Re: Complex Exponential Integral

[Erik L. <eri...@gm...>]

Apologies - I only sent this to Ethan directly.

---------- Forwarded message ---------

Thank you. I can confirm that this works properly. (I only tried it
with/without ZEXINT and CEXINT) Just two small (cosmetic) points:

* There is a closing parenthesis missing in configure. The corrected
configure.ac is attached

* In my setup, configure detects ZEXINT because I set LIBAMOS_LDFLAGS
(before calling configure), which then gets added automatically to LD_FLAGS:

checking for amos libraries using LDFLAGS:  -L/usr/local/lib -lcerf
-lopenspecfun -lgfortran /usr/local/lib/libzexpint.a ...

checking for library containing zairy_... none required

checking for library containing zexint_... none required


I noticed that configure then adds "-L/home/local/lib64" in LIBS in
the Makefile, which may be specific to your system and not serve a purpose
in other systems? (It is hard-wired in configure.)

Best,

Erik

On Thu, Jul 10, 2025 at 3:52 PM Ethan A Merritt <me...@uw...> wrote:

> I have just pushed changes to the configure script and to amos_airy,c
> so that ZEXINT is used if found during configuration in preference to
> CEXINT.
> I tested configure+build against libraries with or without ZEXINT.
>
> commit dda9e5051d
>
> Since both routines were available for testing, I also sampled the
> real/imaginary
> plane for the difference in output between parallel calls to CEXINT and
> ZEXINT
> for orders 2,3,4, and 5.
> The maximum difference found was on the order of    3 x 10^{-15}
> Pure noise level for double precision would be about  2 x 10^{-16}
> Pure noise level for single precision would be about    1 x 10^{-7}.
>
> cheers,
>         Ethan
>
>
>