128-bit integer arithmetic using FP

[jack]

un-optimized times

Code Select Expand

n = 340282366920938463447387429234553266722
d = 18446744073709551615
product = 45370982256125128475310893311622766046
sum of product = 113427455641665582386530236262442737152
assembler version
elapsed time seconds  0.642031399998814

n = 340282366920938463447387429234553266722
d = 18446744073709551615
product = 45370982256125128475310893311622766046
sum of product = 113427455641665582386530236262442737152
high level version
elapsed time seconds  2.85223180000321


[jack]

But wait, something is amiss.
'sum of product = 226854911283331164773060472524885474304' - for the high level version in my test?

and 'sum of product = 113427455641665582386530236262442737152' - for the high level version in your test.

I had forgotten to zero out the sum for the second test, so it just kept adding, it's fixed now if you want to download the binary

[zedd151]

I had forgotten to zero out the sum for the second test, so it just kept adding, it's fixed now if you want to download the binary

Intel(R) Core(TM) i7-7700 CPU @ 3.60GHz  3.60 GHz

Code Select Expand

multiply n times d 100000000 times in a loop
a sum of the product is kept and printed out
in order to prevent gcc from eliminating the loop
and thereby making the timing meaningless

n = 340282366920938463447387429234553266722
d = 18446744073709551615
product = 45370982256125128475310893311622766046
sum of product = 113427455641665582386530236262442737152
assembler version
elapsed time seconds  0.338004100020044

n = 340282366920938463447387429234553266722
d = 18446744073709551615
product = 45370982256125128475310893311622766046
sum of product = 113427455641665582386530236262442737152
high level version
elapsed time seconds  0.0723988000536337

press return to end
perfect. 
Side note: A humongous difference between unoptimised C version and optimized C version. Biterider might have to tweak his code (or you may need to tweak your conversion) to be a contender. 

These huge numbers remind me of my experiments with my 'ascii_adder' routine. It added two humongous ascii decimal strings of any length together, outputting an ascii string on return. May be good for working with the U.S. national debt?