mirror of
https://github.com/NVIDIA/nvbench.git
synced 2026-06-29 18:57:44 +00:00
48b7f61da3
Implemented the clear-gap comparison, with the log-distance-equivalent
algebra and pessimistic SM-clock fallback.
What changed:
- Added TimingInterval and interval construction from summaries:
- robust interval: [min, q3], centered at median
- fallback interval: clipped [mean - stdev, mean + stdev] intersected with [min, max]
- Added CLEAR_GAP_RELATIVE_THRESHOLD = 0.005.
- FAST gap uses:
(ref.lower - cmp.upper) / cmp.upper >= delta
which is equivalent to log(ref.lower / cmp.upper) >= log(1 + delta).
- SLOW gap uses:
(cmp.lower - ref.upper) / ref.upper >= delta
- FAST/SLOW now requires SM clock summaries on both sides and the same clear-gap result after scaling intervals by sm_clock_rate_mean.
- If intervals are missing, overlap, fail the gap threshold, have missing/invalid clock summaries, or time/cycle comparison disagrees, status is UNDECIDED.
- Existing center/noise values are still computed and displayed, but no longer drive FAST/SLOW/SAME classification.
Updated tests to cover:
- center/noise-only comparisons becoming UNDECIDED
- clear FAST/SLOW with matching clock evidence
- missing clock fallback to UNDECIDED
- frequency-shift disagreement becoming UNDECIDED
- regression reporting with robust interval and clock evidence
CUDA Kernel Benchmarking Package
This package provides a Python API to the CUDA Kernel Benchmarking
Library NVBench.
Installation
Install from PyPi
pip install cuda-bench[cu13] # For CUDA 13.x
pip install cuda-bench[cu12] # For CUDA 12.x
Building from source
Ensure recent version of CMake
Since nvbench requires a rather new version of CMake (>=3.30.4), either build CMake from sources, or create a conda environment with a recent version of CMake, using
conda create -n build_env --yes cmake ninja
conda activate build_env
Ensure CUDA compiler
Since building NVBench library requires CUDA compiler, ensure that appropriate environment variables
are set. For example, assuming CUDA toolkit is installed system-wide, and assuming Ampere GPU architecture:
export CUDACXX=/usr/local/cuda/bin/nvcc
export CUDAARCHS=86
Build Python project
Now switch to python folder, configure and install NVBench library, and install the package in editable mode:
cd nvbench/python
pip install -e .
Verify that package works
python test/run_1.py
Run examples
# Example benchmarking numba.cuda kernel
python examples/throughput.py
# Example benchmarking kernels authored using cuda.core
python examples/axes.py
# Example benchmarking algorithms from cuda.cccl.parallel
python examples/cccl_parallel_segmented_reduce.py
# Example benchmarking CuPy function
python examples/cupy_extract.py