tidbit wrote:
my name is all lowercase
I know. It just looks weird in the beginning of a sentence. I'll try to remember to address you as TidbiT the next time. I am terribly sorry
The running time is proportional to the second power of the requested length. That is, for 10 times more bits you have to wait 100 times longer, or more. When your PC runs out of of memory, it starts to swap data to disk, and the computation would take forever.
In my 2.66GHz core i7 PC, with 6GB RAM I got the following running times (ms):
Code:
T n
0 100
31 316
47 1,000
203 3,162
1029 10,000
9235 3,1623
95644 100,000
997002 316,228 (16.6 minutes)
1 million digits of pi could be computed under 3 hours (if the memory was enough - I have not checked). Note that this SW was optimized for code size and complexity, with keeping the computation speed reasonable. With more complicated code you could easily achieve 10 fold speedup at computing 1 million digits of pi. Also, only one of the 4 CPU cores are used in the computations, so another factor of 4 acceleration is possible.