fix compilation error
This commit is contained in:
@@ -124,7 +124,7 @@ def get_plugin(cuda_file):
|
||||
compile_opts += '"%s"' % os.path.join(tf.sysconfig.get_lib(), 'python', '_pywrap_tensorflow_internal.lib')
|
||||
elif os.name == 'posix':
|
||||
compile_opts += '"%s"' % os.path.join(tf.sysconfig.get_lib(), 'python', '_pywrap_tensorflow_internal.so')
|
||||
compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=0\''
|
||||
compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=1\''
|
||||
else:
|
||||
assert False # not Windows or Linux, w00t?
|
||||
compile_opts += ' --gpu-architecture=%s' % _get_cuda_gpu_arch_string()
|
||||
|
||||
Reference in New Issue
Block a user