chore(upscaling): add back upscale multiplier

[doodlum]

Summary by CodeRabbit

• New Features

  • Upscale Preset label now shows the preset name plus the average resolution scale for clearer feedback.

• Bug Fixes

  • Resolution scaling now uses per-axis render-to-screen ratios with a 1:1 shortcut when sizes match.
  • Paused vs. unpaused states and quality modes yield correct per-axis scales.
  • Improved handling for DLSS/FidelityFX scaling to avoid unintended neutral scaling.

[coderabbitai]

Walkthrough

Reworks upscaling to use per-axis resolution scales (float2) and displays the UI preset label with the average scale. Streamline refactors DLSS option setup into SetDLSSOptions(), changes GetInputResolutionScale to return float2 and to return {1.0f,1.0f} on failure, and updates Upscale flow to call the new DLSS setup step.

Changes

Cohort / File(s)	Summary
Upscaling core src/Features/Upscaling.cpp	Switches resolutionScaleBase from scalar to float2; computes renderWidth/renderHeight from .x/.y, uses 1:1 shortcut when render equals screen, otherwise computes per-axis resolutionScale.x/.y; UI “Upscale Preset” label now shows preset name plus average scale using std::format. No exported API changes.
Streamline: DLSS & flow refactor src/Features/Upscaling/Streamline.cpp, src/Features/Upscaling/Streamline.h	Adds SetDLSSOptions() to encapsulate DLSS option setup; changes GetInputResolutionScale(...) return type from float → float2 and semantics (returns {1.0f,1.0f} on failure; computes separate x/y scales depending on paused state); Upscale() now calls CheckFrameConstants(), SetDLSSOptions(), then proceeds with rendering setup; removed direct state access from SetDLSSOptions.
FidelityFX: API type update src/Features/Upscaling/FidelityFX.cpp, src/Features/Upscaling/FidelityFX.h	Changes GetInputResolutionScale(...) return type from float → float2; function now returns a float2 with both components equal to the previously computed scalar scale. Header updated accordingly.

Sequence Diagram(s)

sequenceDiagram
  autonumber
  participant UI as UI
  participant U as Upscaling
  participant R as Renderer

  UI->>U: ConfigureUpscaling(screenW, screenH, renderW, renderH)
  alt render == screen
    U->>U: resolutionScale = {1.0, 1.0}
  else
    U->>U: resolutionScale.x = renderW / screenW
    U->>U: resolutionScale.y = renderH / screenH
  end
  U->>UI: Label = std::format("Preset (avg: {:.2f})", 0.5*(x+y))
  U->>R: Apply per-axis resolutionScale

Loading

sequenceDiagram
  autonumber
  participant S as Streamline
  participant C as Controller (Upscale)
  participant D as DLSS subsystem

  C->>S: Upscale(frame)
  S->>S: CheckFrameConstants()
  S->>S: SetDLSSOptions()   %% new step: configure DLSS options
  S->>D: configure (options)
  S->>S: GetInputResolutionScale(outW,outH,quality)
  alt failed to get optimal settings
    S-->>C: return {1.0,1.0}
  else
    S-->>C: return {scaleX, scaleY}
  end
  S->>S: continue render setup (bind depth, resources)

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• feat(upscaling): add downscale percentages #1506: Overlaps changes to Upscaling UI preset labeling and preset-scale handling in src/Features/Upscaling.cpp.
• fix(upscaling): more upscaling fixes #1498: Related edits to ConfigureUpscaling and render/screen scale logic in Upscaling code paths.
• feat(frame generation): improve frame generation quality #1125: Touches Streamline/Upscaling integration and frame-hook/framebuffer handling that intersect with DLSS/Streamline changes.

Suggested reviewers

• alandtse
• jiayev

Poem

I hop through scales both X and Y,
I format labels, give them a try.
When DLSS wakes I set its tune,
If things go odd I stay at one.
A rabbit’s tweak — the render’s soon. 🐇✨

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 9.09% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ 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 “chore(upscaling): add back upscale multiplier” refers directly to restoring a multiplier in the upscaling logic, which is one of the changes in this pull request, although it does not capture the additional updates to UI labeling and DLSS option handling.

✨ Finishing touches

• 📝 Generate Docstrings

🧪 Generate unit tests

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch upscale-mult

---

📜 Recent review details

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 6ef7cb9 and e2099a0.

📒 Files selected for processing (5)

• src/Features/Upscaling.cpp (2 hunks)
• src/Features/Upscaling/FidelityFX.cpp (1 hunks)
• src/Features/Upscaling/FidelityFX.h (1 hunks)
• src/Features/Upscaling/Streamline.cpp (6 hunks)
• src/Features/Upscaling/Streamline.h (1 hunks)

🧰 Additional context used

📓 Path-based instructions (2)

**/*.{cpp,cxx,cc,c,h,hpp,hxx,hlsl,hlsli,fx,fxh,py}

📄 CodeRabbit inference engine (.github/copilot-instructions.md)

Do not include TODO/FIXME placeholders; provide complete, working solutions

Files:

• src/Features/Upscaling/FidelityFX.h
• src/Features/Upscaling/Streamline.h
• src/Features/Upscaling.cpp
• src/Features/Upscaling/FidelityFX.cpp
• src/Features/Upscaling/Streamline.cpp

src/**/*.{cpp,cxx,cc,h,hpp,hxx}

📄 CodeRabbit inference engine (.github/copilot-instructions.md)

src/**/*.{cpp,cxx,cc,h,hpp,hxx}: Ensure SE/AE/VR runtime compatibility; use runtime detection patterns (e.g., REL::RelocateMember())
Include robust error handling and resource management with graceful degradation in the plugin code

Files:

• src/Features/Upscaling/FidelityFX.h
• src/Features/Upscaling/Streamline.h
• src/Features/Upscaling.cpp
• src/Features/Upscaling/FidelityFX.cpp
• src/Features/Upscaling/Streamline.cpp

🔇 Additional comments (1)

src/Features/Upscaling.cpp (1)

720-743: Clamp render dimensions before jitter calculation.

