[mlir][spirv] Implement SPIR-V lowering for vector.deinterleave (ll…

…vm#95313)

1. Added a conversion for `vector.deinterleave` to the `VectorToSPIRV`
pass.
2. Added LIT tests for the new conversion.

---------

Co-authored-by: Jakub Kuderski <kubakuderski@gmail.com>

mlir/lib/Conversion/VectorToSPIRV/VectorToSPIRV.cpp

@@ -618,6 +618,66 @@ struct VectorInterleaveOpConvert final
   }
 };
 
+struct VectorDeinterleaveOpConvert final
+    : public OpConversionPattern<vector::DeinterleaveOp> {
+  using OpConversionPattern::OpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(vector::DeinterleaveOp deinterleaveOp, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+
+    // Check the result vector type.
+    VectorType oldResultType = deinterleaveOp.getResultVectorType();
+    Type newResultType = getTypeConverter()->convertType(oldResultType);
+    if (!newResultType)
+      return rewriter.notifyMatchFailure(deinterleaveOp,
+                                         "unsupported result vector type");
+
+    Location loc = deinterleaveOp->getLoc();
+
+    // Deinterleave the indices.
+    Value sourceVector = adaptor.getSource();
+    VectorType sourceType = deinterleaveOp.getSourceVectorType();
+    int n = sourceType.getNumElements();
+
+    // Output vectors of size 1 are converted to scalars by the type converter.
+    // We cannot use `spirv::VectorShuffleOp` directly in this case, and need to
+    // use `spirv::CompositeExtractOp`.
+    if (n == 2) {
+      auto elem0 = rewriter.create<spirv::CompositeExtractOp>(
+          loc, newResultType, sourceVector, rewriter.getI32ArrayAttr({0}));
+
+      auto elem1 = rewriter.create<spirv::CompositeExtractOp>(
+          loc, newResultType, sourceVector, rewriter.getI32ArrayAttr({1}));
+
+      rewriter.replaceOp(deinterleaveOp, {elem0, elem1});
+      return success();
+    }
+
+    // Indices for `shuffleEven` (result 0).
+    auto seqEven = llvm::seq<int64_t>(n / 2);
+    auto indicesEven =
+        llvm::map_to_vector(seqEven, [](int i) { return i * 2; });
+
+    // Indices for `shuffleOdd` (result 1).
+    auto seqOdd = llvm::seq<int64_t>(n / 2);
+    auto indicesOdd =
+        llvm::map_to_vector(seqOdd, [](int i) { return i * 2 + 1; });
+
+    // Create two SPIR-V shuffles.
+    auto shuffleEven = rewriter.create<spirv::VectorShuffleOp>(
+        loc, newResultType, sourceVector, sourceVector,
+        rewriter.getI32ArrayAttr(indicesEven));
+
+    auto shuffleOdd = rewriter.create<spirv::VectorShuffleOp>(
+        loc, newResultType, sourceVector, sourceVector,
+        rewriter.getI32ArrayAttr(indicesOdd));
+
+    rewriter.replaceOp(deinterleaveOp, {shuffleEven, shuffleOdd});
+    return success();
+  }
+};
+
 struct VectorLoadOpConverter final
     : public OpConversionPattern<vector::LoadOp> {
   using OpConversionPattern::OpConversionPattern;
@@ -862,9 +922,9 @@ void mlir::populateVectorToSPIRVPatterns(SPIRVTypeConverter &typeConverter,
       VectorReductionFloatMinMax<CL_FLOAT_MAX_MIN_OPS>,
       VectorReductionFloatMinMax<GL_FLOAT_MAX_MIN_OPS>, VectorShapeCast,
       VectorInsertStridedSliceOpConvert, VectorShuffleOpConvert,
-      VectorInterleaveOpConvert, VectorSplatPattern, VectorLoadOpConverter,
-      VectorStoreOpConverter>(typeConverter, patterns.getContext(),
-                              PatternBenefit(1));
+      VectorInterleaveOpConvert, VectorDeinterleaveOpConvert,
+      VectorSplatPattern, VectorLoadOpConverter, VectorStoreOpConverter>(
+      typeConverter, patterns.getContext(), PatternBenefit(1));
 
   // Make sure that the more specialized dot product pattern has higher benefit
   // than the generic one that extracts all elements.

mlir/test/Conversion/VectorToSPIRV/vector-to-spirv.mlir

@@ -507,6 +507,32 @@ func.func @interleave_size1(%a: vector<1xf32>, %b: vector<1xf32>) -> vector<2xf3
 
 // -----
 
+// CHECK-LABEL: func @deinterleave
+// CHECK-SAME: (%[[ARG0:.+]]: vector<4xf32>)
+//       CHECK: %[[SHUFFLE0:.*]] = spirv.VectorShuffle [0 : i32, 2 : i32] %[[ARG0]], %[[ARG0]] : vector<4xf32>, vector<4xf32> -> vector<2xf32>
+//       CHECK: %[[SHUFFLE1:.*]] = spirv.VectorShuffle [1 : i32, 3 : i32] %[[ARG0]], %[[ARG0]] : vector<4xf32>, vector<4xf32> -> vector<2xf32>
+//       CHECK: return %[[SHUFFLE0]], %[[SHUFFLE1]]
+func.func @deinterleave(%a: vector<4xf32>) -> (vector<2xf32>, vector<2xf32>) {
+  %0, %1 = vector.deinterleave %a : vector<4xf32> -> vector<2xf32>
+  return %0, %1 : vector<2xf32>, vector<2xf32>
+}
+
+// -----
+
+// CHECK-LABEL: func @deinterleave_scalar
+// CHECK-SAME: (%[[ARG0:.+]]: vector<2xf32>)
+//       CHECK: %[[EXTRACT0:.*]] = spirv.CompositeExtract %[[ARG0]][0 : i32] : vector<2xf32>
+//       CHECK: %[[EXTRACT1:.*]] = spirv.CompositeExtract %[[ARG0]][1 : i32] : vector<2xf32>
+//       CHECK: %[[CAST0:.*]] = builtin.unrealized_conversion_cast %[[EXTRACT0]] : f32 to vector<1xf32>
+//       CHECK: %[[CAST1:.*]] = builtin.unrealized_conversion_cast %[[EXTRACT1]] : f32 to vector<1xf32>
+//       CHECK: return %[[CAST0]], %[[CAST1]]
+func.func @deinterleave_scalar(%a: vector<2xf32>) -> (vector<1xf32>, vector<1xf32>) {
+  %0, %1 = vector.deinterleave %a: vector<2xf32> -> vector<1xf32>
+  return %0, %1 : vector<1xf32>, vector<1xf32>
+}
+
+// -----
+
 // CHECK-LABEL: func @reduction_add
 //  CHECK-SAME: (%[[V:.+]]: vector<4xi32>)
 //       CHECK:   %[[S0:.+]] = spirv.CompositeExtract %[[V]][0 : i32] : vector<4xi32>