Architecture - Architecture - 2024.2 English

Vitis Libraries

Release Date
2025-05-14
Version
2024.2 English

The GEMM kernel here is an implementation of the operation C = A * B + X, where A, B, X, and C are matrices. This kernel is composed by three major parts, data movers, transpose and buffers, and a systolic array as shown in the following figure.

../../../_images/gemmKernel.png