Quiz submission record for quiz3-2-1 at Tue Jul  6 22:49:26 2004:

Your Answer for Question 1:
I guess the problem comes in when data can be represented as instructions.  Since everything is just 1's and 0's to the computer, it can't tell the difference between a good instruction and a bad instruction.  A person can conceivably reverse-engineer a program, add a few instructions that tell the OS to delete itself or something, and recompile.  Maybe a technique could be some heavy-duty write-protection but that in itself is also just written in data.

Your Answer for Question 2:
Subnormals, or denormalized numbers, are normals with the same exponent as zero but have a nonzero significand.  Subnormals are simply floating-point numbers that are precise than normalized numbers because they treat the leading zeros in the significand as actual zeros in the scientific notation, for example 0.00....01_two x 2^-126.

Your Answer for Question 3:
Back in the old days, there was no guard digit so operations such as 1.0 - x might lose precision due to roundoff errors.  By separating it to (0.5 - x) + 0.5, we can compensate for the lack of guard digit because we do two separate operations that retain the precision.

Your Answer for Question 4:
The IEEE 754 standard has been successful because it's a widely implemented standard with a common floating-point format, and its accuracy of one-half ulp in the least significant bit (from the book).

An argument against the implementation of IEEE 754 would be because it cannot be easily implemented.


Your unique submission ID is quiz3-2-1-cs61c-ck-1089179366-2980.