Add CUDA NGX Ray Reconstruction support

Run the signed CUDA-capable denoiser through feature 13 with independent D/E presets, pre-Init discovery, per-frame guide and camera snapshots, normalized integer guide staging, paired alpha output, and recorded-resource retirement.

Add the required ZLUDA/LLVM patch, behavioral configuration tests, GPU harness coverage, optional denoiser installation and packaging, and tested support limitations. Preserve upstream attribution and distinguish LLVM licensing from the ZLUDA portions of the patch.

docs/architecture.md

@@ -1,3 +1,4 @@
+<!-- Modified in this fork for CUDA Ray Reconstruction support and validation (2026). -->
 # Architecture
 
 d4r keeps NVIDIA's DLSS library unmodified and gives it what it expects: an NGX D3D12 or Vulkan entry point in the game process, a CUDA driver underneath, and CUDA kernels that run. Everything between those points is translation.
@@ -11,7 +12,7 @@ host staging and command-buffer event synchronization; it does not use vkd3d-pro
 |---|---|---|
 | OptiScaler (in GE-Proton) | external | Hooks the game's upscaler call and forwards DLSS requests to an NGX DLL (`NvngxPath`) |
 | NGX shim `d4r_nvngx.dll` | `tools/d4r_nvngx_shim.cpp` | Implements the NGX D3D12 API. Moves inputs and output between the game's D3D12 resources and CUDA, and drives the official NGX core through its CUDA API |
-| NGX core + `nvngx_dlss.dll` | NVIDIA, supplied by you | The real DLSS: parameter handling, network selection, the CUDA kernels |
+| NGX core + `nvngx_dlss.dll` and optional `nvngx_dlssd.dll` | NVIDIA, supplied by you | The real DLSS: parameter handling, Super Resolution or Ray Reconstruction network selection, and CUDA kernels |
 | CUDA bridge `nvcuda.dll` | `tools/wine_nvcuda_bridge.c` | A Wine builtin whose Windows exports forward to ZLUDA's Linux `libcuda.so`, plus d4r helpers (Vulkan memory import, asynchronous 2D copies, GPU-side waits, output redirect, linear textures) |
 | ZLUDA | `patches/zluda` | CUDA driver API on HIP; compiles NVIDIA's PTX to AMDGPU code; serves native kernels in place of PTX kernels |
 | Native kernels | `kernels/` | RDNA3 implementations of the DLSS network layers and parts of texture kernels |
@@ -37,6 +38,16 @@ The watchdog reports delayed output but never signals an unfinished producer as
 
 The shim requires the matching lifetime-aware d3d12.dll and d3d12core.dll, even for host staging. It rejects older runtimes rather than recording references whose lifetime it cannot protect. FrameAge > 0 shows the newest finished result; without VRAM interop, inputs and output are staged through host memory.
 
+### Ray Reconstruction path
+
+Feature 13 uses the NGX core's CUDA lifecycle, not the denoiser's native D3D12 backend. A CUDA-capable `nvngx_dlssd.dll` must sit beside the calling shim; the installer stages it there as well as in the feature directory. Availability follows the core's denoiser availability and initialization result. `[RayReconstruction] Enable = auto` does not manufacture support; `Model` selects its own default/D/E preset independently of the Super Resolution model.
+
+Feature requirements may be queried before NGX initialization. The shim loads the providers without initializing NGX with a provisional application identity; when the core declines feature-13 discovery, it asks the CUDA-capable denoiser's D3D12 requirements export for the adapter's eligibility. This discovery-only call does not select the native D3D12 rendering backend. Initialized capability checks and feature creation still require the core's successful denoiser initialization.
+
+The shim snapshots denoiser guide bindings, scalar settings, camera matrices, and subrect origins per submitted frame. Normals and roughness use the SDK's `GBuffer.Normals` and `GBuffer.Roughness` resource keys; diffuse and specular albedo bindings are translated to the CUDA denoiser's `DLSS.Input.*Albedo` keys. Guide readback and CUDA upload follow the main input marker, and the backing resources remain held until that frame retires. Uploads use per-slot host staging because direct ROCm image copies from Vulkan-mapped memory truncated guide rows on gfx1201. CUDA parameters point at stable texture-handle storage, not D3D12 resource pointers. Normals preserve the caller's component values rather than being renormalized during format conversion.
+
+Separate alpha output follows the same produced-frame slot as RGB, including its output resource and subrect. Requesting separate alpha cuts over through the existing drained host-staging path instead of publishing mismatched VRAM RGB and alpha results.
+
 ## Why each piece exists
 
