Skip to content
Snippets Groups Projects
Commit 26567646 authored by Christophe Geuzaine's avatar Christophe Geuzaine
Browse files

pp

parent 5b175449
No related branches found
No related tags found
No related merge requests found
// $Id: ColorTable.cpp,v 1.1 2007-07-09 13:54:37 geuzaine Exp $
// $Id: ColorTable.cpp,v 1.2 2007-09-13 06:31:37 geuzaine Exp $
//
// Copyright (C) 1997-2007 C. Geuzaine, J.-F. Remacle
//
......
......@@ -35,7 +35,7 @@ typedef struct{
#define COLORTABLE_RGB 1
#define COLORTABLE_HSV 2
// integrer parameters indices
// integer parameters indices
#define COLORTABLE_NUMBER 0 // predefined curve index
#define COLORTABLE_INVERT 1 // invert (rbg<->255-rgb)
......
// $Id: PViewDataGModel.cpp,v 1.3 2007-09-08 21:26:05 geuzaine Exp $
// $Id: PViewDataGModel.cpp,v 1.4 2007-09-13 06:31:37 geuzaine Exp $
//
// Copyright (C) 1997-2007 C. Geuzaine, J.-F. Remacle
//
......@@ -24,6 +24,49 @@
#include "PViewDataGModel.h"
/*
add new optional field in .msh file to store
- node or element post-pro data
- one field per-time step, per view
$Data
name precision-single-double step time-value
type node-or-ele-id num-comp val (num-comp times)
type node-or-ele-id num-comp val (num-comp times)
...
$EndData
Then?
1) in MVertex and MElement, add
std::vector<std::vector<double> > _scalar, _vector, _tensor;
and access using e.g. _scalar[viewIndex][comp & time step]. In GModel
we would store a map with the correspondance between the index and the
unique view id created in PView().
2) Or, we store directly a map indexed with the unique view id
1) and 2) are pretty expensive, due to the overhead and the "micro"
allocations (in each element/node)
3) It might be better to store the data *per entity*?
4) Or just store the big vectors in the GModel?
My preference at the moment is to store per entity (i.e., (3)): not
too much overhead; can do local indexing and checks with
nodes/elements--which are stored per entity too; avoids storing entity
ptr in each value (I think we should have that info in the post-pro:
it's one of the annoying shortcomings of the old format)
f we choose 3, unfortunately, we need to keep track of some local
indices to access the data with a single, continuous index. This is
exactly the same problem as with PViewDataList. Maybe it's worth it?
*/
bool PViewDataGModel::writePOS(std::string name, bool binary, bool parsed,
bool append)
{
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment