On 03/21, Oleg Nesterov wrote:
>
> I don't really understand this code, but it seems that the isNumericalTuple()
> simplification in realeval() is not right. At least the "patch" below seems
> to fix a problem.

Yes, I think this is true... So unless you are going to change the "propogate"
paths, I _think_ that the fix should do something like below. Note also that
both isBoxPrim1 and "xtended" cases have the same problem,

        process = 0,0 : sin;
        process = 0,0 : mem;

but I am sure you understand this much better than me ;)

and I am not sure about isBoxWire()...

Oleg.


--- a/compiler/evaluate/eval.cpp
+++ b/compiler/evaluate/eval.cpp
@@ -306,6 +306,17 @@ static bool isNumericalTuple(Tree box, siglist& L)
  * @return a block diagram in normal form
  */
 
+static bool xxx(const siglist &lsig, Tree a2)
+{
+       xtended* xxt = (xtended*) getUserData(a2);
+       int lsize = lsig.size();
+
+       return  (isBoxWire(a2) /* && ????? */) ||
+                       (xxt && lsize == xxt->arity()) ||
+                       (isBoxPrim1(a2) && lsize == 1) ||
+                       (isBoxPrim2(a2) && lsize == 2);
+}
+
 static Tree realeval (Tree exp, Tree visited, Tree localValEnv)
 {
        //Tree  def;
@@ -342,10 +353,9 @@ static Tree realeval (Tree exp, Tree visited, Tree 
localValEnv)
        } else if (isBoxSeq(exp, e1, e2)) {
         Tree a1 = eval(e1, visited, localValEnv);
         Tree a2 = eval(e2, visited, localValEnv);
-        xtended* xxt = (xtended*) getUserData(a2);
         siglist lsig;
         // try a numerical simplification of expressions of type 2,3:+
-        if (isNumericalTuple(a1,lsig) && (xxt || isBoxWire(a2) || 
isBoxPrim1(a2) || isBoxPrim2(a2))) {
+        if (isNumericalTuple(a1,lsig) && xxx(lsig, a2)) {
             Tree lres = boxPropagateSig(gGlobal->nil, a2, lsig);
             if ( isList(lres) && isNil(tl(lres)) ) {
                 Tree r = simplify(hd(lres));


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Faudiostream-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/faudiostream-users

Reply via email to