:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.

accelerator ai cuda deepseek gpu img-gen kernels llama llm metal scale-out stable-diffusion tenstorrent video-gen
3 Open Issues Need Help Last updated: Feb 27, 2026

Open Issues Need Help

View All on GitHub

AI Summary: This issue proposes the removal of non-posted semaphore functions from the dataflow API. The rationale is that these functions offer no significant benefit and future hardware architectures may not support non-posted atomics, making their deprecation timely. The freed-up resources will be used to implement missing posted atomic functions.

Complexity: 2/5
data_movement good first issue

:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.

C++
#accelerator#ai#cuda#deepseek#gpu#img-gen#kernels#llama#llm#metal#scale-out#stable-diffusion#tenstorrent#video-gen

AI Summary: This issue proposes updating the pull request template to include an optional step for developers to manually run the full clang-tidy code analysis on their branch before submitting a PR. This aims to help developers catch context-dependent linting errors earlier in the development cycle, reducing late-stage rejections and improving the merge queue efficiency.

Complexity: 1/5
infra-ci dragonstrike good first issue area: docs area: infra

:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.

C++
#accelerator#ai#cuda#deepseek#gpu#img-gen#kernels#llama#llm#metal#scale-out#stable-diffusion#tenstorrent#video-gen

AI Summary: This issue proposes refactoring Control Plane tests to leverage a new feature in tt-umd that allows dynamic filtering of mock cluster configurations using the TT_VISIBLE_DEVICES environment variable. The goal is to replace the current practice of using multiple 'split' mock cluster descriptor files with a single, full descriptor and apply filtering dynamically, ultimately simplifying test setup and reducing maintenance.

Complexity: 3/5
good first issue control-plane mock-cluster tech-debt

:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.

C++
#accelerator#ai#cuda#deepseek#gpu#img-gen#kernels#llama#llm#metal#scale-out#stable-diffusion#tenstorrent#video-gen