better error msg and comments for gpusoftmax

theano/sandbox/cuda/nnet.py

@@ -405,7 +405,11 @@ class GpuSoftmax (GpuOp):
             cudaError_t err = cudaGetLastError();
             if( cudaSuccess != err)
             {
-                PyErr_Format(PyExc_RuntimeError, "Cuda error: %%s: %%s.\\n", "kSoftmax_%(nodename)s", cudaGetErrorString(err));
+                PyErr_Format(PyExc_RuntimeError,
+                             "Cuda error: %%s: %%s.\\n Used %%d blocks,"
+                             " %%d threads %%d bytes of shared memory",
+                             "kSoftmax_%(nodename)s", cudaGetErrorString(err),
+                             n_blocks, n_threads, n_shared_bytes);
                 %(fail)s;
             }
         }

theano/sandbox/cuda/tests/test_nnet.py

@@ -227,8 +227,8 @@ def test_softmax():
                       cuda.nnet.GpuSoftmax)
 
     def cmp(n, m, catch=False):
-        """Some old card won't accet the configuration arguments of
-        this implementation."""
+        """Some old card won't accept the configuration arguments of
+        this implementation. For those cases set catch=True to skip those errors."""
         try:
             #print "test_softmax",n,m
             data = numpy.arange(n * m, dtype='float32').reshape(n, m)