[efegokdemir]

Summary

Clarifies how to interpret NVBench timing, sampling, noise, batch, and throughput columns in benchmark output. Fixes #115.

Changes

• Documents adaptive sample counts for cold and batch measurements.
• Defines CPU Time, GPU Time, Noise, Batch GPU, and Batch.
• Explains Elem/s, GlobalMem BW, and BWPeak and how their state declarations affect results.
• Links measurement controls to the CLI stopping-criteria documentation.

Testing

• git diff --check — passed.
• pre-commit run --files docs/benchmarks.md — unavailable locally because pre-commit is not installed.
• No CUDA or CMake build was required for this documentation-only change.

Notes

The wording was checked against the existing measurement summaries in nvbench/detail/measure_cold.cu and nvbench/detail/measure_hot.cu, the CLI documentation, and the README example output. AI assistance was used to prepare this contribution; the changed lines and source references were reviewed before submission. No competing open PR for #115 was found.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.