Error when compiling utensor-mnist-demo

[drajmund]

I followed the instruction, and tried to compile the utensor-mnist-demo with the command:
mbed compile -m DISCO_F413ZH -t GCC_ARM --profile=uTensor/build_profile/develop.json

I received the following message with an error:

Building project utensor-mnist-demo (DISCO_F413ZH, GCC_ARM)
Scan: .
Scan: mbed
Scan: env
Compile [ 4.9%]: main.cpp
[Error] image.h@7,50: expected 'class' before 'TENSOR'
[ERROR] In file included from ./main.cpp:6:0:
./image.h:7:50: error: expected 'class' before 'TENSOR'
template<typename T, template typename TENSOR=RamTensor>
^
[mbed] ERROR: "/usr/bin/python" returned error.
Code: 1
Path: "/home/drajmund/utensor-mnist-demo"
Command: "/usr/bin/python -u /home/drajmund/utensor-mnist-demo/mbed-os/tools/make.py -t GCC_ARM -m DISCO_F413ZH --profile uTensor/build_profile/release.jso
n --source . --build ./BUILD/DISCO_F413ZH/GCC_ARM-RELEASE"
Tip: You could retry the last command with "-v" flag for verbose output

Do you have any idea what is wrong with it?
System parameters:
Ubuntu 16.04.5 LTS
Python 2.7.12
arm-none-eabi-gcc (15:4.9.3+svn231177-1) 4.9.3 20150529 (prerelease)
mbed-cli version number: 1.8.2

[neil-tan]

Hey @drajmund
You might want to update your gcc to version 5 or 6 as described in the mbed-cli doc

Once you got it working, you might also be interested in this branch which does not require an SD card. You can do this by using git checkout f/rom-tensor, mbed deploy follow by a clean compile.

Please let us know how it goes!

[dboyliao]

BTW, we haven't fix the include order yet.
That is, context.hpp must be included first before other header files.
I know I can fix it in the cli but I don't think it's a good approach and leaved it as is then.

[dboyliao]

Please refer to #65 for detail.

[drajmund]

Thank you for your help.

I downloaded the new version of gcc-arm-none-eabi and configured mbed as follows:
$ mbed config -G GCC_ARM_PATH /usr/local/gcc-arm-none-eabi-6-2018-q2-update/bin

I added the support for BSP_DISCO_F413ZH:
$ mbed add https://os.mbed.com/teams/ST/code/BSP_DISCO_F413ZH/
$ mbed deploy

Compilation is OK, but I receive "multiple definition of" errors from the linker:

$ sudo mbed compile -m DISCO_F413ZH -t GCC_ARM --profile=uTensor/build_profile/release.json -c
Building project utensor-mnist-demo (DISCO_F413ZH, GCC_ARM)
Scan: utensor-mnist-demo
Scan: env
Compile [ 0.1%]: ft6x06.c
Compile [ 0.3%]: st7789h2.c
Compile [ 0.4%]: wm8994.c
...
Compile [ 99.7%]: tensorIdxImporter.cpp
Compile [ 99.9%]: quantization_utils.cpp
Compile [100.0%]: uTensor_util.cpp
Link: utensor-mnist-demo
BUILD/DISCO_F413ZH/GCC_ARM-RELEASE/F413ZH_SD_BlockDevice/BSP_DISCO_F413ZH/Drivers/BSP/Components/ft6x06/ft6x06.o: In function ft6x06_TS_ITStatus': ft6x06.c:(.text.ft6x06_TS_ITStatus+0x0): multiple definition of ft6x06_TS_ITStatus'
BUILD/DISCO_F413ZH/GCC_ARM-RELEASE/BSP_DISCO_F413ZH/Drivers/BSP/Components/ft6x06/ft6x06.o:ft6x06.c:(.text.ft6x06_TS_ITStatus+0x0): first defined here
...
BUILD/DISCO_F413ZH/GCC_ARM-RELEASE/F413ZH_SD_BlockDevice/BSP_DISCO_F413ZH/Utilities/Fonts/font8.o:(.rodata.Font8_Table+0x0): multiple definition of `Font8_Table'
BUILD/DISCO_F413ZH/GCC_ARM-RELEASE/BSP_DISCO_F413ZH/Utilities/Fonts/font8.o:(.rodata.Font8_Table+0x0): first defined here
collect2: error: ld returned 1 exit status

