The IO Adapter subgraph delivers the \(\textbf{A}\) matrix from the input PLIO to the QRD subgraph. All downstream MUSIC subgraphs use buffers. These subgraphs can use a single I/O buffer read over an HBM interface, moving from tile to tile in a linear fashion. The folowing section explains this in detail. No bandwidth limitations are present downstream due to the use of the 256-bit AI Engine memory interface. However, two PLIO streams @ 32-bits for 1250 MHz must feed the design to achieve a 1 \(\mu s\) throughput overall. The IO Adapter subgraph sinks two input PLIO streams. It then combines them into a single output buffer containing the input \(\textbf{A}\) matrix for processing by the first QRD subgraph. This requires two streams because the \(128\times 8\) elements of \(\textbf{A}\) cannot transfer over a single PLIO in 1 \(\mu s\). The following block diagram shows the AI Engine physical array view for the IO Adapter subgraph.