[mlir][vector] Fix FlattenGather for scalable vectors

[c-rhodes]

This pattern flattens vector.gather ops by unrolling the outermost
dimension for rank > 2 vectors. There's two issues with this pattern for
scalable vectors:

1. The unrolling doesn't take vscale into account. A constraint is
   added to disable this pattern for vectors with leading scalable
   dims.
2. The scalable dims are dropped when creating the new gather. Fixed
   by propagating the flags.

Depends on #96049.

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-vector

Author: Cullen Rhodes (c-rhodes)

Changes

This pattern flattens vector.gather ops by unrolling the outermost
dimension for rank > 2 vectors. There's two issues with this pattern for
scalable vectors:

1. The unrolling doesn't take vscale into account. A constraint is
   added to disable this pattern for vectors with leading scalable
   dims.
2. The scalable dims are dropped when creating the new gather. Fixed
   by propagating the flags.

Depends on #96049.

---

Full diff: https://github.com/llvm/llvm-project/pull/96074.diff

2 Files Affected:

• (modified) mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp (+9-1)
• (modified) mlir/test/Dialect/Vector/vector-gather-lowering.mlir (+26)

diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp
index dd027d107d16a..1abde32450f1e 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp
@@ -55,6 +55,8 @@ namespace {
 /// ```
 ///
 /// When applied exhaustively, this will produce a sequence of 1-d gather ops.
+///
+/// Supports vector types with trailing scalable dim.
 struct FlattenGather : OpRewritePattern<vector::GatherOp> {
   using OpRewritePattern::OpRewritePattern;
 
@@ -64,6 +66,11 @@ struct FlattenGather : OpRewritePattern<vector::GatherOp> {
     if (resultTy.getRank() < 2)
       return rewriter.notifyMatchFailure(op, "already flat");
 
+    // Unrolling doesn't take vscale into account. Pattern is disabled for
+    // vectors with leading scalable dim(s).
+    if (resultTy.getScalableDims().front())
+      return rewriter.notifyMatchFailure(op, "cannot unroll scalable dim");
+
     Location loc = op.getLoc();
     Value indexVec = op.getIndexVec();
     Value maskVec = op.getMask();
@@ -73,7 +80,8 @@ struct FlattenGather : OpRewritePattern<vector::GatherOp> {
         loc, resultTy, rewriter.getZeroAttr(resultTy));
 
     Type subTy = VectorType::get(resultTy.getShape().drop_front(),
-                                 resultTy.getElementType());
+                                 resultTy.getElementType(),
+                                 resultTy.getScalableDims().drop_front());
 
     for (int64_t i = 0, e = resultTy.getShape().front(); i < e; ++i) {
       int64_t thisIdx[1] = {i};
diff --git a/mlir/test/Dialect/Vector/vector-gather-lowering.mlir b/mlir/test/Dialect/Vector/vector-gather-lowering.mlir
index c2eb88afa4dbf..ff1a92a65c42d 100644
--- a/mlir/test/Dialect/Vector/vector-gather-lowering.mlir
+++ b/mlir/test/Dialect/Vector/vector-gather-lowering.mlir
@@ -74,6 +74,32 @@ func.func @gather_memref_1d_i32_index(%base: memref<?xf32>, %v: vector<2xi32>, %
   return %0 : vector<2x3xf32>
  }
 
+// CHECK-LABEL: @scalable_gather_memref_2d
+// CHECK-SAME:      %[[BASE:.*]]: memref<?x?xf32>,
+// CHECK-SAME:      %[[IDXVEC:.*]]: vector<2x[3]xindex>,
+// CHECK-SAME:      %[[MASK:.*]]: vector<2x[3]xi1>,
+// CHECK-SAME:      %[[PASS:.*]]: vector<2x[3]xf32>
+// CHECK:         %[[C0:.*]] = arith.constant 0 : index
+// CHECK:         %[[C1:.*]] = arith.constant 1 : index
+// CHECK:         %[[INIT:.*]] = arith.constant dense<0.000000e+00> : vector<2x[3]xf32>
+// CHECK:         %[[IDXVEC0:.*]] = vector.extract %[[IDXVEC]][0] : vector<[3]xindex> from vector<2x[3]xindex>
+// CHECK:         %[[MASK0:.*]] = vector.extract %[[MASK]][0] : vector<[3]xi1> from vector<2x[3]xi1>
+// CHECK:         %[[PASS0:.*]] = vector.extract %[[PASS]][0] : vector<[3]xf32> from vector<2x[3]xf32>
+// CHECK:         %[[GATHER0:.*]] = vector.gather %[[BASE]]{{\[}}%[[C0]], %[[C1]]] {{\[}}%[[IDXVEC0]]], %[[MASK0]], %[[PASS0]] : memref<?x?xf32>, vector<[3]xindex>, vector<[3]xi1>, vector<[3]xf32> into vector<[3]xf32>
+// CHECK:         %[[INS0:.*]] = vector.insert %[[GATHER0]], %[[INIT]] [0] : vector<[3]xf32> into vector<2x[3]xf32>
+// CHECK:         %[[IDXVEC1:.*]] = vector.extract %[[IDXVEC]][1] : vector<[3]xindex> from vector<2x[3]xindex>
+// CHECK:         %[[MASK1:.*]] = vector.extract %[[MASK]][1] : vector<[3]xi1> from vector<2x[3]xi1>
+// CHECK:         %[[PASS1:.*]] = vector.extract %[[PASS]][1] : vector<[3]xf32> from vector<2x[3]xf32>
+// CHECK:         %[[GATHER1:.*]] = vector.gather %[[BASE]]{{\[}}%[[C0]], %[[C1]]] {{\[}}%[[IDXVEC1]]], %[[MASK1]], %[[PASS1]] : memref<?x?xf32>, vector<[3]xindex>, vector<[3]xi1>, vector<[3]xf32> into vector<[3]xf32>
+// CHECK:         %[[INS1:.*]] = vector.insert %[[GATHER1]], %[[INS0]] [1] : vector<[3]xf32> into vector<2x[3]xf32>
+// CHECK-NEXT:    return %[[INS1]] : vector<2x[3]xf32>
+func.func @scalable_gather_memref_2d(%base: memref<?x?xf32>, %v: vector<2x[3]xindex>, %mask: vector<2x[3]xi1>, %pass_thru: vector<2x[3]xf32>) -> vector<2x[3]xf32> {
+ %c0 = arith.constant 0 : index
+ %c1 = arith.constant 1 : index
+ %0 = vector.gather %base[%c0, %c1][%v], %mask, %pass_thru : memref<?x?xf32>, vector<2x[3]xindex>, vector<2x[3]xi1>, vector<2x[3]xf32> into vector<2x[3]xf32>
+ return %0 : vector<2x[3]xf32>
+}
+
 // CHECK-LABEL: @gather_tensor_1d
 // CHECK-SAME:    ([[BASE:%.+]]: tensor<?xf32>, [[IDXVEC:%.+]]: vector<2xindex>, [[MASK:%.+]]: vector<2xi1>, [[PASS:%.+]]: vector<2xf32>)
 // CHECK-DAG:     [[M0:%.+]]    = vector.extract [[MASK]][0] : i1 from vector<2xi1>

[MacDue]

LGTM, just a few little nits: