gh-131798: JIT: Optimize _CHECK_IS_NOT_PY_CALLABLE_EX and _CHECK_IS_NOT_PY_CALLABLE_KW

[Wulian233]

While implementing this, I ran into a confusing problem. As shown in the diff below, when I assign f = sorted instead of calling sorted() directly, the JIT fails to recognize the optimization, causing the test to fail.

  def test_check_is_not_py_callable_kw(self):
      def testfunc(n):
          total = 0
-         xs = (3, 1, 2)
-         for _ in range(n):
-             total += sorted(xs, reverse=False)[0]
+         f = sorted
+         xs = (3, 1, 2)
+         for _ in range(n):
+             total += f(xs, reverse=False)[0]
          return total

      res, ex = self._run_with_optimizer(testfunc, TIER2_THRESHOLD)
      self.assertEqual(res, TIER2_THRESHOLD)
      self.assertIsNotNone(ex)
      uops = get_opnames(ex)
      self.assertNotIn("_CHECK_IS_NOT_PY_CALLABLE_KW", uops)

Edit: Please add skip news

• Issue: Better uop coverage in the JIT optimizer #131798

[Fidget-Spinner]

While implementing this, I ran into a confusing problem. As shown in the diff below, when I assign f = sorted instead of calling sorted() directly, the JIT fails to recognize the optimization, causing the test to fail.

That's not really a problem, just how the JIT behaves. We only compile the region inside the for loop first, and then a function only if it's hot enough. As for the code example above you placed the sorted outside the for loop, the JIT has no visibility of it.

[Wulian233]

oh I understand, thank you!

[Fidget-Spinner]

LGTM. Thank you.