[mbed] ERROR: "/usr/bin/python" returned error.
Code: 1
Path: "/home/drajmund/utensor-mnist-demo"
Command: "/usr/bin/python -u /home/drajmund/utensor-mnist-demo/mbed-os/tools/make.py -t GCC_ARM -m DISCO_F413ZH --profile uTensor/build_profile/release.json --source . --build ./BUILD/DISCO_F413ZH/GCC_ARM-RELEASE -c"
Tip: You could retry the last command with "-v" flag for verbose output

[neil-tan]

Hey @drajmund

You shouldn't need to manually add the BSP_DISCO_F413ZH. Is there an error/warning led you to do so?

It would be useful to see the error after this line:
(.text.ft6x06_TS_ITStatus+0x0): first defined here
Also, the output of this comment at your current project root:
tree -L 4 -I '*.idx|mbed-os'

In the meanwhile, can we try a fresh import of the project? Here are the commands I used to get it going:

mbed import https://github.com/uTensor/utensor-mnist-demo
cd utensor-mnist-demo/
mbed compile -m DISCO_F413ZH -t GCC_ARM --profile=uTensor/build_profile/release.json

To use the SD-free version:

mbed import https://github.com/uTensor/utensor-mnist-demo
cd utensor-mnist-demo/
git checkout f/rom-tensor
mbed deploy
mbed compile -m DISCO_F413ZH -t GCC_ARM --profile=uTensor/build_profile/release.json -c

[drajmund]

Oh, the problem was exactly what you mentioned: I manually added BSP_DISCO_F413ZH. (Root cause may be the copy of BSP_DISCO_F413ZH.lib two times?)
After imorting the project and compiling it again I received this :-)
...
Compile [100.0%]: test_env.cpp
Link: utensor-mnist-demo
Elf2Bin: utensor-mnist-demo
...
Total Static RAM memory (data + bss): 19712 bytes
Total Flash memory (text + data): 357041 bytes
Image: ./BUILD/DISCO_F413ZH/GCC_ARM-RELEASE/utensor-mnist-demo.bin

