sin of large x, redux
From: Robert Corbett (robert.corbett_at_sun.com)
Date: 05/27/04
- Next message: Stig Kildegård Andersen: "Re: Setting priority for exe built in DFf6.6"
- Previous message: Antti K: "Setting priority for exe built in DFf6.6"
- Next in thread: Gerry Thomas: "Re: sin of large x, redux"
- Reply: Gerry Thomas: "Re: sin of large x, redux"
- Reply: Dr Chaos: "Re: sin of large x, redux"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 27 May 2004 02:01:03 -0700
A company for which I once worked once received a complaint from a
customer who tried to evaluate cos(exp(x)) for a value of x that
was large enough to cause exp(x) to overflow. He wanted it to
produce the proper appoximate result for the composed function.
He made the valid point that the mathematical value of cos(exp(x))
is defined for all real values and that the approximate result is
representable for most floating-point number systems. We could
have modified the compiler and run-time library to work as he
wanted, but we did not. We shall never know if the company would
have been more successful if we had satisfied that customer's request.
Sincerely,
Bob Corbett
- Next message: Stig Kildegård Andersen: "Re: Setting priority for exe built in DFf6.6"
- Previous message: Antti K: "Setting priority for exe built in DFf6.6"
- Next in thread: Gerry Thomas: "Re: sin of large x, redux"
- Reply: Gerry Thomas: "Re: sin of large x, redux"
- Reply: Dr Chaos: "Re: sin of large x, redux"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|