NVIDIA/nvbench
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvbench.git synced 2026-05-13 17:55:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add decorators for registering benchmarks and adding axis

Browse Source 
cuda.bench.register(fn) continues returning Benchmark, and supports
legacy use.

New signature added:
   cuda.bench.register():
      Returns a decorator

```
@bench.register()
@bench.axis.float64("Duration (s)", [7e-5, 1e-4, 5e-4])
@bench.option.min_samples(120)
def single_float64_axis(state: bench.State):
   ...
```
This commit is contained in:
Oleksandr Pavlyk
2026-05-04 08:21:41 -05:00
parent 9ea77bccaa
commit e07f87910a
15 changed files with 585 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
python/examples/throughput.py
View File

@@ -1,4 +1,4 @@
# Copyright 2025 NVIDIA Corporation
# Copyright 2025-2026 NVIDIA Corporation
#
# Licensed under the Apache License, Version 2.0 with the LLVM exception
# (the "License"); you may not use this file except in compliance with
@@ -39,6 +39,9 @@ def make_throughput_kernel(items_per_thread: int) -> cuda.dispatcher.CUDADispatc
return kernel
@bench.register()
@bench.axis.int64("Stride", [1, 2, 4])
@bench.axis.int64("ItemsPerThread", [1, 2, 3, 4])
def throughput_bench(state: bench.State) -> None:
stride = state.get_int64("Stride")
ipt = state.get_int64("ItemsPerThread")
@@ -69,8 +72,4 @@ def throughput_bench(state: bench.State) -> None:
if __name__ == "__main__":
b = bench.register(throughput_bench)
b.add_int64_axis("Stride", [1, 2, 4])
b.add_int64_axis("ItemsPerThread", [1, 2, 3, 4])
bench.run_all_benchmarks(sys.argv)