Updating cpp standard to C++17 for gnu and nvcc

[access2rohit]

The commit: https://github.com/apache/tvm/pull/12692 has caused DLR build for GPU to Fail

git clone --recursive --branch release-1.13.0 https://github.com/neo-ai/neo-ai-dlr
mkdir neo-ai-dlr/build && cd neo-ai-dlr/build
cmake .. -DUSE_CUDA=ON -DUSE_CUDNN=ON -DUSE_TENSORRT=ON -DTENSORRT_INCLUDE_DIR=/tmp/TensorRT-8.4.3.1/include -DTENSORRT_LIB_DIR=/tmp/TensorRT-8.4.3.1/lib/libnvinfer.so
make -j

Error:

/neo-ai-dlr/3rdparty/tvm/include/tvm/runtime/container/array.h(261): error: namespace "std" has no member class "bool_constant"

Solution: bool_constant is C++17 standard, gcc and nvcc are set to C++14. Two changes are required Change this to 14 -> 17. https://github.com/neo-ai/neo-ai-dlr/blob/release-1.13.0/CMakeLists.txt#L183 (I can create a commit and you can cherry pick and merge it to release branch) C++ 17 for cuda is supported in CMake 3.18 onwards

Testing

CPU

cmake .. && make -j && make test Output:

Running tests...
Test project /neo-ai-dlr/build
      Start  1: dlr_allocator_test
 1/12 Test  #1: dlr_allocator_test ...............   Passed    0.74 sec
      Start  2: dlr_common_test
 2/12 Test  #2: dlr_common_test ..................   Passed    0.00 sec
      Start  3: dlr_elem_test
 3/12 Test  #3: dlr_elem_test ....................   Passed    5.52 sec
      Start  4: dlr_pipeline_test
 4/12 Test  #4: dlr_pipeline_test ................   Passed    0.01 sec
      Start  5: dlr_relayvm_elem_test
 5/12 Test  #5: dlr_relayvm_elem_test ............   Passed   46.65 sec
      Start  6: dlr_relayvm_test
 6/12 Test  #6: dlr_relayvm_test .................   Passed   43.65 sec
      Start  7: dlr_test
 7/12 Test  #7: dlr_test .........................   Passed   14.83 sec
      Start  8: dlr_treelite_test
 8/12 Test  #8: dlr_treelite_test ................   Passed    0.01 sec
      Start  9: dlr_tvm_elem_test
 9/12 Test  #9: dlr_tvm_elem_test ................   Passed    2.81 sec
      Start 10: dlr_tvm_test
10/12 Test #10: dlr_tvm_test .....................   Passed    3.15 sec
      Start 11: dlr_dlsym_test
11/12 Test #11: dlr_dlsym_test ...................   Passed    4.79 sec
      Start 12: dlr_multiple_lib_test
12/12 Test #12: dlr_multiple_lib_test ............   Passed    1.16 sec

100% tests passed, 0 tests failed out of 12

Total Test time (real) = 123.31 sec

GPU

cmake .. -DUSE_CUDA=ON -DUSE_CUDNN=ON -DUSE_TENSORRT=ON -DTENSORRT_INCLUDE_DIR=/tmp/TensorRT-8.4.3.1/include -DTENSORRT_LIB_DIR=/tmp/TensorRT-8.4.3.1/lib/libnvinfer.so && make -j && make test Output:

Running tests...
Test project /neo-ai-dlr/build
      Start  1: dlr_allocator_test
 1/12 Test  #1: dlr_allocator_test ...............   Passed    0.78 sec
      Start  2: dlr_common_test
 2/12 Test  #2: dlr_common_test ..................   Passed    0.03 sec
      Start  3: dlr_elem_test
 3/12 Test  #3: dlr_elem_test ....................   Passed    5.54 sec
      Start  4: dlr_pipeline_test
 4/12 Test  #4: dlr_pipeline_test ................   Passed    0.03 sec
      Start  5: dlr_relayvm_elem_test
 5/12 Test  #5: dlr_relayvm_elem_test ............   Passed   46.51 sec
      Start  6: dlr_relayvm_test
 6/12 Test  #6: dlr_relayvm_test .................   Passed   43.70 sec
      Start  7: dlr_test
 7/12 Test  #7: dlr_test .........................   Passed   14.88 sec
      Start  8: dlr_treelite_test
 8/12 Test  #8: dlr_treelite_test ................   Passed    0.04 sec
      Start  9: dlr_tvm_elem_test
 9/12 Test  #9: dlr_tvm_elem_test ................   Passed    2.87 sec
      Start 10: dlr_tvm_test
10/12 Test #10: dlr_tvm_test .....................   Passed    3.20 sec
      Start 11: dlr_dlsym_test
11/12 Test #11: dlr_dlsym_test ...................   Passed    3.83 sec
      Start 12: dlr_multiple_lib_test
12/12 Test #12: dlr_multiple_lib_test ............   Passed    1.23 sec

100% tests passed, 0 tests failed out of 12

Total Test time (real) = 122.64 sec

[access2rohit]

@jsheng-jian please review

[ghost]

LGTM, thanks!