In this chapter
This chapter describes the port merging process, where frames from multiple ports on a
single SmartNIC are merged.
Each MAC outputs frames from a port on the SmartNIC. The port merger combines the frames from different ports and forwards them to the subsequent blocks.
The frames are time-stamped with the end-of-frame or start-of-frame time when forwarded by
the port merger. This is done with an internal resolution of 1 ns. If two frames have
exactly the same time stamp, they are forwarded in port order (0, 1,
Note: Start-of-frame time-stamping does not apply to NT40E3-4-PTP SmartNICs running
on the capture/replay image, to NT40A01 SmartNICs running on the 4 × 10/1 Gbit/s SLB image
nor to the Intel® Programmable Acceleration Card with Intel® Arria®
10 GX FPGA.
Line-rate performance based on frame length
Line-rate frame handling is provided for frames according to the table in Performance
Undersized frames (shorter than 64 bytes) might be discarded in the port merger, if such
frames are received frequently. However, the port merger contains a small buffer for each
port, so bursts of short frames can be handled and forwarded in the processing pipe.
NT200A01 running at 2 × 100 Gbit/s ignores all frames shorter than 64 bytes.