[Build] Fail to pass AutoEpSelection and OrtEpLibrary tests in Windows x64 QNN build

[quic-hungjuiw]

Describe the issue

1. I failed on the six AutoEpSelection testcases with the command:

   [  FAILED  ] AutoEpSelection.CpuEP
   [  FAILED  ] AutoEpSelection.MiscApiTests
   [  FAILED  ] AutoEpSelection.PreferCpu
   [  FAILED  ] AutoEpSelection.PreferGpu
   [  FAILED  ] AutoEpSelection.PreferNpu
   [  FAILED  ] AutoEpSelection.PolicyDelegate
   
   with the command
   onnxruntime_autoep_test.exe --gtest_filter=AutoEpSelection.*
   
   and the two OrtEpLibrary testcases:
   [  FAILED  ] OrtEpLibrary.LoadUnloadPluginLibrary
   [  FAILED  ] OrtEpLibrary.LoadUnloadPluginLibraryCxxApi
   
   with the command
   onnxruntime_autoep_test.exe --gtest_filter=OrtEpLibrary.*
   

2. I have tried investigated the testcase AutoEpSelection.CpuEP. In DefaultDeviceSelection,

   • c_api->GetEpDevices(*ort_env, &ep_devices, &num_devices) cannot get devices for CpuExecutionProvider
     • ep_devices is still empty
     • num_devices is still 0
   • devices.empty() cause the error

3. Note: I am using QNN sdk version 2.33.2.250410, which should match the version in Azure CI pipeline.

Urgency

No response

Target platform

Windows x64 QNN

Build script

cd onnxruntime
.\build.bat --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --cmake_generator "Visual Studio 17 2022" --use_qnn --qnn_home qairt\2.33.2.250410 --build_dir build\build_pybind --enable_pybind --build_wheel

Error / output

I failed on the six AutoEpSelection testcases:

[ RUN      ] AutoEpSelection.CpuEP
onnxruntime\onnxruntime\test\autoep\test_autoep_selection.cc(43): error: Value of: !devices.empty()
  Actual: false
Expected: true
No devices found with EP name of CPUExecutionProvider
Stack trace:
...
unknown file: error: C++ exception with description "Must provide one or more OrtEpDevice instances." thrown in the test body.
[  FAILED  ] AutoEpSelection.CpuEP (33 ms)
[ RUN      ] AutoEpSelection.MiscApiTests
unknown file: error: SEH exception with code 0xc0000005 thrown in the test body.
Stack trace:
...
[  FAILED  ] AutoEpSelection.MiscApiTests (1 ms)
[ RUN      ] AutoEpSelection.PreferCpu
unknown file: error: C++ exception with description "onnxruntime\onnxruntime\core\session\providime_autoep_test.exe --gter_policy_context.cc:342 onnxruntime::PreferCpuEpPolicy::SelectProvidersForDevices first_cpu != sorted_devices.end() was false. No CPU based execution providers were found." thrown in the test body.
[  FAILED  ] AutoEpSelection.PreferCpu (1 ms)
[ RUN      ] AutoEpSelection.PreferGpu
unknown file: error: C++ exception with description "onnxruntime\onnxruntime\core\session\provider_policy_context.cc:342 onnxruntime::PreferCpuEpPolicy::SelectProvidersForDevices first_cpu != sorted_devices.end() was false. No CPU based execution providers were found." thrown in the test body.
[ RUN      ] AutoEpSelection.PreferNpu
unknown file: error: SEH exception with code 0xc0000005 thrown in the test body.
Stack trace:
...
[  FAILED  ] AutoEpSelection.PreferNpu (7 ms)
[ RUN      ] AutoEpSelection.PolicyDelegate
unknown file: error: SEH exception with code 0xc0000005 thrown in the test body.
Stack trace:
...
[  FAILED  ] AutoEpSelection.PolicyDelegate (3 ms) 

and the two OrtEpLibrary testcases:

[ RUN      ] OrtEpLibrary.LoadUnloadPluginLibrary
onnxruntime\onnxruntime\test\autoep\test_autoep_selection.cc(538): error: Expected equality of these values:
  num_test_ep_devices
    Which is: 0
  1
Expected an OrtEpDevice to have been created by the test library.
Stack trace:
...

[  FAILED  ] OrtEpLibrary.LoadUnloadPluginLibrary (102 ms)
[ RUN      ] OrtEpLibrary.LoadUnloadPluginLibraryCxxApi
unknown file: error: C++ exception with description "library is already registered under example_ep" thrown in the test body.

[  FAILED  ] OrtEpLibrary.LoadUnloadPluginLibraryCxxApi (0 ms)

Visual Studio Version

17.12.3

GCC / Compiler Version

No response

[skottmckay]

There should always be a CPU device. In the loop here when guid is GUID_DEVCLASS_PROCESSOR does it return any entries?

[quic-hungjuiw]

Hi @skottmckay, thank you for the information. After investigating, it appears that I have vendor AuthenticAMD, which is not yet covered by the GetVendorId function. Please let me know if I misunderstand anything.

[skottmckay]

It should be covered by that. Potentially it's exiting here if it can't get a vendor or device id from SetupApi.

onnxruntime/onnxruntime/core/platform/windows/device_discovery.cc

Line 172 in 03a2544

if (vendor_id == 0 && device_id == 0) {

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.

[microsoft-github-policy-service]

Applying stale label due to no activity in 30 days

[microsoft-github-policy-service]

Closing issue due to no activity in 30 days