Thank you for your help. I think I have learned a lot from this.
The tree command output was (before reimporting the project):
.
├── BSP_DISCO_F413ZH
│ ├── Drivers
│ │ └── BSP
│ │ ├── Components
│ │ └── STM32F413H-Discovery
│ └── Utilities
│ └── Fonts
│ ├── font12.c
│ ├── font16.c
│ ├── font20.c
│ ├── font24.c
│ ├── font8.c
│ └── fonts.h
├── BSP_DISCO_F413ZH.lib
├── BUILD
│ └── DISCO_F413ZH
│ ├── ARM-RELEASE
│ │ ├── BSP_DISCO_F413ZH
│ │ ├── F413ZH_SD_BlockDevice
│ │ ├── mbed_config.h
│ │ ├── models
│ │ └── uTensor
│ └── GCC_ARM-RELEASE
│ ├── BSP_DISCO_F413ZH
│ ├── F413ZH_SD_BlockDevice
│ ├── main.d
│ ├── main.o
│ ├── mbed_config.h
│ ├── models
│ ├── uTensor
│ └── utensor-mnist-demo.map
├── constants
│ └── deep_mlp
├── docs
│ └── images
│ ├── jupyter.png
│ ├── kernel.png
│ └── train_mlp.png
├── F413ZH_SD_BlockDevice
│ ├── BSP_DISCO_F413ZH
│ │ ├── Drivers
│ │ │ └── BSP
│ │ └── Utilities
│ │ └── Fonts
│ ├── BSP_DISCO_F413ZH.lib
│ ├── F413ZH_SD_BlockDevice.cpp
│ ├── F413ZH_SD_BlockDevice.h
│ └── README.md
├── F413ZH_SD_BlockDevice.lib
├── image.h
├── LICENSE
├── main.cpp
├── mbed_app.json
├── mbed-os.lib
├── mbed_settings.py
├── mbed_settings.pyc
├── models
│ ├── deep_mlp.cpp
│ ├── deep_mlp.hpp
│ └── deep_mlp_weight.hpp
├── README.md
├── requirements.txt
├── simconfig.json
├── tensorflow-models
│ ├── deep_mlp.ipynb
│ └── deep_mlp.py
├── uTensor
│ │ └── release.json
│ ├── CMakeLists.txt
│ ├── contribution_guide.md
│ ├── deprecated
│ │ └── TESTS
│ │ ├── ArrayTests.hpp
│ │ ├── context_test.hpp
│ │ ├── main.cpp
│ │ ├── MathTests.hpp
│ │ ├── MatrixTests.hpp
│ │ ├── mlp_test.hpp
│ │ ├── NnTests.hpp
│ │ ├── scripts
│ │ ├── sdtest.hpp
│ │ ├── tensorIdxImporterTests.hpp
│ │ ├── TensorImport
│ │ ├── tensor_test.hpp
│ │ ├── test.cpp
│ │ ├── test.hpp
│ │ └── vmtest.hpp
│ ├── docs
│ │ ├── img
│ │ │ └── uTensorFlow.jpg
│ │ └── newProject.md
│ ├── examples
│ │ ├── deep_mnist_mlp.cpp
│ │ └── deep_mnist_mlp.hpp
│ ├── LICENSE
│ ├── mbed_app.json
│ ├── mbed_lib.json
│ ├── README.md
│ ├── reference
│ │ ├── models
│ │ │ ├── miscOps
│ │ │ └── qntMatMulDeqnt
│ │ ├── README.md
│ │ └── tf_src
│ │ ├── CXX
│ │ └── python
│ ├── REFERENCES.md
│ ├── requirements.txt
│ ├── TESTS
│ │ ├── constants
│ │ │ ├── ArgMax
│ │ │ ├── deep_mlp
│ │ │ ├── deQ
│ │ │ ├── idxImport
│ │ │ ├── import-MatMul_eightbit_requantize
│ │ │ ├── mlpTest
│ │ │ ├── qA
│ │ │ ├── qB
│ │ │ ├── qMatMul
│ │ │ ├── ref_add
│ │ │ ├── ref_max
│ │ │ ├── ref_min
│ │ │ ├── ref_qRelu
│ │ │ ├── ref_reshape
│ │ │ ├── rQ
│ │ │ ├── rqRange
│ │ │ ├── sdtmp
│ │ │ ├── test
│ │ │ └── test_quantized_add
│ │ ├── core
│ │ │ ├── context
│ │ │ ├── tensor
│ │ │ ├── tensorIdxImporter
│ │ │ └── virtualmem
│ │ ├── dummy
│ │ │ └── sanity
│ │ ├── gtest_main.cpp
│ │ ├── operators
│ │ │ ├── arrayOps
│ │ │ ├── basicMath
│ │ │ ├── matrix
│ │ │ └── quantizedAdd
│ │ ├── README.md
│ │ └── test_helper.h
│ ├── unity
│ │ └── CMakeLists.txt
│ └── uTensor
│ ├── core
│ │ ├── context.cpp
│ │ ├── context.hpp
│ │ ├── sdtensor.hpp
│ │ ├── tensor.cpp
│ │ ├── tensor.hpp
│ │ ├── uTensorBase.cpp
│ │ ├── uTensorBase.hpp
│ │ ├── vm.cpp
│ │ └── vm.hpp
│ ├── loaders
│ │ ├── tensorIdxImporter.cpp
│ │ └── tensorIdxImporter.hpp
│ ├── ops
│ │ ├── ArrayOps.hpp
│ │ ├── MathOps.hpp
│ │ ├── MatrixOps.hpp
│ │ └── NnOps.hpp
│ └── util
│ ├── math_utils.hpp
│ ├── quantization_utils.cpp
│ ├── quantization_utils.hpp
│ ├── uTensor_util.cpp
│ └── uTensor_util.hpp
└── uTensor.lib

[neil-tan]

@drajmund Np : ). I highly recommend using the f/rom-tensor branch. We are looking to merged it into master during the next release.