TosaToTensor: Support reshape on tensors of unsigned integer (#91734)

This adds 
- `mlir::tosa::populateTosaToLinalgTypeConversion` which converts
tensors of unsigned integers into tensors of signless integers
- modifies the `tosa.reshape` lowering in TosaToTensor to use the type
converter correctly

I choose to implement the type converter in
`mlir/Conversion/TosaToLinalg/TosaToLinalg.h` instead of
`mlir/Conversion/TosaToTensor/TosaToTensor.h` because I need the same
type converter in the TosaToLinalg lowerings (future PR).
Alternatively, I could duplicate the type converter so it exists both in
TosaToLinalg and TosaToTensor. Let me know if you prefer that.

Author: mgehre-amd

mlir/include/mlir/Conversion/TosaToTensor/TosaToTensor.h

@@ -16,6 +16,7 @@
 #include "mlir/Pass/Pass.h"
 
 namespace mlir {
+class TypeConverter;
 
 #define GEN_PASS_DECL_TOSATOTENSOR
 #include "mlir/Conversion/Passes.h.inc"
@@ -24,7 +25,8 @@ namespace tosa {
 
 std::unique_ptr<Pass> createTosaToTensor();
 
-void populateTosaToTensorConversionPatterns(RewritePatternSet *patterns);
+void populateTosaToTensorConversionPatterns(TypeConverter &converter,
+                                            RewritePatternSet *patterns);
 
 } // namespace tosa
 } // namespace mlir

mlir/include/mlir/Dialect/Tosa/Transforms/Passes.h

@@ -18,6 +18,7 @@
 #include "mlir/Pass/Pass.h"
 
 namespace mlir {
+class TypeConverter;
 namespace tosa {
 
 #define GEN_PASS_DECL
@@ -38,6 +39,8 @@ void populateTosaConstantReduction(MLIRContext *ctx,
                                    RewritePatternSet &patterns,
                                    bool aggressiveReduceConstant);
 
+void populateTosaTypeConversion(TypeConverter &converter);
+
 std::unique_ptr<Pass> createTosaLayerwiseConstantFoldPass();
 std::unique_ptr<Pass> createTosaLayerwiseConstantFoldPass(
     const TosaLayerwiseConstantFoldPassOptions &options);

mlir/lib/Conversion/TosaToTensor/TosaToTensor.cpp

@@ -224,8 +224,17 @@ class ReshapeConverter : public OpConversionPattern<tosa::ReshapeOp> {
   matchAndRewrite(tosa::ReshapeOp reshape, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = reshape.getLoc();
-    auto resultType = reshape.getResult().getType();
-    auto input = reshape.getInput1();
+    auto resultType = cast_if_present<ShapedType>(
+        getTypeConverter()->convertType(reshape.getType()));
+    if (!resultType) {
+      return rewriter.notifyMatchFailure(reshape.getLoc(),
+                                         "could not convert result type");
+    }
+    auto input = dyn_cast<TypedValue<TensorType>>(adaptor.getInput1());
+    if (!input) {
+      return rewriter.notifyMatchFailure(reshape.getLoc(),
+                                         "expected input type to be tensor");
+    }
     auto newShape = reshape.getNewShape();
 
     // Infer all intermediate types
@@ -288,12 +297,13 @@ class SliceConverter : public OpConversionPattern<tosa::SliceOp> {
   }
 };
 
-class PadConverter : public OpRewritePattern<tosa::PadOp> {
+class PadConverter : public OpConversionPattern<tosa::PadOp> {
 public:
-  using OpRewritePattern<tosa::PadOp>::OpRewritePattern;
+  using OpConversionPattern::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(tosa::PadOp padOp,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(tosa::PadOp padOp, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     auto loc = padOp.getLoc();
     auto input = padOp.getInput1();
     auto padding = padOp.getPadding();
@@ -428,11 +438,8 @@ struct ConcatConverter : public OpConversionPattern<tosa::ConcatOp> {
 } // namespace
 
 void mlir::tosa::populateTosaToTensorConversionPatterns(
-    RewritePatternSet *patterns) {
-  patterns->add<
-    ConcatConverter,
-    PadConverter,
-    ReshapeConverter,
-    SliceConverter
-  >(patterns->getContext());
+    TypeConverter &converter, RewritePatternSet *patterns) {
+  patterns
+      ->add<ConcatConverter, PadConverter, ReshapeConverter, SliceConverter>(
+          converter, patterns->getContext());
 }

mlir/lib/Conversion/TosaToTensor/TosaToTensorPass.cpp

@@ -42,7 +42,10 @@ struct TosaToTensor : public impl::TosaToTensorBase<TosaToTensor> {
     target.addLegalDialect<arith::ArithDialect>();
     target.addLegalDialect<tensor::TensorDialect>();
 
