Merge pull request #3895 from nouiz/debug

doc/tutorial/extending_theano_c.txt

@@ -983,6 +983,27 @@ macros defined by ``COp``. Finally, the function
 macros and also because it calls ``vector_elemwise_mult()`` which is
 an apply-specific function.
 
+
+Using GDB to debug Op's C code
+==============================
+
+When debugging C code, it can be useful to use GDB for code compiled
+by Theano.
+
+For this, you must enable this Theano: `cmodule.remove_gxx_opt=True`.
+For the GPU, you must add in this second flag `nvcc.flags=-g` (it slow
+down computation on the GPU, but it is enabled by default on the CPU).
+
+Then you must start Python inside GDB and in it start your Python
+process (e.g. theano-nose):
+
+.. code-block::
+
+    $gdb python
+    (gdb)r bin/theano-nose theano/
+
+`Quick guide to GDB <https://www.cs.cmu.edu/~gilpin/tutorial/>`_.
+
 Final Note
 ==========
 

theano/sandbox/cuda/__init__.py

@@ -194,15 +194,16 @@ if compile_cuda_ndarray and cuda_available:
                         tmpdir = os.environ['TMPDIR']
                         if not os.path.exists(tmpdir):
                             os.makedirs(tmpdir)
-
                     compiler = nvcc_compiler.NVCC_compiler()
+                    preargs = ['-O3'] + compiler.compile_args()
+                    preargs += [f for f in config.nvcc.flags.split(' ') if f]
                     compiler.compile_str(
                             'cuda_ndarray',
                             code,
                             location=cuda_ndarray_loc,
                             include_dirs=[cuda_path],
                             libs=[config.cublas.lib],
-                            preargs=['-O3'] + compiler.compile_args(),
+                            preargs=preargs,
                     )
                     from cuda_ndarray.cuda_ndarray import *
             except Exception as e:

theano/sandbox/cuda/nvcc_compiler.py

@@ -203,6 +203,9 @@ class NVCC_compiler(Compiler):
             preargs = list(preargs)
         if sys.platform != 'win32':
             preargs.append('-fPIC')
+        if config.cmodule.remove_gxx_opt:
+            preargs = [p for p in preargs if not p.startswith('-O')]
+
         cuda_root = config.cuda.root
 
         # The include dirs gived by the user should have precedence over