Allow to pass optimization option to nvcc and document this.

doc/tutorial/using_gpu.txt

@@ -284,6 +284,14 @@ Tips for Improving Performance on GPU
   Check the line similar to *Spent Xs(X%) in cpu op, Xs(X%) in gpu op and Xs(X%) in transfer op*.
   This can tell you if not enough of your graph is on the GPU or if there
   is too much memory transfer.
+* Use nvcc options. nvcc support those options to speed up some
+  computations: `-ftz=true` to `flush denormals values to
+  zeros. <https://developer.nvidia.com/content/cuda-pro-tip-flush-denormals-confidence>`_,
+  `--prec-div=false` and `--prec-sqrt=false` option to speed up
+  division and square root operation by being less precise. You can
+  enable all of them with with the `nvcc.flags=--use_fast_math` Theano
+  flags or you can enable them individually as in this example
+  `nvcc.flags=-ftz=true --prec-div=false`.
 
 .. _gpu_async:
 

theano/sandbox/cuda/nvcc_compiler.py

@@ -255,10 +255,15 @@ class NVCC_compiler(object):
         # compute capability? '--gpu-architecture=compute_13',
         # '--gpu-code=compute_13',
         #nvcc argument
-        preargs1 = [pa for pa in preargs
-                    if pa.startswith('-O') or
-                    pa.startswith('--maxrregcount=') or
-                    pa.startswith('-arch=')]
+        preargs1 = []
+        for pa in preargs:
+            for pattern in ['-O', '-arch=',
+                            '--fmad', '--ftz', '--maxrregcount',
+                            '--prec-div', '--prec-sqrt',  '--use_fast_math',
+                            '-fmad', '-ftz', '-maxrregcount',
+                            '-prec-div', '-prec-sqrt', '-use_fast_math']:
+                if pa.startswith(pattern):
+                    preargs1.append(pa)
         preargs2 = [pa for pa in preargs
                     if pa not in preargs1]  # other arguments