Takes about 15 - 20KB flash and 1KB ram (heap + variables).
I have to say that after this, I am immensely impressed with how those programmers did this back then on meager hardware and software, and they managed to deliver far more than I did here: I couldn't even get the decNumber library to compile on PIC24 or STM32F030

BTW, decNumber is quite impressive. I think implementing a high precision calculator on a fairly enough chip using decNumber should be fairly easy. Certainly wouldn't be a problem on a STM32F1/F4 chip. Not sure about the 8-bit Arduinos however.