Re: Fwd: 'for' loop vectorization

[David Bateman]

Jordi Gutiérrez Hermoso wrote:
> On 24/10/2007, David Bateman <address@hidden> wrote:
>> I went to a lot of effort a while back to get a memory efficient and
>> fast triu/tril in an oct-file. However, at the time I proposed it John
>> thought that it was too hard to maintain relative to the m-file version
>> and the amount of use that triu/tril see doesn't justify this additional
>> maintenance .
> 
> So what does the competition do here? triu seems like a reasonable
> function to be used in vectorisation, and I really want my "clever"
> vectorisation of the cummax function to work. If triu isn't the way to
> vectorise, then I'll have to come up with another solution to the
> cummax vectorisation puzzle.
> 

Matlab uses builtin versions of triu and tril.. Have you tried the code
for triu/tril I sent? Does it give a significant advantage?

D.