Describe the issue

identifier "__hmul2_rn" is undefined

Urgency

No response

Target platform

aarch64, cuda11.3

Build script

./build.sh --use_cuda --cudnn_home /PATH/TO/CUDNN_HOME --cuda_home /PATH/TO/CUDA_HOME --build_shared_lib --enable_pybind --build_wheel --update --build --compile_no_warning_as_error --skip_tests

Error / output

/home/share/cimc/home/lizhongxin/softwares/onnx/onnxruntime/onnxruntime/contrib_ops/cuda/bert/relative_attn_bias_impl.cu(159):
error: identifier "__hmul2_rn" is undefined

I modified the file "relative_attn_bias_impl.cu", and change "__hmul2_rn" to "__huml2_sat" , it seems that this problem is solved。

Visual Studio Version

No response

GCC / Compiler Version

GCC 9.3.0, cuda11.3.0