Skip to content

P2P tensors must be dense #2161

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 4 commits into
base: main
Choose a base branch
from
Open

P2P tensors must be dense #2161

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9aa908b
P2P tensors must be dense
Chao1Han Oct 13, 2025
b02bc49
Merge branch 'main' into xccl/dense
Chao1Han Oct 14, 2025
900897c
Merge branch 'main' into xccl/dense
Chao1Han Oct 15, 2025
2db1604
Merge branch 'main' into xccl/dense
Chao1Han Oct 16, 2025
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
8 changes: 7 additions & 1 deletion src/xccl/ProcessGroupXCCL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,9 +99,15 @@ void checkSingleTensor(
C10_THROW_ERROR(ValueError, "Tensors must be XPU and dense");
}

// Skip the following requirements for P2P operations
// Check memory format
if (!tensor.is_contiguous(tensor.suggest_memory_format())) {
// P2P is a bit relaxed, supporting transfer of a transposed tensor
if (p2p) {
// But must be dense still
if (!tensor.is_non_overlapping_and_dense()) {
C10_THROW_ERROR(
ValueError, "Tensors for P2P must be non-overlapping and dense");
}
TORCH_WARN_ONCE(
"Detected non-contiguous tensor in P2P operations. It is user "
"responsibility to guarantee that source and destination tensors have "
Expand Down
15 changes: 15 additions & 0 deletions test/xpu/distributed/test_c10d_xccl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -248,6 +248,21 @@ def rank_to_GPU(self):
# return rank to GPU map
return init_multigpu_helper(self.world_size, "xccl")

@requires_xccl()
@skip_if_lt_x_gpu(2)
def test_send_recv_non_dense_tensor(self):
pg = self._create_process_group_xccl()
device = self.rank_to_GPU[self.rank][0]
full = torch.empty((64, 64), device=device).fill_(self.rank)
# Take a slice in col dimension, making it non-dense
Copy link

Copilot AI Oct 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The comment mentions 'col dimension' but should be more precise. The slice [:, 16:32] creates a non-contiguous view by selecting columns 16-31, which results in a non-dense tensor due to the stride pattern in memory.

Suggested change
# Take a slice in col dimension, making it non-dense
# Take a slice along columns 16 to 31 (inclusive), resulting in a non-contiguous (non-dense) tensor due to the stride pattern in memory

Copilot uses AI. Check for mistakes.

block = full[:, 16:32]
if self.rank == 0:
with self.assertRaises(ValueError):
dist.send(block, dst=1)
elif self.rank == 1:
with self.assertRaises(ValueError):
dist.recv(block, src=0)

@requires_xccl()
@skip_but_pass_in_sandcastle_if(
torch.xpu.device_count() < 2, "XCCL test requires 2+ GPUs"
Expand Down