-    mlir::tosa::populateTosaToTensorConversionPatterns(&patterns);
+    TypeConverter converter;
+    mlir::tosa::populateTosaTypeConversion(converter);
+
+    mlir::tosa::populateTosaToTensorConversionPatterns(converter, &patterns);
 
     if (failed(applyPartialConversion(getOperation(), target,
                                       std::move(patterns))))

mlir/lib/Dialect/Tosa/Transforms/CMakeLists.txt

@@ -7,6 +7,7 @@ add_mlir_dialect_library(MLIRTosaTransforms
   TosaLayerwiseConstantFoldPass.cpp
   TosaMakeBroadcastable.cpp
   TosaOptionalDecompositions.cpp
+  TosaTypeConverters.cpp
   TosaValidation.cpp
 
   ADDITIONAL_HEADER_DIRS

mlir/lib/Dialect/Tosa/Transforms/TosaTypeConverters.cpp

@@ -0,0 +1,52 @@
+
+//===- TosaTypeConverters.cpp ---------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Type converters for lowering TOSA to linalg/arith.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/Tosa/Transforms/Passes.h"
+
+#include "mlir/Transforms/DialectConversion.h"
+
+using namespace mlir;
+
+void mlir::tosa::populateTosaTypeConversion(TypeConverter &converter) {
+  converter.addConversion([&](Type type) -> std::optional<Type> {
+    if (type.isUnsignedInteger()) {
+      return IntegerType::get(type.getContext(), type.getIntOrFloatBitWidth(),
+                              IntegerType::SignednessSemantics::Signless);
+    }
+    return type;
+  });
+  converter.addConversion([&](TensorType type) -> std::optional<Type> {
+    auto converted = converter.convertType(type.getElementType());
+    if (!converted)
+      return {};
+    return type.clone(converted);
+  });
+  converter.addSourceMaterialization([&](OpBuilder &builder, Type resultType,
+                                         ValueRange inputs,
+                                         Location loc) -> std::optional<Value> {
+    if (inputs.size() != 1)
+      return std::nullopt;
+
+    return builder.create<UnrealizedConversionCastOp>(loc, resultType, inputs)
+        .getResult(0);
+  });
+  converter.addTargetMaterialization([&](OpBuilder &builder, Type resultType,
+                                         ValueRange inputs,
+                                         Location loc) -> std::optional<Value> {
+    if (inputs.size() != 1)
+      return std::nullopt;
+
+    return builder.create<UnrealizedConversionCastOp>(loc, resultType, inputs)
+        .getResult(0);
+  });
+}

mlir/test/Conversion/TosaToTensor/tosa-to-tensor.mlir

@@ -420,6 +420,20 @@ func.func @test_reshape_6d_down_s2s_explicit(%arg0: tensor<1x2x3x5x7x11xf32>) ->
 
 // -----
 
+// CHECK-LABEL: @test_reshape_samerank_unsigned
+//  CHECK-SAME: (%[[ARG0:.*]]: tensor<3x2xui8>)
+func.func @test_reshape_samerank_unsigned(%arg0: tensor<3x2xui8>) -> tensor<2x3xui8> {
+  // CHECK-NEXT: %[[CAST1:.*]] = builtin.unrealized_conversion_cast %[[ARG0]] : tensor<3x2xui8> to tensor<3x2xi8>
+  // CHECK-NEXT: %[[RESHAPE1:.*]] = tensor.collapse_shape %[[CAST1]] {{\[}}[0, 1]] : tensor<3x2xi8> into tensor<6xi8>
+  // CHECK-NEXT: %[[RESHAPE2:.*]] = tensor.expand_shape %[[RESHAPE1]] {{\[}}[0, 1]] output_shape {{\[}}2, 3] : tensor<6xi8> into tensor<2x3xi8>
+  // CHECK-NEXT: %[[CAST2:.*]] = builtin.unrealized_conversion_cast %[[RESHAPE2]] : tensor<2x3xi8> to tensor<2x3xui8
+  %0 = "tosa.reshape"(%arg0) {new_shape = array<i64: 2, 3>} : (tensor<3x2xui8>) -> tensor<2x3xui8>
+  // CHECK-NEXT: return %[[CAST2]]
+  return %0 : tensor<2x3xui8>
+}
+
+// -----
+
 // CHECK-LABEL: func @slice
 func.func @slice(%arg0: tensor<6xf32>) ->() {
   // CHECK: [[SLICE:%.+]] = tensor.extract_slice %arg0[2] [1] [1]