[FFmpeg-devel] [PATCH] lls: move the Cholesky decomposition in a separate function

[Michael Niedermayer]

On Sun, Nov 01, 2009 at 12:54:33AM +0100, Nicolas George wrote:
> Le decadi 10 brumaire, an CCXVIII, Michael Niedermayer a ?crit?:
> > The question i have is, why?
> 
> I intend to isolate the code that can be useful in other context that just
> the LPC computations. The Cholesty decomposition seemed a good place to
> start, since it is a very well defined and delimited part of the code.

In what other context would we want to use it?


> 
> And the comments were really missing for anyone who wanted to understand
> this code.

agree


> 
> > Also you completely rewrite the code during the move, making it much more
> > complex, harder to read (for me) and bigger (in terms of lines of code)
> 
> I rewrote the code to be sure that I really understood it. And in the end, I
> chose to submit my version because it did not have the hardcoded limit on
> the size of the matrice, although I admit it comes at the cost of harder to
> read pointer arithmetic.
> 
> I can re-submit the same with exactly your code if you prefer.
> 
> Or I can write code using a VLA prototype:
> 
> void function(... unsigned n, double m[n][])
> 
> if such code is not frowned upon (it is standard C99 but I expect it may
> make quite a few compilers chock).

Iam not convinced yet that the compile time limit is a problem
so iam not in favor of replacing it by more complex or less portable code.
what is a problem is that this limit that should be internal is in the
public header (it makes changing it tricky ...)

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The greatest way to live with honor in this world is to be what we pretend
to be. -- Socrates
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20091101/60ec9c48/attachment.pgp>