbers. This occurs when calculating determinants in barycentric coordinates calculations. This fixes such overflows by using a power of 2 scaling factor to clamp the exponent at 60. If x is the largest floating-point with exponent 62, then x*x + x*x is just barely below infinity. Here 60 is used instead of 62 in order to have a small safety buffer. An alternative solution would be to just discard fragments if barycentric coordinates are NaN. At least Radeon RX 6600 and also Intel UHD 770 hardware fail to properly rasterize anyway when coordinates are very large.
Collaborator
|
Haven't look into this yet but might fix this as well. |
Collaborator
|
Here a video on masters. SEIZURE WARNING2025-11-03.20-02-30.mp4Edit: This pr does reduce the issue but still occurs but is less way less frequent. Pr build 2025-11-03.20-17-24.mp4 |
Collaborator
|
Confirmation that does fix Capcom vs. SNK 2 EO 2025-11-03.20-08-23.mp4PS: hope audio issue get fixed someday as it affect some games. |
Collaborator
|
Also fixes #2539 not sure about first issue |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
If output screen x,y coordinates from vertex shader are large, e.g. 1e22, multiplying them will overflow the maximum exponent of single-precision floating point numbers. This occurs when calculating determinants in barycentric coordinates calculations.
This fixes such overflows by using a power of 2 scaling factor to clamp the exponent at 60. If x is the largest floating-point with exponent 62, then xx + xx is just barely below infinity. Here 60 is used instead of 62 in order to have a small safety buffer.
An alternative solution would be to just discard fragments if barycentric coordinates are NaN. At least Radeon RX 6600 and also Intel UHD 770 hardware fail to properly rasterize anyway when coordinates are very large.
Fixes #2501