 - **CUDA NGX instead of D3D12 NGX.** NVIDIA's D3D12 path needs NVIDIA's driver. The CUDA path only needs a CUDA driver API, which ZLUDA provides.
@@ -72,7 +83,7 @@ The release zip (`scripts/package_release.sh`) is unpacked into the folder that
 | `d4r\zluda\libcuda.so` | ZLUDA |
 | `d4r\kernels\<gfx target>\` | native kernels and their manifest `d4r-kernels.txt` |
 | `d4r\d4r.ini` | this game's settings |
-| `d4r\nvngx_dlss.dll`, `d4r\ngx\_nvngx.dll` | NVIDIA's files, downloaded by `install.sh` (`D4R_BUNDLE_NVIDIA=installer`, the default), bundled (`=1`) or added by the user (`=0`) |
+| `d4r\nvngx_dlss.dll`, `d4r\ngx\_nvngx.dll`, optional `d4r\nvngx_dlssd.dll` | NVIDIA's files, downloaded by `install.sh` (`D4R_BUNDLE_NVIDIA=installer`, the default), bundled (`=1`) or added by the user (`=0`); the optional denoiser is bundled only through `D4R_BUNDLE_DLSSD` or added by the user |
 
 **NGX routing.** OptiScaler tries `_nvngx.dll` before `nvngx.dll`, with system fallback for each name. `OptiDllPath=d4r` alone can therefore select Proton's system `_nvngx.dll` before reaching the shim. A file-valued `NvngxPath` takes priority on the first probe, regardless of the probe's name. It must select `d4r\nvngx.dll`, never `d4r\ngx\_nvngx.dll`: the latter is NVIDIA's core, which the shim loads internally. See OptiScaler 0.9.4's [NGX loader](https://github.com/optiscaler/OptiScaler/blob/v0.9.4/OptiScaler/proxies/NVNGX_Proxy.h) and [override handling](https://github.com/optiscaler/OptiScaler/blob/v0.9.4/OptiScaler/Util.cpp).
 

docs/building.md

@@ -1,3 +1,4 @@
+<!-- Modified in this fork for CUDA Ray Reconstruction support and validation (2026). -->
 # Building d4r
 
 These steps produce the pieces the launcher needs: a patched ZLUDA, a patched vkd3d-proton, the NGX shim and CUDA bridge, a staged runtime directory with your NVIDIA files, and the native kernels. The versions below are the ones that were tested; others may need changes.
@@ -45,6 +46,7 @@ What the patches add:
 - **0006**: `m16n8k8` f16 MMAs (DLSS 3 CNN, presets E/F) on RDNA3 WMMA like the k16 ones (`D4R_ZLUDA_WMMA_K8=0` disables it); weight-image slots for native prep kernels (`d4r_prep_key_at`, `d4r_prep_key_slots`); and wave64 compilation for offline texture-kernel builds (`D4R_ZLUDA_WAVE64=1`, with wave64 builds of the helper bitcode). `scripts/build_zluda_ptx_helpers.sh` now writes the `_w64` helper variants too.
 - **0007**: gfx12 WMMA layout lowering, optional native e4m3 FP8 WMMA (`D4R_ZLUDA_WMMA_FP8_NATIVE=1`), and an architecture argument for `d4r_emit`. `D4R_ZLUDA_WMMA_LAYOUT=12` on gfx11 is a validation shim, not a release setting.
 - **0008**: post-phase kernels for native overrides (`NAME_post1` … `NAME_post8`, see [native-kernels.md](native-kernels.md)). Preset K's dec5 layer runs as four phase kernels, so the K set needs this patch: an older `libnvcuda.so` would skip the phases.
+- **0008**: CUDA denoiser PTX support: four-half surface stores, floating-coordinate half texture sampling, streaming-cache store spellings, byte-vector trapping stores (including low-byte truncation from wider registers), finite-saturating half conversion, correctly typed half exponentiation, half unordered comparisons, and atomic reductions. CUDA block barriers use a workgroup release fence before the hardware barrier and an acquire fence afterward. It also fixes strict floating-point division operand selection in the LLVM AMDGPU backend and Cargo build-profile detection. Two LLVM sinking fixes remove repeated instruction-order rebuilding and avoid alias scans when no sink target exists; write tracking and memory-dependency checks remain intact. Initialize the LLVM submodule before applying this patch; rebuild device helpers and ZLUDA afterward.
 
