Expand Windows GPU support, add native FSR RR, and improve SR interop

Add a native D3D12 FSR Ray Regeneration route for gfx1200/gfx1201,
using the game's device and queue with embedded HLSL guide preparation
and output composition. Preserve the vkd3d compatibility route for
RDNA3 and expose explicit route overrides and diagnostics.

Improve Windows runtime compatibility:
- Retry rejected application-ID initialization with a project identity.
- Defer expensive RR backend creation until feature creation.
- Recover from transient submission failures without disabling later frames.
- Add shared-buffer export fallbacks through explicit heaps and a private
  device on the same adapter.
- Safely evict old RR descriptor sets when input views change.
- Treat requested RR presets as hints when a configured model is available.
- Expand SR/RR input formats and handle diffuse/specular hit distances,
  packed normals, depth-stencil resources, and unused FSR guides.

Reduce SR overhead by batching output downloads and format conversion
under one completion wait. Optimize strict M enc2/tube attention with
bounded FP32 accumulation and the original FP64 fallback for ambiguous
rounding cases.

Extend RDNA4 tooling with gfx1200/gfx1201 native FP8 RR model builds,
capability checks, architecture-aware deployment, and experimental K/M
kernel paths. Require matching model and runtime arithmetic families.

Add optional embedded FSR shader optimizations keyed by shader hash and
exact original SPIR-V contents. Preserve reflected metadata and record
optimizer provenance, artifact hashes, and cache identity.

Simplify Windows installation with automatic backend dependency downloads,
download progress and ETA, backed-up absolute OptiScaler library paths,
and recovery from interrupted RR model builds.

Add build-kernels.bat and a kernels-only build matrix for all eleven
registered RDNA targets, with code-object target validation and SHA256
manifests. Expand conversion probes, numerical stress tests, temporal
benchmark tooling, installer/deployment tests, and setup documentation.

Document synthetic regression and timing results, experimental build
options, and remaining RDNA4 hardware validation requirements.

tools/fsr_rr_probe.cpp

@@ -37,6 +37,24 @@
 #include "windows/fsr_rr_identity.h"
 #include "fsr_rr_scene.h"
 
+// Per-dispatch GPU timing of AMD's denoiser (D4R_FSR_RR_DISPATCH_PROFILE=1), shared with the Linux shim.
+inline void logf(const char* format, ...)
+{
+    va_list args;
+    va_start(args, format);
+    std::vfprintf(stderr, format, args);
+    va_end(args);
+    std::fputc('\n', stderr);
+}
+inline std::string env_string(const char* name) { const char* v = std::getenv(name); return v ? v : ""; }
+inline uint32_t env_uint(const char* name, uint32_t fallback)
+{
+    const char* v = std::getenv(name);
+    return v && *v ? static_cast<uint32_t>(std::strtoul(v, nullptr, 0)) : fallback;
+}
+inline uint32_t profile_log_every() { return env_uint("D4R_PROFILE_LOG_EVERY", 1000000); }
+#include "d4r_fsr_rr_dispatch_profile.h"
+
 namespace
 {
 using namespace fsr_scene;
@@ -425,6 +443,7 @@ int wmain(int argc, wchar_t** argv)
     // 3. Context (shader pipelines are created here: the slow and the most likely to fail part).
     ffxContext context = nullptr;
     const auto createStart = std::chrono::steady_clock::now();
+    fsr_rr_dp::hook_device(g_device);
     const ffxReturnCode_t created = fsr_rr_create_context(runtime, g_device, width, height, signals, 0, context);
     const double createMs =
         std::chrono::duration<double, std::milli>(std::chrono::steady_clock::now() - createStart).count();
@@ -574,7 +593,9 @@ int wmain(int argc, wchar_t** argv)
         }
         list->EndQuery(timestamps, D3D12_QUERY_TYPE_TIMESTAMP, 0);
         const auto cpuStart = std::chrono::steady_clock::now();
+        fsr_rr_dp::begin(g_device, list, f);
         const ffxReturnCode_t dispatched = runtime.dispatch(&context, &desc.header);
+        fsr_rr_dp::end(list);
         cpuMs.push_back(std::chrono::duration<double, std::milli>(std::chrono::steady_clock::now() - cpuStart).count());
         list->EndQuery(timestamps, D3D12_QUERY_TYPE_TIMESTAMP, 1);
         list->ResolveQueryData(timestamps, D3D12_QUERY_TYPE_TIMESTAMP, 0, 2, timestampBuffer, 0);
@@ -666,7 +687,9 @@ int wmain(int argc, wchar_t** argv)
         }
         previousCamera = camera;
     }
-    json_field("frame_log", frameLog + "\n  ]");
+    if (fsr_rr_dp::enabled())
+        fsr_rr_dp::log_stats();
+    json_field("frame_log", frameLog +"\n  ]");
     if (dispatchOk)
     {
         std::vector<double> warm(gpuMs.begin() + std::min<size_t>(gpuMs.size(), 3), gpuMs.end());