Access to shared GPU cublas/cublasLt/cudnn library handles from custom_op kernels

[renderist]

The GPU custom_op examples only shows direct CUDA programming examples, where the CUDA stream handle is accessible via the API. The provider and contrib_ops show access to cublas, cublasLt, and cudnn NVidia library handles. How might one gain access to these library handles from a custom_op?

[pranavsharma]

Can you file this as an issue? This is not the first time this has been asked. #14284. Added to our backlog.

[renderist]

Thanks Pranav, I added my comments on the feature request that you pointed out to me, is there any utility to filing it as a new issue in terms of escalation? Thanks, Kevin On Mar 30, 2023, at 2:37 PM, Pranav Sharma ***@***.******@***.***>> wrote: Can you file this as an issue? This is not the first time this has been asked. #14284<#14284>. Added to our backlog. — Reply to this email directly, view it on GitHub<#15286 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABI2FOASDGXSFSZ7IU3HUPTW6X4LJANCNFSM6AAAAAAWNUKGG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[pranavsharma]

Just easier to measure the demand :-).

[renderist]

Understood. Issue #15296 On Mar 30, 2023, at 9:32 PM, Pranav Sharma ***@***.******@***.***>> wrote: Just easier to measure the demand :-). — Reply to this email directly, view it on GitHub<#15286 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABI2FOADR75UNU5DVE5NEELW6ZM4JANCNFSM6AAAAAAWNUKGG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>