2001/11/17
Now Rob's STL is got into the SCALP framework. I'll make the Floorplanner work
tomorrow.
******************************************************************************`
2001/11/18
The problem is how to mainain the connection information efficiently during iterative improvement.
I decided to ignore mux and register at the time being. The floorplan only have funcational unit modules and the interconnection is regarded as module to module and center to center.

Now it's working with the SCALP -A option( Area optimization) without considering the interconnection. It uses the floorplanned area as the compute_areacost(). Iierative improvements  actually reduced floorplanned areas. It looks good.
Current assumption: 1)modules are hard and square.2) No register or mux. 2) No interconnection.

Next step: incorporate interconnection into the floorplanner. 
*******************************************************************************
2001/11/22
I compared the area optimization of the original SCALP and the flooplanner augmented one. The original SCALP is better. The reason is at this moment, the floorplanner is only used when a series of moves are found . It is used to decide whether to carry out these moves or not. However, each move is found by raw area reduction. So, this might be the reason. I'll try to use floorplanner to find each move. But this is not a good idea, since floorplanner will be called much more times. The time consumption will increase significantly. But how could we do incremental floorplanner?

I'm not going to incorporate the interconnection before I can get some gain from using floorplanner for area optimization.

I reorganized the codes. FloorPlan is moved intto Datapath. Within each object of Datapath, there is a FloorPlan pointer, a module_size and a module_link. I'm trying to reimplement the compute_areagain() with re-floorplanning for each move. This is not a good solution. I need a incremental floorplanner which can handle local binding/allocation changes efficiently.

Going to finish the current version tomorrow.
*******************************************************************************
2001/11/23
I finished the merge_two_link_units for compute_areagain()

Still need to update the copy_dp_and_flowgraph() used in iterative improvement since two members are added into class Datapath
*****************************************************************************

*******************************************************************
2002/01/26
The old way demands too many changes to the high level synthesis tool. I'm going to make 
the physical things an tool independent package.

Started to write the iscalp/interface.C

CVS on devry is working.
