fix(water): hdr water taa blend

[jiayev]

set water blend buffers to hdr format
removed water taa override because don't need anymore

Summary by CodeRabbit

• Refactor

  • Simplified water shading and color output; removed legacy log-space conversions.

• New Features

  • Added dedicated render targets to support water rendering buffers.

• Bug Fixes

  • TAA/upscaler interaction adjusted so water TAA is enabled when an upscaler is active and only disabled when upscaling is off.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 1d654aa8-ba50-45a5-aa4f-d67b8f1cd118

📥 Commits

Reviewing files that changed from the base of the PR and between 83a1747 and 8c4ce13.

📒 Files selected for processing (1)

• src/Features/Upscaling.cpp

🚧 Files skipped from review as they are similar to previous changes (1)

• src/Features/Upscaling.cpp

---

📝 Walkthrough

Walkthrough

Shader and renderer changes switch water history/output from 4-component to 3-component colors, remove log-linear conversions, add two water TAA render targets, and simplify Upscaling's TAA enablement logic.

Changes

Cohort / File(s)	Summary
Water Blend Shader package/Shaders/ISWaterBlend.hlsl	Changed PS_OUTPUT.Color1 from float4 to float3; removed LogToLinear / LinearToLog; converted waterHistory usage from float4 to float3, replaced waterHistory.w == 1 gating with any(waterHistory), and removed log-space conversions and alpha-based history gating.
Render Target Setup src/Deferred.cpp	Added SetupRenderTarget calls to register two new TAA water targets (RE::RENDER_TARGETS::kWATER_1, kWATER_2) using DXGI_FORMAT_R11G11B10_FLOAT and bind flags for shader resource/render target usage.
Upscaling / TAA Flag src/Features/Upscaling.cpp	Simplified TAA enablement: taaEnabled is now set based on upscaleMethod != NONE (removes prior early-return/conditional paths and explicit water-TAA overrides).

Sequence Diagram(s)

sequenceDiagram
  participant Upscale as Upscaling Config
  participant Deferred as Deferred Renderer
  participant GPU as GPU / ISWaterBlend
  participant RT as Water Render Targets

  Upscale->>Deferred: set disableWaterTAA / taaEnabled
  Deferred->>RT: SetupRenderTarget(kWATER_1 / kWATER_2)
  Deferred->>GPU: draw water pass (sample history, compute)
  GPU->>RT: write finalColor (float3) to Color / Color1
  RT-->>Deferred: stored water buffers for next-frame TAA

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• fix: remove water clamp #1497: Directly modifies ISWaterBlend shader history and log/linear handling — strong overlap with shader edits.
• fix(upscaling): more upscaling fixes #1498: Changes Upscaling::ConfigureTAA and water-TAA enablement — closely related to the Upscaling changes.
• fix(upscaling): disable water TAA instead of enable #1726: Alters Upscaling TAA enable/disable behavior for water TAA — related to TAA flag handling here.

Suggested reviewers

• doodlum
• alandtse
• davo0411

Poem

🐰 I hopped through pixels, soft and free,

Trimmed one channel from the sea.
No more logs to twist the light,
Two new pools to hold the sight.
A rabbit cheers for clearer gleam ✨

🚥 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 'fix(water): hdr water taa blend' accurately summarizes the main change—implementing HDR water with TAA blending by refactoring the shader and rendering pipeline.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

📝 Coding Plan

• Generate coding plan for human review comments

---

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.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@package/Shaders/ISWaterBlend.hlsl`:
- Around line 46-47: The reprojection logic currently treats an all-zero RGB
sample as "no history" (any(waterHistory)), which incorrectly discards valid
black/dark history; change the history texture/sampling to use an explicit
validity channel instead of relying on RGB: sample a float4 from waterHistoryTex
(keep waterHistorySampler and motionAdjustedScreenPosition) and use a dedicated
component (e.g., waterHistory.a) as the history-valid sentinel; update the
producer that writes history to set that alpha to 1 when history is valid (and 0
when invalid) and replace any(waterHistory) checks with checks against that
alpha (e.g., waterHistory.a > 0.5).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 9689a982-0459-40c3-b1cd-48711c901934

📥 Commits

Reviewing files that changed from the base of the PR and between 491e83e and 4935935.

📒 Files selected for processing (2)

• package/Shaders/ISWaterBlend.hlsl
• src/Deferred.cpp

[github-actions]

No actionable suggestions for changed features.

[github-actions]

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

[coderabbitai]

🧹 Nitpick comments (1)

src/Features/Upscaling.cpp (1)

1042-1043: Redundant condition after early return.

The condition (upscaleMethod == UpscaleMethod::kNONE) at line 1043 will always evaluate to false because kNONE causes an early return at line 1037. This could be simplified for clarity.

✨ Optional simplification

 	bool* disableWaterTAA = reinterpret_cast<bool*>(reinterpret_cast<uintptr_t>(BSImagespaceShaderISTemporalAA) + 0x38LL);
-	*disableWaterTAA = (upscaleMethod == UpscaleMethod::kNONE);
+	*disableWaterTAA = false;  // Enable water TAA for TAA/FSR/DLSS methods

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/Features/Upscaling.cpp` around lines 1042 - 1043, The assignment to
*disableWaterTAA uses (upscaleMethod == UpscaleMethod::kNONE) which is
impossible here because UpscaleMethod::kNONE returns early; update the code
around disableWaterTAA / BSImagespaceShaderISTemporalAA so the assignment is
simplified—either remove the redundant assignment entirely or replace it with
the constant *disableWaterTAA = false; to reflect that this branch only runs
when upscaleMethod != kNONE.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@src/Features/Upscaling.cpp`:
- Around line 1042-1043: The assignment to *disableWaterTAA uses (upscaleMethod
== UpscaleMethod::kNONE) which is impossible here because UpscaleMethod::kNONE
returns early; update the code around disableWaterTAA /
BSImagespaceShaderISTemporalAA so the assignment is simplified—either remove the
redundant assignment entirely or replace it with the constant *disableWaterTAA =
false; to reflect that this branch only runs when upscaleMethod != kNONE.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: e706d624-8e4d-449f-bb35-59a319d76ce8

📥 Commits

Reviewing files that changed from the base of the PR and between 4935935 and 83a1747.

📒 Files selected for processing (1)

• src/Features/Upscaling.cpp

[alandtse]

Are you reverting #1986? Make sure you didn't break that fix.

[Dlizzio]

Tested with all upscaler options and see no flicker. #1986 never really fixed it.

[jiayev]

Are you reverting #1986? Make sure you didn't break that fix.

that one didn't really fix, as far as i and dlizzio tested

[alandtse]

Wasn't the issue with no upscaler?

[Dlizzio]

Wasn't the issue with no upscaler?

Possibly. I tried with upscaler set to off and upscaling disabled at boot with this PR and see no flickering.

[alandtse]

Let's see if @jturnley can test and confirm whether this causes a regression or not.

[jiayev]

Wasn't the issue with no upscaler?

it wasn't. lack of watertaa cause flickering, because taa/upscaler don't like its pattern. (as water changes with no motion)
i was the one who disabled watertaa in the first place. because vanilla uses two unorm textures for water history buffer, i disabled it for hdr. this pr change the buffers to hdr, so disabling watertaa is not needed anymore.