Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Dear P1788 members: Many thanks for all the comments!It was perhaps a mistake that the text was coupled too much to IEEE 754. The motion is not intended to overload IEEE 1788 with complicated questions.
I prepared a revised text, see the attachment. The motion is now numbered 24.02.
The only concern of the motion is interval arithmetic and simplification of its realization. The question is whether the motion is benificial for this. The question is not how practical the motion is for existing architectures.
I think the old way of IEEE 754, first to set the rounding mode and then to call the operation, is well established and motion 24.02 does not request its elimination. I see a need for this mechanism as well, in particular for those applications where the rounding mode is to be selected randomly.
The demand for the 8 operations with the directed roundings is older than IEEE 754 arithmetic. At Karlsruhe these operations were realized in hardware on an Electrologica X8 already in 1967. Not to require them in IEEE 1788 would be fatal for interval arithmetic. The rounding should be an integral part of the arithmetic operation. Every one of the arithmetic operations with rounding to nearest, downwards or upwards should be equally fast. In the end an interval operation should be as fast as the corresponding floating-point operation. Modern hardware allows this easily.
Best wishes Ulrich -- Karlsruher Institut für Technologie (KIT) Institut für Angewandte und Numerische Mathematik (IANM2) D-76128 Karlsruhe, Germany Prof. Ulrich Kulisch Telefon: +49 721 608-42680 Fax: +49 721 608-46679 E-Mail:ulrich.kulisch@xxxxxxx www.kit.edu www.math.kit.edu/ianm2/~kulisch/ KIT - Universität des Landes Baden-Württemberg und nationales Großforschungszentrum in der Helmholtz-Gemeinschaft
Attachment:
RoundOper1.pdf
Description: Adobe PDF document