 When linking on a system with ROCm libraries outside the default search path, include their library directory in `LIBRARY_PATH`. The build also needs the appropriate ROCm link libraries. `CARGO_BUILD_JOBS=8` caps parallel Rust compilation if memory is limited.
 
@@ -54,7 +56,9 @@ When linking on a system with ROCm libraries outside the default search path, in
 scripts/build_vkd3d_proton_d4r.sh ~/.cache/d4r-vkd3d-d4r
 ```
 
-Patches `0001` and `0002` provide same-frame command-list splitting and resource-lifetime tracking; both are required by the shim. Patch `0003` adds experimental exportable linear textures, enabled only by `VKD3D_D4R_LINEAR_TEXTURES=1`. Point `D4R_VKD3D_DIR` (or `VkD3DDir` in d4r.ini) at the output and update `d3d12.dll` and `d3d12core.dll` together.
+Patches `0001` and `0002` provide same-frame command-list splitting and resource-lifetime tracking; both are required by the shim, which rejects feature creation without the lifetime interface, even when CPU staging or previous-frame output is selected. Patch `0003` adds experimental exportable linear textures, enabled only by `VKD3D_D4R_LINEAR_TEXTURES=1`. Point `D4R_VKD3D_DIR` (or `VkD3DDir` in d4r.ini) at the output and update `d3d12.dll` and `d3d12core.dll` together.
+
+Game-local `d3d12.dll` and `d3d12core.dll` override the prefix's copies. Update both if the game already contains an older d4r runtime; changing only `D4R_VKD3D_DIR` does not replace those local overrides.
 
 ## 3. Shim, bridge and NVAPI identity
 
@@ -69,9 +73,13 @@ The launcher builds the small NVAPI identity bridge (`scripts/build_d4r_nvapi_id
 
