split in chunks? - 128-bit? - parallel computation?

[newbie-02]

hi,

seeing the last digit of 32-bit and 64-bit computations slow,
and would like to use for 128-bit too,
and as speedups are always welcome ...

Is it a feasible/sensible idea to split the value into smaller chunks
in a first step? For example, into a maximum of 9 digits in order
to remain within 32-bit arithmetic?

And let them write into adequate memory locations , the smaller
magnitudes "left zero padded", to form a complete string in the end.

As most machines today are "vectorized", perhaps it would also
be possible to convert these in parallel, since they are independent?
E.g.:

   340,282,366,920,938,463,463,374,607,431,768,211,455 split in: 
 340  | 282,366,920 | 938,463,463 | 374,607,431 | 768,211,455 
  v          v             v             v             v 
 p-4       proc-3        proc-2        proc-1        proc-0 
  v          v             v             v             v 
"340" | "282366920" | "938463463" | "374607431" | "768211455" 

Or compute a value fitting as "similar length", in above example 7/8/8/8/8:

  340,282,366,920,938,463,463,374,607,431,768,211,455 split in: 
3,402,823 | 66,920,938 | 46,346,337 | 46,074,317 | 68,211,455 
    v           v            v            v            v 
  proc-4      proc-3       proc-2       proc-1       proc-0 
    v           v            v            v            v 
"3402823" | "66920938" | "46346337" | "46074317" | "68211455"