-
Notifications
You must be signed in to change notification settings - Fork 8
Expand file tree
/
Copy pathperformance_baselines.json
More file actions
53 lines (53 loc) · 2.48 KB
/
performance_baselines.json
File metadata and controls
53 lines (53 loc) · 2.48 KB
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
43
44
45
46
47
48
49
50
51
52
53
{
"version": "1.13.0",
"updated": "2026-03-20",
"baselines": {
"SimpleSelect": {
"ns_per_op": 700,
"tolerance_percent": 60,
"description": "Basic SELECT query: SELECT id, name FROM users",
"current_performance": "~712 ns/op on Apple Silicon; ~800-1200 ns/op on CI (shared runners)",
"note": "Benchmarks now use ParseFromModelTokens (production path) directly - eliminates the legacy token-wrapping overhead. Baseline set at 1.5x M4 measurement to accommodate CI runner variability."
},
"ComplexQuery": {
"ns_per_op": 2660,
"tolerance_percent": 65,
"description": "Complex SELECT with JOIN, WHERE, ORDER BY, LIMIT",
"current_performance": "~2660 ns/op on Apple Silicon; ~3000-4500 ns/op on CI (shared runners)",
"note": "Benchmarks now use ParseFromModelTokens (production path) directly. Baseline set at 1.5x M4 measurement; wide tolerance for CI runner variability."
},
"WindowFunction": {
"ns_per_op": 1180,
"tolerance_percent": 60,
"description": "Window function query: ROW_NUMBER() OVER (PARTITION BY ... ORDER BY ...)",
"current_performance": "~1180 ns/op on Apple Silicon; ~1300-2000 ns/op on CI (shared runners)",
"note": "Benchmarks now use ParseFromModelTokens (production path) directly. Baseline set at 1.5x M4 measurement."
},
"CTE": {
"ns_per_op": 1200,
"tolerance_percent": 60,
"description": "Common Table Expression with WITH clause",
"current_performance": "~1200 ns/op on Apple Silicon; ~1300-2000 ns/op on CI (shared runners)",
"note": "Benchmarks now use ParseFromModelTokens (production path) directly. Baseline set at 1.5x M4 measurement."
},
"INSERT": {
"ns_per_op": 1000,
"tolerance_percent": 60,
"description": "Simple INSERT statement",
"current_performance": "~1008 ns/op on Apple Silicon; ~1100-1600 ns/op on CI (shared runners)",
"note": "Benchmarks now use ParseFromModelTokens (production path) directly. Baseline set at 1.5x M4 measurement."
},
"TokenizationThroughput": {
"tokens_per_sec": 8000000,
"tolerance_percent": 20,
"description": "Tokenizer throughput in tokens per second",
"note": "Measured separately via tokenizer benchmarks"
},
"EndToEndSustained": {
"ops_per_sec": 1380000,
"tolerance_percent": 20,
"description": "End-to-end sustained throughput in operations per second",
"note": "Measured via sustained load tests"
}
}
}