By: Michael S (already5chosen.delete@this.yahoo.com), November 5, 2006 5:03 pm
Room: Moderated Discussions
Eric Bron (eric.bron@adeptdevelopment.com) on 11/5/06 wrote:
---------------------------
>>I don't understand. Do you mean 43ns/iter on C2D? Why so slow?
>
>ooops, yes, I wasn't very clear
>
>the loops are something like :
>
>
>
>so the vectorized loop has 200 iterations
>
>the bench runs 10'000 x each test loop., for a total of 2'000'000 iterations
>
>
>hope it makes more sense
>
Nop. It's worse than before. No we have 175ns/iter.
---------------------------
>>I don't understand. Do you mean 43ns/iter on C2D? Why so slow?
>
>ooops, yes, I wasn't very clear
>
>the loops are something like :
>
>
>
>float *a,*b,*c,*d; // alloc size = 3200 bytes each
>[..]
>for (i=0; i<800; i++)
>d[i] = a[i]*b[i]+c[i];
>
>
>
>so the vectorized loop has 200 iterations
>
>the bench runs 10'000 x each test loop., for a total of 2'000'000 iterations
>
>
>hope it makes more sense
>
Nop. It's worse than before. No we have 175ns/iter.
