Silvio a Beccara wrote:

Dear Monsieur Lombardi,

thank you for your suggestion about the CEPHES library. I used your wrapper unit, modified for double types, and it works! I do not get that foolish error any longer. Do you know anything about its speed as compared to native GPC functions?

No way to compare with djgpp. There are no LongReal (long double: 80 bits) math routines there. Even those included in the x87 coprocessor are not declared in math.h nor in libm/math.h, and thus unknown in C (I think) and in gpc (I am sure). If I understand correctly, at least with the option -ffast-math they are used in intermediate calculations in C, which are done in the FP, but the final results are converted to double (real). Anyway after some unconclusive trials with these options when I begun to use gpc, I use CEPHES with a import unit each time I need mathematical accuracy.

Silvio a Beccara

ps: are your roots italian?

What are roots ? The father of my father was italian, born in france, turned back to italy afterwards, and I still know cousins in Italy (two of them have migrated meanwhile to Spain and USA, and married there). The mother of my father came from corsica, and thus was french to begin with, but migrated to italy afterwards. My father was born in france, and chose the french nationality when doing military service. My mother was french, and I have also cousins round here. The husband of my daughter is malian. And there is presumably a drop of yellow somewhere in the ancestors of my mother.

Maurice