Re: Odd behavior of time-domain convolutions

[Quentin Spencer]

John W. Eaton wrote:
> On  5-Sep-2007, Quentin Spencer wrote:
>
> | John W. Eaton wrote:
> | > On  5-Sep-2007, Quentin Spencer wrote:
> | >
> | > | I've used the wrapper 
> | > | approach before where the only difference was the interface code, and 
> | > | even then the mex code was a little faster, but not by the large margins 
> | > | you are observing.
> | >
> | > I'm surprised by that becuase I think more work has to be done to set
> | > up and recover from calling a function with the MEX interface.  Also,
> | > in some cases copies of the data passed to and from a MEX file are
> | > required when no copy is needed with the .oct file interface.
> | 
> | It may be that my problem was what David pointed out--using the indexing 
> | functions that bring in a lot of extra overhead.
>
> You say that the only difference is the inteface code.  Did the
> interface code perform indexing on Array/Matrix values?
>   

Yes. I was taking double values and recasting one at a time into an 
array of integers for the low-level code. I was using the X(i,j) 
indexing to get the double values. I take it that extracting the array 
pointers and doing the low-level indexing directly would be faster.

Quentin