Skip to content

PCIE bandwidth measure #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

PCIE bandwidth measure #2

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
879e84c
bandwidth measure
chsasank Feb 17, 2024
3179a6d
add better message
chsasank Feb 19, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 13 additions & 6 deletions benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@
num_trails = args.num_trails


def flops_benchmark(device):
def flops_benchmark():
test_range = 2 ** np.arange(8, 13, 0.25)

print('size, elapsed_time, flops')
Expand Down Expand Up @@ -57,9 +57,11 @@ def synchronize(device):
pass


def memory_bandwidth_benchmark(device):
def memory_bandwidth_benchmark(from_device=device):
from_device = torch.device(from_device)
test_range = 2 ** (np.arange(20, 27, 0.5))

print(f'measuring bw from {from_device} to {device}')
print('size (GB), elapsed_time, bandwidth')
for size in test_range:
elapsed_time = 0
Expand All @@ -68,7 +70,7 @@ def memory_bandwidth_benchmark(device):

# Create random tensors
a = torch.rand(size, device=device)
b = torch.rand(size, device=device)
b = torch.rand(size, device=from_device)

# Warm-up to ensure CUDA kernel is initialized if using GPU
synchronize(device)
Expand All @@ -93,7 +95,12 @@ def memory_bandwidth_benchmark(device):
elapsed_time = elapsed_time / num_trails
# Calculate Bandwidth in GB/s
bytes_copied = a.nelement() * a.element_size() # bytes
bandwidth = 2 * bytes_copied / elapsed_time / 1e9 # GB/s
if from_device == device:
# because data has transferred from and back
factor = 2
else:
factor = 1
bandwidth = factor * bytes_copied / elapsed_time / 1e9 # GB/s

print(bytes_copied / 1e9, elapsed_time, bandwidth, sep=', ')

Expand All @@ -102,5 +109,5 @@ def memory_bandwidth_benchmark(device):

if __name__ == "__main__":
print(f'benchmarking {device}')
flops_benchmark(device)
memory_bandwidth_benchmark(device)
flops_benchmark()
memory_bandwidth_benchmark()