Does slangtorch support uint64_t scalar type in TorchTensor?

[ozchango]

Thank you for the awesome work. I want to allocate a new uint64_t type tensor.

uint P = conic.size(0);
TorchTensor<uint64_t3> dL_duvd_int = TorchTensor<uint64_t3>.alloc(P, 3);
TorchTensor<uint64_t3> dL_dconic_int = TorchTensor<uint64_t3>.alloc(P, 3);
TorchTensor<uint64_t> dL_dopacity_int = TorchTensor<uint64_t>.alloc(P);
TorchTensor<uint64_t3> dL_dfeature_int = TorchTensor<uint64_t3>.alloc(P, 3);
TorchTensor<uint64_t2> dL_dndc_int = TorchTensor<uint64_t2>.alloc(P, 2);

However, it seems that the scalar type is not supported.

I am using CUDA 12.2, pytorch 2.1.1 and slangtorch 1.2.6

[bmillsNV]

Thanks for raising this. Is this issue blocking you? We plan to get to this one in Q1 2025. Any concerns with that timeline?

[bmillsNV]

No feedback, so planed for this quarter.