Specification Details

5. Numbers

[Disagreement with ANS]

5.1.5
If the absolute value of a numeric constant is extremely large, an error occurs on compiling.
5.4.4
On the explanation of EPS(x), x, x', and x'' are chosen on the accuracy of variables.

[Implementation-defined]

5.1.4
Numerical constants have a precision between 19 digits and 27 digits.
5.2.4
The initial value of a variable is zero.
5.3.6
An underflow is replaced by zero.
5.3.6
Operations of the same priority are performed from left to right.
5.4.4
If arithmetic option is DECIMAL,
EPS(0)=1.E-99
MAXNUM=1.E99.
The values of RND functions are pseudo-random numbers.
5.4.6
Randomize-statement uses the real-time clock.
5.6.4
m=15, i.e. a numeric variable has precision of 15 digits.
5.6.4
Numeric constants and results of addition, subtraction, multiplication and division have 19 to 27 digits of precision. The result of a transcendental function, including a power operation, has 17 digits of precision, where the 17th digit is even. The other supplied functions have same digits of precision as above.
5.6.4
If arithmetic option is NATIVE, numbers are double precision numbers of Intel 80x87 FPU, the positive minimum number is about 4.9E-324C and the maximum number is about 1.8E308.

[Note]
Accuracy of numerical operations is not guaranteed.