* j. romildo:

> That is, the expression p.x has different performances for the types
> { x:int } and { mutable x:int } for p?
>
> In summary, what additional cost does the use of mutable imposes on a
> record field?

Internally, record fields are always mutable.  Adding a mutable
declaration should not incur a performance cost.  perhaps with the
exception that the compiler will see less opportunities for
elimination of common subexpression (but I'm not sure if it performs
this optimization at all).