[llvm] [AMDGPU] Reschedule loads in clauses to improve throughput (RFC) (PR #102595)

[Carl Ritson via llvm-commits]

perlfu wrote:

Implementation is not entirely complete; however, I would be interested in feedback, specifically if we think this is "OK" from a memory model perspective.
My opinion is currently "yes", with a few additional checks on address spaces, cache bypass flags and atomicity that are currently missing from the code.
This draft only targets MIMG, but I think it makes sense for all VMEM clauses.

https://github.com/llvm/llvm-project/pull/102595