make lib wrong

[10183308]

the info is bellow

4 errors detected in the compilation of "/tmp/tmpxft_0000762e_00000000-6_deform_psroi_pooling_op_gpu.cu.cpp1.ii".

could you help me?

[FakerYFX]

you can follw this one:https://github.com/Zardinality/TF_Deformable_Net,just need to run the lib/make.sh

@10183308 I had teh same error ,How do I solve this problem? Thank you .
@xieyufei1993

tensorflow's version ==1.4.1

[FakerYFX]

follow this One https://github.com/Zardinality/TF_Deformable_Net

@xieyufei1993
[https://github.com/Zardinality/TF_Deformable_Net/issues/20]
10 errors detected in the compilation of "/tmp/tmpxft_00003e41_00000000-7_deform_conv.cu.cpp1.ii".
请问你是怎么解决的？我还是报这个错误，实在搞不懂。

[FakerYFX]

what is your gcc version and python version?

[10183308]

我也是遇到这个问题，没有解决掉。

@xieyufei1993
cuda 8.0 cuDNN 5.x
gcc version 4.9.2 20150212 (Red Hat 4.9.2-6) (GCC)
gcc version 5.3.1 20160406 (Red Hat 5.3.1-6) (GCC)
I am learning your code, but I am getting an error during compilation.
line 64 in make.sh:
nvcc -std=c++11 -c -o deform_psroi_pooling_op.cu.o deform_psroi_pooling_op_gpu.cu.cc
-I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC -D GOOGLE_CUDA -arch=$ARCH
If remove -D GOOGLE_CUDA=1 , the compilation can pass,But get errors when running , the file could not be found.
can't find deform_psroi_pooling_op.cu.o
undefined symbol: _ZTIN10tensorflow8OpKernelE
In short, the 64 lines are wrong, I don't know how to change.

[FakerYFX]

just follow this one CharlesShang/TFFRCNN#108

@xieyufei1993
Thank you for your advice, I have solved the previous problem ,mainly the latter error :
4 errors detected in the compilation of "/tmp/tmpxft_00005a43_00000000-7_deform_psroi_pooling_op_gpu.cu.cpp1.ii".
g++: error: deform_psroi_pooling_op.cu.o: No such file or directory

[lqyiii]

@ccnankai @10183308 have you solved this problem? i also meet this error...

[MrXieJ]

@ccnankai it bother me serveral days ,i finally solved it by myself ,only need to modify the make.sh file .have you solved the problem, i,d love to share my solution

[lqyiii]

i added this command (--expt-relaxed-constexpr) after which line has nvcc. the error disappear.
eg,
nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu.cc
-I $TF_INC -I $NSYNC_INC -D GOOGLE_CUDA=1 -L $CUDA_HOME/lib64 -x cu -Xcompiler -fPIC -D GOOGLE_CUDA -arch=$ARCH --expt-relaxed-constexpr

[aditipanda]

i added this command (--expt-relaxed-constexpr) after which line has nvcc. the error disappear.
eg,
nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu.cc
-I $TF_INC -I $NSYNC_INC -D GOOGLE_CUDA=1 -L $CUDA_HOME/lib64 -x cu -Xcompiler -fPIC -D GOOGLE_CUDA -arch=$ARCH --expt-relaxed-constexpr

@lqyiii Even after adding this command (--expt-relaxed-constexpr), at the end of every nvcc, I am still getting the following error:

[aditipanda]

@MrXieJ what changes did you do in make.sh?