Move pytest-run-parallel setup (and hopefully actually make it work)

Author: seberg

.github/workflows/test-wheel-linux.yml

@@ -283,10 +283,6 @@ jobs:
           # we use self-hosted runners on which setup-python behaves weirdly (Python include can't be found)...
           AGENT_TOOLSDIRECTORY: "/opt/hostedtoolcache"
 
-      - name: Install pytest-run-parallel for free-threaded Python
-        if: endsWith(matrix.PY_VER, 't')
-        run: pip install pytest-run-parallel
-
       - name: Set up mini CTK
         if: ${{ matrix.LOCAL_CTK == '1' }}
         uses: ./.github/actions/fetch_ctk

.github/workflows/test-wheel-windows.yml

@@ -257,10 +257,6 @@ jobs:
         with:
           python-version: ${{ matrix.PY_VER }}
 
-      - name: Install pytest-run-parallel for free-threaded Python
-        if: endsWith(matrix.PY_VER, 't')
-        run: pip install pytest-run-parallel
-
       - name: Verify LongPathsEnabled
         run: |
           $val = (Get-ItemProperty -Path 'HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem' -Name 'LongPathsEnabled').LongPathsEnabled

ci/tools/run-tests

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 
-# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # SPDX-License-Identifier: Apache-2.0
 
@@ -20,6 +20,14 @@ fi
 
 test_module=${1}
 
+FREE_THREADING=""
+PYTEST_PARALLEL_ARGS=()
+if python -c 'import sys; assert not sys._is_gil_enabled()' 2> /dev/null; then
+  FREE_THREADING="-ft"
+  PYTEST_PARALLEL_ARGS=(--parallel-threads=4)
+  pip install pytest-run-parallel
+fi
+
 # For standard modes, install pathfinder up front (it is a direct dependency
 # of bindings, and a transitive dependency of core).  Nightly modes install
 # all wheels together in a single pip call further below.
@@ -36,7 +44,7 @@ if [[ "${test_module}" == "pathfinder" ]]; then
     "LD:${CUDA_PATHFINDER_TEST_LOAD_NVIDIA_DYNAMIC_LIB_STRICTNESS} " \
     "FH:${CUDA_PATHFINDER_TEST_FIND_NVIDIA_HEADERS_STRICTNESS} " \
     "BC:${CUDA_PATHFINDER_TEST_FIND_NVIDIA_BITCODE_LIB_STRICTNESS}"
-  pytest -ra -s -v --durations=0 tests/ |& tee /tmp/pathfinder_test_log.txt
+  pytest -ra -s -v --durations=0 "${PYTEST_PARALLEL_ARGS[@]}" tests/ |& tee /tmp/pathfinder_test_log.txt
   # Report the number of "INFO test_" lines (including zero)
   # to support quick validations based on GHA log archives.
   line_count=$(awk '/^INFO test_/ {count++} END {print count+0}' /tmp/pathfinder_test_log.txt)
@@ -51,21 +59,16 @@ elif [[ "${test_module}" == "bindings" ]]; then
     pip install $(ls "${CUDA_BINDINGS_ARTIFACTS_DIR}"/*.whl)[all] --group test
   fi
   echo "Running bindings tests"
-  ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize tests/
+  ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize "${PYTEST_PARALLEL_ARGS[@]}" tests/
   if [[ "${SKIP_CYTHON_TEST}" == 0 ]]; then
-    ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize tests/cython
+    ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize "${PYTEST_PARALLEL_ARGS[@]}" tests/cython
   fi
   popd
 elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
   # Shared setup for core and nightly modes.
   TEST_CUDA_MAJOR="$(cut -d '.' -f 1 <<< ${CUDA_VER})"
   CUDA_VER_MINOR="$(cut -d '.' -f 1-2 <<< "${CUDA_VER}")"
 
-  FREE_THREADING=""
-  if python -c 'import sys; assert not sys._is_gil_enabled()' 2> /dev/null; then
-    FREE_THREADING+="-ft"
-  fi
-
   # Resolve bindings based on BINDINGS_SOURCE (set by env-vars):
   #   main/backport → local wheel from artifacts dir
   #   published     → install from PyPI by version
@@ -106,11 +109,11 @@ elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
     echo "Installed packages before core tests:"
     pip list
     echo "Running core tests"
-    ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize tests/
+    ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize "${PYTEST_PARALLEL_ARGS[@]}" tests/
     # Currently our CI always installs the latest bindings (from either major version).
     # This is not compatible with the test requirements.
     if [[ "${SKIP_CYTHON_TEST}" == 0 ]]; then
-      ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize tests/cython
+      ${SANITIZER_CMD} pytest -rxXs -v --durations=0 --randomly-dont-reorganize "${PYTEST_PARALLEL_ARGS[@]}" tests/cython
     fi
   else
     # Nightly optional-dependency testing.

scripts/run_tests.sh

@@ -57,11 +57,7 @@ RUN_CYTHON=0
 RUN_EXAMPLES=1
 RUN_PTDS=1
 INSTALL_MODE=auto  # auto|force|skip
-DEFAULT_PARALLEL_THREADS=0
-if python -mpip show pytest-run-parallel >/dev/null 2>&1; then
-  DEFAULT_PARALLEL_THREADS=4
-fi
-PARALLEL_THREADS=${DEFAULT_PARALLEL_THREADS}
+PARALLEL_THREADS=0
 while [[ $# -gt 0 ]]; do
   case "$1" in
     -h|--help)