Created by: LihangLiu

请问用安装了paddlepaddle_gpu_1.1.0.post97，from paddle import fluid最后会报illegal instruction的错误是什么原因： 

*** Aborted at 1541212980 (unix time) try "date -d @1541212980" if you are using GNU date ***
PC: @                0x0 (unknown)
*** SIGILL (@0x7f4adaac38a6) received by PID 31500 (TID 0x7f4b1e5ca700) from PID 18446744073083304102; stack trace: ***
    @     0x7f4b1e397500 (unknown)
    @     0x7f4adaac38a6 std::deque<>::~deque()
    @     0x7f4adc00b97a ThreadPool::ThreadPool()
    @     0x7f4adc00c6aa paddle::platform::CUDADeviceContext::CUDADeviceContext()
    @     0x7f4adc00da08 paddle::platform::DeviceContextPool::DeviceContextPool()
    @     0x7f4adc0102d8 paddle::framework::InitDevices()
    @     0x7f4adc01050d paddle::framework::InitDevices()
    @     0x7f4ada9cea0a _ZZN8pybind1112cpp_function10initializeIZN6paddle6pybindL13pybind11_initEvEUlbE65_vIbEINS_4nameENS_5scopeENS_7siblingEEEEvOT_PFT0_DpT1_EDpRKT2_ENUlRNS_6detail13function_callEE1_4_FUNESL_
    @     0x7f4adaa06414 pybind11::cpp_function::dispatcher()
    @     0x7f4b1e6c8ea4 PyEval_EvalFrameEx
    @     0x7f4b1e6ca6c9 PyEval_EvalCodeEx
    @     0x7f4b1e6c7b98 PyEval_EvalFrameEx
    @     0x7f4b1e6ca6c9 PyEval_EvalCodeEx
    @     0x7f4b1e6ca8ea PyEval_EvalCode
    @     0x7f4b1e6da41e PyImport_ExecCodeModuleEx
    @     0x7f4b1e6da69d load_source_module
    @     0x7f4b1e6dbb1f load_package
    @     0x7f4b1e6db323 import_submodule
    @     0x7f4b1e6db578 load_next
    @     0x7f4b1e6dbecc PyImport_ImportModuleLevel
    @     0x7f4b1e6bf006 builtin___import__
    @     0x7f4b1e62e973 PyObject_Call
    @     0x7f4b1e6c0c38 PyEval_CallObjectWithKeywords
    @     0x7f4b1e6c6638 PyEval_EvalFrameEx
    @     0x7f4b1e6ca6c9 PyEval_EvalCodeEx
    @     0x7f4b1e6ca8ea PyEval_EvalCode
    @     0x7f4b1e6e3bad run_mod
    @     0x7f4b1e6e4d28 PyRun_FileExFlags
    @     0x7f4b1e6e5f48 PyRun_SimpleFileExFlags
    @     0x7f4b1e6f816c Py_Main
    @     0x7f4b1d989cdd __libc_start_main
    @     0x7f4b1e7c807f (unknown)
Illegal instruction

1. 机器是centos 6.3,
2. cuda 9.0, cudnn7,
3. if cat /proc/cpuinfo | grep -i avx; then echo Yes; else echo No; fi 输出yes
4. paddlepaddle非gpu版本是可以运行的。