libtorch win-arm64: inject BSD uint typedefs for ATen aarch64 vec headers

Run 6: MAX_JOBS=2 fixed the OOM — win-arm64 compiled cleanly to [563/1448] of
torch_cpu, then a real compile error:
  vec128_uint_aarch64.h: error: unknown type name 'uint'

PyTorch's aarch64 NEON vec headers use the BSD integer typedefs (uint/ushort/ulong/
uchar) that <sys/types.h> supplies on Linux/macOS but MSVC/Windows does not — win-arm64
is a new, under-tested PyTorch target. Inject the typedefs into the ATen vec headers
that use them (idempotent marker, survives the cached source tree).

Author: vackva

TODO.md

@@ -80,11 +80,12 @@ Green (incl. find_package(Torch) smoke): macOS arm64, macOS x86_64, **macOS univ
 smoke ✅), Linux aarch64, Linux x86_64, Windows x86_64. **7/8 archives green.** Only Windows arm64
 left. The `if: !cancelled()` decoupling worked — universal now validated.
 
-Windows arm64 (run 5): clang-cl arm64 targeting fixed — it compiled cleanly to [433/1448] of
-torch_cpu, then the **windows-11-arm runner was OOM-killed** ("hosted runner lost communication …
-starves it for CPU/Memory"; logs truncated, step stuck in_progress). Emulated x64 clang-cl on big
-ATen TUs at MAX_JOBS=nproc exhausted the 16 GB runner. Fix: `MAX_JOBS=2` on Windows (trade time for
-memory; 6h budget). If it still OOMs, drop to 1 or use native arm64 LLVM (no emulation).
+Windows arm64 progress (each round clears a layer): clone (longpaths) → pip (requirements-build) →
+clang-cl x64→arm64 target → OOM (MAX_JOBS=2) → run 6 reached [563/1448] then a real compile error:
+`vec128_uint_aarch64.h: unknown type name 'uint'`. PyTorch's aarch64 NEON vec headers use the BSD
+`uint`/`ushort`/`ulong`/`uchar` typedefs (from `<sys/types.h>` on Linux/macOS; absent on MSVC) —
+win-arm64 is under-tested upstream. Fix: inject those typedefs into the ATen vec headers that use
+them (idempotent, survives the source cache). If more non-vec TUs hit it, widen the patch scope.
 
 ### Still open (LibTorch)
 - **From-source recipes need CI iteration** (first-pass `build-libtorch.sh`):

engines/libtorch/build-libtorch.sh

@@ -107,6 +107,16 @@ case "$PLATFORM" in
     # ~30% into torch_cpu — emulated x64 clang-cl on PyTorch's large ATen TUs is memory
     # -hungry. Fewer concurrent compiles trades build time (we have 6h) for not OOM-ing.
     export MAX_JOBS=2
+    # PyTorch's aarch64 NEON vec headers use the BSD integer typedefs (uint/ushort/ulong/
+    # uchar) that <sys/types.h> provides on Linux/macOS but MSVC/Windows does NOT ->
+    # "unknown type name 'uint'" in vec128_uint_aarch64.h. win-arm64 is a new, under-tested
+    # PyTorch target. Inject the typedefs into the ATen vec headers that use them
+    # (idempotent via the marker, so it survives the cached source tree).
+    while IFS= read -r h; do
+      grep -q '__win_uint_fix__' "$h" 2>/dev/null && continue
+      printf '// __win_uint_fix__\ntypedef unsigned int uint;\ntypedef unsigned short ushort;\ntypedef unsigned long ulong;\ntypedef unsigned char uchar;\n' \
+        | cat - "$h" > "$h.__t" && mv "$h.__t" "$h"
+    done < <(grep -rlwE 'uint|ushort|ulong|uchar' "$SRC/aten/src/ATen/cpu/vec" 2>/dev/null || true)
     ;;
   *) echo "ERROR: unknown platform '$PLATFORM'"; exit 1 ;;
 esac