<div>Is the following a fair summary?</div><div><br></div><div>The idea is to divide the real axis in three parts.</div><div> x < y < z </div><div><br></div><div>Here x is a real (double), y is a slightly larger than +max.0, </div>
<div>and z is a very large.</div><div><br></div><div>Now sqrt(x) is just a real (double).</div><div><br></div><div>Let y be an exact integer only slightly larger than +max.0.</div><div>Even though y is larger than +max.0 the true sqrt of y is</div>
<div>actually smaller than +max.0 and thus representable as a float.</div><div>It is therefore possible to extend the standard sqrt function</div><div>to numbers slightly above +max.0. </div><div><br></div><div>For very large numbers z where sqrt(x) is larger than +max.0</div>
<div>return +inf.0 unless z happens to be a perfect square.</div><div><br></div><div><br></div><div>The sqrt function is a sublinear function. This extension is also</div><div>possible for other members of this class such as the logarithms.</div>
<div>How to extend the function depend on the function.</div><div><br></div><div><br></div><div>The other example, the periodic trigonometric functions is </div><div>slighly different.</div><div><br></div><div>Here the problem is that the argument must be reduced before</div>
<div>the actual computation can take place. The reduction is</div><div>simple in principle - just subtract the period repeatedly</div><div>until the argumet is so small the standard function works</div><div><br></div><div>
However the period irrational, so in order to avoid loss of</div><div>precision, one must compute the period with an appropriate</div><div>number of decimals compared to the argument before the</div><div>reduction can take place.</div>
<div><br></div><div>/Jens Axel</div><div><br></div>