Handling floating-point exceptions in numeric programs
ACM Transactions on Programming Languages and Systems (TOPLAS)
A Simple but Realistic Model of Floating-Point Computation
ACM Transactions on Mathematical Software (TOMS)
Termination of Floating-Point Computations
Journal of Automated Reasoning
| Hi-index | 0.00 |
While the denormalized numbers extend the effective range of a floating point format for a given number format, we feel that the benefits do not justify the expense. This expense can be seen not only in the implementation of floating point but in the use of the resulting system. We feel that they further increase the difficulty of understanding the standard and its implications. Specifically:a. The feature increases the complexity of implementing the standard.b. The cost of producing high-quality software will increase because of this feature.c. Unexpected errors may be introduced into results due to denormalized numbers.d. The added effective range is quite small, especially in Double Format.e. Denormalized numbers may be added at a later time if desired, losing little of the proposed facility.