CHAPTER 13
Buffered Flow Control
Adding buffers to our networks results in significantly more efficient flow control.
This is because a buffer decouples the allocation of adjacent channels. Without a
buffer, the two channels must be allocated to a packet (or flit) during consecutive
cycles, or the packet must be dropped or misrouted. There is nowhere else for the
packet to go. Adding a buffer gives us a place to store the packet (or flit) while waiting
for the second channel, allowing the allocation of the second channel to be delayed
without complications.
Once we add buffers to an interconnection network,our flow control mechanism
must allocate buffers

Get Principles and Practices of Interconnection Networks now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.