Arch Linux

LieslW

Member

Registered: 2014-01-10

Posts: 2

Cuda compile error

error: kernel launches from templates are not allowed in system files
Similar to:
https://code.google.com/p/thrust/issues/detail?id=359
There are a few dozen versions of this question, but I couldn't find any that did not involve thrust.

I am trying to build caffe (http://caffe.berkeleyvision.org/installation.html) on Arch Linux. Which I haven't been using for very long. I installed cuda, and have run some samples, including another neural network library, convnet. These built and ran, and used the GPU correctly.

But I cannot get this one to build:

/usr/bin/g++ -pthread -fPIC -DNDEBUG -O2 -I/usr/local/include -I/usr/include/python2.7 -I/usr/local/lib/python2.7/dist-packages/numpy/core/include -I./src -I./include -I/home/lieslw/cuda/include -I/opt/intel/mkl/include   -c -o src/caffe/solver.o src/caffe/solver.cpp
/home/lieslw/cuda/bin/nvcc -ccbin=/usr/bin/g++ -Xcompiler -fPIC -DNDEBUG -O2 -I/usr/local/include -I/usr/include/python2.7 -I/usr/local/lib/python2.7/dist-packages/numpy/core/include -I./src -I./include -I/home/lieslw/cuda/include -I/opt/intel/mkl/include -arch=sm_30 -c src/caffe/layers/lrn_layer.cu -o src/caffe/layers/lrn_layer.cuo
/home/lieslw/cuda/bin/nvcc -ccbin=/usr/bin/g++ -Xcompiler -fPIC -DNDEBUG -O2 -I/usr/local/include -I/usr/include/python2.7 -I/usr/local/lib/python2.7/dist-packages/numpy/core/include -I./src -I./include -I/home/lieslw/cuda/include -I/opt/intel/mkl/include -arch=sm_30 -c src/caffe/layers/padding_layer.cu -o src/caffe/layers/padding_layer.cuo

/home/lieslw/cuda/bin/nvcc -ccbin=/usr/bin/g++ -Xcompiler -fPIC -DNDEBUG -O2 -I/usr/local/include -I/usr/include/python2.7 -I/usr/local/lib/python2.7/dist-packages/numpy/core/include -I./src -I./include -I/home/lieslw/cuda/include -I/opt/intel/mkl/include -arch=sm_30 -c src/caffe/layers/pooling_layer.cu -o src/caffe/layers/pooling_layer.cuo
src/caffe/layers/pooling_layer.cu(163): error: kernel launches from templates are not allowed in system files

src/caffe/layers/pooling_layer.cu(168): error: kernel launches from templates are not allowed in system files

src/caffe/layers/pooling_layer.cu(289): error: kernel launches from templates are not allowed in system files

src/caffe/layers/pooling_layer.cu(295): error: kernel launches from templates are not allowed in system files

src/caffe/layers/pooling_layer.cu(301): error: kernel launches from templates are not allowed in system files

5 errors detected in the compilation of "/tmp/tmpxft_0000094e_00000000-4_pooling_layer.cpp4.ii".
Makefile:130: recipe for target 'src/caffe/layers/pooling_layer.cuo' failed
make: *** [src/caffe/layers/pooling_layer.cuo] Error 2

So, does it think that my $HOME directory is a system file location? I investigated env, and removed my home directory from $PATH, but that didn't fix this issue.  I moved all of cuda into $HOME. I moved caffe out of it. Is there some way that I may have configured (or failed to configure) gcc or nvcc, something that wouldn't show up in the environment?

Interestingly, it fails in this way on three of the files and none of the others: dropout_layer.cu, softmax_layer.cu and pooling_layer.cu. Looking at the kernels in each I could not tell how they are different from the others.

Last edited by LieslW (2014-01-10 20:16:31)

LieslW

Member

Registered: 2014-01-10

Posts: 2

Re: Cuda compile error

Fixed it: installed gcc version 4.4 (4.6 is out of date). Used 4.4 to build my cuda code instead of 4.8.

Now I am past the cuda errors and working on mkl issues.

Last edited by LieslW (2014-01-11 01:41:05)