[TD] Enable td on cpu windows (#125049)

yolo

Also
* Ensure that at least 1 test always gets run (`//` does truncation which results in 0 if you have too few tests discovered)
* Don't run test removal on slow tests - I'm not touching that yet

I am avoid everything other than pull + trunk workflows, so not doing this on windows CUDA, which runs on periodic
Pull Request resolved: #125049
Approved by: https://github.com/huydhn, https://github.com/ZainRizvi

test/run_test.py

@@ -30,6 +30,7 @@
     IS_CI,
     IS_LINUX,
     IS_MACOS,
+    IS_WINDOWS,
     parser as common_parser,
     retry_shell,
     set_cwd,
@@ -1192,9 +1193,12 @@ def parse_args():
                 and os.getenv("TEST_CONFIG") == "distributed"
                 and TEST_CUDA
             )
+            or (IS_WINDOWS and not TEST_CUDA)
         )
         and os.getenv("BRANCH", "") != "main"
-        and not strtobool(os.environ.get("NO_TD", "False")),
+        and not strtobool(os.environ.get("NO_TD", "False"))
+        and "slow" not in os.getenv("TEST_CONFIG", "")
+        and "slow" not in os.getenv("BUILD_ENVIRONMENT", ""),
     )
     parser.add_argument(
         "additional_unittest_args",

tools/testing/target_determination/heuristics/interface.py

@@ -116,7 +116,8 @@ def get_top_per_tests(self, n: int) -> Tuple[List[TestRun], List[TestRun]]:
         """Divides list of tests into two based on the top n% of scores.  The
         first list is the top, and the second is the rest."""
         tests = [x[1] for x in self._traverse_scores()]
-        return tests[: n * len(tests) // 100], tests[n * len(tests) // 100 :]
+        index = n * len(tests) // 100 + 1
+        return tests[:index], tests[index:]
 
     def get_info_str(self, verbose: bool = True) -> str:
         info = ""