CUBLAS Sgemm help

Hi!

I’ve set up a simple matrix multiplier based on the simple CUBLAS example. This is great for returning a multiplied matrix but now I need the intermediate results (i.e for x = (1,2,3,4).(a,b,c,d) = 1xa + 2xb + 3xc + 4xd. I need to make a matrix containing the 1xa, 2xb, 3xc, 4xd values).

Is there a way to get Sgemm to return other data? Or will I have to wait until Nvidia gets back to me on my Registered Developer application so that I can look through the source code and edit it for my purposes?

Tom W