[mlir][vector] Decouple unrolling gather and gather to llvm lowering (#132206)

This patch decouples unrolling vector.gather and lowering vector.gather
to llvm.masked.gather.

This is consistent with how vector.load, vector.store,
vector.maskedload, vector.maskedstore lower to LLVM.

Some interesting test changes from this patch:

- 2D vector.gather lowering to llvm tests are deleted. This is
consistent with other memory load/store ops.
- There are still tests for 2D vector.gather, but the constant mask for
these test is modified. This is because with the updated lowering, one
of the unrolled vector.gather disappears because it is masked off (also
demonstrating why this is a better lowering path)

Overall, this makes vector.gather take the same consistent path for
lowering to LLVM as other load/store ops.

Discourse Discussion:
https://discourse.llvm.org/t/rfc-improving-gather-codegen-for-vector-dialect/85011/13

1 files changed, 1 insertions, 0 deletions

diff --git a/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp b/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp
index c9bb34b692dd..a54ae816570a 100644
--- a/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp
+++ b/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp
@@ -782,6 +782,7 @@ struct TestVectorGatherLowering
   void runOnOperation() override {
     RewritePatternSet patterns(&getContext());
     populateVectorGatherLoweringPatterns(patterns);
+    populateVectorGatherToConditionalLoadPatterns(patterns);
     (void)applyPatternsGreedily(getOperation(), std::move(patterns));
   }
 };