============================================================================
OmniBuf-Py Benchmark Suite (Python 3.14.5 / darwin-arm64 / Apple M1 Pro)
Workload: 11-field PeerUpdate (366B OmniBuf, 341B Protobuf) + >16 KiB stress cases
Methodology: Full decode reads all 11 fields; mutations check city and increment/grow.
============================================================================

### 1. Full Message Encode Speed

| Codec / Operation | Latency (ns/op) | Throughput (ops/sec) | Wire Size (B) | vs Baseline |
|---|---:|---:|---:|---:|
| OmniBuf (Generated) | 527 ns | 1,899,162 | 366 | 1.00x (baseline) |
| OmniBuf (@dataclass) | 501 ns | 1,995,069 | 366 | 1.05x faster |
| protobuf-py | 1,237 ns | 808,125 | 341 | 2.35x slower |
| google.protobuf (upb) | 565 ns | 1,768,456 | 341 | 1.07x slower |
| msgspec | 827 ns | 1,209,346 | 634 | 1.57x slower |
| msgpack | 2,225 ns | 449,503 | 634 | 4.23x slower |
| json (stdlib) | 6,294 ns | 158,878 | 874 | 11.95x slower |

### 2. Full Message Decode Speed

| Codec / Operation | Latency (ns/op) | Throughput (ops/sec) | vs Baseline |
|---|---:|---:|---:|
| OmniBuf (Generated) | 2,662 ns | 375,620 | 1.00x (baseline) |
| OmniBuf (@dataclass) | 2,725 ns | 366,987 | 1.00x (baseline) |
| protobuf-py | 4,110 ns | 243,313 | 1.54x slower |
| google.protobuf (upb) | 6,078 ns | 164,539 | 2.28x slower |
| msgspec | 3,182 ns | 314,224 | 1.20x slower |
| msgpack | 5,669 ns | 176,407 | 2.13x slower |
| json (stdlib) | 8,795 ns | 113,702 | 3.30x slower |

### 3. Read a Single Nested Field

| Codec / Operation | Latency (ns/op) | Throughput (ops/sec) | vs Baseline |
|---|---:|---:|---:|
| OmniBuf (Generated) | 564 ns | 1,772,128 | 1.00x (baseline) |
| protobuf-py | 3,016 ns | 331,620 | 5.34x slower |
| google.protobuf (upb) | 1,028 ns | 972,376 | 1.82x slower |
| msgspec | 1,531 ns | 653,001 | 2.71x slower |
| msgpack | 3,667 ns | 272,676 | 6.50x slower |
| json (stdlib) | 6,222 ns | 160,722 | 11.03x slower |

### 4. Selective Single-Field Mutation

| Codec / Operation | Latency (ns/op) | Throughput (ops/sec) | vs Baseline |
|---|---:|---:|---:|
| OmniBuf (Generated) | 1,298 ns | 770,407 | 1.00x (baseline) |
| protobuf-py | 4,211 ns | 237,481 | 3.24x slower |
| google.protobuf (upb) | 1,792 ns | 558,173 | 1.38x slower |
| msgspec | 2,409 ns | 415,033 | 1.86x slower |
| msgpack | 6,003 ns | 166,577 | 4.62x slower |
| json (stdlib) | 12,943 ns | 77,260 | 9.97x slower |

### 5. Large Embedded Message Encode

| Codec / Operation | Latency (ns/op) | Throughput (ops/sec) | Wire Size (B) | vs Baseline |
|---|---:|---:|---:|---:|
| OmniBuf (Generated) | 3,274 ns | 305,464 | 39256 | 1.00x (baseline) |
| OmniBuf (@dataclass) | 3,277 ns | 305,125 | 39256 | 1.00x (baseline) |
| protobuf-py | 4,456 ns | 224,430 | 39227 | 1.36x slower |
| google.protobuf (upb) | 2,889 ns | 346,167 | 39227 | 1.13x faster |
| msgspec | 2,367 ns | 422,555 | 39516 | 1.38x faster |
| msgpack | 4,100 ns | 243,882 | 39516 | 1.25x slower |
| json (stdlib) | 89,592 ns | 11,162 | 39752 | 27.37x slower |

### 6. Embedded Message Growth

| Codec / Operation | Latency (ns/op) | Throughput (ops/sec) | vs Baseline |
|---|---:|---:|---:|
| OmniBuf (Generated) | 5,202 ns | 192,227 | 1.00x (baseline) |
| protobuf-py | 18,255 ns | 54,778 | 3.51x slower |
| google.protobuf (upb) | 5,912 ns | 169,144 | 1.14x slower |
| msgspec | 9,434 ns | 106,003 | 1.81x slower |
| msgpack | 24,925 ns | 40,121 | 4.79x slower |
| json (stdlib) | 83,964 ns | 11,910 | 16.14x slower |
