[llvm] [LV] Convert gather loads with invariant stride into strided loads (PR #147297)

[Mel Chen via llvm-commits]

Mel-Chen wrote:

This is a port of the approach used in RISCVGatherScatterLowering, implemented entirely in VPlan. If you'd prefer to switch to LLVM IR–based analysis using SCEV, please let me know.

That said, the current patch faces an awkward situation: all existing strided access patterns in our lit tests are currently not converted to strided accesses, because the cost returned by TTI.getStridedMemoryOpCost is higher than that of a gather.

As a result, I'm marking this patch as Draft for now, until we have test cases that can demonstrate the intended functionality.

https://github.com/llvm/llvm-project/pull/147297