Improved _cast_to_d2o when flatted d2o is inserted.

Made the steepest descent work.

demos/demo_wf2.py

@@ -32,6 +32,23 @@
 
 """
 from __future__ import division
+
+#from pycallgraph import PyCallGraph
+#from pycallgraph import Config
+#from pycallgraph import GlobbingFilter
+#from pycallgraph.output import GraphvizOutput
+#
+#config = Config()
+#config.trace_filter = GlobbingFilter(exclude=[
+#    'pycallgraph.*',
+#    #'*.secret_function',
+#])
+#
+#graphviz = GraphvizOutput(output_file='steepest_profiling.png')
+#
+#
+
+
 from nifty import *                                              # version 0.8.0
 
 
@@ -43,10 +60,10 @@ k_space = x_space.get_codomain()                                 # get conjugate
 # some power spectrum
 power = (lambda k: 42 / (k + 1) ** 3)
 
-S = power_operator(k_space, spec=power)                          # define signal covariance
-s = S.get_random_field(domain=x_space)                           # generate signal
+S = power_operator(k_space, codomain=x_space, spec=power)                          # define signal covariance
+s = S.get_random_field(domain=x_space, codomain=k_space)                           # generate signal
 
-R = response_operator(x_space, sigma=0.0, mask=1.0, assign=None) # define response
+R = response_operator(x_space, codomain=k_space, sigma=0.0, mask=1.0, assign=None) # define response
 d_space = R.target                                               # get data space
 
 # some noise variance; e.g., signal-to-noise ratio of 1
@@ -70,11 +87,13 @@ def eggs(x):
     return H, g
 
 
-m = field(x_space, target=k_space)                               # reconstruct map
+m = field(x_space, codomain=k_space)                               # reconstruct map
+
+#with PyCallGraph(output=graphviz, config=config):
 m, convergence = steepest_descent(eggs=eggs, note=True)(m, tol=1E-3, clevel=3)
 
-s.plot(title="signal")                                           # plot signal
-d_ = field(x_space, val=d.val, target=k_space)
-d_.plot(title="data", vmin=s.min(), vmax=s.max())                # plot data
-m.plot(title="reconstructed map", vmin=s.min(), vmax=s.max())    # plot map
+#s.plot(title="signal")                                           # plot signal
+#d_ = field(x_space, val=d.val, target=k_space)
+#d_.plot(title="data", vmin=s.min(), vmax=s.max())                # plot data
+#m.plot(title="reconstructed map", vmin=s.min(), vmax=s.max())    # plot map
 

nifty_core.py

@@ -1167,13 +1167,17 @@ class point_space(space):
             if np.any(np.array(x.shape) != np.array(self.get_shape())):
                 # Check if at least the number of degrees of freedom is equal
                 if x.get_dim() == self.get_dim():
-                    # If the number of dof is equal or 1, use np.reshape...
-                    about.warnings.cflush(
-                        "WARNING: Trying to reshape the data. This " +
-                        "operation is expensive as it consolidates the " +
-                        "full data!\n")
-                    temp = x.get_full_data()
-                    temp = np.reshape(temp, self.get_shape())
+                    try:
+                        temp = x.copy_empty(global_shape=self.get_shape())
+                        temp.set_local_data(x.get_local_data(), copy=False)
+                    except:
+                        # If the number of dof is equal or 1, use np.reshape...
+                        about.warnings.cflush(
+                            "WARNING: Trying to reshape the data. This " +
+                            "operation is expensive as it consolidates the " +
+                            "full data!\n")
+                        temp = x.get_full_data()
+                        temp = np.reshape(temp, self.get_shape())
                     # ... and cast again
                     return self._cast_to_d2o(temp,
                                              dtype=dtype,

nifty_mpi_data.py

@@ -773,7 +773,7 @@ class distributed_data_object(object):
         """
         self.hermitian = hermitian
         if copy is True:
-            self.data[:] = data
+            self.data[:] = data.reshape(self.local_shape)
         else:
             self.data = np.array(data,
                                  dtype=self.dtype,

operators/nifty_minimization.py

@@ -581,7 +581,8 @@ class steepest_descent(object):
                 self.note.cprint("\niteration : %08u   alpha < 1.0E-13\n... dead."%ii)
                 break
             else:
-                delta = np.absolute(g.val).max()*(alpha/norm)
+                delta = abs(g).max()*(alpha/norm)
+                #delta = np.absolute(g.val).max()*(alpha/norm)
                 self.note.cflush("\niteration : %08u   alpha = %3.1E   delta = %3.1E"%(ii,alpha,delta))
                 ## update
                 self.x = x_