[Build] ORT 1.24 Windows x86 targeting Windows ARM64 build fails with unknown opcode: whilelt, incw

[john-dance]

Describe the issue

Using the following build settings:
python3 tools/ci_build/build.py
--cmake_generator "Visual Studio 17 2022"
--build_dir "$build_dir"
--build_shared_lib
--config $build_type
--compile_no_warning_as_error
--parallel
--use_cache
--skip_tests
--skip_submodule_sync
$@

Results in errors while building the kleidiai library:

\intermediates\arm64\RelWithDebInfo_deps\kleidiai-src\kai\ukernels\matmul\matmul_clamp_f32_f32_f32p\kai_matmul_clamp_f32_f32_f32p4vlx1b_6x4vl_sve_mla_asm.S(69): error A2034: unknown opcode: whilelt [C:\dev\tetracode\build\third_party.build\onnxruntime\intermediates\arm64\RelWithDebInfo_deps\kleidiai-build\kleidiai.vcxproj]
\intermediates\arm64\RelWithDebInfo_deps\kleidiai-src\kai\ukernels\matmul\matmul_clamp_f32_f32_f32p\kai_matmul_clamp_f32_f32_f32p4vlx1b_6x4vl_sve_mla_asm.S(70): error A2034: unknown opcode: incw [C:\dev\tetracode\build\third_party.build\onnxruntime\intermediates\arm64\RelWithDebInfo_deps\kleidiai-build\kleidiai.vcxproj]

Urgency

No response

Target platform

Windows ARM64

Build script

python3 tools/ci_build/build.py
--cmake_generator "Visual Studio 17 2022"
--build_dir "$build_dir"
--build_shared_lib
--config $build_type
--compile_no_warning_as_error
--parallel
--use_cache
--skip_tests
--skip_submodule_sync
$@

Error / output

\intermediates\arm64\RelWithDebInfo_deps\kleidiai-src\kai\ukernels\matmul\matmul_clamp_f32_f32_f32p\kai_matmul_clamp_f32_f32_f32p4vlx1b_6x4vl_sve_mla_asm.S(69): error A2034: unknown opcode: whilelt [C:\dev\tetracode\build\third_party.build\onnxruntime\intermediates\arm64\RelWithDebInfo_deps\kleidiai-build\kleidiai.vcxproj]
\intermediates\arm64\RelWithDebInfo_deps\kleidiai-src\kai\ukernels\matmul\matmul_clamp_f32_f32_f32p\kai_matmul_clamp_f32_f32_f32p4vlx1b_6x4vl_sve_mla_asm.S(70): error A2034: unknown opcode: incw [C:\dev\tetracode\build\third_party.build\onnxruntime\intermediates\arm64\RelWithDebInfo_deps\kleidiai-build\kleidiai.vcxproj]

[... and a bunch more]

Visual Studio Version

VS 17 2022

GCC / Compiler Version

n/a

[edgchen1]

can you please share the actual build command (after variable expansion, etc)? it should be displayed pretty early on in the build.py output.

[john-dance]

