[benchmarks] Microbenchmark for dynamic_routing_table (drt_node perf) by cong1920 · Pull Request #3 · cong1920/lithium

cong1920 · 2026-03-16T05:14:52Z

What

Standalone microbenchmark for dynamic_routing_table to establish a performance baseline for the routing table's insert and lookup operations.

Benchmark covers

Route registration (insert) - 152 realistic REST API routes (static, parameterized, nested)
Route lookup (hit) - 1,044 concrete URLs x 1,000 reps = ~1M lookups
Route lookup (miss) - 7 non-matching URLs x 10,000 reps
RSS memory delta (Linux only)

How to run

# From repo root (Linux/WSL)
g++ -O2 -std=c++20 -I libraries/http_server -o bench_drt benchmarks/bench_drt.cc
./bench_drt 7   # 7 iterations, reports median

Current baseline results (WSL2, g++ 9.4, -O2)

Metric	master
Insert (ns/op)	285.8
Lookup hit (ns/op)	77.7
Lookup miss (ns/op)	45.4

This serves as the baseline to measure against when optimizing the routing table in the simplify_drt_node branch.

cong1920

added two comments

cong1920 · 2026-03-16T05:29:24Z

What

Standalone microbenchmark for dynamic_routing_table to measure the performance impact of the simplify_drt_node branch refactoring.

Benchmark covers

Route registration (insert) - 152 realistic REST API routes (static, parameterized, nested)

Route lookup (hit) - 1,044 concrete URLs x 1,000 reps = ~1M lookups

Route lookup (miss) - 7 non-matching URLs x 10,000 reps

RSS memory delta (Linux only)

How to run
g++ -O2 -std=c++20 -I libraries/http_server -o bench_drt benchmarks/bench_drt.cc
./bench_drt 7
Initial results (WSL2, g++ 9.4, -O2)

Metric master simplify_drt_node Delta
Insert (ns/op) 285.8 347.4 +21.5% slower
Lookup hit (ns/op) 77.7 80.7 +3.9% slower
Lookup miss (ns/op) 45.4 45.2 ~same
Note: simplify_drt_node wins on code clarity, ownership semantics, and O(1) table copy - raw lookup/insert throughput is comparable but not faster at this scale.

Can you update the result the initial results table to not include "simplify_drt_node" data and the delta? And update the comment to say this is current test result and will use this as baseline to improve in branch "simplify_drt_node"? I don't want to have the PR description to confuse repo owner, and other readers. Thanks!

cong1920

Commented again. Will let my local LLM do the work.

…ollision)

[benchmarks] Add microbenchmark for dynamic_routing_table insert/lookup

eced4d4