TDA4VM: Building edgeai-tidl-tools CPP Examples

Gildardo

Prodigy 210 points

Part Number: TDA4VM

HI,

I have problems when building CPP examples on Target TDA4M

Using latest commit on

https://github.com/TexasInstruments/edgeai-tidl-tools/

Following guide on

https://github.com/TexasInstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/README.md

First when building with make, it complains about not finding itidl_rt.h. Then I have added the tidl_tools path

export TIDL_TOOLS_PATH="/home/root/edgeai-tidl-tools/tidl_tools/"

Now I got the following error:

[ 77%] Linking CXX executable /home/root/edgeai-tidl-tools/bin/Release/ort_main
/usr/lib/gcc/aarch64-linux/9.2.1/../../../../aarch64-linux/bin/ld: CMakeFiles/ort_main.dir/onnx_main.cpp.o: in function `onnx::main::runInference(tidl::modelInfo::ModelInfo*, tidl::arg_parsing::Settings*)':
onnx_main.cpp:(.text+0x3db0): undefined reference to `OrtSessionsOptionsSetDefault_Tidl'

I kindly request your support to solve the issue

Below the complete log of cmake and make

Best

Gildardo

root@j7-evm:~/edgeai-tidl-tools# mkdir build && cd build
root@j7-evm:~/edgeai-tidl-tools/build# cmake ../examples/
-- The C compiler identification is GNU 9.2.1
-- The CXX compiler identification is GNU 9.2.1
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = edgeai_tidl_examples
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = edgeai_tidl_examples
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = edgeai_tidl_examples
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = tfl_main
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = ort_main
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = dlr_main
-- Detected processor: aarch64
-- CMAKE_BUILD_TYPE = Release PROJECT_NAME = tidlrt_clasification
-- Configuring done
-- Generating done
-- Build files have been written to: /home/root/edgeai-tidl-tools/build
root@j7-evm:~/edgeai-tidl-tools/build# make
make[1]: Entering directory '/home/root/edgeai-tidl-tools/build'
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
Scanning dependencies of target post_process
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
[ 5%] Building CXX object osrt_cpp/post_process/CMakeFiles/post_process.dir/post_process.cpp.o
[ 11%] Linking CXX static library /home/root/edgeai-tidl-tools/lib/Release/libpost_process.a
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
[ 11%] Built target post_process
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
Scanning dependencies of target pre_process
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
[ 16%] Building CXX object osrt_cpp/pre_process/CMakeFiles/pre_process.dir/pre_process.cpp.o
[ 22%] Linking CXX static library /home/root/edgeai-tidl-tools/lib/Release/libpre_process.a
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
[ 22%] Built target pre_process
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
Scanning dependencies of target utils
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
[ 27%] Building CXX object osrt_cpp/utils/CMakeFiles/utils.dir/src/utility_functs.cpp.o
[ 33%] Building CXX object osrt_cpp/utils/CMakeFiles/utils.dir/src/model_info.cpp.o
[ 38%] Building CXX object osrt_cpp/utils/CMakeFiles/utils.dir/src/edgeai_classnames.cpp.o
[ 44%] Building CXX object osrt_cpp/utils/CMakeFiles/utils.dir/src/ti_logger.cpp.o
[ 50%] Building CXX object osrt_cpp/utils/CMakeFiles/utils.dir/src/arg_parsing.cpp.o
[ 55%] Linking CXX static library /home/root/edgeai-tidl-tools/lib/Release/libutils.a
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
[ 55%] Built target utils
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
Scanning dependencies of target tfl_main
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
[ 61%] Building CXX object osrt_cpp/tfl/CMakeFiles/tfl_main.dir/tfl_main.cpp.o
/home/root/edgeai-tidl-tools/examples/osrt_cpp/tfl/tfl_main.cpp: In function 'int tflite::main::runInference(tidl::modelInfo::ModelInfo*, tidl::arg_parsing::Settings*)':
/home/root/edgeai-tidl-tools/examples/osrt_cpp/tfl/tfl_main.cpp:254:25: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
254 |         char *keys[] = {"artifacts_folder", "num_tidl_subgraphs", "debug_level"};
      |                         ^~~~~~~~~~~~~~~~~~
/home/root/edgeai-tidl-tools/examples/osrt_cpp/tfl/tfl_main.cpp:254:45: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
254 |         char *keys[] = {"artifacts_folder", "num_tidl_subgraphs", "debug_level"};
      |                                             ^~~~~~~~~~~~~~~~~~~~
/home/root/edgeai-tidl-tools/examples/osrt_cpp/tfl/tfl_main.cpp:254:67: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
254 |         char *keys[] = {"artifacts_folder", "num_tidl_subgraphs", "debug_level"};
      |                                                                   ^~~~~~~~~~~~~
/home/root/edgeai-tidl-tools/examples/osrt_cpp/tfl/tfl_main.cpp:255:81: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
255 |         char *values[] = {(char *)modelInfo->m_infConfig.artifactsPath.c_str(), "16", "0"};
      |                                                                                 ^~~~
/home/root/edgeai-tidl-tools/examples/osrt_cpp/tfl/tfl_main.cpp:255:87: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
255 |         char *values[] = {(char *)modelInfo->m_infConfig.artifactsPath.c_str(), "16", "0"};
      |                                                                                       ^~~
[ 66%] Linking CXX executable /home/root/edgeai-tidl-tools/bin/Release/tfl_main
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
[ 66%] Built target tfl_main
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
Scanning dependencies of target ort_main
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make[2]: Entering directory '/home/root/edgeai-tidl-tools/build'
[ 72%] Building CXX object osrt_cpp/ort/CMakeFiles/ort_main.dir/onnx_main.cpp.o
[ 77%] Linking CXX executable /home/root/edgeai-tidl-tools/bin/Release/ort_main
/usr/lib/gcc/aarch64-linux/9.2.1/../../../../aarch64-linux/bin/ld: CMakeFiles/ort_main.dir/onnx_main.cpp.o: in function `onnx::main::runInference(tidl::modelInfo::ModelInfo*, tidl::arg_parsing::Settings*)':
onnx_main.cpp:(.text+0x3db0): undefined reference to `OrtSessionsOptionsSetDefault_Tidl'
collect2: error: ld returned 1 exit status
make[2]: *** [osrt_cpp/ort/CMakeFiles/ort_main.dir/build.make:87: /home/root/edgeai-tidl-tools/bin/Release/ort_main] Error 1
make[2]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make[1]: *** [CMakeFiles/Makefile2:332: osrt_cpp/ort/CMakeFiles/ort_main.dir/all] Error 2
make[1]: Leaving directory '/home/root/edgeai-tidl-tools/build'
make: *** [Makefile:130: all] Error 2
root@j7-evm:~/edgeai-tidl-tools/build#

