|
Authored by: bugstomper on Friday, March 29 2013 @ 09:04 PM EDT |
What you described would make some sense, but it doesn't match what I see in the
description. The register format has more bits than the memory format. You
described having a choice of four modes for the necessarily lossy transformation
from register to memory format, with your choice made according to the
trade-offs of each. However, the description says that four modes correspond to
different formats of the memory representation, each of which has different
interpretation of some of the bits, and that there is no way to determine from
the data itself which representation was used in storing the number into the
memory.
As far as I can tell the only way to use it is to have machine instructions to
set the choice of four modes and make sure that your program is always in the
proper mode when loading and storing the particular floating point numbers so
that you never load a number using a different mode than you stored it in. But
there is no way from looking at the data to know the correct mode to use.
[ Reply to This | Parent | # ]
|
|
|
|
|