This is the mail archive of the gsl-discuss@sources.redhat.com mailing list for the GSL project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
I also noticed something else about this code - although Legendre polynomials are "hard coded" up until P_3, in the general function, only up to P_2 is hard coded. Anyway, if you let me know which of the two error analyses you think is more approriate, I'll happily cook up a patch to sort it out.(a) Multiplication: Looking at coupling.c, it looks like the _relative_ generated error for multplication (A*B) is (2 * GSL_DBL_EPSILON). [See lines 73 and 74 of coupling.c]. However, I then looked at legendre_poly.c, as this contains some very simple algebra. Looking at line 84, I see that my assertion must be wrong. But then, comparing line 122, which is the same algebraic expression as line 84, gives an entirely different generated error analysis. What gives?
Yes, these are inconsistent by a factor of 3. This is a code
duplication problem, and I have very little idea how extensive
it is. Rubber mallet time?
What do you mean by truncation error?
thanks for the reference, I'll have a look at that.By the way, lately I have been influenced by a nice little book by Muller called "Elementary Functions: Algorithms and Implementation". I wish I had seen it before I started working on the special functions for GSL.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |