Documentation
¶
Overview ¶
Package benchmarks ports the dmonad/crdt-benchmarks B1-B4 workload suite to ygo. The benchmark IDs (B1.1 .. B4) match the upstream spec verbatim so cross-implementation comparison stays apples-to- apples — see README in https://github.com/dmonad/crdt-benchmarks for the canonical descriptions.
Methodology per benchmark:
- "ops" sub-benchmark times the workload itself (mutations into a fresh Doc), no encode.
- "encode_v1" / "encode_v2" sub-benchmarks time the wire encoder against a pre-built Doc; report doc-size in bytes via b.ReportMetric.
- "parse_v1" / "parse_v2" sub-benchmarks time DecodeUpdate + Apply against a fresh receiver Doc.
Run: go test -bench=. -benchmem ./benchmarks/ Run one: go test -bench=B1_1 -benchmem ./benchmarks/
B4 expects the upstream automerge-paper trace at testdata/b4-trace.json.gz — skipped if absent. See testdata/gen/fetch-b4-trace.sh.
Index ¶
Constants ¶
const N = 6000
N is the canonical operation count per dmonad/crdt-benchmarks B1 workloads (6,000 ops). Kept as a const so all B1.x scenarios use the same scale for cross-comparison.
const N2 = N / 2
N2 is the per-peer op count for B2 concurrent workloads. Each of two peers does N2 ops, so total ops is 2*N2 = 6000 (matches B1).
const N3 = 489
N3 is the client count for B3 high-conflict workloads. Spec says "20 * sqrt(N)" where N=600 → 489 clients. We use the spec value directly rather than recomputing.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
This section is empty.
Source Files