[llvm] [AMDGPU][GlobalISel] Reduce instruction count for store in global addrspace for non-powers-of-2 (PR #206043)

[Rajveer Singh Bharadwaj via llvm-commits]

Rajveer100 wrote:

Also, simply widening doesn't seem safe enough since we are introducing extra bits which are unused.

https://github.com/llvm/llvm-project/pull/206043