fnrpow.gms : Test correctness of rpower intrinsic
rpower(x,y) = x**y
these points are the "easy" ones, i.e. there should not be any
exceptions as all the inputs are of reasonable size.
Small Model of Type: GAMS Includes: fnset_xy.inc fntest_xy.inc
<![CDATA[
$title 'Test correctness of rpower intrinsic' (FNRPOW,SEQ=194)
* rpower(x,y) = x**y
* these points are the "easy" ones, i.e. there should not be any
* exceptions as all the inputs are of reasonable size.
$include fnset_xy.inc
$gdxin fnrpow
$load aeps reps relToInput T data
$gdxin
* hack, reps in GDX file to narrow for Delphi math libs
reps = reps * 2;
* hack for GDX bug
relToInput = max(0,relToInput);
data(T,'fyx_') = data(T,'fxy_');
loop {T,
data(T, 'f') = rpower.value( data(T,'x'),data(T,'y'));
data(T, 'fx') = rpower.grad(1: data(T,'x'),data(T,'y'));
data(T, 'fy') = rpower.grad(2: data(T,'x'),data(T,'y'));
data(T,'fxx') = rpower.hess(1:1:data(T,'x'),data(T,'y'));
data(T,'fxy') = rpower.hess(1:2:data(T,'x'),data(T,'y'));
data(T,'fyx') = rpower.hess(2:1:data(T,'x'),data(T,'y'));
data(T,'fyy') = rpower.hess(2:2:data(T,'x'),data(T,'y'));
data(T, 'rc') = mathlastrc;
data(T, 'ec') = mathlastec;
};
* SGI flushes underflow to zero by default.
* This makes for less accurate results for some derivs if x < 1, c > 1
* $if %system.platform% == SGI $if defined aeps aeps = max(aeps,1e-285);
display relToInput;
$include fntest_xy.inc
]]></plaintext></body></div>