tests pass on 32bit

_test_tensor.py

@@ -536,14 +536,15 @@ DotTester = make_tester(name = 'DotTester',
 
 
 #useful mostly for unit tests
-def _approx_eq(a,b,eps=1.0e-9):
+def _approx_eq(a,b,eps=1.0e-4):
     a = numpy.asarray(a)
     b = numpy.asarray(b)
     if a.shape != b.shape:
         if _approx_eq.debug:
             print a.shape, b.shape
         return False
-    if numpy.max(numpy.abs(a-b)) >= eps:
+    abs_rel_err = numeric_grad.abs_rel_err(a,b)
+    if numpy.max(abs_rel_err) >= eps:
         if _approx_eq.debug:
             print a, b
         return False
@@ -861,7 +862,6 @@ class T_subtensor(unittest.TestCase):
         self.failUnless(tval.shape == ())
         self.failUnless(numpy.all(tval == 0))
 
-
     def test_grad_1d(self):
         n = as_tensor(numpy.random.rand(2,3))
         z = scal.constant(0)
@@ -872,7 +872,6 @@ class T_subtensor(unittest.TestCase):
         s1 = 'array([ 1.55009327,  0.        ,  0.        ])'
         self.failUnless(repr(gval[0,:]) == s0)
         self.failUnless(repr(gval[1,:]) == s1)
-
     def test_grad_0d(self):
         n = as_tensor(numpy.random.rand(2,3))
         t = n[1,0]
@@ -1521,7 +1520,8 @@ class t_gemm(unittest.TestCase):
 
             f = function([tz,ta,tx,ty,tb], gemm(tz,ta,tx,ty,tb), mode = compile.Mode(optimizer = None, linker=l))
             f(z, a, x, y, b)
-            self.failUnless(_approx_eq(z_after, z), (z_orig, z_after, z))
+            self.failUnless(_approx_eq(z_after, z), (z_orig, z_after, z, z_after - z))
+
             f(z.T, a, y.T, x.T, b)
             self.failUnless(_approx_eq(z_after, z))
 
@@ -1778,7 +1778,7 @@ if __name__ == '__main__':
     if 0:
         unittest.main()
     else:
-        testcase =  AbsInplaceTester
+        testcase =  T_Join_and_Split
 
         suite = unittest.TestLoader()
         suite = suite.loadTestsFromTestCase(testcase)

tensor.py

@@ -369,6 +369,7 @@ def _multi(*fns):
 fscalar = Tensor('float32', ())
 dscalar = Tensor('float64', ())
 bscalar = Tensor('int8', ())
+wscalar = Tensor('int16', ())
 iscalar = Tensor('int32', ())
 lscalar = Tensor('int64', ())
 def scalar(name = None, dtype = 'float64'):
@@ -376,9 +377,13 @@ def scalar(name = None, dtype = 'float64'):
     return type(name)
 scalars, fscalars, dscalars, iscalars, lscalars = _multi(scalar, fscalar, dscalar, iscalar, lscalar)
 
+int_types = bscalar, wscalar, iscalar, lscalar
+float_types = fscalar, dscalar
+
 fvector = Tensor('float32', (False, ))
 dvector = Tensor('float64', (False, ))
 bvector = Tensor('int8', (False,))
+wvector = Tensor('int16', (False,))
 ivector = Tensor('int32', (False, ))
 lvector = Tensor('int64', (False, ))
 def vector(name = None, dtype = 'float64'):
@@ -386,9 +391,13 @@ def vector(name = None, dtype = 'float64'):
     return type(name)
 vectors, fvectors, dvectors, ivectors, lvectors = _multi(vector, fvector, dvector, ivector, lvector)
 
+int_vector_types = bvector, wvector, ivector, lvector
+float_vector_types = fvector, dvector
+
 fmatrix = Tensor('float32', (False, False))
 dmatrix = Tensor('float64', (False, False))
 bmatrix = Tensor('int8', (False, False))
+wmatrix = Tensor('int16', (False, False))
 imatrix = Tensor('int32', (False, False))
 lmatrix = Tensor('int64', (False, False))
 def matrix(name = None, dtype = 'float64'):
@@ -396,9 +405,13 @@ def matrix(name = None, dtype = 'float64'):
     return type(name)
 matrices, fmatrices, dmatrices, imatrices, lmatrices = _multi(matrix, fmatrix, dmatrix, imatrix, lmatrix)
 
+int_matrix_types = bmatrix, wmatrix, imatrix, lmatrix
+float_matrix_types = fmatrix, dmatrix
+
 frow = Tensor('float32', (True, False))
 drow = Tensor('float64', (True, False))
 brow = Tensor('int8', (True, False))
+wrow = Tensor('int16', (True, False))
 irow = Tensor('int32', (True, False))
 lrow = Tensor('int64', (True, False))
 def row(name = None, dtype = 'float64'):
@@ -409,6 +422,7 @@ rows, frows, drows, irows, lrows = _multi(row, frow, drow, irow, lrow)
 fcol = Tensor('float32', (False, True))
 dcol = Tensor('float64', (False, True))
 bcol = Tensor('int8', (False, True))
+wcol = Tensor('int16', (False, True))
 icol = Tensor('int32', (False, True))
 lcol = Tensor('int64', (False, True))
 def col(name = None, dtype = 'float64'):
@@ -1375,9 +1389,9 @@ class Split(Op):
         axis = as_tensor(axis)
         splits = as_tensor(splits)
 
-        if splits.type != lvector: 
+        if splits.type not in int_vector_types: 
             raise TypeError('splits must have type tensor.lvector', splits.type)
-        if axis.type != lscalar: 
+        if axis.type not in int_types: 
             raise TypeError('axis must have type lscalar', axis.type)
 
         inputs = [x, axis, splits]
@@ -1464,8 +1478,8 @@ class Join(Op):
             bcastable[axis] = False
 
         inputs = [as_tensor(axis)] + as_tensor_args
-        if inputs[0].type != lscalar: 
-            raise TypeError('Axis could not be cast to lscalar', axis)
+        if inputs[0].type not in int_types: 
+            raise TypeError('Axis could not be cast to an integer type', axis, inputs[0].type, int_types)
 
         outputs = [tensor(dtype = dtypes[0],
                           broadcastable = bcastable)]
@@ -1506,6 +1520,7 @@ class Join(Op):
                 for k in range(len(sizes_along_axis))]
 
     def vec_length(self, node):
+        """Guess the length of a Join Result"""
         assert isinstance(node.owner.op, Join)
         if node.ndim != 1:
             raise TypeError('argument must be symbolic vector')
@@ -1515,6 +1530,7 @@ class Join(Op):
         # the question is whether all the inputs are broadcastable.
         if all(i.broadcastable[0] for i in tensors):
             return len(tensors)
+        raise ValueError("could not determine vector length")
 
 @_redefine_asRoutine(Join())
 def join(axis, *tensors):