[QST] A100 double-precision Tensor Cores ?

[touisteur]

Hi,

is there any support for DP Tensor Cores in cutlass, available or foreseen ?

Thanks in advance

[jackkosaian]

Yes.

For an example, you can modify the example 14_ampere_tf32_tensorop_gemm to use double precision.

To do so, you can change ElementAccumulator, ElementInputA, and ElementInputB on these lines to be of type double, and change ShapeMMAOp here to be 8x8x4.

[hwu36]

We have many f64 or complex f64 unit tests here: https://github.com/NVIDIA/cutlass/tree/master/test/unit/gemm/device

[touisteur]

Thanks a lot to both of you, I have everything I need.

Le jeu. 23 juin 2022 à 05:53, Haicheng Wu @.***> a écrit :

We have many f64 or complex f64 unit tests here: https://github.com/NVIDIA/cutlass/tree/master/test/unit/gemm/device

— Reply to this email directly, view it on GitHub https://github.com/NVIDIA/cutlass/issues/536#issuecomment-1163895707, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGBRHDTQVZRYYP2U2PF4JNLVQPNTFANCNFSM5ZMJV2HQ . You are receiving this because you authored the thread.Message ID: @.***>

[github-actions[bot]]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.