Hi Team,
I noticed that many CUDA APIs have introduced 64-bit versions starting with CUDA 12.0, which is greatly appreciated for handling large datasets. However, I couldn’t find a 64-bit version of the cublasSgelsBatched API in the documentation. Could you please clarify if there are plans to provide 64-bit support for this function in future releases? If not, could you elaborate on the reason behind this omission?
Thank you for your assistance.
Before you take any other action, I highly recommend retrieval of the latest CUDA version (12.8) to ensure that you are acting on the latest relevant information.
Although some documentation can be generated automatically, it is always possible for documentation to get out of sync with implementation. Have you checked what is available according to the relevant header file?
(1) If the header file indicates the existence of the function (which could be double-checked by simple experiment) but it is absent from the documentation, file a bug to get the documentation updated.
(2) If the header file does not indicate the existence of the desired function, file a feature request to get that function added. You would file such a feature request via the bug reporting mechanism as well.
Generally speaking, based on historically observed behavior, NVIDIA does not discuss future functionality in public settings, and will not provide a rationale for inclusion or exclusion of any particular feature.