Using Threads package I've got some promising results, because it was the first time when my CPU was working at 83%:
Add function managed to consume 83% of CPU

But the time it used surprised me:

> restart;
   with(Threads):
   t:=time():
   Add((-i)^3, i=1..10^7);
   time()-t;
                                   -2500000500000025000000000000
                                                           15.078
> t:=time():
   add((-i)^3, i=1..10^7);
   time()-t;
                                   -2500000500000025000000000000
                                                            7.188
>

So, what's the point to use Add if it works slower? Moreover, add used only 25% of CPU (Add used up to 83% of CPU).

Despite the fact that my goal was achieved (for using more than 25% of CPU) it was all in vain. :)

I've got even more disappointment when I tried using threads myself:

> restart;
   with(Threads):
   loop:=proc(m)
      global n;
      local i,s;
      s:=0;
      for i from n*(m-1)/4 to n*m/4-1 do
         s:=s+i;
      end do;
      s;
   end proc:

> n:=10^7;
   id1 := Create(loop(1), s1);
   id2 := Create(loop(2), s2);
   id3 := Create(loop(3), s3);
   id4 := Create(loop(4), s4);
                                                            n := 10000000
                                                                 id1 := 1
                                                                 id2 := 2
                                                                 id3 := 3
                                                                 id4 := 4

> t:=time():
   Wait(id1,id2,id3,id4); 
   s1+s2+s3+s4+n;
   time()-t;
                                                          50000005000000
                                                                  20.063

> t:=time():
   add(k,k=0..n);
   time()-t;
                                                         50000005000000
                                                                    1.906
>
 

Maple 12 / help states this:
LinearAlgebra[IsDefinite] - test for positive or negative definite or indefinite Matrices

Description
The IsDefinite(A, query = 'positive_definite') returns true if A is a real symmetric or a complex Hermitian Matrix and all the eigenvalues are determined to be positive. Because the default query is query = 'positive_definite', this command is equivalent to IsDefinite(A).


This description doesn't imply that IsDefinite(A) will always give true when matrix A is positive definite and not symmetric.

Question:
Is it just a bug in help description or we can't rely on function IsDefinite() for non symmetric matrices?
(I didn't find a counterexample, i.e. a non symmetric positive definite matrix A for which IsDefinite(A) = false)