Fix build failure when both CUDA and MiGraphX execution providers are enabled

[Copilot]

Description

arena_extend_strategy was defined in separate #if defined(USE_CUDA) and #if defined(USE_MIGRAPHX) preprocessor blocks, causing a redefinition error when both providers are enabled.

Consolidates the definition and extern declaration into a single combined guard:

#if defined(USE_MIGRAPHX) || defined(USE_CUDA) || defined(USE_CUDA_PROVIDER_INTERFACE)
onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;
#endif

Both files updated:

• onnxruntime/python/onnxruntime_pybind_state_common.cc — variable definition
• onnxruntime/python/onnxruntime_pybind_state_common.h — extern declaration

No other duplicate symbols found across CUDA/MiGraphX conditional blocks.

Motivation and Context

Building with --use_cuda --use_migraphx fails with:

error: redefinition of 'onnxruntime::ArenaExtendStrategy onnxruntime::python::arena_extend_strategy'

Both providers share this config variable; it just needs a unified guard.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>[Build] Cannot build with cuda and migraphx</issue_title>
<issue_description>### Describe the issue

My laptop has an nvidia dgpu and an amd igpg. To build with cuda and migraphx I had to edit onnxruntime/python/onnxruntime_pybind_state_common.cc to remove one of the two onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;

#if defined(USE_CUDA) || defined(USE_CUDA_PROVIDER_INTERFACE)
// TODO remove deprecated global config
OrtCudnnConvAlgoSearch cudnn_conv_algo_search = OrtCudnnConvAlgoSearchExhaustive;
// TODO remove deprecated global config
bool do_copy_in_default_stream = true;
// TODO remove deprecated global config
onnxruntime::cuda::TunableOpInfo tunable_op{};
onnxruntime::CUDAExecutionProviderExternalAllocatorInfo external_allocator_info{};
// TODO remove deprecated global config
onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;
#endif

#if defined(USE_MIGRAPHX)
// TODO remove deprecated global config
onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;
#endif

I think it could be changed to

#if defined(USE_CUDA) || defined(USE_CUDA_PROVIDER_INTERFACE)
// TODO remove deprecated global config
OrtCudnnConvAlgoSearch cudnn_conv_algo_search = OrtCudnnConvAlgoSearchExhaustive;
// TODO remove deprecated global config
bool do_copy_in_default_stream = true;
// TODO remove deprecated global config
onnxruntime::cuda::TunableOpInfo tunable_op{};
onnxruntime::CUDAExecutionProviderExternalAllocatorInfo external_allocator_info{};
// TODO remove deprecated global config
#endif

#if defined(USE_MIGRAPHX) || defined(USE_CUDA) || defined(USE_CUDA_PROVIDER_INTERFACE)
// TODO remove deprecated global config
onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;
#endif

Urgency

No response

Target platform

AMD + Nvidia setups

Build script

./build.sh --config Release --build_shared_lib --build_wheel --parallel 12 --skip_tests --use_cuda --cuda_home /opt/cuda --cudnn_home /opt/cuda --use_tensorrt --tensorrt_home /usr --use_openvino CPU --use_migraphx --migraphx_home /opt/rocm --cmake_extra_defines CMAKE_CXX_FLAGS="-Wno-maybe-uninitialized -Wno-free-nonheap-object" CMAKE_CUDA_ARCHITECTURES="120"

Error / output

/home/someone/onnxruntime-3.13/onnxruntime/onnxruntime/python/onnxruntime_pybind_state_common.cc:36:34: error: redefinition of ‘onnxruntime::ArenaExtendStrategy onnxruntime::python::arena_extend_strategy’
36 | onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;
| ^~~~~~~~~~~~~~~~~~~~~
/home/someone/onnxruntime-3.13/onnxruntime/onnxruntime/python/onnxruntime_pybind_state_common.cc:31:34: note: ‘onnxruntime::ArenaExtendStrategy onnxruntime::python::arena_extend_strategy’ previously defined here
31 | onnxruntime::ArenaExtendStrategy arena_extend_strategy = onnxruntime::ArenaExtendStrategy::kNextPowerOfTwo;
| ^~~~~~~~~~~~~~~~~~~~~

Visual Studio Version

na

GCC / Compiler Version

gcc version 15.2.1 20260209 (GCC)</issue_description>

<agent_instructions>Please create a pull request, and check whether there are other places that enabling both cuda and migraphx might cause build errors.</agent_instructions>

Comments on the Issue (you are @copilot in this section)

• Fixes [Build] Cannot build with cuda and migraphx #27797

---

💬 Send tasks to Copilot coding agent from Slack and Teams to turn conversations into code. Copilot posts an update in your thread when it's finished.

[tianleiwu]

/azp run Linux QNN CI Pipeline, Win_TRT_Minimal_CUDA_Test_CI, Windows ARM64 QNN CI Pipeline, Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).