We’re still flooring the scaled dimensions to int without any rounding or lower bound. For extreme downscale multipliers this drops renderWidth/renderHeight to zero, which then feeds into GetJitterPhaseCount and the projectionPosScale math, leading to divide-by-zero and NaNs. This is the same correctness issue flagged in the earlier review—please add rounding and a std::max(1, …) guard so the DLSS/FSR path can’t produce zero-sized render surfaces.

-        auto renderWidth = static_cast<int>(screenWidth * resolutionScaleBase.x);
-        auto renderHeight = static_cast<int>(screenHeight * resolutionScaleBase.y);
+        auto renderWidth = std::max(1, static_cast<int>(screenWidth * resolutionScaleBase.x + 0.5f));
+        auto renderHeight = std::max(1, static_cast<int>(screenHeight * resolutionScaleBase.y + 0.5f));

Ensure the subsequent resolutionScale calculations use these clamped integers. As per coding guidelines.

---

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: 8047cd3
Using base ref: 8047cd3
Base commit date: 2025-09-27T16:58:33+01:00 (Saturday, September 27, 2025 04:58 PM)
No actionable suggestions for changed features.

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (2)

src/Features/Upscaling/Streamline.cpp (1)

401-404: Return early for DLAA; avoid unnecessary SDK query and potential error log.

When dlssMode == eDLAA, you can return 1.0f before calling slDLSSGetOptimalSettings. This removes an avoidable external call and prevents spurious critical logs if the SDK query fails under DLAA.

Also add a guard for zero outputWidth/Height to avoid divide-by-zero in rare edge cases.

Apply:

 float Streamline::GetInputResolutionScale(uint32_t outputWidth, uint32_t outputHeight, uint32_t qualityMode)
 {
   sl::DLSSMode dlssMode;
   switch (qualityMode) {
@@
     break;
   }
 
+  // DLAA is 1:1; bypass SDK queries entirely.
+  if (dlssMode == sl::DLSSMode::eDLAA) {
+    return 1.0f;
+  }
+  if (outputWidth == 0 || outputHeight == 0) {
+    logger::warn("[Streamline] GetInputResolutionScale called with zero output dimension ({}x{}), defaulting to 1.0", outputWidth, outputHeight);
+    return 1.0f;
+  }
+
   sl::DLSSOptions dlssOptions{};
   dlssOptions.mode = dlssMode;
   dlssOptions.outputWidth = outputWidth;
   dlssOptions.outputHeight = outputHeight;
@@
-  // For DLAA, ensure exactly 1.0 scale
-  if (dlssMode == sl::DLSSMode::eDLAA) {
-    return 1.0f;
-  }

As per coding guidelines (robust error handling).

src/Features/Upscaling.cpp (1)

206-211: UI: consider showing X/Y scales when anisotropic.

Current label shows average scale only. If X≠Y, optionally append "(Xx, Yx)" for clarity. Low priority.

Example:

-std::string labelWithScale = std::format("{} ( {:.2f}x )", baseLabel, (resolutionScale.x + resolutionScale.y) * 0.5f);
+const float avg = (resolutionScale.x + resolutionScale.y) * 0.5f;
+std::string labelWithScale = std::format("{} ( {:.2f}x )", baseLabel, avg);
+if (fabs(resolutionScale.x - resolutionScale.y) > 0.001f) {
+  labelWithScale = std::format("{} ( {:.2f}x, {:.2f}x )", baseLabel, resolutionScale.x, resolutionScale.y);
+}

📜 Review details

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 8047cd3 and 6ef7cb9.

📒 Files selected for processing (2)

• src/Features/Upscaling.cpp (2 hunks)
• src/Features/Upscaling/Streamline.cpp (1 hunks)

🧰 Additional context used

📓 Path-based instructions (2)

**/*.{cpp,cxx,cc,c,h,hpp,hxx,hlsl,hlsli,fx,fxh,py}

📄 CodeRabbit inference engine (.github/copilot-instructions.md)

Do not include TODO/FIXME placeholders; provide complete, working solutions

Files:

• src/Features/Upscaling/Streamline.cpp
• src/Features/Upscaling.cpp

src/**/*.{cpp,cxx,cc,h,hpp,hxx}

📄 CodeRabbit inference engine (.github/copilot-instructions.md)

src/**/*.{cpp,cxx,cc,h,hpp,hxx}: Ensure SE/AE/VR runtime compatibility; use runtime detection patterns (e.g., REL::RelocateMember())
Include robust error handling and resource management with graceful degradation in the plugin code

Files:

• src/Features/Upscaling/Streamline.cpp
• src/Features/Upscaling.cpp

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)

• GitHub Check: Build plugin and addons

[github-actions]

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