[with patch, positive review] complex double fast callable interpreter

[robertwb]

After RDF and RR, CDF would be very handy to have too.

[robertwb]

I started thinking about this as I was refereeing the original ticket, trying to make sure I understood how it all worked.

[cwitty]

Code looks good, doctests pass. Positive review, unless Michael complains about the portability issues.

Michael, I see two potential portability issues with the code:

1) it uses C99 complex numbers

2) it adds the compiler argument -std=c99

I'm guessing the latter will only work with gcc (unless other compilers specifically copy gcc command-line arguments). Does Sage currently build with non-gcc compilers?

[mabshoff]

Replying to cwitty:

Code looks good, doctests pass. Positive review, unless Michael complains about the portability issues.

Michael, I see two potential portability issues with the code:

1) it uses C99 complex numbers

That is not a problem. FLINT mandates that we use C99 anyway.

2) it adds the compiler argument -std=c99

We can work around that.

I'm guessing the latter will only work with gcc (unless other compilers specifically copy gcc command-line arguments). Does Sage currently build with non-gcc compilers?

Not at the moment, but there is work under way to support the pathscale compiler for SiCortex?.

Cheers,

Michael

[robertwb]

Thanks.

BTW, It compiles fine with gcc without the c99 flag, but I figured I'd put it there to be explicit. (I actually only knew about that flag because of flint.)

[mabshoff]

Merged in Sage 3.4.1.rc0.

Cheers,

Michael

[jason]

The work around for the bug in Cython < 0.11 can be removed because we upgraded to Cython 0.11 in this release of Sage, right?

from patch above:

# This is to work around a header ordering bug in Cython < 0.11 
# (Pari is included from sage.rings.complex_double.)