distributions/non_central_chi_squared.hpp fails in gcc 3.2.3

[joequant@…]

non_central_chi_squared.hpp fails in 3.2.3 for two reasons

gcc 3.2.3 does not have the emmintrin.h file. This can be fixed by disabiling SSE2 mode

a second problem is that igamma_large.hpp has problems with the static evaluate_polynomial and returns with an error

make[1]: * [eq1products.static-lib] Error 2 make[1]: Target `all' not remade because of errors. 1808: matching function for call to `evaluate_polynomial(long double[13], long 1808: double)' 1808: make[2]: * lib-linux.opt/volBEModel.o Error

This can be worked around by adding the polynomial size to the evaluate_polynomial functions that take the workspaces and figure a final value (line 260). This seems to be a template evaluation bug, since the static evaluate_polynomials seem to work fine in calculating the evaluating the elements of the workspace.

This problem was found in 1.39, but the code in trunk is identical

[John Maddock]

(In [64789]) Fix gcc-3.2.x compiler errors. Fixes #4507.