Add ROCm 6.4.3+ support

[Looong01]

Because ROCm 6.4.3 implements stricter type checking in the HIP warp shuffle API, code that previously compiled in 6.4.2 now triggers a static_assert at compile time.

The error message reads:
static assertion failed due to requirement 'sizeof(unsigned int) == 8': The mask must be a 64-bit integer

This is because ROCm 6.4.3's __shfl_sync / __shfl_up_sync / __shfl_down_sync function templates now require the mask parameter to be a 64-bit integer (uint64_t), but your code is passing an unsigned int (32-bit), resulting in a compilation failure.

Reason for Change
In the HIP header file (amd_warp_sync_functions.h) of ROCm 6.4.3, AMD added the following:
static_assert(sizeof(MaskT) == 8, "The mask must be a 64-bit integer...");

This was added to unify the behavior and avoid potential errors caused by implicit type promotion.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.09%. Comparing base (2d55981) to head (2340737).
⚠️ Report is 12 commits behind head on master.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #405   +/-   ##
=======================================
  Coverage   73.09%   73.09%           
=======================================
  Files          29       29           
  Lines        1171     1171           
=======================================
  Hits          856      856           
  Misses        315      315           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.