Downstairs

Pi Test
PI test is a benchmarking toy that I build up from first trying to calculate PI as many decimals as possible. Benchmarking results can be uploaded to here for comparing with others from the toy application. See and try. Anything in the calculations done is not compiled for some specific CPU.

Updated	3.12.2006 15:36:38
File	PiTest.zip
Size	338,81 KB

Below picture presents the amount of Intel processors versus AMD processors on top 20 on all four different pi calculation routines.

TOP1000 Results 1, machin's formula. Machin's formula is used for calculating 10000 first decimals of pi. Method uses only integer math and stresses most memory of these three tests.

TOP1000 Results 2, leibniz serie. This is very slow method for calculating pi value, it takes about 1 billion cycles to get, if i remember this correctly, 10 digits calculated. Method uses only floating point math.

TOP1000 Results 3, monte carlo estimation. Is really only one way to estimate pi, and not even too accurate after 20000000 iterations. Euclid's method, bases on integer comparision and math is left almost totally out. Very simple integer math.

TOP1000 Results 4, anonymous formula. Uses floating point math and large number of iterations, routine itself is not too complex.

TOP1000 results Total Time, of total calculation time of all above pi calculation tests. This is where we are able to see the benefits of multiple cpus in one system. A bit modified pi test now available to download will use one thread for each cpu that is installed on to the system. However, in two cpu system will first do 2 first tests, then remaining two. If four cpus, then all threads are done simultaneously.