problem with lgamma function in f90

[Krzysztof Kielak]

Hi,

I have problems calling lgamma function from fortran 90. In the following example the returned value should be 0 (because lgamma(1) = 0), but it's not:

rcc:fit# cat test.f
double precision x, r
x = 1.0
write (*,*) lgamma(x)
end
rcc:fit# f90 -lm test.f
test.f
main program

6 Lines Compiled
/usr/ccs/bin/ld: (Warning) At least one PA 2.0 object file (test.o) was detected. The linked output may not run on a PA 1.x system.
rcc3:fit# ./a.out
1082562470
rcc:fit#


Does anyone has any ideas why?
best regards
Chris

[Nicolas Dumeige]

Hello,

My guess is that the linker is telling you that the builted binary is 64 bits and won't work on a 32 system.

Cheers,

Nicolas

[A. Clay Stephenson · Accepted solution]

No, that is a normal warning simply indicating that this code may not run on all HP boxes.

I just wrote a quickie C program and linked with -lm; the underlying lgamma function appears fine. It would be good to know if this is a problem with the write function (or perhaps the implicit format) or with the compiler itself. I would first assign the result of lgamma to your double precision variable, r. Write r using your existing format; if that fails use an explicit format and see if the results are any better. This will help separate out the "real" problem. You might want to stop the f90 compile at the assembly step and have a look at the .s source code.


It's always a good idea to look for cumulative compiler/linker/library patches but since you didn't bother to mention your version of compiler or HP-UX, I'm not going to bother looking.

[Nicolas Dumeige]

I've googled it :

http://h21007.www2.hp.com/dspp/tech/tech_TechSingleTipDetailPage_IDX/1,2366,399,00.html

[Mike Stroyan]

You need to specify the return type and the parameter passing convention before you call a function with a "C" language binding. The alias directive can be used to tell f90 that the parameter is passed by value instead of the fortran convention of pass by reference.

!$HP$ ALIAS lgamma(%VAL)
double precision lgamma
double precision x, r
x = 1.0
write (*,*) lgamma(x)
x = 3.0
write (*,*) lgamma(x)
x = 3.06
write (*,*) lgamma(x)
end