result of erf is never more precise than double

3 results were exact to double the precision

20 largest ulp errors (stored in a double)
   0.55029 ulp for erf(0.173003437415309991) = 0.19328324616993478)
   0.54448 ulp for erf(0.532200885538677615) = 0.54833715828603613)
   0.54040 ulp for erf(-0.489341802678973925) = -0.51108390576853624)
   0.52689 ulp for erf(-0.237989233146952112) = -0.263557116890746901)
   0.52652 ulp for erf(-0.220614686665097565) = -0.244956651014343191)
   0.52414 ulp for erf(0.789433081039224605) = 0.735760585524578681)
   0.52372 ulp for erf(-0.237989233146952528) = -0.263557116890747345)
   0.51817 ulp for erf(1.45550972655508826) = 0.960448401281772002)
   0.51604 ulp for erf(-1.29245567966885266) = -0.932421680527195207)
   0.51470 ulp for erf(-0.11950033357842324) = -0.134202566094570735)
   0.51280 ulp for erf(1.33777652839433969) = 0.9414959220645438)
   0.51243 ulp for erf(0.644448162086819432) = 0.637908673562691209)
   0.51150 ulp for erf(1.0962664969880267) = 0.878943654476177727)
   0.51148 ulp for erf(-1.29245567966885422) = -0.932421680527195651)
   0.50985 ulp for erf(1.29335219460623585) = 0.932611808380637064)
   0.50972 ulp for erf(1.97479464988596676) = 0.994774300306054737)
   0.50840 ulp for erf(1.10546442291395453) = 0.882032710438650058)
   0.50762 ulp for erf(0.393287002759313054) = 0.421920264755931529)
   0.50750 ulp for erf(1.09626649698801426) = 0.878943654476173508)
   0.50749 ulp for erf(-1.2238890926331345) = -0.916519793341092348)

<tr><td align=left><a href=erf.res>erf</a></td><td align=right>1277</td><td align=right>0.55029</td><td>0.173003437415309991</td><td>0.19328324616993478</td></tr>