Engine: native presets K and M in the D3D12 shim, wider game coverage

A standalone Vulkan inference engine (engine/) for DLSS presets K and M that
runs from weights extracted from the user's own nvngx_dlss.dll, with d4r's
own input, network and reconstruction shaders or the native HIP network
layers, recorded into the game's command list by the shim. Opt-in
([Engine] Enabled / D4R_ENGINE=1); anything uncovered falls back to the
CUDA/ZLUDA path with the reason logged.

Byte-identical to NGX's CUDA backend under the phase-capable reference
(1706x960 -> 2560x1440, native HIP network, eleven frames, RX 9070 XT):
- K render- and display-resolution motion vectors, HDR and LDR
- K game-supplied exposure and automatic exposure with any pre-exposure
  and exposure scale (one residual: effective exposure exactly 2 or 4)
- K and M with the DLSS indicator's inverted-axis parameters
- M with any exposure settings (NGX ignores them for M)
K packages need the current compile_k.py (marker D4RO0003).

Also: HIP shared buffers are pooled across feature recreation (ROCm leaks
destroyed imports), hipnet.bin must match the preset, M-only games do not
load the K model, install_d4r_runtime.sh stages k, k-ldr, m, m-ldr with
hip/, the activation table has a generator (gen_activation_m.py), and
package_release.sh can bundle the engine's shaders (never its weights).

Signed-off-by: woutermans <[email protected]>

README.md

@@ -12,6 +12,13 @@ See [supported games](SUPPORTED_GAMES.md) for the tested games and DLSS models.
 
 The development shim also has [experimental native Vulkan support](docs/vulkan.md).
 
+A [standalone native inference engine](engine/README.md) runs presets K and M from weights extracted from your own
+`nvngx_dlss.dll`, with d4r's own Vulkan input, network and reconstruction shaders (optionally the native HIP network
+layers) and GPU-resident temporal history. The game shim uses the CUDA/ZLUDA path described below by default;
+`[Engine] Enabled = true` (`D4R_ENGINE=1`) records supported K and M evaluations with the native engine into the
+game's own command list instead and hands anything else to the CUDA path (see the engine's README). Presets E and L
+remain CUDA-only.
+
 > **Not affiliated with NVIDIA or AMD.**
 
 
@@ -160,6 +167,7 @@ Gameplay recordings of each DLSS model at the Quality, Performance and Ultra Per
 |---|---|
 | `tools/` | the NGX shim, the Wine CUDA bridge, a D3D12 DLSS harness, probes and kernel replay tools |
 | `kernels/` | native RDNA3/RDNA4 kernels (`k/` DLSS 4, `m/` DLSS 4.5, `tex/` texture-kernel parts, `common/` WMMA layouts), their build script, validation tools and numpy reference models |
+| `engine/` | standalone weight-based preset K and M inference runtime (Vulkan shaders, optional native HIP network), model packaging and full-frame validation |
 | `patches/` | ZLUDA and vkd3d-proton patches |
 | `scripts/` | build, install, launch and probe scripts |
 | `config/` | the default `d4r.ini` for the developer launcher |