Merge branch 'master' into netgen-1.5

"*.GLOBAL" as equivalent to ".GLOBAL".  Corrected the property
matching to handle property combination when no "critical"
property is given.  Critical properties exist when one property
must remain constant and equal for other properties to combine,
such as transistor length.  But, for example, capacitors can
combine area without any restriction based on another property.
Also, corrected the property matching code to allow more than
one property to be additive (example:  capacitor area and
perimeter).  Corrected the equation for adding properties in
parallel combination.

VERSION

@@ -1 +1 @@
-1.5.300
+1.5.301

base/netcmp.c

@@ -5218,14 +5218,23 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 
    if (comb == TRUE) {
       double pd;
+      int hascrit = FALSE;
       int mult, cidx = -1;
-      struct valuelist *avl, *cvl = NULL;
+      struct valuelist **avl, **cvl;
+      struct valuelist *cvlp;
+
       critval.type = PROP_ENDLIST;
       critval.value.dval = 0.0;
+
+      /* Track properties separately so that multiple properties can be
+       * added together (e.g., area and perimeter)
+       */
+      avl = (struct valuelist **)CALLOC(pcount, sizeof(struct valuelist *));
+      cvl = (struct valuelist **)CALLOC(pcount, sizeof(struct valuelist *));
+
       for (i = 0; i < run; i++) {
-         avl = NULL;
-	 // if (vlist[0][i] == NULL) continue;
-	 // mult = vlist[0][i]->value.ival;
+	 for (p = 0; p < pcount; p++) avl[p] = NULL;
+
 	 if (vlist[0][i] == NULL)
 	    mult = 1;
 	 else
@@ -5242,6 +5251,7 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 
 	    /* critical properties never combine, but track them */
 	    if ((series == TRUE) && (ctype & MERGE_S_CRIT)) {
+		hascrit = TRUE;
 	        pd = 2 * fabs(vl->value.dval - critval.value.dval) /
 				(vl->value.dval + critval.value.dval);
 		if ((vl->type != critval.type) || (pd > kl->slop.dval))
@@ -5253,6 +5263,7 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 		continue;
 	    }
 	    if ((series == FALSE) && (ctype & MERGE_P_CRIT)) {
+		hascrit = TRUE;
 	        pd = 2 * fabs(vl->value.dval - critval.value.dval) /
 				(vl->value.dval + critval.value.dval);
 		if ((vl->type != critval.type) || (pd > kl->slop.dval))
@@ -5286,19 +5297,30 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 		}
 	    }
 	    if (ctype & (MERGE_S_ADD | MERGE_P_ADD | MERGE_S_PAR | MERGE_P_PAR))
-		avl = vl;
+		avl[p] = vl;
          }
-	 if (cidx == i) cvl = avl;
+	 if (hascrit == FALSE) cidx = 0;	/* No critical property */
+
+	 /* Each time a new critical value is found, set the location	*/
+	 /* of the property record that will collect the summation or	*/
+	 /* parallel combination of values.				*/
+	 if (cidx == i) {
+	    for (p = 1; p < pcount; p++) {
+	 	cvl[p] = avl[p];
+	    }
+	 }
 
 	 /* Sorting should have put all records with the same critical	*/
 	 /* value together sequentially.  So if there are still		*/
 	 /* multiple property records, then merge them into the first	*/
-	 /* record with the same critical property value.		*/
+	 /* record with the same critical property value.  If no 	*/
+	 /* critical value exists, then all records can be merged.	*/
 
 	 if ((i > 0) && (cidx >= 0) && (cidx < i)) {
 	    for (p = 1; p < pcount; p++) {
 		vl = vlist[p][i];
 		ctype = clist[p][i];
+		cvlp = cvl[p];
 
