1 files changed, 88 insertions, 0 deletions
diff --git a/tools/cuda-ab/ab.sh b/tools/cuda-ab/ab.sh
new file mode 100755
index 0000000..07f6b24
--- /dev/null
+++ b/tools/cuda-ab/ab.sh
@@ -0,0 +1,88 @@
+#!/bin/bash
+# SPDX-License-Identifier: BSD-3-Clause
+#
+# ab.sh - CUDA A/B harness: run the same workload on the host (bare metal,
+# real Turing GPU) and in the guest (Open-IOV emulated vGPU), then compare the
+# machine-readable results. Mirrors the nvidia-smi host/guest comparison used
+# earlier: the host run is the oracle.
+#
+# Usage:
+#   tools/cuda-ab/ab.sh host     # build + run on the host, save run/host.txt
+#   tools/cuda-ab/ab.sh guest    # push the binary + run in the running VM
+#   tools/cuda-ab/ab.sh compare  # diff the stable result lines
+#   tools/cuda-ab/ab.sh all      # host + guest + compare
+#
+# The binary links the CUDA runtime statically, so the guest needs no CUDA
+# toolkit - only the guest driver's libcuda.so.
+set -e
+
+HERE="$(cd "$(dirname "$0")" && pwd)"
+REPO="$(cd "$HERE/../.." && pwd)"
+VMEXEC="$REPO/scripts/vgpu_vm.py"
+RUN="$HERE/run"
+SRC="$HERE/vector_add.cu"
+BIN="$RUN/vector_add"
+
+mkdir -p "$RUN"
+
+build_host() {
+	command -v nvcc >/dev/null || { echo "nvcc not found (install nvidia-cuda-toolkit)"; exit 1; }
+	nvcc -O2 --cudart=static -o "$BIN" "$SRC"
+}
+
+run_host() {
+	build_host
+	echo "[*] host run (oracle)"
+	"$BIN" | tee "$RUN/host.txt"
+}
+
+push_bin() {
+	[ -x "$BIN" ] || build_host
+	local b64 chunk off total size
+	b64=$(base64 -w0 "$BIN")
+	total=${#b64}
+	off=0
+	size=70000
+	echo "[*] pushing $(stat -c%s "$BIN") bytes to guest"
+	"$VMEXEC" exec "rm -f /tmp/vadd.b64" >/dev/null 2>&1
+	while [ $off -lt $total ]; do
+		chunk=${b64:$off:$size}
+		"$VMEXEC" exec "echo -n '$chunk' >> /tmp/vadd.b64" >/dev/null 2>&1
+		off=$((off + size))
+	done
+	"$VMEXEC" exec "base64 -d /tmp/vadd.b64 > /tmp/vector_add && chmod +x /tmp/vector_add && rm -f /tmp/vadd.b64 && echo PUSH_OK"
+}
+
+run_guest() {
+	push_bin
+	echo "[*] guest run (emulated vGPU)"
+	"$VMEXEC" exec "/tmp/vector_add" 2>&1 | tee "$RUN/guest.txt"
+}
+
+compare() {
+	local h g
+	[ -f "$RUN/host.txt" ] || { echo "missing $RUN/host.txt (run host first)"; exit 1; }
+	[ -f "$RUN/guest.txt" ] || { echo "missing $RUN/guest.txt (run guest first)"; exit 1; }
+	h=$(grep -E '^(DEVICE|RESULT|STATUS)' "$RUN/host.txt")
+	g=$(grep -E '^(DEVICE|RESULT|STATUS)' "$RUN/guest.txt")
+	echo "--- host:";  echo "$h"
+	echo "--- guest:"; echo "$g"
+	# DEVICE lines carry the GPU name (differs by design); compare the rest.
+	h_norm=$(echo "$h" | sed 's/^DEVICE .*/DEVICE <gpu>/')
+	g_norm=$(echo "$g" | sed 's/^DEVICE .*/DEVICE <gpu>/')
+	if [ "$h_norm" = "$g_norm" ]; then
+		echo "A/B: MATCH"
+	else
+		echo "A/B: MISMATCH"
+		diff <(echo "$h_norm") <(echo "$g_norm") || true
+		exit 1
+	fi
+}
+
+case "${1:-all}" in
+host)    run_host ;;
+guest)   run_guest ;;
+compare) compare ;;
+all)     run_host; run_guest; compare ;;
+*)       echo "usage: $0 [host|guest|compare|all]"; exit 1 ;;
+esac