refactor(upscaling): standardize behavior and tune settings

[doodlum]

Summary by CodeRabbit

• Refactor
  • Unified input-to-render resolution scaling into a single pathway across all upscaling backends for consistent behavior.
  • Consolidated jitter and render-size calculations to follow the unified scaling.
• Chore
  • Removed several backend-specific scaling helpers and streamlined control flow.
• Style
  • Lowered default sharpness for upscaled output.
• Quality
  • Normalized mip bias for temporal upscaling to produce more uniform detail.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

📝 Walkthrough

Walkthrough

This change unifies upscaling resolution calculation into a single pathway (removing backend-specific GetInputResolutionScale), adjusts default sharpness values to 0.0 for FSR and DLSS, and normalizes MipBias computation for temporal non-TAA upscaling.

Changes

Cohort / File(s)	Summary
Core Upscaling src/Features/Upscaling.cpp, src/Features/Upscaling.h	Unified resolution scaling via a single resolutionScaleBase derived from quality mode; renderWidth/renderHeight now computed from screen dimensions; removed per-backend scaling branches. Updated default sharpness: sharpnessFSR 1.0 → 0.0, sharpnessDLSS 0.5 → 0.0.
FidelityFX Backend src/Features/Upscaling/FidelityFX.cpp, src/Features/Upscaling/FidelityFX.h	Removed public GetInputResolutionScale(...) implementation and declaration (backend no longer exposes per-backend input-scale helper).
Streamline / DLSS Backend src/Features/Upscaling/Streamline.cpp, src/Features/Upscaling/Streamline.h	Removed GetInputResolutionScale(...); added explicit DLSS preset assignments and removed DLSS-specific scale logic.
State / Shared Data src/State.cpp	Normalized MipBias computation: uses log2f(renderSize.x / screenSize.x) - 1.0f for temporal non-TAA paths instead of a DLSS-only branch.

Sequence Diagram(s)

(omitted)

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• chore(upscaling): revert fsr to typical settings #1523 — Modifies the same MipBias computation in State::UpdateSharedData for temporal non-TAA upscaling.
• fix(upscaling): more upscaling fixes #1498 — Refactors upscaling subsystem, removing/replacing backend-specific GetInputResolutionScale implementations.
• chore(upscaling): add back upscale multiplier #1532 — Changes upscaling scaling API/logic and interacts with GetInputResolutionScale removal across backends.

Suggested reviewers

• davo0411
• alandtse

Poem

🐰
I nibbled at branches, made them one,
Hopped through pixels till the work was done,
Sharpness hushed to silence, scales aligned,
MipBias tuned so renderings bind,
A happy rabbit’s patch — smooth and fine 🌿✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main changes: refactoring upscaling logic to standardize behavior across backends and tuning default quality settings (sharpness parameters).

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch upscaling-tweaks

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

Using provided base ref: f6ab97c
Using base ref: f6ab97c
Base commit date: 2026-02-01T20:03:05Z (Sunday, February 01, 2026 08:03 PM)
No actionable suggestions for changed features.

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@src/Features/Upscaling/Streamline.cpp`:
- Around line 285-293: Update the DLSS preset assignments so high-quality modes
use NVIDIA's current recommended preset: change dlssOptions.dlaaPreset,
dlssOptions.ultraQualityPreset, dlssOptions.qualityPreset, and
dlssOptions.balancedPreset to use sl::DLSSPreset::ePresetK (leave
dlssOptions.performancePreset as ePresetM and dlssOptions.ultraPerformancePreset
as ePresetL), ensuring the four high-quality fields are aligned with the new
ePresetK mapping.

[github-actions]

✅ A pre-release build is available for this PR:
Download

[alandtse]

This looks like much more than settings tweaks. What are you trying to do here? Chore should not be used unless there's really no way to describe the point of the changes.

[alandtse]

This looks like much more than settings tweaks. What are you trying to do here? Chore should not be used unless there's really no way to describe the point of the changes.

[doodlum]

This looks like much more than settings tweaks. What are you trying to do here? Chore should not be used unless there's really no way to describe the point of the changes.

These are all surface level changes to the presets. They have no deep modifications or breaking changes.

[alandtse]

This looks like much more than settings tweaks. What are you trying to do here? Chore should not be used unless there's really no way to describe the point of the changes.

These are all surface level changes to the presets. They have no deep modifications or breaking changes.

If it was settings tweaks you wouldn't be restructuring stuff. What is the purpose so we can write a real commit message instead of chore slop?

[vanni-giachin]

@doodlum I tested the branch, 4k quality preset. With M I see the same issues I saw with the old F, so introduction of artifacts for example on highly reflective materials like glass. And you are right, M solves a lot of ghosting issues we see with K. Personally I feel K is still better for quality/balanced on advanced setups with PBR mods, while M is probably better on something more close to vanilla. Perhaps we could allow users to choose the preset , similar to how the NVIDIA App handles DLSS overrides (excluding outdated presets like F, as you previously suggested). Regarding sharpness I confirm M introduces some extra sharpness (I like more something close to 0.5, but with K I usually increase too, like to 1, so it is personal preference).

[doodlum]

Preset selection probably makes sense for us to add back. I guess as an override value that changes every quality setting.