TL;DR: Add an experimental FSR Ray Regeneration backend on Windows, run the K

network natively on HIP with Vulkan->HIP shared memory, extend K kernels to
gfx11 (RDNA3), and add dependency/packaging/test tooling around them.

Windows: FSR RR backend, native HIP K runner, gfx11 K kernels, tooling

FSR Ray Regeneration (experimental)
- Add fsr_rr_backend.{h,cpp}, fsr_rr_passes.h and the fsr_rr/ prepare and
  compose compute shaders. They run AMD's FidelityFX denoiser
  (amd_fidelityfx_denoiser_dx12.dll, SDK 2.3, not bundled) on a private
  vkd3d-proton device.
- Add the amdxc64 shim (tools/windows/amdxc64_shim) and two vkd3d-proton
  patches for libcxx includes. AMD's amdxc64.dll is never loaded.
- d4r_nvngx_windows: RrFeature now selects between the native DLSS RR
  backend and the FSR backend (D4R_RR_BACKEND=fsr). After the denoiser it
  runs an internal Super Sampling stage (feature 1) to upscale. If SR is
  unavailable and output != render size, the frame fails; it is never
  returned noisy.
- d4r.ini / player_config: new [RayReconstruction] FsrDenoiser,
  FsrStabilityBias (auto or >= 0) and FsrDetail (0-1), with validation, and
  matching D4R_FSR_* environment variables.
- Add the FSR RR fixture, output-check and ngx test scripts, and a shader
  compile script.

Native HIP K runner (Windows)
- CMake: add d4r_engine_k_hip (d4r-k-hip) plus separate volk/engine copies,
  because VK_USE_PLATFORM_WIN32_KHR changes volk's entry points.
- Share Vulkan memory with HIP via Win32 opaque handles on Windows
  (VK_KHR_external_memory_win32) and via fds on Linux. Handle types are now
  intptr_t, and the platform choice is in D4R_EXT_MEMORY_* macros.
- HipNetwork: add debugRepeatHashes and debugNodeMicros, exposed as
  D4R_HIP_REPEAT_CHECK, D4R_HIP_REPEAT_ALL and D4R_HIP_NODE_PROFILE in
  replay.cpp.
- replay.cpp: display-resolution-motion frames (FrameFlagDisplayMotion) no
  longer require near2rnd == out2rnd. Both values must still be in (0, 1].

K kernels on gfx11
- Allow position-token and enc1 tiling on __GFX11__ as well as gfx12.
- Add an opt-in ds_bpermute cross-half exchange
  (D4R_WINDOWS_GFX11_DS_EXCHANGE, enabled for K on Windows gfx11). M keeps
  its LDS path.
- Fix a block-barrier race in the position layer. Without POS_VSHARE a fast
  wave could overwrite its hb rows while a slower wave still read them for a
  later head, giving rare wrong results on gfx1101 (enc0 outputs differing
  between identical launches).

Packaging and install
- install.bat, d4r.ini, WINDOWS_README: FSR option, and
  install_dependencies.py to install pinned NVIDIA dependencies with staged,
  hash-verified replacement and rollback.
- Update the release package, local package refresh and test scripts.
- Add the InteropBatchInputCopies option (default false). Batching is faster
  on some synthetic runs, but player temporal exactness is unresolved.

Tooling and tests
- K model: capture_k_weights, build_k_model, compare_k_outputs,
  benchmark_k_hip, build_capture_runtime, test_k_model and the
  zluda-k-output-capture patch. Captured weights and models are private
  local assets and must not enter release archives.
- Synthetic probes: test_m_ngx, test_sr_dynamic,
  prepare_rr_dynamic_fixture and the d3d12/rr_ngx probe updates.
- Add tests/windows/test_install_dependencies.py.

Docs
- Update the windows-fsr-rr, k-performance, local-package, rr-interop and
  rr-model-cache docs. Add windows-parity-handoff.md and
  windows-parity-continuation-prompt.md.

packaging/windows/d4r.ini

@@ -27,6 +27,11 @@ ShowWatermark = false
 ; faster depends on the game and has not been measured broadly; turn it off if the game crashes,
 ; stutters or shows a wrong image.
 AsyncInterop = false
+; Experimental: submit input conversions/array copies as one ordered GPU batch, retaining
+; whole-context completion before NGX. Some synthetic runs are faster, but player temporal
+; exactness is unresolved. Keep false until that investigation is complete.
+; Missing/invalid values retain the older unbatched behavior. Set false to disable batching.
+BatchInputCopies = false
 
 [Debug]
 ; Check every DLSS output for invalid values (NaN/Inf) on the GPU. Costs some speed; use it only
@@ -51,8 +56,20 @@ NativeFp8 = false
 
 [RayReconstruction]
 ; off is the default. native uses the local Vulkan RR engine through native D3D12.
-; fsr is currently unavailable and rejects RR requests. No fallback to noisy upscaling.
+; fsr (experimental) runs AMD's FSR Ray Regeneration denoiser on a private vkd3d-proton device
+; (d4r\fsr) and upscales with Super Sampling. install.bat [F] sets it up. A frame that cannot be
+; denoised fails; it is never returned noisy.
 Backend = off
 ; auto selects d4r/engine/rr-d. Relative paths resolve beside this d4r.ini.
 ; Generate the D/E model locally; model data is not bundled in the ZIP.
 ModelDir = auto
+; Backend = fsr only: AMD's amd_fidelityfx_denoiser_dx12.dll. auto = d4r\amd_fidelityfx_denoiser_dx12.dll.
+; AMD's file is not bundled; install.bat downloads the pinned FidelityFX SDK 2.3 binary.
+FsrDenoiser = auto
+; Backend = fsr only. AMD's temporal stability bias: auto keeps AMD's default of 1, which weights history heavily.
+; Measured on d4r's synthetic moving scene (error against the noise-free image): without upscaling 0 is about 3x
+; closer than auto (0.036 vs 0.107); after a 2x Super Sampling upscale auto was about 16% closer (0.214 vs 0.248).
+; Linux testing on a game capture found 0 closer to DLSS RR. Pick per game; a number >= 0 is allowed.
+FsrStabilityBias = auto
+; Backend = fsr only. 0-1: share of the colour that skips the denoiser (keeps hair/water/foliage detail, adds grain).
+FsrDetail = 0