GoSQLX/performance_baselines.json at main · ajitpratap0/GoSQLX · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
{
  "version": "1.17.0",
  "updated": "2026-10-22",
  "purpose": "SMOKE TEST ONLY. These ns/op ceilings and their 60-65% tolerances exist so TestPerformanceRegression catches order-of-magnitude breakage on noisy CI. They are not a performance gate and must not be tightened to act as one. Regressions are judged by the interleaved A/B gate (scripts/perf/ab.sh, task bench:gate) and the deterministic allocs/op and B/op baselines in benchmarks/baselines/. See docs/PERFORMANCE_METHODOLOGY.md.",
  "baselines": {
    "SimpleSelect": {
      "ns_per_op": 800,
      "tolerance_percent": 60,
      "description": "Basic SELECT query: SELECT id, name FROM users",
      "current_performance": "~349 ns/op on Apple M4 (median of 20 runs, re-measured 2026-10-22); ~467-918 ns/op typical on CI (GitHub ubuntu-latest Benchmark job, range over the last 9 Tests runs on main, 0715bb9..61ca6c8)",
      "note": "Measured with BenchmarkPerformanceBaseline (ParseFromModelTokens production path), -count=10 -benchmem, median. Baseline set at ~2.5x the M4 median: GitHub ubuntu-latest runners measure about 1.5-3.2x slower than M4 on these queries (up to ~3.2x), so a 1.5x baseline (limit ~2.4x M4 at this tolerance) would fail CI's slowest runs."
    },
    "ComplexQuery": {
      "ns_per_op": 2900,
      "tolerance_percent": 65,
      "description": "Complex SELECT with JOIN, WHERE, ORDER BY, LIMIT",
      "current_performance": "~1234 ns/op on Apple M4 (median of 20 runs, re-measured 2026-10-22); ~1842-3386 ns/op typical on CI (GitHub ubuntu-latest Benchmark job, range over the last 9 Tests runs on main, 0715bb9..61ca6c8)",
      "note": "Measured with BenchmarkPerformanceBaseline (ParseFromModelTokens production path), -count=10 -benchmem, median. Baseline set at ~2.5x the M4 median: GitHub ubuntu-latest runners measure about 1.5-3.2x slower than M4 on these queries (up to ~3.2x), so a 1.5x baseline (limit ~2.4x M4 at this tolerance) would fail CI's slowest runs."
    },
    "WindowFunction": {
      "ns_per_op": 1500,
      "tolerance_percent": 60,
      "description": "Window function query: ROW_NUMBER() OVER (PARTITION BY ... ORDER BY ...)",
      "current_performance": "~631 ns/op on Apple M4 (median of 20 runs, re-measured 2026-10-22); ~940-1876 ns/op typical on CI (GitHub ubuntu-latest Benchmark job, range over the last 9 Tests runs on main, 0715bb9..61ca6c8)",
      "note": "Measured with BenchmarkPerformanceBaseline (ParseFromModelTokens production path), -count=10 -benchmem, median. Baseline set at ~2.5x the M4 median: GitHub ubuntu-latest runners measure about 1.5-3.2x slower than M4 on these queries (up to ~3.2x), so a 1.5x baseline (limit ~2.4x M4 at this tolerance) would fail CI's slowest runs."
    },
    "CTE": {
      "ns_per_op": 1400,
      "tolerance_percent": 60,
      "description": "Common Table Expression with WITH clause",
      "current_performance": "~644 ns/op on Apple M4 (median of 20 runs, re-measured 2026-10-22); ~824-1692 ns/op typical on CI (GitHub ubuntu-latest Benchmark job, range over the last 9 Tests runs on main, 0715bb9..61ca6c8)",
      "note": "Measured with BenchmarkPerformanceBaseline (ParseFromModelTokens production path), -count=10 -benchmem, median. Baseline set at ~2.5x the M4 median: GitHub ubuntu-latest runners measure about 1.5-3.2x slower than M4 on these queries (up to ~3.2x), so a 1.5x baseline (limit ~2.4x M4 at this tolerance) would fail CI's slowest runs."
    },
    "INSERT": {
      "ns_per_op": 950,
      "tolerance_percent": 60,
      "description": "Simple INSERT statement",
      "current_performance": "~406 ns/op on Apple M4 (median of 20 runs, re-measured 2026-10-22); ~605-1169 ns/op typical on CI (GitHub ubuntu-latest Benchmark job, range over the last 9 Tests runs on main, 0715bb9..61ca6c8)",
      "note": "Measured with BenchmarkPerformanceBaseline (ParseFromModelTokens production path), -count=10 -benchmem, median. Baseline set at ~2.5x the M4 median: GitHub ubuntu-latest runners measure about 1.5-3.2x slower than M4 on these queries (up to ~3.2x), so a 1.5x baseline (limit ~2.4x M4 at this tolerance) would fail CI's slowest runs."
    }
  }
}