Route old-format CUDA ELF to nvdisasm 12.9 so CFGs work

nvdisasm 13.x disables flow analysis on pre-CUDA 13 ELF (OSABI 0x33), so
control flow graphs fail. Fetch nvdisasm 12.9 from NVIDIA's redist site at
startup and use it for those files.

README.md

@@ -12,5 +12,7 @@ short_description: Disassemble CUDA cubins with cuobjdump and nvdisasm
 
 Upload a CUDA cubin, fatbin or ELF object and run NVIDIA's `cuobjdump` or `nvdisasm` on it.
 Both tools are installed from PyPI (`nvidia-cuda-cuobjdump`, `nvidia-cuda-nvdisasm`); no GPU required.
+nvdisasm 13.x cannot build control flow graphs for old-format CUDA ELF (pre-CUDA 13, OSABI `0x33`), so those files are
+routed to nvdisasm 12.9, downloaded from NVIDIA's redist site at startup.
 
 `examples/` holds two cubins assembled with `ptxas` from the included PTX sources.

app.py

@@ -1,5 +1,9 @@
+import platform
 import shlex
 import subprocess
+import tarfile
+import tempfile
+import urllib.request
 from pathlib import Path
 
 import gradio as gr
@@ -12,7 +16,23 @@ def find_tool(name):
     return str(path)
 
 
+def fetch_legacy_nvdisasm(version):
+    arch = "linux-sbsa" if platform.machine() == "aarch64" else "linux-x86_64"
+    name = f"cuda_nvdisasm-{arch}-{version}-archive"
+    path = Path(tempfile.gettempdir()) / name / "nvdisasm"
+    if not path.exists():
+        url = f"https://developer.download.nvidia.com/compute/cuda/redist/cuda_nvdisasm/{arch}/{name}.tar.xz"
+        with urllib.request.urlopen(url) as r, tarfile.open(fileobj=r, mode="r|xz") as tar:
+            member = next(m for m in tar if m.name == f"{name}/bin/nvdisasm")
+            path.parent.mkdir(exist_ok=True)
+            path.write_bytes(tar.extractfile(member).read())
+        path.chmod(0o755)
+    return str(path)
+
+
 TOOLS = {name: find_tool(name) for name in ("cuobjdump", "nvdisasm")}
+# nvdisasm 13.x can't run flow analysis (CFGs) on old-format CUDA ELF (OSABI 0x33, pre-CUDA 13); 12.9 can.
+LEGACY_NVDISASM = fetch_legacy_nvdisasm("12.9.88")
 OUTPUTS = {
     "cuobjdump": ["--dump-sass", "--dump-ptx", "--dump-elf", "--dump-resource-usage", "--dump-elf-symbols",
                   "--list-elf", "--list-ptx"],
@@ -37,13 +57,23 @@ d3.select(graph).graphviz({useWorker: false, zoom: true}).renderDot(props.value,
 EXAMPLES = sorted(str(p) for p in Path(__file__).with_name("examples").glob("*.cubin"))
 
 
-def version(tool):
-    return subprocess.run([TOOLS[tool], "--version"], capture_output=True, text=True).stdout.split("release ")[-1].splitlines()[0]
+def version(exe):
+    return subprocess.run([exe, "--version"], capture_output=True, text=True).stdout.split("release ")[-1].splitlines()[0]
+
+
+def is_old_elf(file):
+    with open(file, "rb") as f:
+        head = f.read(8)
+    return head[:4] == b"\x7fELF" and head[7] == 0x33
+
+
+def executable(tool, file):
+    return LEGACY_NVDISASM if tool == "nvdisasm" and is_old_elf(file) else TOOLS[tool]
 
 
 def run_one(tool, args, file):
     try:
-        r = subprocess.run([TOOLS[tool], *args, file], capture_output=True, text=True, timeout=60)
+        r = subprocess.run([executable(tool, file), *args, file], capture_output=True, text=True, timeout=60)
     except subprocess.TimeoutExpired:
         return "Timed out after 60s."
     return (r.stdout + r.stderr or "(no output)") + (f"\n[exit code {r.returncode}]" if r.returncode else "")
@@ -65,8 +95,9 @@ def select_tool(tool):
 
 with gr.Blocks(title="cuobjdump / nvdisasm") as demo:
     gr.Markdown("# cuobjdump / nvdisasm\n"
-                "Disassemble CUDA cubins, fatbins and ELF objects with NVIDIA's tools from PyPI ("
-                + ", ".join(f"`{t}` {version(t)}" for t in TOOLS) + ").")
+                "Disassemble CUDA cubins, fatbins and ELF objects with NVIDIA's tools ("
+                + ", ".join(f"`{t}` {version(exe)}" for t, exe in TOOLS.items())
+                + f"; old-format CUDA ELF goes to `nvdisasm` {version(LEGACY_NVDISASM)}).")
     with gr.Row(equal_height=True):
         with gr.Column(scale=1):
             file = gr.File(label="cubin / fatbin / CUDA ELF", type="filepath")