Re: oct file: slower than expected using fortran_vec

[Seb Astien]

On Mon, May 16, 2011 at 7:29 AM, Andy Buckle <address@hidden> wrote:
> On Sun, May 15, 2011 at 9:10 PM, Seb Astien <address@hidden> wrote:
>> Hi,
>>
>> I have been experimenting a bit with oct files to see how much faster
>> it would be.
>> I wrote a trivial example to sum the elements of an array.
>> To my surprised, this was still much slower than built-in functions:
>>
>> octave:1> A=rand(10000,10000);
>> octave:2> tic; s1=sum(sum(A)); toc
>> Elapsed time is 0.222649 seconds.
>> octave:3> tic; s2=sumit(A); toc
>> Elapsed time is 1.5032 seconds.
>>
>> The C++ code is below. I added a const in front of *p to prevent
>> copying, but I guess it is still happening.
>>
>> Any comment, pointer, explanation would be greatly appreciated.
>>
>> Regards,
>>
>> Seb
>>
>> ===
>> #include <octave/oct.h>
>>
>> DEFUN_DLD (sumit, args, , "Sum elements of an array")
>> {
>>        int nargin = args.length();
>>
>>        if (nargin != 1){
>>                print_usage();
>>                return octave_value_list();
>>        }
>>
>>        NDArray A = args(0).array_value();
>>        const double *p = A.fortran_vec();
>>        double sum = 0;
>>        octave_idx_type N = args(0).nelem();
>>
>>        if (! error_state){
>>                for(octave_idx_type i=0; i<N; i++)
>>                        sum += *(p++);
>>                return octave_value (sum);
>>        }
>> }
>> ===
>
> The first time you run sumit, it will have to be loaded. This will ad
> overhead. If you run sumit a few times, what is the comparison like?
>
>
>
> --
> /* andy buckle */
>

It is a bit faster the second time, but it does not explain the gap
between the two:
The bigger the matrix, the bigger the gap between built-in function
and the oct one.
I suspect a copying taking it place somehwere.

octave:12> a=rand(1000,1000);
octave:13> tic; sum(sum((a))); toc
Elapsed time is 0.00299295 seconds.
octave:14> tic; sumit(a); toc
Elapsed time is 0.020828 seconds.
octave:15> a=rand(10000,10000);
octave:16> tic; sum(sum((a))); toc
Elapsed time is 0.208262 seconds.
octave:17> tic; sumit(a); toc
Elapsed time is 1.49816 seconds.

Seb