sub x{(caller())[ob-//]}
[download]

This defines the sub called x .
It invokes the function caller and takes a subscript of its last element because ob-// always evaluates to -1 .
This element of the result of caller happens to be the line number of where it is called !

Thus all the following subs: ob , fu , sc, etc..., that are only calling this x sub are in fact returning their own line numbers .

At the end comes the definition of the _ function :

sub _{
    vec($_,$%,d_)=eval"$%x$_[$%]";
    print chr vec($_,$%,d_)
}
[download]

$% is equal to zero .
d_ is a call to a previously defined sub and if you count the lines you'll see it returns 8 .
This _ function is passed one argument.
For example in the first call :

_(&fu.a)
[download]

this is really 4a beacause &fu is defined at line 4 .
So :

eval"$%x$_[$%]";
[download]

is really in this case eval"0x4a" , since $_[$%] is subscript zero to @_, which is hex number 4a which is the ascii code for "J" ...

Lastly comes the vec stuff which is really just plain obfu , since what _ really does is print chr of 4a : you can replace the first vec call by a simple assign to $_, and then just print chr $_ . The vec is a binary way of storing the hex value to print.

The _ is then called multiple times on the appropriate hex values, each time printing the correct letter with chr :

_(&fu.a) calls on 4a -> J
_(&te.&sc) calls on 75 -> u
_(&te.&ob) calls on 73 -> s
_(&te.&fu) calls on 74 -> t
_(&ob-//.$%) calls on (3-1).'0' that is 0x20 -> Space
[download]

Note that in this last line &ob is used, while in the definition of the sub x it is plain ob that is used in ob-// returning -1 and not 2 for some unclear reason ...