It seems rather petty to criticize the use of uint64_t in the 64-bit versions in multiple places when the original does the same:

32-bit version:

  const unsigned b = 65536; // Number base (16 bits).
  unsigned un1, un0,        // Norm. dividend LSD's.
           vn1, vn0,        // Norm. divisor digits.
           q1, q0,          // Quotient digits.
           un32, un21, un10,// Dividend digit pairs.
           rhat;            // A remainder.

So a properly adapted 64-bit version should use uint64_t and a 32-bit number base.

I did not attempt to follow the author's arguments why he advocates for 32-bit variables or why that deviation from the original would be correct. Using 64-bit does not seem to be any bottleneck at all here, so why bother.