[AMDGPU] Eliminate unnecessary packing in wider f16 vectors for sdwa/opsel-able instruction#137137

Open

vg0204 wants to merge 8 commits intollvm:mainfrom

vg0204:vg0204/optimize-unnecessary-packing-wider-f16-vectors-sdwa

Commits on Jun 3, 2025

[AMDGPU][PeepholeOpt] Eliminate unnecessary packing in fp16 vector
vg0204
committed
Update the LIT tests to accomodate the patch effects.
vg0204
committed
[AMDGPU][NFC] Added Pre-commit tests for #137137
vg0204
committed
Refactored the code in order to encapsulate redundant code as lambda function.
vg0204
committed
Added reviewed changes addressing redundant code & complex logic.
vg0204
committed
Added MIR test to demonstrate the specific MIR pattern handling to eliminate packing for fp16.
vg0204
committed
Reduced duplicate code length & added a new MIR test in existing testFile.
vg0204
committed
Updated MIR test case, and addressed code reviews.
vg0204
committed