Error building faster-rnnlm

[anavc94]

Hello,

I am having some issues installing faster-RNNLM in Ubuntu 16.04. As the instrucction suggests, I run the script build.sh but the output says:

Installing Faster RNNLM
Already up-to-date.
Folder 'eigen3' already exists. Exiting
nvcc cuda_softmax.cu -c -Xcompiler "-O3 -march=native -funroll-loops" -o cuda_softmax.o
/usr/include/string.h: In function ‘void* __mempcpy_inline(void*, const void*, size_t)’:
/usr/include/string.h:652:42: error: ‘memcpy’ was not declared in this scope
return (char *) memcpy (__dest, __src, __n) + __n;
^
Makefile:60: recipe for target 'cuda_softmax.o' failed
make: *** [cuda_softmax.o] Error 1

I've been reading something about the compiler, so my version is:

gcc --version

gcc (Ubuntu 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609
Copyright (C) 2015 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

I've installed CUDA toolkit 9.0 (with the deb local option because the .run was giving me a lot of problems).
Could you please help me?

Thanks!

Edit: I've tried to compile with gcc-4.8 and g++-4.8 but the same error appears on screen.

[akhti]

Try to add -D_FORCE_INLINES in NVCC_FLAGS in Makefile:
https://github.com/yandex/faster-rnnlm/blob/master/faster-rnnlm/Makefile#L7

NVCC_CFLAGS = -D_FORCE_INLINES -O3 -march=native -funroll-loops