From f250c8cca503c7c372ed156f6608d1599cf6f925 Mon Sep 17 00:00:00 2001
From: morluto <76467478+morluto@users.noreply.github.com>
Date: Fri, 23 Oct 2026 17:59:09 +0800
Subject: [PATCH] fix(gpu): accept PTX compute target aliases

Preserve the source spelling of NVIDIA's documented compute_xx target synonym.

Fixes #6494
---
 docs/gpu-code-objects.md                   |  2 ++
 src/gpu/PtxKernels.ts                      |  2 +-
 tests/acceptance/gpuCodeObjectsCli.test.ts | 19 +++++++++++++++++++
 3 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/docs/gpu-code-objects.md b/docs/gpu-code-objects.md
index ead60041c9..0c377c216c 100644
--- a/docs/gpu-code-objects.md
+++ b/docs/gpu-code-objects.md
@@ -81,6 +81,8 @@ Kernel representations have different meanings:
 - PTX reports parameter declarations and virtual registers. Shared bytes cover
   supported static declarations inside that entry; they do not establish module
   shared memory, dynamic allocation or physical device resources.
+  A `.target compute_xx` spelling is retained as written; NVIDIA PTX defines it
+  as a synonym for `.target sm_xx`.
   Scalar register declarations count each bare or percent-prefixed name and
   each parameterized range. Unsupported declaration forms or overflowing totals
   retain an unknown count and partial decoding coverage. Unsupported parameter
diff --git a/src/gpu/PtxKernels.ts b/src/gpu/PtxKernels.ts
index 008f52942b..b1edccf59c 100644
--- a/src/gpu/PtxKernels.ts
+++ b/src/gpu/PtxKernels.ts
@@ -406,7 +406,7 @@ export const readPtxTarget = (text: string): string | null => {
       if (token.value === ".target") {
         const value = stream.next();
         target =
-          !value.done && /^sm_[0-9]+[a-z]?$/u.test(value.value.value)
+          !value.done && /^(?:sm|compute)_[0-9]+[a-z]?$/u.test(value.value.value)
             ? value.value.value
             : null;
       }
diff --git a/tests/acceptance/gpuCodeObjectsCli.test.ts b/tests/acceptance/gpuCodeObjectsCli.test.ts
index 1579d6ad05..4483c6a445 100644
--- a/tests/acceptance/gpuCodeObjectsCli.test.ts
+++ b/tests/acceptance/gpuCodeObjectsCli.test.ts
@@ -111,6 +111,25 @@ cliTest(
   },
 );
 
+cliTest("accepts the PTX compute_xx .target synonym verbatim", async ({ cli }) => {
+  const root = await createTestTempDirectory("rea-ptx-target-cli-");
+  const text = `.version 8.8\n.target compute_80\n.address_size 64\n.entry kernel() {\nret;\n}\n`;
+  const path = join(root, "target.ptx");
+  await writeFile(path, text);
+  const response = await cli.run({
+    arguments: ["inspect-gpu-code-objects", path, "--json"],
+  });
+  expect(response.exitCode, response.stderr).toBe(0);
+  const evidence = parseEvidence(
+    toolContract("inspect_gpu_code_objects").outputSchema.parse(response.json),
+  );
+  const report = gpuCodeObjectsSchema.parse(evidence.normalized_result);
+  expect(report.objects[0]!.target).toBe("compute_80");
+  expect(report.ptx_artifacts[0]!.sha256).toBe(
+    createHash("sha256").update(text).digest("hex"),
+  );
+});
+
 cliTest(
   "attributes malformed selector JSON to its actual CLI option",
   async ({ cli }) => {