On 1/11/2010 7:54:58 AM, gnu wrote:
>need more than the usual
>(Pentium?) 16-digit precision,
>at least 32, 64 would be
>better (working with 10**100
>numbers ...) - anybody know
>how to set?
____________________________
Results with more digits than the numbers processed is meaningless. The conventional Pentium 16 machine precision does process at the floating point 20/21. Whether symbolic will display > 21 digits or not ... the pentium will take only 20. The 16 "true" Pentium machine precision represents the radius in cm of the electron seating at the top of a tower 2818 m height. Assuming all your numbers are known at 40 digits, about no symbolic results will be > 20 (floating arithmetic).
"The Accuracy of a machine real number depends on the value of $MachinePrecision on your computer."
Guide line:
There are plenty more culprits in numaccuracy !
jmG
>need more than the usual
>(Pentium?) 16-digit precision,
>at least 32, 64 would be
>better (working with 10**100
>numbers ...) - anybody know
>how to set?
____________________________
Results with more digits than the numbers processed is meaningless. The conventional Pentium 16 machine precision does process at the floating point 20/21. Whether symbolic will display > 21 digits or not ... the pentium will take only 20. The 16 "true" Pentium machine precision represents the radius in cm of the electron seating at the top of a tower 2818 m height. Assuming all your numbers are known at 40 digits, about no symbolic results will be > 20 (floating arithmetic).
"The Accuracy of a machine real number depends on the value of $MachinePrecision on your computer."
Guide line:
There are plenty more culprits in numaccuracy !
jmG