2026-02-11 10:52:06,812 build [INFO] - C:\Users\jdance\scoop\shims\cmake.EXE C:\dev\tetracode\build\third_party.build\onnxruntime\src\cmake --compile-no-warning-as-error -Donnxruntime_USE_DML=ON -Donnxruntime_USE_WINML=OFF -Donnxruntime_USE_TELEMETRY=OFF -Donnxruntime_ENABLE_PIX_FOR_WEBGPU_EP=OFF -Donnxruntime_ENABLE_MEMLEAK_CHECKER=ON -Donnxruntime_RUN_ONNX_TESTS=OFF -Donnxruntime_GENERATE_TEST_REPORTS=ON -DPython_EXECUTABLE=C:\Users\jdance\scoop\apps\python310\current\python.exe -Donnxruntime_USE_VCPKG=OFF -Donnxruntime_USE_MIMALLOC=OFF -Donnxruntime_ENABLE_PYTHON=OFF -Donnxruntime_BUILD_CSHARP=OFF -Donnxruntime_BUILD_JAVA=OFF -Donnxruntime_BUILD_NODEJS=OFF -Donnxruntime_BUILD_OBJC=OFF -Donnxruntime_BUILD_SHARED_LIB=ON -Donnxruntime_BUILD_APPLE_FRAMEWORK=OFF -Donnxruntime_USE_DNNL=OFF -Donnxruntime_USE_NNAPI_BUILTIN=OFF -Donnxruntime_USE_VSINPU=OFF -Donnxruntime_USE_RKNPU=OFF -Donnxruntime_ENABLE_MICROSOFT_INTERNAL=OFF -Donnxruntime_USE_VITISAI=OFF -Donnxruntime_USE_TENSORRT=OFF -Donnxruntime_USE_NV=OFF -Donnxruntime_USE_TENSORRT_BUILTIN_PARSER=ON -Donnxruntime_USE_TENSORRT_INTERFACE=OFF -Donnxruntime_USE_CUDA_INTERFACE=OFF -Donnxruntime_USE_NV_INTERFACE=OFF -Donnxruntime_USE_OPENVINO_INTERFACE=OFF -Donnxruntime_USE_VITISAI_INTERFACE=OFF -Donnxruntime_USE_QNN_INTERFACE=OFF -Donnxruntime_USE_MIGRAPHX_INTERFACE=OFF -Donnxruntime_USE_MIGRAPHX=OFF -Donnxruntime_DISABLE_CONTRIB_OPS=OFF -Donnxruntime_DISABLE_ML_OPS=OFF -Donnxruntime_DISABLE_RTTI=OFF -Donnxruntime_DISABLE_EXCEPTIONS=OFF -Donnxruntime_MINIMAL_BUILD=OFF -Donnxruntime_EXTENDED_MINIMAL_BUILD=OFF -Donnxruntime_MINIMAL_BUILD_CUSTOM_OPS=OFF -Donnxruntime_REDUCED_OPS_BUILD=OFF -Donnxruntime_CLIENT_PACKAGE_BUILD=OFF -Donnxruntime_BUILD_MS_EXPERIMENTAL_OPS=OFF -Donnxruntime_ENABLE_LTO=OFF -Donnxruntime_USE_ACL=OFF -Donnxruntime_USE_ARMNN=OFF -Donnxruntime_ARMNN_RELU_USE_CPU=ON -Donnxruntime_ARMNN_BN_USE_CPU=ON -Donnxruntime_USE_JSEP=OFF -Donnxruntime_USE_WEBGPU=OFF -Donnxruntime_USE_EXTERNAL_DAWN=OFF -Donnxruntime_WGSL_TEMPLATE=static -Donnxruntime_ENABLE_NVTX_PROFILE=OFF -Donnxruntime_ENABLE_TRAINING=OFF -Donnxruntime_ENABLE_TRAINING_OPS=OFF -Donnxruntime_ENABLE_TRAINING_APIS=OFF -Donnxruntime_ENABLE_CPU_FP16_OPS=OFF -Donnxruntime_USE_NCCL=OFF -Donnxruntime_BUILD_BENCHMARKS=OFF -Donnxruntime_GCOV_COVERAGE=OFF -Donnxruntime_ENABLE_MEMORY_PROFILE=OFF -Donnxruntime_ENABLE_CUDA_LINE_NUMBER_INFO=OFF -Donnxruntime_USE_CUDA_NHWC_OPS=OFF -Donnxruntime_BUILD_WEBASSEMBLY_STATIC_LIB=OFF -Donnxruntime_ENABLE_WEBASSEMBLY_EXCEPTION_CATCHING=ON -Donnxruntime_ENABLE_WEBASSEMBLY_API_EXCEPTION_CATCHING=OFF -Donnxruntime_ENABLE_WEBASSEMBLY_EXCEPTION_THROWING=ON -Donnxruntime_WEBASSEMBLY_RUN_TESTS_IN_BROWSER=OFF -Donnxruntime_ENABLE_WEBASSEMBLY_JSPI=OFF -Donnxruntime_ENABLE_WEBASSEMBLY_THREADS=OFF -Donnxruntime_ENABLE_WEBASSEMBLY_DEBUG_INFO=OFF -Donnxruntime_ENABLE_WEBASSEMBLY_PROFILING=OFF -Donnxruntime_ENABLE_LAZY_TENSOR=OFF -Donnxruntime_ENABLE_CUDA_PROFILING=OFF -Donnxruntime_USE_XNNPACK=OFF -Donnxruntime_USE_WEBNN=OFF -Donnxruntime_USE_CANN=OFF -Donnxruntime_DISABLE_FLOAT8_TYPES=OFF -Donnxruntime_DISABLE_FLOAT4_TYPES=OFF -Donnxruntime_DISABLE_SPARSE_TENSORS=OFF -Donnxruntime_DISABLE_OPTIONAL_TYPE=OFF -Donnxruntime_CUDA_MINIMAL=OFF -Donnxruntime_BUILD_CACHE=ON -Donnxruntime_QNN_HOME=C:/dev/tetracode/build/third_party/qnn_sdk.tungsten/cmake -Donnxruntime_USE_QNN=ON -Donnxruntime_BUILD_QNN_EP_STATIC_LIB=ON -Donnxruntime_USE_KLEIDIAI=ON -Donnxruntime_USE_SVE=ON -DCMAKE_VS_GLOBALS=UseMultiToolTask=true;EnforceProcessCountAcrossBuilds=true -Donnxruntime_CROSS_COMPILING=ON -DONNX_USE_MSVC_STATIC_RUNTIME=OFF -Dprotobuf_MSVC_STATIC_RUNTIME=OFF -DABSL_MSVC_STATIC_RUNTIME=OFF -Dgtest_force_shared_crt=ON -DCMAKE_MSVC_DEBUG_INFORMATION_FORMAT=$<$CONFIG:Debug,RelWithDebInfo:Embedded> -A ARM64 -G Visual Studio 17 2022 -DCMAKE_C_FLAGS=/MP -DCMAKE_CXX_FLAGS=/MP -DCMAKE_BUILD_TYPE=RelWithDebInfo

