What complexity have a log(sum) shape?

[Andrei Alexandrescu via Digitalmars-d]

On 12/08/2015 04:55 PM, Timon Gehr wrote:
> The context is asymptotic runtime bounds. The special cases for small
> values of continuous logarithms can just be defined away. They have no
> relevance for asymptotic runtime analysis (even though defining big-O
> adequately for multiple variables is more tricky than for a single
> variable).

 From the cycle "Andrei's esprits d'escalier", the inequality can be 
derived from the somewhat notorious log sum inequality if you make all 
bi equal: 
http://ocw.usu.edu/Electrical_and_Computer_Engineering/Information_Theory/lecture3.pdf

Andrei