Segfault when running CxxWrap example

[rossviljoen]

When running CxxWrapExample/src/example.jl:

signal (11): Segmentation fault
in expression starting at REPL[4]:1
_ZN9mindspore7compile7FinalVM4EvalERKNS_9VectorRefE at /home/ross/msproj/mindspore/mindspore/_c_expression.cpython-37m-x86_64-linux-gnu.so (unknown line)
_ZNSt17_Function_handlerIFN9mindspore7BaseRefERNS0_7compile7FinalVMERKNS0_9VectorRefEEZN5jlcxx11TypeWrapperIS3_E6methodIS1_S3_JS7_EEERSB_RKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEMT0_FT_DpT1_EEUlS4_S7_E_E9_M_invokeERKSt9_Any_dataS4_S7_ at /home/ross/msproj/MindSpore.jl/build/lib/libcxx_graph_api.so (unknown line)
_ZN5jlcxx6detail11CallFunctorIN9mindspore7BaseRefEJRNS2_7compile7FinalVMERKNS2_9VectorRefEEE5applyEPKvNS_13WrappedCppPtrESD_ at /home/ross/msproj/MindSpore.jl/build/lib/libcxx_graph_api.so (unknown line)
Eval at /home/ross/.julia/packages/CxxWrap/ptbgM/src/CxxWrap.jl:619
_jl_invoke at /buildworker/worker/package_linux64/build/src/gf.c:2247 [inlined]
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2429
jl_apply at /buildworker/worker/package_linux64/build/src/julia.h:1788 [inlined]
do_call at /buildworker/worker/package_linux64/build/src/interpreter.c:126
eval_value at /buildworker/worker/package_linux64/build/src/interpreter.c:215
eval_stmt_value at /buildworker/worker/package_linux64/build/src/interpreter.c:166 [inlined]
eval_body at /buildworker/worker/package_linux64/build/src/interpreter.c:587
jl_interpret_toplevel_thunk at /buildworker/worker/package_linux64/build/src/interpreter.c:731
jl_toplevel_eval_flex at /buildworker/worker/package_linux64/build/src/toplevel.c:885
jl_toplevel_eval_flex at /buildworker/worker/package_linux64/build/src/toplevel.c:830
jl_toplevel_eval_flex at /buildworker/worker/package_linux64/build/src/toplevel.c:830
jl_toplevel_eval_in at /buildworker/worker/package_linux64/build/src/toplevel.c:944
eval at ./boot.jl:373 [inlined]
eval_user_input at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:150
repl_backend_loop at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:246
start_repl_backend at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:231
#run_repl#47 at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:364
run_repl at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:351
_jl_invoke at /buildworker/worker/package_linux64/build/src/gf.c:2247 [inlined]
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2429
#930 at ./client.jl:394
jfptr_YY.930_45169.clone_1 at /home/ross/julia-1.7.2/lib/julia/sys.so (unknown line)
_jl_invoke at /buildworker/worker/package_linux64/build/src/gf.c:2247 [inlined]
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2429
jl_apply at /buildworker/worker/package_linux64/build/src/julia.h:1788 [inlined]
jl_f__call_latest at /buildworker/worker/package_linux64/build/src/builtins.c:757
#invokelatest#2 at ./essentials.jl:716 [inlined]
invokelatest at ./essentials.jl:714 [inlined]
run_main_repl at ./client.jl:379
exec_options at ./client.jl:309
_start at ./client.jl:495
jfptr__start_38732.clone_1 at /home/ross/julia-1.7.2/lib/julia/sys.so (unknown line)
_jl_invoke at /buildworker/worker/package_linux64/build/src/gf.c:2247 [inlined]
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2429
jl_apply at /buildworker/worker/package_linux64/build/src/julia.h:1788 [inlined]
true_main at /buildworker/worker/package_linux64/build/src/jlapi.c:559
jl_repl_entrypoint at /buildworker/worker/package_linux64/build/src/jlapi.c:701
main at julia (unknown line)
__libc_start_main at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
unknown function (ip: 0x400808)
Allocations: 6386357 (Pool: 6383760; Big: 2597); GC: 6
./runjulia: line 10:  5855 Segmentation fault      (core dumped)

This appears to be a segfault when attempting to run the VM that is produced after compiling the graph. In this case, the graph is empty (has no operations) - but it is possible to execute a blank graph directly from C++ (see CxxWrapExample/ccsrc/main.cc)

[agdkyang]

Hi, rossviljoen, have you ever compiled and run ./cxx_example successfully
(https://gitee.com/ginfung/ms_cxx_graph_api_example/tree/master)?

[rossviljoen]

Hi! Yes, I am able to run that example using version 1.2.1 of MindSpore

[agdkyang]

@rossviljoen Errors may occur when you run an empty graph. You need add non-empty graph just like the function TensorAddExample and CreateTensorAddGraph in https://github.com/cambridge-mlg/MindSpore.jl/blob/master/ms_cxx_graph_api_example/main.cc. Also you can try an empty graph with c++ interface and check if the error is created from this problem.

[rossviljoen]

I can confirm that running an empty graph with the c++ interface doesn't error and seems to work fine.

See the code for this here

I think that the error comes from limitations in Julia's C++ FFI CxxWrap.jl - I think that this is probably not a MindSpore issue. Many thanks for the help though!

[yebai]

@rossviljoen can you try to call these C++ functions via extern c instead of using CxxWrap.jl?

[agdkyang]

@yebai @rossviljoen The link is how to realize the Julia AD https://rogerluo.dev/Brochure.jl/dev/automatic-differentiation/#%E5%8A%A8%E6%80%81%E5%9B%BE-VS-%E9%9D%99%E6%80%81%E5%9B%BE-1