> -----Original Message----- > From: IBM Mainframe Assembler List > [mailto:[email protected]] On Behalf Of Miklos > Szigetvari > Sent: 23 March 2011 07:43 > To: [email protected] > Subject: IEEE floating points > > Hi > > Can I expect with IEEE floating points, to get the same > results in Windows as in z/OS ? (Seems to us till now, it is > not the case, we are using LE C++ in zOS and Windows Visual > C++, but got differences) >
Ah the nuances of english, <pedant mode> Many folks would EXPECT the results to be the same, but many may also be dissapointed, </pedant mode> <irony> Personally I am suprisd if I get the same answers displayed from two runs of any reasonably complex floating point code on the same hardware. </irony> Seriosly moving any floating point code that's reasonably complex to a different compiler can often produce slight differences in the output. In general (i.e. not in the C++ case) expressions such as "0.1 * 0.1" can be evaluated as single or double precesion. Which the compiler chooses can have an effect on the answers.... Dave
