The new command ArrayTools[GeneralOuterProduct] (introduced in Maple 2021) computes the generalized outer product of two rtables, and again, there exists a similar function Outer in Mma (cf. the end of this question). But in practice, it appears that this Maple command is not so fast as Mma's one. To begin with, we need to generate four lists: w, x, y, and z. Our goal is forming all possible combinations of the lowest‐level elements in a nested structure (rather than a flat structure). Now let us start the test.

In Mathematica (the real time is about ): 

And in Maple (the real time is about ): 
 

Download Outer.mw

As you can see, Maple and Mathematica returns identical results (∵p₁＝p₃); nevertheless, Maple consumes too much time: the ratio is 199.880/0.784176 ≈ 254.892. (What a wide gap between them!) 
So, is there any possibility of speeding up Maple's ArrayTools:-GeneralOuterProduct? Or any ideas of obtaining the same results in Maple efficiently?

Explanatory notes. Here is an illustrative animation: 

That is to say, a generalized map. 
E.g., here is a nested list: 

nl := [[[[s, t]], [u, [v, w]]], [[x, [y, z]]]]:

We can use map to apply the mapped function F to "each operand" (i.e., the first‐level parts) of : 

:-map(F, nl);
 = 
         [F([[[s, t]], [u, [v, w]]]), F([[x, [y, z]]])]

But in Mathematica, we can make further explorations: 

In[1]:= nl = {{{{s, t}}, {u, {v, w}}}, {{x, {y, z}}}}; 

In[2]:= Map[F, nl, {1}] (*Maple's result*)

Out[2]= {F[{{{s, t}}, {u, {v, w}}}], F[{{x, {y, z}}}]}

In[3]:= Map[F, nl, {2, -2}]

Out[3]= {{F[{F[{s, t}]}], F[{u, F[{v, w}]}]}, {F[{x, F[{y, z}]}]}}

In[4]:= Map[F, nl, {-3, 3}]

Out[4]= {{F[{F[{s, t}]}], F[{F[u], F[{v, w}]}]}, {F[{F[x], F[{y, z}]}]}}

In[5]:= Map[F, nl, {0, \[Infinity]}, Heads -> \[Not] True]

Out[5]= F[{F[{F[{F[{F[s], F[t]}]}], F[{F[u], F[{F[v], F[w]}]}]}], F[{F[{F[x], F[{F[y], F[z]}]}]}]}]

Note that the last case has been implemented in Maple as MmaTranslator[Mma][MapAll]:  

MmaTranslator:-Mma:-MapAll(F,nl);
 = 
   F([F([F([F([F(s), F(t)])]), F([F(u), F([F(v), F(w)])])]), 

     F([F([F(x), F([F(y), F(z)])])])])

Naturally, how to reproduce the other two results in Maple programmatically? (The output may not be easy to read or understand; I have added an addendum below.)

Addendum. It is also possible to display in "tree" structure (like dismantle) manually: 

`[]`
(
    `[]`
    (
        `[]`
        (
            `[]`
            (
                s
            ,
                t
            )
        )
    ,
        `[]`
        (
            u
        ,
            `[]`
            (
                v
            ,
                w
            )
        )
    )
,
    `[]`
    (
        `[]`
        (
            x
        ,
            `[]`
            (
                y
            ,
                z
            )
        )
    )
)

As you can see, the "depth" of  is five (0, 1, 2, 3, and 4), while the classical map just maps at the first "level". (Moreover, such descriptions may lead to a confusion.)

Supplement. Unfortunately, there remains a bug in the MmaTranslator[Mma][Level]. Compare: 

MmaTranslator:-Mma:-Level(nl, [4]); (*Maple*)
                             [v, w]

MmaTranslator:-Mma:-Level(nl, [-1]); (*Maple*)
          [s, t, u, v, w, x, y, z, -1, x, c, r, y, 2]

In[6]:= Level[nl, {4}] (*Mathematica*)

Out[6]= {s, t, v, w, y, z}

In[7]:= Level[nl, {-1}] (*Mathematica*)

Out[7]= {s, t, u, v, w, x, y, z}

For instance, here is a list of valid "rules" (given as equations): 

(* restart; *)
rules := [g = d, e = a, a = b, f = d, c = g, b = b, d = c, c = f, a = e]:

But these rules can be represented more compactly as a list of "cycles": 

cycles := [[g, d, c], [e, a], [f, d, c], [b]]:

Is there a efficient way to convert into ? 

Let a, b be arbitrary real parameters. I intend to compute something like: (with exact piecewise output) 

Optimization:-Maximize(8*x + 7*y, {5*y <= 6 - 9*b, -6*x - 4*y <= 8 - 5*a - 7*b, -4*x + 7*y <= -1 - 2*a - 7*b, -x + y <= 6 + 4*a - 5*b, 7*x + 5*y <= a + 4*b}, variables = {x, y}): # Error
Optimization:-Minimize((x - 1)^2 + (2*y - 1)^2, {x - 2*y <= 2*a - b + 1, x + 2*y <= a + b, 2*x - y <= a - b + 1}, variables = {x, y}): # Error

Unfortunately, these Maple codes are virtually invalid, and the relevant commands minimize, maximize, extrema, and Student[MultivariateCalculus][LagrangeMultipliers] do not support general inequality constraints. Is it possible to tackle these small-scale constrained parametric problems in Maple?

For instance, I'd like to find the integer solutions of the following system (unknowns: k[1], k[2], k[3], k[4], k[5], and k[6]): 

eqs:=eval~(k[1]*x^3+k[2]*x^2*y+k[5]*x^2*z+k[3]*x*y^2+k[6]*x*y*z+k[8]*x*z^2+k[4]*y^3+k[7]*y^2*z+k[9]*y*z^2+k[10]*z^3,{{x=1,y=1,z=1},{x=RootOf(_Z^3-4*_Z^2+_Z+1,index=1),y=RootOf(_Z^3-4*_Z^2+_Z+1,index=2),z=RootOf(_Z^3-4*_Z^2+_Z+1,index=3)},{x=RootOf(_Z^3-4*_Z^2+_Z+1,index=2),y=RootOf(_Z^3-4*_Z^2+_Z+1,index=3),z=RootOf(_Z^3-4*_Z^2+_Z+1,index=1)},{x=RootOf(_Z^3-4*_Z^2+_Z+1,index=3),y=RootOf(_Z^3-4*_Z^2+_Z+1,index=1),z=RootOf(_Z^3-4*_Z^2+_Z+1,index=2)}}):
{isolve}(eqs=~0);

It seems that the isolve command cannot solve such a system of linear equations, but its integer solutions do exist: 

sol := [k[1] = _Z1, k[2] = _Z2, k[3] = _Z3, k[4] = _Z4, k[5] = _Z5, k[6] = _Z1 + 2*_Z3 + 2*_Z4 + _Z5 + 3*_Z6, k[7] = 2*_Z1 + _Z3 + 8*_Z4 - _Z6, k[8] = -21*_Z1 + 2*_Z2 - _Z4 - 4*_Z5 + _Z6, k[9] = 24*_Z1 - 4*_Z2 - 5*_Z3 - 12*_Z4 + 3*_Z5 - 4*_Z6, k[10] = -7*_Z1 + _Z2 + _Z3 + 2*_Z4 - _Z5 + _Z6]: (*results from Mathematica*)
evala(subs(sol, eqs));
 = 
                              {0}

Here _Z1, _Z2, _Z3, _Z4, _Z5, and _Z6 are integer parameters. However, how do I get this result in Maple?