[john-dance]

I also saw this in the output...
CMake Warning at CMakeLists.txt:527 (message):
onnxruntime_USE_SVE was set but it is not supported on this platform. It
will be disabled.

[edgchen1]

sorry, I meant the output from this line:

onnxruntime/tools/ci_build/build.py

Line 2292 in 2cf5bbd

log.debug("Command line arguments:\n {}".format(" ".join(shlex.quote(arg) for arg in sys.argv[1:]))) # noqa: G001

it would be helpful for reproducing the issue.

[john-dance]

2026-02-11 10:55:28,362 build [DEBUG] - Command line arguments:
--cmake_generator 'Visual Studio 17 2022' --build_dir C:/dev/tetracode/build/third_party.build/onnxruntime/intermediates/arm64 --build_shared_lib --config RelWithDebInfo --compile_no_warning_as_error --parallel --use_cache --skip_tests --skip_submodule_sync --arm64 --use_qnn static_lib --use_dml --qnn_home=C:/dev/tetracode/build/third_party/qnn_sdk.tungsten/cmake

[Colm-in-Arm]

Hello John.

Can tell me what version of the MSVC compiler you're using? I've successfully built 1.24.1 with:

Microsoft (R) C/C++ Optimizing Compiler Version 19.44.35217 for ARM64

My understanding is recognizing SME intrinsics appeared a couple of years ago around version 19.40.

Colm.