Add a host map option for a UVM tensor alloc

pytorch / FBGEMM

FB (Facebook) + GEMM (General Matrix-Matrix Multiplication) - https://code.fb.com/ml-applications/fbgemm/

Other

1.17k stars 474 forks source link

Closed sryap closed 3 weeks ago

sryap commented 4 weeks ago

Summary: FBGEMM-GPU provides two options to allocate a UVM buffer: (1) malloc

cudaHostRegister and (2) cudaMallocManaged. Each one of them has a different performance implication (this is platform specific). This diff adds an option for a user to choose a UVM buffer allocation method in SSD-TBE.

Differential Revision: D61362852

netlify[bot] commented 4 weeks ago

Name	Link
Latest commit	9e4f07c1edc3444e64fa923c6e36a68e2f57c957
Latest deploy log	https://app.netlify.com/sites/pytorch-fbgemm-docs/deploys/66bfabd3017fe900086f77bf
Deploy Preview	https://deploy-preview-3003--pytorch-fbgemm-docs.netlify.app
Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify site configuration.

facebook-github-bot commented 4 weeks ago

This pull request was exported from Phabricator. Differential Revision: D61362852

facebook-github-bot commented 3 weeks ago

This pull request was exported from Phabricator. Differential Revision: D61362852

facebook-github-bot commented 3 weeks ago

This pull request was exported from Phabricator. Differential Revision: D61362852

facebook-github-bot commented 3 weeks ago

This pull request has been merged in pytorch/FBGEMM@1ec08ad61c71617186c2a1a6774eda0db4c9705b.