I computed the digits of pi using Borwein's method. I used a divide-and-conquer multiply routine, hand coded in 68020 assembly language. It was capable of multiplying two 1.25+ million digit numbers in about 20 minutes on an HP 9000/370 (a 25MHz 68030?). The computation took a little over three days, at which point I had the answer in *binary*. :-( The binary to decimal conversion was no simple task.
I checked my results by performing the same calculation to 2.5+ million digit precision, (9 days) and compared the binaries. The only independent check has come from David Bailey, whose results agree with mine to at least 1 million digits (probably.... The last 100 digits are the same.)