rr: add FSR Ray Regeneration, exact RDNA3 inference and motion experiments

Add an experimental FSR Ray Regeneration backend for D3D12 RR requests:
- Convert depth, motion, normals and albedos into AMD denoiser inputs.
- Split and recompose diffuse/specular signals, then run Super Sampling.
- Add stability/detail controls, demodulation guards and halo limiting.
- Handle AMD provider discovery under NVIDIA GPU spoofing.
- Add runtime probes, synthetic reference scenes and per-dispatch profiling.

Add exact gfx11 DLSS-RR inference using f16 cooperative matrices:
- Reproduce FP8 conversions, accumulation and camera reprojection arithmetic.
- Widen model weights and avoid requiring VK_EXT_shader_float8.
- Extend model building, replay tools and descriptor metadata.

Expand motion preprocessing and experimental AW2 native-motion support:
- Add tiled LDS dilation, padded allocation handling and thin-only sky fill.
- Order dilation with input waits and NGX on the selected stream.
- Add jitter-aligned native composition, hybrid routing and wire history repair.
- Add prepared masks, sparse repair and independent motion/repair controls.

Add optional fused R11G11B10 input/output conversion, sampled stage timings
and profile summarization. Extend configuration, packaging, documentation,
FidelityFX headers and CPU/GPU regression coverage.

tests/test_native_motion.py

@@ -0,0 +1,65 @@
+"""Native compositor ABI and rejection of unsafe buffer contracts."""
+from pathlib import Path
+import subprocess
+import tempfile
+import unittest
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+class NativeMotionTests(unittest.TestCase):
+    def test_buffer_ranges_and_pitches(self):
+        source = r'''
+#include "d4r_native_motion.h"
+#include <assert.h>
+#include <string.h>
+int main(void) {
+    D4rNativeMotionParams p; memset(&p,0,sizeof(p));
+    p.width=2560; p.height=1440; p.render_width=853; p.render_height=480;
+    p.depth=0x10000000; p.geometry=0x20000000; p.scene=0x30000000;
+    p.matrix=0x40000000; p.output=0x50000000;
+    p.depth_pitch=10240; p.geometry_pitch=10244; p.output_pitch=10248;
+    assert(sizeof(p)==80 && d4r_native_motion_parameters_valid(&p));
+    assert(!d4r_native_motion_parameters_valid(0));
+    D4rNativeMotionParams q=p;
+    q.output=q.geometry+4; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.output=q.scene+16; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.output=q.matrix-16; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.depth=UINT64_MAX-16; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.output_pitch=UINT64_MAX-3; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.geometry_pitch=10239; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.depth_pitch=10241; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.render_width=2561; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.height=0; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.scene=0; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.matrix+=2; assert(!d4r_native_motion_parameters_valid(&q));
+    q=p; q.height=1; q.render_height=1; q.output_pitch=10240; assert(d4r_native_motion_parameters_valid(&q));
+    /* Adjacent, disjoint ranges are valid; output padding is not written. */
+    q=p; q.output=q.geometry+(q.height-1)*q.geometry_pitch+q.width*4;
+    assert(d4r_native_motion_parameters_valid(&q));
+    D4rNativeMotionHybridParams hybrid={{p,0,0},0x60000000,853*4};
+    assert(sizeof(hybrid)==104 && d4r_native_motion_hybrid_parameters_valid(&hybrid));
+    D4rNativeMotionHybridParams bad=hybrid;bad.game_motion=bad.native.base.output;
+    assert(!d4r_native_motion_hybrid_parameters_valid(&bad));
+    bad=hybrid;bad.game_motion_pitch=853*4-1;assert(!d4r_native_motion_hybrid_parameters_valid(&bad));
+    bad=hybrid;bad.native.jitter_x=2;assert(!d4r_native_motion_hybrid_parameters_valid(&bad));
+    bad=hybrid;bad.game_motion=UINT64_MAX-16;assert(!d4r_native_motion_hybrid_parameters_valid(&bad));
+    bad=hybrid;bad.game_motion_pitch=UINT64_MAX-3;assert(!d4r_native_motion_hybrid_parameters_valid(&bad));
+    D4rNativeMotionPreparedParams prep={{p,0,0},0x70000000,0x80000000,1,0};
+    assert(sizeof(prep)==112 && d4r_native_motion_prepared_parameters_valid(&prep));
+    D4rNativeMotionPreparedParams no=prep;no.thin_mask=0;assert(!d4r_native_motion_prepared_parameters_valid(&no));
+    no=prep;no.wire_axes=no.thin_mask;assert(!d4r_native_motion_prepared_parameters_valid(&no));
+    no=prep;no.wire_axes=no.native.base.output;assert(!d4r_native_motion_prepared_parameters_valid(&no));
+    no=prep;no.thin_mask=UINT64_MAX-3;assert(!d4r_native_motion_prepared_parameters_valid(&no));
+    no=prep;no.wire_flags=4;assert(!d4r_native_motion_prepared_parameters_valid(&no));
+    no=prep;no.reserved=1;assert(!d4r_native_motion_prepared_parameters_valid(&no));
+    return 0;
+}
+'''
+        with tempfile.TemporaryDirectory() as tmp:
+            runner = Path(tmp) / "native-motion"
+            for compiler, language, standard in [("cc", "c", "c11"), ("c++", "c++", "c++17")]:
+                subprocess.run([compiler, "-x", language, "-std=" + standard, "-Wall", "-Wextra", "-Werror",
+                                "-I", str(ROOT / "tools"), "-o", str(runner), "-"],
+                               input=source, text=True, check=True, capture_output=True)
+                subprocess.run([str(runner)], check=True, timeout=5)