 		if (ctype & (MERGE_S_ADD | MERGE_P_ADD)) {
 		    if (!vlist[0][i]) {
@@ -5307,21 +5329,21 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 				sizeof(struct valuelist));
 		    }
 		    vlist[0][i]->value.ival = 0;	/* set M to 0 */
-		    if (cvl && (cvl->type == PROP_INTEGER))
+		    if (cvlp && (cvlp->type == PROP_INTEGER))
 		    {
 			if (vl->type == PROP_INTEGER)
-			    cvl->value.ival += vl->value.ival;
+			    cvlp->value.ival += vl->value.ival;
 			else {
-			    cvl->type = PROP_DOUBLE;
-			    cvl->value.dval = (double)cvl->value.ival + vl->value.dval;
+			    cvlp->type = PROP_DOUBLE;
+			    cvlp->value.dval = (double)cvlp->value.ival + vl->value.dval;
 			}
 		    }
-		    else if ((cvl && vl->type == PROP_DOUBLE))
+		    else if ((cvlp && vl->type == PROP_DOUBLE))
 		    {
 			if (vl->type == PROP_INTEGER)
-			    cvl->value.dval += (double)vl->value.ival;
+			    cvlp->value.dval += (double)vl->value.ival;
 			else
-			    cvl->value.dval += vl->value.dval;
+			    cvlp->value.dval += vl->value.dval;
 		    }
 		}
 		else if (ctype & (MERGE_S_PAR | MERGE_P_PAR)) {
@@ -5331,21 +5353,22 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 				sizeof(struct valuelist));
 		    }
 		    vlist[0][i]->value.ival = 0;    /* set M to 0 */
-		    /* To do parallel combination, both types need to
+		    /* Parallel value combination: (X * Y) / (X + Y)
+		     * To do parallel combination, both types need to
 		     * be double, so recast them if they are integer.
 		     */
 		    if (vl->type == PROP_INTEGER) {
 			vl->type = PROP_DOUBLE;
 			vl->value.dval = (double)(vl->value.ival);
 		    }
-		    if (cvl && (cvl->type == PROP_INTEGER)) {
-			cvl->type = PROP_DOUBLE;
-			cvl->value.dval = (double)cvl->value.ival;
+		    if (cvlp && (cvlp->type == PROP_INTEGER)) {
+			cvlp->type = PROP_DOUBLE;
+			cvlp->value.dval = (double)cvlp->value.ival;
 		    }
-		    if ((cvl && (vl->type == PROP_DOUBLE))) {
-		        cvl->value.dval =
-				sqrt(cvl->value.dval * cvl->value.dval
-				+ vl->value.dval * vl->value.dval);
+		    if ((cvlp && (vl->type == PROP_DOUBLE))) {
+			cvlp->value.dval =
+				(vl->value.dval * cvlp->value.dval) /
+				(vl->value.dval + cvlp->value.dval);
 		    }
 		}
 	    }
@@ -5357,6 +5380,9 @@ int PropertyOptimize(struct objlist *ob, struct nlist *tp, int run, int series,
 		Printf("Combined %d parallel devices.\n", changed);
 	 }
       }
+
+      FREE(avl);
+      FREE(cvl);
    }
 
    // Remove entries with M (S) = 0

base/spice.c

@@ -534,7 +534,12 @@ void ReadSpiceFile(char *fname, int filenum, struct cellstack **CellStackPtr,
     if ((EndParseFile()) && (nexttok == NULL)) break;
     if (nexttok == NULL) break;
 
-    if (nexttok[0] == '*') SkipNewLine(NULL);
+    /* Handle comment lines.  Note that some variants of CDL format
+     * use "*." for information that is transparent to SPICE simulators.
+     * Handle "*.GLOBAL" entries.  All others are ignored.
+     */
+    if ((nexttok[0] == '*') && (!matchnocase(nexttok, "*.GLOBAL")))
+	SkipNewLine(NULL);
 
     else if (matchnocase(nexttok, ".SUBCKT")) {
       SpiceTokNoNewline();
@@ -822,7 +827,10 @@ skip_ends:
 
     // Handle some commonly-used cards
 
-    else if (matchnocase(nexttok, ".GLOBAL")) {
+    /* .GLOBAL and *.GLOBAL.  Note that *.GLOBAL is excepted from comment-line
+     * handling, above, so any line starting with '*' is "*.GLOBAL".
+     */
+    else if (matchnocase(nexttok, ".GLOBAL") || (nexttok[0] == '*')) {
       while (nexttok != NULL) {
 	 int numnodes = 0;
          SpiceTokNoNewline();