Embed Notice

@futurebird

The TI calculator does what mathematics normally means (round to nearest).

A programming language will do one of the things specified in IEEE 754 floating-point rounding modes. The default for Python (and C), I believe, is 'round towards zero', also known as truncation. This is implemented in floating point as a very simple normalisation operation (shift the mantissa so that the exponent is 1, which either shifts some zeroes in from the right or shifts some bits off to the right and loses them).

The extra fun on the calculator might be that it's may be using some binary-coded decimal representation internally, because otherwise you have fun rounding errors from not being able to represent a lot of decimal fractions precisely in a binary floating-point value. More advanced calculators will store rational numbers fractions, which lets them avoid any precision loss until the final convert-to-decimal step that they do for display.