Test Name: BUG_834
Owner: Eric Keiter
Test netlists:
GS files:
Mode:  N/A
Comparator:  
Version of Xyce: Release_3.0.1
                                                                                                                                              
Description:
============
Transition from homotopy-based tranOP to transient phase broken.

This is a blocker for bug 822.

The bug had to do with the device package setParam function, which is normally
called throughout a homotopy solve, being called afterwards during  the
transient phase, even though the nonlinear solves for transient steps were
no longer using homotopy.  The result was that if homotopy was used for
a DCOP calculation, then setParam was called during the subsequent transient,
with the *initial* values of the various homotopy values being imposed.

So, if the homotopy was a natural parameter homotopy, in which a voltage source
was ramped from (say) 0.0 to 4.0 volts, then at the end of the DCOP the
value of that voltage source was 4.0 Volts.  Then, when the transient started,
the setParam function was called during the standard newton solve, resetting
the voltage source value to 0.0.  As a standard newton solve doesn't sweep any
parameters, the voltage source just stays at zero from that point onward.

So, the result is that there is a very abrupt change in the solution,
immediately after the DCOP, which is very obviously incorrect.

================================================================================

Test Procedure:
There are two circuits here.  inv_loca.cir, and inv.cir.  They are the same
transient inverter circuit, except that inv_loca.cir uses a homotopy 
algorithm for the DCOP phase, while inv.cir does not.  

Run both of these circuits:

Xyce inv.cir >& inv.out &  <return>
Xyce inv_loca.cir >& inv_loca.out &  <return>

Then do a "diff" of these two *prn files:

diff inv.cir.prn  inv_loca.cir.prn <return>

This test should be run on every serial platform.  It is too small for
parallel, so don't run in parallel.

Test Verification:

The diff should reveal NO differences.  The 2 prn files should be 100% 
identical.

Test Notes:

There is no gold standard.  This is a comparison of Xyce on a given platform
to Xyce on that same platform.


