Repository navigation
docs: Model L is under test; Model M's arithmetic is WrongGPU's own (Refs #89) - #223
Merged
Merged
Conversation
phantoslayer21
marked this pull request as ready for review
November 15, 2026 03:43
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Refs #89. Wording only: release blocker A10's rule for Model L, and oversight's rule after spec Q10 for Model M (the shipped M follows the functional form: its constants match the original, its arithmetic does not, so no public text may call it exact or identical to the original or to DLSS). Until the Q7-form oracle is re-run, Model L is described as included, off by default and under test. Nothing says it is exact, matches the original or passes the oracle.
README.md(section "Model L"): "in development for 0.7.0" becomes "planned for 0.6.5, off by default and still under test". The timing figures with records (2.25 / 1.17 ms,clean/l/SPEED.md) stay.README.md(opening, lines 5-8): "runs the neural networks behind DLSS ..." becomes "runs its own implementation of the neural networks behind DLSS ...", plus one sentence: the weights come from the player's own DLSS files, the arithmetic is WrongGPU's, so the output is close to DLSS's but not identical. No other public text called Model M exact or identical to the original or to DLSS. The byte-identity statements (HIP vs D3D12, files, restores) are true and stay.clean/l/README.md(intro):model=L.Checked with no change needed: the RR "Exact" text in the notes (it already says it can't match DLSS pixel for pixel),
clean/equiv/README.md(tool docs, no verdicts),clean/l/SPEED.md(timing only, every figure with its record),docs/TESTING-OTHER-GAMES.md,BUILDING.md,COMPATIBILITY.md,ROADMAP.md, andclean/FORMATS.md(one measured arena size for Model L at 4K native).Gates: the must-pass gates pass. Leak hits 42 / 29 and numbers 133 match main's baseline.