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]>

scripts/install_d4r_runtime.sh

@@ -7,6 +7,11 @@ set -euo pipefail
 #   bin/d4r_nvngx.dll   D3D12-to-CUDA NGX core shim (OptiScaler NvngxPath target)
 #   ngx/_nvngx.dll      official NGX core
 #   dlss/nvngx_dlss.dll official DLSS SR feature DLL (also copied into bin/)
+#   engine/k/, k-ldr/   models of preset K for the native engine ([Engine] in d4r.ini), when D4R_ENGINE_MODEL /
+#                       D4R_ENGINE_MODEL_LDR name the folders engine/compile_k.py wrote (--ldr for the second)
+#   engine/m/, m-ldr/   models of preset M, when D4R_ENGINE_MODEL_M / D4R_ENGINE_MODEL_M_LDR name the folders
+#                       engine/compile_m.py wrote
+# Each model folder is copied whole, with the native HIP network's hipnet.bin and hip/ when built with --hip.
 # usage: install_d4r_runtime.sh PATH_TO_NGX_CORE_DLL PATH_TO_NVNGX_DLSS_DLL
 if [[ $# -ne 2 ]]; then
   printf 'usage: %s PATH_TO_NGX_CORE_DLL PATH_TO_NVNGX_DLSS_DLL\n' "$0" >&2
@@ -25,5 +30,22 @@ copy_unless_same "$2" "$RUNTIME/dlss/nvngx_dlss.dll"
 # The NGX core's CUDA init ignores the feature search paths it is given and
 # scans the directory of the module that calls it, i.e. the shim's.
 cp -f "$RUNTIME/dlss/nvngx_dlss.dll" "$RUNTIME/bin/nvngx_dlss.dll"
+# K and M models; the -ldr variants serve games without the HDR flag
+for variant in k k-ldr m m-ldr; do
+  case "$variant" in
+    k) source_dir="${D4R_ENGINE_MODEL:-}" required="weights.bin offsets.bin lut.bin input_k.spv output_k.spv exposure_k0.spv exposure_k1.spv" compiler=compile_k.py ;;
+    k-ldr) source_dir="${D4R_ENGINE_MODEL_LDR:-}" required="weights.bin offsets.bin lut.bin input_k.spv output_k.spv" compiler="compile_k.py --ldr" ;;
+    m) source_dir="${D4R_ENGINE_MODEL_M:-}" required="offsets.bin enc0_m.spv" compiler=compile_m.py ;;
+    m-ldr) source_dir="${D4R_ENGINE_MODEL_M_LDR:-}" required="offsets.bin enc0_m.spv" compiler="compile_m.py --ldr" ;;
+  esac
+  [[ -n "$source_dir" ]] || continue
+  for f in $required; do
+    [[ -f "$source_dir/$f" ]] || { printf '%s lacks %s (run engine/%s)\n' "$source_dir" "$f" "$compiler" >&2; exit 2; }
+  done
+  rm -rf "$RUNTIME/engine/$variant"
+  mkdir -p "$RUNTIME/engine"
+  cp -r "$source_dir" "$RUNTIME/engine/$variant"
+  printf 'Installed the native engine model in %s\n' "$RUNTIME/engine/$variant"
+done
 sha256sum "$RUNTIME"/ngx/_nvngx.dll "$RUNTIME"/dlss/nvngx_dlss.dll
 printf 'Installed d4r runtime in %s\n' "$RUNTIME"