User guide

CertainnumericalerrorswillcausethemathcoprocessortocreateaNaNcalledan
"indefinite."Thiswillbeprintedasa-NaNwhenusingtheprintcommand.These
valuesarecreatedbythefollowingoperations:
l +∞plus-∞
l +∞minus+∞
l -∞minus-∞
l 0*∞
l ∞/∞
l 0/0
l OperationswhereoneorbothoperandsisaNaN
l Trigonometricfunctionsinvolving∞
INF
Whenthemathcoprocessoroverflows,theresultwillbeaproperlysignedinfinity.Sub-
sequentcalculationswillnotdealwellwithaninfinity;itusuallysignalsanerrorinyour
program.TheresultofanoperationinvolvinganinfinityismostoftenaNaN.
DEN, UNN
Whensomemathcoprocessorsunderflow,theymaydosograduallybyshiftingthesig-
nificandofthenumberasnecessarytokeeptheexponentinrange.Theresultofthisisa
denormal(DEN).Whendenormalsareusedincalculations,theyareusuallyhandled
automaticallyinanappropriateway.Theresultwilleitherbeanunnormal(UNN),
whichlikethedenormalrepresentsanumberveryclosetozero,oranormal,depending
onhowsignificanttheeffectofthedenormalwasinthecalculation.Insomecasesthe
resultwillbeaNaN.
Followingaresomeproceduresfordealingwiththesevalues.Theseproceduresarenot
definedintheRun-Time Library.Ifyouwanttousethem,youwillneedtodefine
themyourself.
9-28
GAUSSUser Guide