Skip to content

nccl_ep: Remove unused tensor type enum#2042

Open
artpol84 wants to merge 1 commit intoNVIDIA:masterfrom
artpol84:fix/nccl_ep/remove_unused_enum
Open

nccl_ep: Remove unused tensor type enum#2042
artpol84 wants to merge 1 commit intoNVIDIA:masterfrom
artpol84:fix/nccl_ep/remove_unused_enum

Conversation

@artpol84
Copy link
Copy Markdown
Collaborator

@artpol84 artpol84 commented Mar 9, 2026

Clean the codebase removing the deadcode

Clean the codebase removing the deadcode

Signed-off-by: Artem Y. Polyakov <artemp@nvidia.com>
// Tensor containing tokens received per expert (pinned host memory)
NCCL_EP_TENSOR_TAG_RECV_EXPERT_COUNTER_HOST = 6,
// Tensor containing per-expert token counts
NCCL_EP_TENSOR_TAG_TOKENS_PER_EXPERTS = 7,
Copy link
Copy Markdown
Collaborator

@sb17v sb17v Mar 11, 2026

Choose a reason for hiding this comment

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

I think we need to fix the following in the dispatch API if we are removing this:

//   local_tensors - [IN,OUT] Array of pointers to preallocated tensors, with information that is local to the rank.
//                            LL mode: accepts 1 optional local tensor:
//                                    NUM_TOKENS_PER_EXPERTS: [OUT] a 1D tensor of unsigned int [num_experts]
//                                    that contains the number of tokens received by each expert on this rank.

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

This enum is a deadcode, the two enums above duplicate it and are actually used:
https://github.com/NVIDIA/nccl/pull/2042/changes#diff-0ed431b7dc072f7ce49c9470b66954b5beba57b7fdf066e641d0b9867012fe0eL32

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

No API changes required. This PR is a code cleanup only

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants