Thanks very much for the timings, Victor.

Just out of interest, could you try the pydigits script with the 
+optimizations patch on 32-bit?

As mentioned above, there's a significant (for 30-bit digits) problem 
with x_divrem:  the inner loop does a 32 x 64-bit multiply when it 
should be doing a 32 x 32-bit multiply  (the variable q is declared as 
twodigits, but always fits into a digit).  This is fixed in the 
+optimizations patch, and pi_digits is heavy on the divisions, so I
wonder whether this might make a difference.