over 3 years ago

0 Takuma Fujiwara over 3 years ago

TI__Guru 50193 points

Hello Gildardo,

Which SDK is this using? For reference, I will link the SDK download page https://www.ti.com/tool/PROCESSOR-SDK-J721E (e.g. PROCESSOR-SDK-LINUX-SK-TDA4VM or PROCESSOR-SDK-LINUX-J721E, etc)?

And is this on a SK-TDA4VM, EVM, or custom board?

Regards,

Takuma

0 Gildardo over 3 years ago in reply to Takuma Fujiwara

Prodigy 210 points

Hi Takuma,

I am using the latest PROCESSOR-SDK-RTOS-J721E Version: 08.01.00.13 and PROCESSOR-SDK-LINUX-J721E Version: 08.01.00.07 on a EVM -TDA4VM.

Best

Gildardo

0 Gildardo over 3 years ago in reply to Gildardo

Prodigy 210 points

Hi Takuma,

Looking at the provided link, I realized that there is a 3rd PSDK "PROCESSOR-SDK-LINUX-SK-TDA4VM Version: 08.01.00.02". Is it a complement or replacement to PROCESSOR-SDK-LINUX-J721E? Could you please explain the use case for both? I am interested in infering CNNs using booth TIOVX 100% on C7x (imported by TIDL) and the OSRT as heterogeneous execution (Onnx, TFLite, etc).

Thanks in advance

Gildardo

0 Takuma Fujiwara over 3 years ago in reply to Gildardo

TI__Guru 50193 points

Hi Gildardo,

The PROCESSOR-SDK-LINUX-SK-TDA4VM Version: 08.01.00.02 is a replacement for the PROCESSOR-SDK-LINUX-J721E. When using the set up script provided with the SK-TDA4VM version of PSDK Linux, it should download all the necessary dependencies for edge AI related tools and applications.

Could you try using PROCESSOR-SDK-LINUX-SK-TDA4VM instead of PROCESSOR-SDK-LINUX-J721E?

Regards,

Takuma

0 Takuma Fujiwara over 3 years ago in reply to Takuma Fujiwara

TI__Guru 50193 points

Hi Gildardo,

I had overlooked that we are using the EVM for edgeai-tidl-tools. For the EVM, we do not recommend customers to use any of the Edge AI components such as edgeai-tidl-tools, because we only provide testing and support on SK board for Edge AI.

I would recommend looking at our FAQ that explains the difference between SK and EVM boards. This FAQ also covers which SDK is supported for SK and EVM: https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1083612/faq-tda4vm-what-is-the-difference-between-sk-tda4vm-and-the-j721e-evm?tisearch=e2e-sitesearch&keymatch=difference%20sk%20evm#

Could you tell us what the final use case for your application is (e.g. surround view, auto valet parking, autonomous navigation, image classification, automotive, industrial robot, performance benchmarking for models, etc)? Depending on your use case we may be able to point you to a better platform, SDK, or demo.

Regards,

Takuma

Processors

Processors forum

TDA4VM: Building edgeai-tidl-tools CPP Examples