compatibility with Jetson Nano

[jobesu14]

I built taichi on the Jetson Nano successfully (after doing that modif on taichi-dev/glfw repo) with the following cmake command:
cmake .. -DCMAKE_C_COMPILER=clang-8 -DCMAKE_CXX_COMPILER=clang-8 -DCUDA_VERSION=10.0 -DTI_WITH_CUDA:BOOL=True

When I try the to run the fractal.py example (or any other example) I get this runtime error (both with archi==ti.cpu or archi==ti.gpu):

jetbot@jetson-4-3:~/dev/taichi/examples$ python3 fractal.py 
[Taichi] mode=development
[Taichi] preparing sandbox at /tmp/taichi-j3ui2gqi
[Taichi] <dev mode>, llvm 8.0.1, commit 477b1e1d, python 3.6.9
[Taichi] Starting on arch=arm64
[E 06/26/20 21:56:23.972] [jit_cpu.cpp:create_llvm_jit_session_cpu@373] LLVM TargetMachineBuilder has failed when getting data layout.


***********************************
* Taichi Compiler Stack Traceback *
***********************************
/tmp/taichi-j3ui2gqi/taichi_core.so: taichi::Logger::error(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, bool)
/tmp/taichi-j3ui2gqi/taichi_core.so: taichi::lang::create_llvm_jit_session_cpu(taichi::lang::Arch)
/tmp/taichi-j3ui2gqi/taichi_core.so: taichi::lang::TaichiLLVMContext::TaichiLLVMContext(taichi::lang::Arch)
/tmp/taichi-j3ui2gqi/taichi_core.so: taichi::lang::Program::Program(taichi::lang::Arch)
/tmp/taichi-j3ui2gqi/taichi_core.so(+0x74c24c) [0x7fa809324c]
/tmp/taichi-j3ui2gqi/taichi_core.so(+0x6156e4) [0x7fa7f5c6e4]
python3(_PyCFunction_FastCallDict+0x1ac) [0x5b9b2c]
python3(_PyObject_FastCallDict+0x1c0) [0x6058d0]
python3() [0x5f2bcc]
python3(PyObject_Call+0x50) [0x606c40]
python3() [0x595cd8]
python3() [0x5a52b4]
python3(_PyObject_FastCallDict+0xd0) [0x6057e0]
python3() [0x529738]
python3(_PyEval_EvalFrameDefault+0x4770) [0x52e5a8]
python3() [0x527860]
python3() [0x5297dc]
python3(_PyEval_EvalFrameDefault+0x4770) [0x52e5a8]
python3() [0x528ff0]
python3() [0x529584]
python3() [0x5297dc]
python3(_PyEval_EvalFrameDefault+0x49c8) [0x52e800]
python3() [0x528ff0]
python3(PyRun_FileExFlags+0x118) [0x630438]
python3(PyRun_SimpleFileExFlags+0xfc) [0x635acc]
python3(Py_Main+0xe78) [0x6202c8]
python3(main+0x1cc) [0x420d3c]
/lib/aarch64-linux-gnu/libc.so.6: __libc_start_main
python3() [0x420e94]

Internal Error occurred, check this page for possible solutions:
https://taichi.readthedocs.io/en/stable/install.html#troubleshooting
Expected<T> must be checked before access or destruction.
Unchecked Expected<T> contained error:
Unable to find target for this triple (no targets are registered)[E 06/26/20 21:56:23.978] Received signal 6 (Aborted)


***********************************
* Taichi Compiler Stack Traceback *
***********************************
/tmp/taichi-j3ui2gqi/taichi_core.so: taichi::Logger::error(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, bool)
/tmp/taichi-j3ui2gqi/taichi_core.so: taichi::signal_handler(int)
linux-vdso.so.1(__kernel_rt_sigreturn+0) [0x7fac0cb6c0]
/lib/aarch64-linux-gnu/libc.so.6: raise

Internal Error occurred, check this page for possible solutions:
https://taichi.readthedocs.io/en/stable/install.html#troubleshooting

The issue seems to come from that line in jit_cpu.cpp@371: auto data_layout = jtmb->getDefaultDataLayoutForTarget();
That returns the errors message: Unable to find target for this triple (no targets are registered)
I printed the triple that jtmb return: aarch64-unknown-linux-gnu which seems to be fine...

I saw in this closed issue that @hgnan has successfully built and run Taichi on Jetson Nano for Taichi v0.3.x...

I'd be happy to help debug if needed.

[archibate]

Hi, thank for purposing this! But @yuanming-hu seems to be busy with his online lecture these days. I'm willing to help, but I don't have much knowledge on the LLVM backend.. sorry. If you found the solution, feel free to open a PR and request @yuanming-hu for review.

[bobcao3]

Redirecting this issue to #4711