7834Fudging large value arithmetic in unfriendly languages
- Jul 2, 2002With 80-bit floats, and a 32*32->64 integer multiply, the intel-using
C/C++ and Asm programmers can without too much hassle write 62*62%62
and it would be nice to know quite how far JS can be taken and remain
precise, without having to resort to multi-word operations, or
Does this sound about right?
Floating point multiplication with 52 bits of accuracy
Integer mutliplication with 32 bits of accuracy
=> maybe 84-bit intermediate exactly representable
=> 42*42%42 is probably pushing it.
Which is fairly poor for everything apart from squfof (which uses
half-length values almost the whole time).
I'm not intending the applets to be used in anger, obviously, but for
them to not be utterly useless would be quite nice.
I don't relish the thought of having to write a 'bignum' library in
to do such things?
"One cannot delete the Web browser from KDE without
losing the ability to manage files on the user's own
hard disk." - Prof. Stuart E Madnick, MIT.
So called "expert" witness for Microsoft. 2002/05/02
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
- Next post in topic >>