[Issue]: DeviceFor::Bulk support

[tpopp]

Problem Description

As far as I can tell, device_for.hpp support is incomplete. See examples for issues 2/3 here: https://github.com/ROCm/hipCUB/commit/f342111197dd020f1c4210b16aa550b08992e97b

1. cub's symbol is DeviceFor::Bulk as a static method in a struct, but hipCUB doesn't even put it in a namespace: https://github.com/NVIDIA/cccl/blob/main/cub/cub/device/device_for.cuh#L230
2. It seems that device_for.hpp isn't included for the rocprim backend
3. One implementation of Bulk calls the other before it is defined in the header.

Operating System

Ubuntu 22.04.5 LTS (Jammy Jellyfish)

CPU

AMD Ryzen Threadripper PRO 7975WX 32-Cores

GPU

AMD Instinct MI210 x4

ROCm Version

ROCm 6.3.1

ROCm Component

hipCUB

Steps to Reproduce

No response

(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support

No response

Additional Information

No response

[ppanchad-amd]

Hi @tpopp. Internal ticket has been created for investigating this issue. Thanks!

[assistant-librarian[bot]]

This issue has been migrated to: https://github.com/ROCm/rocm-libraries/issues/94

[jayhawk-commits]

Closing the issue in this repo. Please refer to the migrated issue for updates.