svdcmp: float to double accuracy

[evgeny]

Hi,

I need to adopt NR functions such as svdcmp(), qrdcmp(), tred(2), tqli() and pythag() to run with double accuracy.

What is a simplest way to do this?

evgeny

[yinyang]

Hi,

don't you just replace float with double? Will it affect something?

Greets
m.

[evgeny]

I am afraid that this is not so simple procedure:

1) svdcmp() uses exit from itertaions as comparison with zero in if operator

2) this may affected on small singular values and closed pairs of singular values

2) a few operations shoud be done with higher accuracy (double) also in the case of float function version

Evgeny