Packet Quantization Calculator

The pkt-quantize tool calculates bandwidth and packet rate metrics for packet-based interfaces where packets are quantized to bus beats. Given bus width, clock frequency, and minimum processing cycles, it computes performance across a range of packet sizes and generates tables, CSV files, and plots. The tool assumes packets cannot share a bus beat (e.g., AXI-Stream). There are protocols like AMBA CXS that allow multiple packets per beat. Such protocols are not subject to the quantization effects modeled by this tool.

---

Getting Started

Set Up and Install the Environment

See ABE Python Development for Python environment setup details.

make py-venv-all
source .venv/bin/activate
make py-install-all

Run Examples

pkt-quantize
pkt-quantize --bus-width 64 --clk-freq 1.5e9 --min-cycles 2 --min-size 64
--max-size 512

Examine Outputs

out_arch_pkt_quantize/*

Explore Relevant Directory Layout

├── mk
│   ├── 80-uarch-pq.mk
├── src
│   └── abe
│       ├── uarch
│       │   └── pkt_quantize.py
│       └── utils.py

---

Command Line Options

Option	Description	Default
--bus-width	Bus width (bytes)	128
--clk-freq	Clock frequency (Hz)	1.1e9
--min-cycles	Minimum clock cycles to process a packet	1
--min-size	Minimum packet size (bytes)	64
--max-size	Maximum packet size (bytes)	1518
--outdir	Output directory	out_uarch_pkt_quantize/
--no-plot-save	Skip saving plots to PNG	Off
--no-plot-show	Skip displaying plots in GUI	Off
--log-level	Logging level (debug, info, warning, error)	info

---

Outputs

File	Description
table_<config>_abbrev.txt	Abbreviated table of results (Markdown-style)
table_<config>_full.txt	Full table of results (Markdown-style)
table_<config>.csv	CSV-formatted results
plot_bw_<config>.png	Bandwidth vs packet size plot
plot_pps_<config>.png	PPS vs packet size plot

---

Import as a Library

You can import the PktQuantize class into another Python program.

from abe.uarch.pkt_quantize import PktQuantize

# Create instance with custom parameters
pq = PktQuantize(
    bus_width=64,
    clk_freq=1.5e9,
    min_cycles=2,
    min_size=64,
    max_size=512
)

# Run calculations
pq.calc()

# Access results
for size, stats in pq.pkt.items():
    print(f"Size: {size}, BW: {stats.bw / 1e9:.3f} Gbps")

---

Licensing

See the LICENSES directory at the repository root.