Data Layout in AMD GPU

Discussion created by gpgpu on Feb 20, 2010
Latest reply on Feb 26, 2010 by gaurav.garg

Dear all:

I wonder how data is stored in GPU. ie, if I have a 2-D array A[3][4] on CPU, after using streamRead (Agpu, A), what is the data layout? Should it be:

A[0,0]  A[0,1] A[0,2] A[0,3]

A[1,0]  A[1,1] A[1,2] A[1,3]

A[2,0]  A[2,1] A[2,2] A[2,3]

as in CPU?

Making sure this problem will help me to understand the mat_mul code in user guide, becoz that code gives me the impression that C = A's col * B's row. Thanks in advance.