 ```sh
 scripts/install_d4r_runtime.sh /path/to/_nvngx.dll /path/to/nvngx_dlss.dll
+# With the CUDA-capable Ray Reconstruction feature library:
+scripts/install_d4r_runtime.sh /path/to/_nvngx.dll /path/to/nvngx_dlss.dll /path/to/nvngx_dlssd.dll
 ```
 
-This stages the shim, the bridge, the NGX core and the DLSS feature library in `D4R_RUNTIME_DIR` (default `~/.local/share/d4r-dlss`, or `RuntimeDir` in d4r.ini). The NVIDIA files come from your own sources, for example an NVIDIA driver package (NGX core) and a game or the DLSS SDK (feature library).
+This stages the shim, the bridge, the NGX core and the DLSS feature libraries in `D4R_RUNTIME_DIR` (default `~/.local/share/d4r-dlss`, or `RuntimeDir` in d4r.ini). The NVIDIA files come from your own sources, for example an NVIDIA driver package (NGX core) and a game or the DLSS SDK (feature libraries). Ray Reconstruction requires a CUDA-capable `nvngx_dlssd.dll`, such as 310.7; the older D3D12-only denoiser is not a replacement. The installer also places the denoiser beside the calling shim, where the NGX core discovers it.
+
+RR D/E temporal evaluation and separate alpha output have been verified on gfx1201 with native FP8 WMMA (`[Kernels] NativeFp8 = true`). For developer launches without an INI, set `D4R_ZLUDA_WMMA=1`, `D4R_ZLUDA_WMMA_FP8=1` and `D4R_ZLUDA_WMMA_FP8_NATIVE=1`. The scalar FP8 fallback produced nonfinite temporal RR output and is not a validated RR path. Other GPU targets and image-quality parity with RTX hardware remain unverified.
 
 ## 5. Native kernels
 
@@ -110,7 +118,7 @@ The script:
 - builds the shim, the bridge and both fast and accuracy network-layer kernels for gfx1100–gfx1103 and gfx1200–gfx1201 by default (`D4R_GPU_ARCHS` can select fewer targets); each gfx12 target also gets an `<arch>-fp8` variant;
 - writes the kernel manifest from the DLLs you list (it records hashes of their PTX, nothing else);
 - stages OptiScaler as `dxgi.dll` with the settings in `packaging/optiscaler.settings`, applied by `scripts/configure_optiscaler.py`; both full and clean ZIPs set `NvngxPath=d4r\nvngx.dll` explicitly, retaining `OptiDllPath=d4r` for other libraries;
-- includes `install.sh` instead of NVIDIA's two DLLs by default, retaining any supplied texture kernels (built from NVIDIA's PTX by `kernels/build.sh tex`); builds accuracy texture sets with `D4R_ZLUDA_EMIT` unless marked prebuilt sets are supplied in `D4R_BUNDLE_TEX/accuracy/<target>`;
+- includes `install.sh` instead of NVIDIA's two DLLs by default, retaining any supplied texture kernels (built from NVIDIA's PTX by `kernels/build.sh tex`); builds accuracy texture sets with `D4R_ZLUDA_EMIT` unless marked prebuilt sets are supplied in `D4R_BUNDLE_TEX/accuracy/<target>`; the full variant adds the optional `nvngx_dlssd.dll` supplied through `D4R_BUNDLE_DLSSD`;
 - with `D4R_BUNDLE_ENGINE`, adds the native engine's shaders as `d4r/engine/k`; the model's weights and tables are never packaged;
 - zips the result together with the ZLUDA and vkd3d-proton builds, the ROCm runtime (as `d4r/rocm`), `packaging/d4r.ini`, the licenses and the patches.
 
@@ -141,7 +149,7 @@ Mount the repository at `/work`, ROCm's compiler/headers/device libraries at `/o
 
 Run `bash scripts/build_release_glibc241.sh` inside the container. The script checks that the build system actually reports `glibc 2.41`, rejects external source-cache checkouts, builds with the locked Cargo dependencies offline, and runs the configuration tests before packaging. The ZIP and checksum go in `$D4R_RELEASE_BUILD_ROOT/dist/`; logs and `toolchain.txt` remain in the build directory. AMD's bundled ROCm runtime, OptiScaler and NVIDIA's libraries remain supplied binaries; the Linux ABI check covers the bundled ROCm libraries too.
 
-`D4R_BUNDLE_TEX` accepts the older flat directory for gfx1101, or a directory with per-target subdirectories (`gfx1100/`–`gfx1103/`, `gfx1200/`, `gfx1201/`, plus `gfx1200-fp8/` and `gfx1201-fp8/`). Build each texture set with `D4R_GPU_ARCH` and `D4R_ZLUDA_EMIT`; the `-fp8` sets also need `D4R_NATIVE_FP8=1`. `d4r_emit` targets those GPUs offline. Missing texture kernels fall back to ZLUDA on that target; the network-layer kernels are still included. Only the RX 7700 XT has been tested on real hardware.
+`D4R_BUNDLE_TEX` accepts the older flat directory for gfx1101, or a directory with per-target subdirectories (`gfx1100/`–`gfx1103/`, `gfx1200/`, `gfx1201/`, plus `gfx1200-fp8/` and `gfx1201-fp8/`). Build each texture set with `D4R_GPU_ARCH` and `D4R_ZLUDA_EMIT`; the `-fp8` sets also need `D4R_NATIVE_FP8=1`. `d4r_emit` targets those GPUs offline. Missing texture kernels fall back to ZLUDA on that target; the network-layer kernels are still included. The published Super Resolution benchmarks use RX 7700 XT; separate native-FP8 Ray Reconstruction D/E temporal validation and Cyberpunk gameplay use RX 9070 XT (gfx1201). This does not validate every packaged native-kernel target.
 
 ## Checks