Re: Total order on floating point numbers


In article <2ZWdnaP0_9h88pTanZ2dnUVZ_v-hnZ2d@xxxxxxxxxxx>,
glen herrmannsfeldt <gah@xxxxxxxxxxxxxxxx> writes:
|>
|> > For reasons that a hardware person might be able to explain, the
|> > exception when the quotient overflows is sometimes divide by zero.
|> > It can also be unexpected in other ways.
|>
|> http://publibfp.boulder.ibm.com/cgi-bin/bookmgr/BOOKS/dz9zr003/6.5.2.18

Yes, but why THAT exception?

|> For S/360 the usual way to do divide was to load the dividend into
|> the high word and then do an arithmetic right shift of 32 bits.

Yes. Not all architectures support that.


Regards,
Nick Maclaren.
.

Relevant Pages

  • Re: fexception handling in C
    ... > I am using C on linux and performed a 1/0 operation i.e a divide by ... > zero using integer variables/constants.it gives me a floating point ... Interesting that you got a floating-point exception ...
    (comp.lang.c)
  • Re: Returing Exception Message - ExceptionService.zip (0/1)
    ... The first 'just tries to execute the divide of the operants comming ... custom message for the SOAP exception thrown. ... >>>I am using Visual Studio and I implemented my webservice through Visual Studio. ...
    (microsoft.public.dotnet.framework.webservices)
  • Re: Exceptions are not raised
    ... you're expecting a "divide by zero" exception, but it will not occur, ... dividing by zero is not an illegal operation. ... Dim num1, num2 As Short ...
    (microsoft.public.dotnet.general)
  • Re: C to Delphi set problems
    ... In other words, it's undefined, and should throw an exception. ... to divide by zero. ... could be said for trying to find the lowest element in a set that is empty. ...
    (comp.lang.pascal.delphi.misc)
  • Re: Total order on floating point numbers
    ... exception when the quotient overflows is sometimes divide by zero. ... IBM calls it fixed point divide exception. ...
    (comp.arch.arithmetic)