-
Book Overview & Buying
-
Table Of Contents
Modern Computer Architecture and Organization - Third Edition
By :
The GA102 tensor cores are designed to accelerate computations of geometric transformations in graphics display and the tensor operations in artificial intelligence applications. Each tensor core is a programmable matrix multiply and accumulate unit.
Unlike the multiplication of two scalars, the order of the operands in matrix multiplication is significant, just as it is when dividing one scalar by another. Furthermore, the multiplication of two matrices is only possible if the number of columns in the first matrix is equal to the number of rows in the second. These dimensions are referred to as M (the number of rows in the first matrix), N (the number of columns in the second matrix), and K (the number of columns in the first matrix and the number of rows in the second matrix). The result of a matrix multiplication with these dimensions is a matrix with M rows and N columns.
The computational units in the GA102 tensor core support a limited number of relatively small...