Measuring CUDA performance with a vector addition

Measure the time between two additions, one with CUDA, one with numpy. The script can be profiled with Nsight.

nsys profile python _doc/examples/plot_bench_cuda_vector_add.py

Vector Add

from tqdm import tqdm
import numpy
import matplotlib.pyplot as plt
from pandas import DataFrame
from teachcompute.ext_test_case import measure_time, unit_test_going
import torch

has_cuda = torch.cuda.is_available()

try:
    from teachcompute.validation.cuda.cuda_example_py import vector_add
except ImportError:
    has_cuda = False


def cuda_vector_add(values):
    torch.cuda.nvtx.range_push(f"CUDA dim={values.size}")
    res = vector_add(values, values, 0)
    torch.cuda.nvtx.range_pop()
    return res


obs = []
dims = [2**10, 2**15, 2**20, 2**25]
if unit_test_going():
    dims = [10, 20, 30]
for dim in tqdm(dims):
    values = numpy.ones((dim,), dtype=numpy.float32).ravel()

    if has_cuda:
        diff = numpy.abs(vector_add(values, values, 0) - (values + values)).max()
        res = measure_time(lambda: cuda_vector_add(values), max_time=0.5)

        obs.append(
            dict(
                dim=dim,
                size=values.size,
                time=res["average"],
                fct="CUDA",
                time_per_element=res["average"] / dim,
                diff=diff,
            )
        )

    diff = 0
    res = measure_time(lambda: values + values, max_time=0.5)

    obs.append(
        dict(
            dim=dim,
            size=values.size,
            time=res["average"],
            fct="numpy",
            time_per_element=res["average"] / dim,
            diff=0,
        )
    )


df = DataFrame(obs)
piv = df.pivot(index="dim", columns="fct", values="time_per_element")
print(piv)

  0%|          | 0/4 [00:00<?, ?it/s]
 25%|██▌       | 1/4 [00:01<00:04,  1.46s/it]
 50%|█████     | 2/4 [00:02<00:02,  1.29s/it]
 75%|███████▌  | 3/4 [00:04<00:01,  1.34s/it]
100%|██████████| 4/4 [00:06<00:00,  1.82s/it]
100%|██████████| 4/4 [00:06<00:00,  1.65s/it]
fct               CUDA         numpy
dim
1024      1.271343e-06  6.823135e-10
32768     5.698531e-08  1.918299e-10
1048576   1.995813e-08  6.269388e-10
33554432  1.787547e-08  1.253388e-09

Plots

piv_diff = df.pivot(index="dim", columns="fct", values="diff")
piv_time = df.pivot(index="dim", columns="fct", values="time")

fig, ax = plt.subplots(1, 3, figsize=(12, 6))
piv.plot(ax=ax[0], logx=True, title="Comparison between two summation")
piv_diff.plot(ax=ax[1], logx=True, logy=True, title="Summation errors")
piv_time.plot(ax=ax[2], logx=True, logy=True, title="Total time")
fig.tight_layout()
fig.savefig("plot_bench_cuda_vector_add.png")

/home/xadupre/.local/lib/python3.10/site-packages/pandas/plotting/_matplotlib/core.py:822: UserWarning: Data has no positive values, and therefore cannot be log-scaled.
  labels = axis.get_majorticklabels() + axis.get_minorticklabels()

CUDA seems very slow but in fact, all the time is spent in moving the data from the CPU memory (Host) to the GPU memory (device).