improve performance of multiidx2n (#173)

* improve performance of multiidx2n

* minor cleanup

* minor fix

include/qpp/internal/util.hpp

@@ -105,15 +105,11 @@ template <typename V, typename T = V, typename U = T>
 #endif
     // no error checks in release version to improve speed
 
-    // static allocation for speed!
-    // allocate twice the size for matrices reshaped as vectors
-    T part_prod[2 * internal::maxn];
-
+    T part_prod = 1;
     T result = 0;
-    part_prod[numdims - 1] = 1;
     for (std::size_t i = 1; i < numdims; ++i) {
-        part_prod[numdims - i - 1] = part_prod[numdims - i] * dims[numdims - i];
-        result += midx[numdims - i - 1] * part_prod[numdims - i - 1];
+        part_prod *= dims[numdims - i];
+        result += midx[numdims - i - 1] * part_prod;
     }
 
     return result + midx[numdims - 1];