merge

theano/sandbox/cuda/__init__.py

@@ -21,8 +21,8 @@ def debug(*msg):
 # printed and this module will not be working properly (we set `cuda_available`
 # to False).
 
-# This variable is True by default, and set to False if something goes wrong
-# when trying to initialize cuda.
+# This variable is True by default, and set to False if nvcc is not available or
+# their is no cuda card or something goes wrong when trying to initialize cuda.
 cuda_available = True
 
 # Global variable to avoid displaying the same warning multiple times.
@@ -89,6 +89,9 @@ except Exception, e:
     error( "Failed to compile cuda_ndarray.cu: %s" % str(e))
     set_cuda_disabled()
 
+if cuda_available:
+    cuda_available=device_available()
+
 if cuda_available:
     #check if their is an old cuda_ndarray that was loading instead of the one we compiled!
     import cuda_ndarray.cuda_ndarray

theano/sandbox/cuda/cuda_ndarray.cu

@@ -1604,6 +1604,35 @@ static PyTypeObject CudaNdarrayType =
     CudaNdarray_new,           /* tp_new */
 };
 
+
+//This fct return True it is able to find a cuda card and query its properti
+//Otherwise we return False
+PyObject *
+device_available(PyObject* _unsed, PyObject * args)
+{
+  int deviceCount;
+
+  cudaError err = cudaGetDeviceCount(&deviceCount);
+  if( cudaSuccess != err) {
+    Py_RETURN_FALSE;
+  }
+  if (deviceCount <= 0) {
+    Py_RETURN_FALSE;
+  }
+
+  cudaDeviceProp deviceProp;
+  err=cudaGetDeviceProperties(&deviceProp, 0);
+  if( cudaSuccess != err) {
+    Py_RETURN_FALSE;
+  }
+
+  if(deviceProp.major == 9999 && deviceProp.minor == 9999 ){
+    Py_RETURN_FALSE;
+  }
+  
+  Py_RETURN_TRUE;
+}
+
 PyObject *
 CudaNdarray_gpu_init(PyObject* _unsed, PyObject * args)
 {
@@ -1810,6 +1839,7 @@ filter(PyObject* __unsed_self, PyObject *args) // args = (data, broadcastable, s
 
 static PyMethodDef module_methods[] = {
     {"dot", CudaNdarray_Dot, METH_VARARGS, "Returns the matrix product of two CudaNdarray arguments."},
+    {"device_available", device_available, METH_VARARGS, "Return Py_True if a cuda card is available."},
     {"gpu_init", CudaNdarray_gpu_init, METH_VARARGS, "Allow to select the gpu card to use."},
     {"filter", filter, METH_VARARGS, "filter(obj, broadcastable, strict, storage) returns a CudaNdarray initialized to obj if it matches the constraints of broadcastable.  strict=True prevents any numeric casting. If storage is a CudaNdarray it may be overwritten and used as the return value."},    
     {"outstanding_mallocs", outstanding_mallocs, METH_VARARGS, "how many more mallocs have been called than free's"},

theano/sandbox/test_rng_mrg.py

@@ -22,7 +22,6 @@ from theano.tests import unittest_tools as utt
 #TODO: make tests work when no flags gived. Now need: THEANO_FLAGS=device=gpu0,floatX=float32
 # Partly done, in test_consistency_GPU_{serial,parallel}
 
-#TODO: bug fix test_normal0, in normal() fct, n_samples currently need to be numpy.prod(size) not self.n_streams(size)
 
 mode = config.mode
 mode_with_gpu = theano.compile.mode.get_default_mode().including('gpu')
@@ -324,7 +323,7 @@ def test_uniform():
         sample_size = (10,100)
         steps = 50
     else:
-        sample_size = (500,100)
+        sample_size = (500,50)
         steps = int(1e3)
     
     x = tensor.matrix()
@@ -383,7 +382,7 @@ def test_binomial():
         sample_size = (10,50)
         steps = 70
     else:
-        sample_size = (500,100)
+        sample_size = (500,50)
         steps = int(1e3)
     
     x = tensor.matrix()
@@ -430,9 +429,9 @@ def test_normal0():
 
     steps = 50
     if mode in ['DEBUG_MODE','FAST_COMPILE']:
-        sample_size = (99,100)
+        sample_size = (99,30)
     else:
-        sample_size = (999,100)
+        sample_size = (999,50)
 
     print ''
     print 'ON CPU:'