Re: [Gnucap-devel] Problem with logic elements, patch proposal

[Felix Salfelder]

On Tue, Nov 06, 2012 at 08:21:24AM +0100, Felix Salfelder wrote:
> 3) add a node_t to MODEL_LOGIC (_m=_ttt=0, _lv="something that
>    interprets gnd according to logic"), use it as "0" during
>    DEV_LOGIC::map_nodes.

since the logicvalue is stored in _sim, it doesnt make sense to
put nodes into MODEL_LOGIC. i have now implemented this by adding two
more LOGIC_NODEs to _nstat in u_sim_data:

-  _nstat = new LOGIC_NODE[_total_nodes+1];
+  _nstat = new LOGIC_NODE[_total_nodes+3];
   for (uint_t ii=0;  ii <= _total_nodes;  ++ii) {
     _nstat[_nm[ii]].set_user_number(ii);
   }
+  _nstat[_total_nodes+1].set_lv(lvSTABLE0);
+  _nstat[_total_nodes+2].set_lv(lvSTABLE1);
+  _nstat[_total_nodes+1].store_old_lv();
+  _nstat[_total_nodes+2].store_old_lv();

which are mapped to in

+void DEV_LOGIC::map_nodes() {
+  ELEMENT::map_nodes();
+
+  const COMMON_LOGIC* c = prechecked_cast<const COMMON_LOGIC*>(common());
+  assert(c);
+  const MODEL_LOGIC* m = prechecked_cast<const MODEL_LOGIC*>(c->model());
+  assert(m);
+
+  for (uint_t i = 3; i < ext_nodes()+int_nodes(); ++i) {
+    if(!_n[i].m_()){
+      assert(m->th0 < m->th1); //for now
+      if(m->th0 > 0.){
+        _n[i].map_to_logic(lvSTABLE0);
+      } else if(m->th1 < 0.) {
+        _n[i].map_to_logic(lvSTABLE1);
+      } else {
+        // some sort of warning?
+      }
+    }
+  }
+}
+/*--------------------------------------------------------------------------*/

node_t::map_to_logic just sets _m correspondingly. since reading _lv via
data() is all the logic devices do, i doubt there are any negative side
effects.

i will merge this into -uf some time. any warnings or better ideas are welcome.

regards
felix