diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -1376,6 +1376,129 @@
   }
 };
 
+// Lowerings the TableOp to a series of gathers and numerica operations. This
+// includes interpolation between the high/low values. For the I8 varient, this
+// simplifies to a single gather operation.
+class TableConverter : public OpRewritePattern<tosa::TableOp> {
+public:
+  using OpRewritePattern<tosa::TableOp>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(tosa::TableOp op,
+                                PatternRewriter &rewriter) const final {
+    auto loc = op.getLoc();
+    Value input = op.input();
+    Value table = op.table();
+    auto inputTy = input.getType().template cast<ShapedType>();
+    auto tableTy = table.getType().template cast<ShapedType>();
+    auto resultTy = op.getType().template cast<ShapedType>();
+
+    auto inputElementTy = inputTy.getElementType();
+    auto tableElementTy = tableTy.getElementType();
+    auto resultElementTy = resultTy.getElementType();
+
+    if (!inputTy.hasStaticShape())
+      return rewriter.notifyMatchFailure(
+          op, "No initial value found for table operation");
+
+    auto initTensor =
+        rewriter
+            .create<linalg::InitTensorOp>(loc, ArrayRef<Value>({}),
+                                          resultTy.getShape(), resultElementTy)
+            .result();
+
+    SmallVector<AffineMap, 2> affineMaps = {
+        rewriter.getMultiDimIdentityMap(resultTy.getRank()),
+        rewriter.getMultiDimIdentityMap(resultTy.getRank())};
+
+    bool bodySucceeded = true;
+
+    rewriter.replaceOpWithNewOp<linalg::GenericOp>(
+        op, resultTy, ValueRange({input}), ValueRange{initTensor}, affineMaps,
+        getNParallelLoopsAttrs(resultTy.getRank()),
+        [&](OpBuilder &nestedBuilder, Location nestedLoc, ValueRange args) {
+          if (inputElementTy.isInteger(8) && tableElementTy.isInteger(8) &&
+              resultElementTy.isInteger(8)) {
+            Value index = nestedBuilder.create<IndexCastOp>(
+                nestedLoc, rewriter.getIndexType(), args[0]);
+            Value extract = nestedBuilder.create<tensor::ExtractOp>(
+                nestedLoc, table, ValueRange{index});
+            nestedBuilder.create<linalg::YieldOp>(nestedLoc, extract);
+            return;
+          }
+
+          if (inputElementTy.isInteger(16) && tableElementTy.isInteger(16) &&
+              resultElementTy.isInteger(32)) {
+            Value extend = rewriter.create<SignExtendIOp>(
+                nestedLoc, rewriter.getI32Type(), args[0]);
+
+            auto offset = nestedBuilder.create<ConstantIntOp>(
+                nestedLoc, 32768, nestedBuilder.getI32Type());
+
+            auto seven = nestedBuilder.create<ConstantIntOp>(
+                nestedLoc, 7, nestedBuilder.getI32Type());
+
+            auto one = nestedBuilder.create<ConstantIntOp>(
+                nestedLoc, 1, nestedBuilder.getI32Type());
+            auto b1111111 = nestedBuilder.create<ConstantIntOp>(
+                nestedLoc, 127, nestedBuilder.getI32Type());
+
+            // Compute the index and fractional part from the input value:
+            // value = value + 32768
+            // index = value >> 7;
+            // fraction = 0x01111111 & value
+            auto extendAdd =
+                nestedBuilder.create<AddIOp>(nestedLoc, extend, offset);
+            Value index = nestedBuilder.create<UnsignedShiftRightOp>(
+                nestedLoc, extendAdd, seven);
+            Value fraction = nestedBuilder.create<mlir::AndOp>(
+                nestedLoc, extendAdd, b1111111);
+
+            // Extract the base and next values from the table.
+            // base = (int32_t) table[index];
+            // next = (int32_t) table[index + 1];
+            Value indexPlusOne =
+                nestedBuilder.create<AddIOp>(nestedLoc, index, one);
+
+            index = nestedBuilder.create<IndexCastOp>(
+                nestedLoc, nestedBuilder.getIndexType(), index);
+            indexPlusOne = nestedBuilder.create<IndexCastOp>(
+                nestedLoc, nestedBuilder.getIndexType(), indexPlusOne);
+
+            Value base = nestedBuilder.create<tensor::ExtractOp>(
+                nestedLoc, table, ValueRange{index});
+            Value next = nestedBuilder.create<tensor::ExtractOp>(
+                nestedLoc, table, ValueRange{indexPlusOne});
+
+            base = nestedBuilder.create<SignExtendIOp>(
+                nestedLoc, nestedBuilder.getI32Type(), base);
+            next = nestedBuilder.create<SignExtendIOp>(
+                nestedLoc, nestedBuilder.getI32Type(), next);
+
+            // Use the fractional part to interpolate between the input values:
+            // result = (base << 7) +  (next - base) * fraction
+            Value baseScaled =
+                nestedBuilder.create<ShiftLeftOp>(nestedLoc, base, seven);
+            Value diff = nestedBuilder.create<SubIOp>(nestedLoc, next, base);
+            Value diffScaled =
+                nestedBuilder.create<MulIOp>(nestedLoc, diff, fraction);
+            Value result =
+                nestedBuilder.create<AddIOp>(nestedLoc, baseScaled, diffScaled);
+
+            nestedBuilder.create<linalg::YieldOp>(nestedLoc, result);
+            return;
+          }
+
+          bodySucceeded = false;
+        });
+
+    if (!bodySucceeded)
+      return rewriter.notifyMatchFailure(
+          op, "unable to create body for tosa.table op.");
+
+    return success();
+  }
+};
+
 } // namespace
 
