[Compile] identifier "__shfl_down" is undefined

[JungHY]

i compiled CD, grouping, interpolation, sampling successfully

failed at emd

this is my tf_auctionmatch_compile.sh

#!/usr/bin/env bash
nvcc=/usr/local/cuda-8.0/bin/nvcc
cudalib=/usr/local/cuda-8.0/lib64/
TF_INC=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_include())')
TF_LIB=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_lib())')

$nvcc tf_auctionmatch_g.cu  -c -o tf_auctionmatch_g.cu.o -std=c++11  -I $TF_INC -DGOOGLE_CUDA=1\
 -x cu -Xcompiler -fPIC -O2

g++ tf_auctionmatch.cpp tf_auctionmatch_g.cu.o -o tf_auctionmatch_so.so -std=c++11 -shared -fPIC -I $TF_INC \
-I$TF_INC/external/nsync/public -L$TF_LIB -ltensorflow_framework -lcudart -L $cudalib -O2

error message

tf_auctionmatch_g.cu(220) : error : identifier "__shfl_down" is undefined

tf_auctionmatch_g.cu(243) : error : identifier "__shfl_down" is undefined

2 errors detected in the comilation of "/tmp/tmpxft_00003e79_00000000-7_tf_auctionmatch_g.cpp1.li

i did on cuda 8.0 and cuda 9.2

both same result

[ahmedwael19]

Can you let me know the commands that you used to compile them? As I am having a hard time doing so

[sunkaiand]

I have the same question as yours. Do you fix it at the end, can I have the specific solution.

[ahmedwael19]

I did not fix it.

I hope someone can help both of us as I need to run the code urgently.

[yulequan]

Hi, I do not have the same questions. After some research, it seems that changing the cuda and cudnn version would be a possible solution. Adding some parameters would also be possible. See MichalBusta/DeepTextSpotter#26

[sunkaiand]

Thanks, I will try

[mdane-x]

Same issue on CUDA 11.6.