rdkafka_performance: Add warmup flag

[Stephan Dollberg (StephanDollberg)]

Add flag to produce/consume additional N warmup messages. Stats are only
accounted after the warmup messages have been produced/consumed.

This makes the stats a lot more useful as it avoids noise from initial
message bursts.

Basically this prevents patterns like these:

~/librdkafka/examples/rdkafka_performance -P -t foo -c 4000000 ... -u
% Sending 4000000 messages of size 100 bytes
|    elapsed |       msgs |      bytes |        rtt |         dr |     dr_m/s |    dr_MB/s |     dr_err |     tx_err |       outq |     offset
|       1000 |      53244 |    5324400 |          0 |      52247 |      52246 |       5.22 |          0 |      39854 |        997 |    1676611
|       2000 |     156593 |   15659300 |          0 |     155594 |      77796 |       7.78 |          0 |     119926 |       1000 |    1679692
|       3000 |     261605 |   26160500 |          0 |     260606 |      86868 |       8.69 |          0 |     204584 |        999 |    1688160
|       4000 |     367411 |   36741100 |          0 |     366413 |      91602 |       9.16 |          0 |     291071 |        999 |    1692104
|       5000 |     471801 |   47180100 |          0 |     470804 |      94160 |       9.42 |          0 |     374877 |        998 |    1700897
|       6000 |     575822 |   57582200 |          0 |     574823 |      95803 |       9.58 |          0 |     458438 |        999 |    1704330
|       7000 |     679726 |   67972600 |          0 |     678728 |      96960 |       9.70 |          0 |     542196 |       1000 |    1710758

Now we immediately get proper steady state stats:

~/librdkafka/examples/rdkafka_performance -P -t foo -c 4000000 ...  -u -w 500000
% Sending 4500000 messages of size 100 bytes
|    elapsed |       msgs |      bytes |        rtt |         dr |     dr_m/s |    dr_MB/s |     dr_err |     tx_err |       outq |     offset
|       1000 |     106054 |   10605400 |          0 |     106052 |     106047 |      10.60 |          0 |      95080 |       1002 |    1780260
|       2000 |     212116 |   21211600 |          0 |     212114 |     106054 |      10.61 |          0 |     189620 |        999 |    1785319
|       3000 |     318316 |   31831600 |          0 |     318314 |     106102 |      10.61 |          0 |     284414 |       1001 |    1791383
|       4000 |     424627 |   42462700 |          0 |     424625 |     106154 |      10.62 |          0 |     379057 |       1002 |    1798530
|       5000 |     531569 |   53156900 |          0 |     531567 |     106311 |      10.63 |          0 |     474389 |       1000 |    1801665
|       6000 |     637679 |   63767900 |          0 |     637678 |     106278 |      10.63 |          0 |     569161 |        998 |    1810564
|       7000 |     743458 |   74345800 |          0 |     743456 |     106206 |      10.62 |          0 |     663704 |       1000 |    1818493

[confluent-cla-assistant]

🎉 All Contributor License Agreements have been signed. Ready to merge.
✅ StephanDollberg
_{Please push an empty commit if you would like to re-run the checks to verify CLA status for all contributors.}

[Copilot]

Pull request overview

This PR adds a -w <cnt> warmup flag to the rdkafka_performance benchmarking tool. It produces/consumes an initial batch of "warmup" messages, then resets all stats counters so that only the steady-state throughput is reflected in the reported numbers. This avoids misleading initial startup noise in benchmark output.

Changes:

• Adds warmup_cnt global and the warmup reset logic in print_stats, which zeros cnt and adjusts msgcnt once warmup messages have been processed.
• Adds the -w CLI option (getopt string, argument handler, and help text).
• Adjusts msgcnt before the run loop to include warmup messages in the total.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.