 void mlir::tosa::populateTosaToLinalgOnTensorsConversionPatterns(
@@ -1404,8 +1527,8 @@
       IdentityNConverter<tosa::IdentityOp>,
       IdentityNConverter<tosa::IdentityNOp>, ReduceConverter<tosa::ReduceMinOp>,
       ReduceConverter<tosa::ReduceMaxOp>, ReduceConverter<tosa::ReduceSumOp>,
-      ReduceConverter<tosa::ReduceProdOp>, ArgMaxConverter, ConcatConverter, PadConverter,
-      ReshapeConverter, RescaleConverter, ReverseConverter, TileConverter,
-      TransposeConverter, MatMulConverter, FullyConnectedConverter>(
-        patterns->getContext());
+      ReduceConverter<tosa::ReduceProdOp>, ArgMaxConverter, ConcatConverter,
+      PadConverter, ReshapeConverter, RescaleConverter, ReverseConverter,
+      TableConverter, TileConverter, TransposeConverter, MatMulConverter,
+      FullyConnectedConverter>(patterns->getContext());
 }
diff --git a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
--- a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
+++ b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
@@ -793,3 +793,47 @@
 
   return
 }
+
+// -----
+
+// CHECK-LABEL: @table8
+func @table8(%arg0: tensor<6xi8>, %arg1: tensor<513xi8>) -> () {
+  // CHECK: [[INIT:%.+]] = linalg.init_tensor [6]
+  // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#map, #map], iterator_types = ["parallel"]} ins(%arg0 : tensor<6xi8>) outs([[INIT]] : tensor<6xi8>)
+  // CHECK: ^bb0([[ARG_IN:%.+]]: i8, [[ARG_INIT:%.+]]: i8)
+  // CHECK:   [[CAST:%.+]] = index_cast [[ARG_IN]]
+  // CHECK:   [[EXTRACT:%.+]] = tensor.extract %arg1{{\[}}[[CAST]]]
+  // CHECK:   linalg.yield [[EXTRACT]]
+  %0 = "tosa.table"(%arg0, %arg1)  : (tensor<6xi8>, tensor<513xi8>)  -> (tensor<6xi8>)
+  return
+}
+
+// CHECK-LABEL: @table16
+func @table16(%arg0: tensor<6xi16>, %arg1: tensor<513xi16>) -> () {
+  // CHECK: [[INIT:%.+]] = linalg.init_tensor [6]
+  // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#map, #map], iterator_types = ["parallel"]} ins(%arg0 : tensor<6xi16>) outs(%0 : tensor<6xi32>)
+  // CHECK: ^bb0(%arg2: i16, %arg3: i32)
+  // CHECK: [[EXT_IN:%.+]] = sexti %arg2
+  // CHECK: [[C32768:%.+]] = constant 32768
+  // CHECK: [[C7:%.+]] = constant 7
+  // CHECK: [[C1:%.+]] = constant 1
+  // CHECK: [[C127:%.+]] = constant 127
+  // CHECK: [[INADD:%.+]] = addi [[EXT_IN]], [[C32768]]
+  // CHECK: [[IDX:%.+]] = shift_right_unsigned [[INADD]], [[C7]]
+  // CHECK: [[FRACTION:%.+]] = and [[INADD]], [[C127]]
+  // CHECK: [[IDXPLUS1:%.+]] = addi [[IDX]], [[C1]]
+  // CHECK: [[IDX_CAST:%.+]] = index_cast [[IDX]]
+  // CHECK: [[IDXPLUS1_CAST:%.+]] = index_cast [[IDXPLUS1]]
+  // CHECK: [[BASE:%.+]] = tensor.extract %arg1{{\[}}[[IDX_CAST]]]
+  // CHECK: [[NEXT:%.+]] = tensor.extract %arg1{{\[}}[[IDXPLUS1_CAST]]]
+  // CHECK: [[BASE_EXT:%.+]] = sexti [[BASE]]
+  // CHECK: [[NEXT_EXT:%.+]] = sexti [[NEXT]]
+  // CHECK: [[BASE_MUL:%.+]] = shift_left [[BASE_EXT]], [[C7]]
+  // CHECK: [[DIFF:%.+]] = subi [[NEXT_EXT]], [[BASE_EXT]]
+  // CHECK: [[DIFF_MUL:%.+]] = muli [[DIFF]], [[FRACTION]]
+  // CHECK: [[RESULT:%.+]] = addi [[BASE_MUL]], [[DIFF_MUL]]
+  // CHECK: linalg.yield [[RESULT]]
+  %0 = "tosa.table"(%arg0, %arg1)  : (tensor<6xi16>, tensor<513xi16>)  -> (tensor<6xi32>)
+  return
+}
+