improve doc (advection)

docs/sphinx/users_guide/advection.rst

@@ -12,30 +12,42 @@ Advection of one or more fields, i.e. find field X which solves
 
 for a given velocity field and assuming incompressible flow.
 
+All solvers (except pure python) are able to solve advection in a
+bi-level configuration : velocity is known on a coarser grid than the
+advected field. In that case, the velocity is linearly interpolated
+before move particles. Linear interpolation is a default method and
+the only one implemented for the moment.
 
-Multi-CPU (scales) advection
-----------------------------
 
-Scales :
+Scales (Multi-CPU) :
+--------------------
+
+Scales library is iterfaced into HySoP (using :class:`~hysop.operator.advection.Advection` operator).
 
 * Fortran routines
 * 3D only
 
 Available solvers:
 ^^^^^^^^^^^^^^^^^^
+
 * 'p_O2' : order 4 method, corrected to allow large CFL number,
   untagged particles
 * 'p_O4' : order 4 method, corrected to allow large CFL number,
   untagged particles
-* 'p_L2' : limited and corrected lambda 2
-* 'p_M4' : Lambda_2,1 (=M'4) 4 point formula
-* 'p_M6' (default) : Lambda_4,2 (=M'6) 6 point formula
-* 'p_M8' : M8prime formula
-* 'p_44' : Lambda_4,4 formula
-* 'p_64' : Lambda_6,4 formula
-* 'p_66' : Lambda_6,6 formula
-* 'p_84' : Lambda_8,4 formula
+* 'p_L2' : limited and corrected :math:`\Lambda_2`
+* 'p_M4' : :math:`\Lambda_{2,1} (=M'_4)` 4 point formula
+* 'p_M6' (default) : :math:`\Lambda_{4,2} (=M'_6)` 6 point formula
+* 'p_M8' : :math:`M8prime` formula
+* 'p_44' : :math:`\Lambda_{4,4}` formula
+* 'p_64' : :math:`\Lambda_{6,4}` formula
+* 'p_66' : :math:`\Lambda_{6,6}` formula
+* 'p_84' : :math:`\Lambda_{8,4}` formula
+
 
+Time integration:
+^^^^^^^^^^^^^^^^^
+
+* Runge-Kutta 2nd order
 
 
 Splitting
@@ -44,7 +56,7 @@ Splitting
 Computations are performed with a dimensional splitting as follows:
 
 * 'strang' (2nd order):
-  
+
   * X-dir, half time step
   * Y-dir, half time step
   * Z-dir, full time step
@@ -52,6 +64,36 @@ Computations are performed with a dimensional splitting as follows:
   * X-dir, half time step
 
 * 'classic' (1st order):
+
   * X-dir, full time step
   * Y-dir, full time step
   * Z-dir, full time step
+
+
+HySoP (Multi-CPU an Multi-GPU) :
+--------------------------------
+
+* 1D, 2D or 3D advection
+* Pure python code
+* Generated OpenCL code
+* Ghosts layers are computed from a maximum CFL number
+
+
+Available solvers:
+^^^^^^^^^^^^^^^^^^
+
+* All :math:`\Lambda_{p,q}` kernels are computed
+
+Time integration:
+^^^^^^^^^^^^^^^^^
+
+Explicit time intergration solvers:
+
+* Euler
+* Runge-Kutta (orders 1, 2, 3, 4, 4_38)
+
+
+Splitting
+^^^^^^^^^
+
+Splitting is handled as a higer level from HySoP :class:`~hysop.numerics.splitting.directional_splitting.DirectionalSplitting` operators.

hysop/backend/device/opencl/operator/directional/advection_dir.py

@@ -95,7 +95,7 @@ class OpenClDirectionalAdvection(DirectionalAdvectionBase, OpenClDirectionalOper
         kwds = {}
         kwds['velocity'] = self.dvelocity
         kwds['position'] = self.dposition
-        kwds['is_multiscale'] = self.is_multiscale
+        kwds['is_bilevel'] = self.is_bilevel
 
         kwds['direction']       = self.splitting_direction
         kwds['velocity_cfl']    = self.velocity_cfl

hysop/operator/advection.py

@@ -16,6 +16,45 @@ class Advection(ComputationalGraphNodeFrontend):
     """
     Interface the Scales fortran advection solver.
     Available implementations are: Fortran
+
+    Available remeshing formulas:
+
+    * 'p_O2' : order 4 method, corrected to allow large CFL number,
+      untagged particles
+    * 'p_O4' : order 4 method, corrected to allow large CFL number,
+      untagged particles
+    * 'p_L2' : limited and corrected lambda 2
+    * 'p_M4' : Lambda_2,1 (=M'4) 4 point formula
+    * 'p_M6' (default) : Lambda_4,2 (=M'6) 6 point formula
+    * 'p_M8' : M8prime formula
+    * 'p_44' : Lambda_4,4 formula
+    * 'p_64' : Lambda_6,4 formula
+    * 'p_66' : Lambda_6,6 formula
+    * 'p_84' : Lambda_8,4 formula
+
+    Time integration:
+
+    * Runge-Kutta 2nd order
+
+    Splitting:
+
+    Computations are performed with a dimensional splitting as follows:
+
+    * 'strang' (2nd order):
+
+      * X-dir, half time step
+      * Y-dir, half time step
+      * Z-dir, full time step
+      * Y-dir, half time step
+      * X-dir, half time step
+
+    * 'classic' (1st order):
+
+      * X-dir, full time step
+      * Y-dir, full time step
+      * Z-dir, full time step
+
+
     """
 
     @classmethod