[tosa] Implement Argmax support (#485)

Signed-off-by: Suraj Sudhir <[email protected]>

Author: sjarus

lib/Conversion/TorchToTosa/TorchToTosa.cpp

@@ -12,6 +12,8 @@
 #include "torch-mlir/Conversion/TorchToTosa/TosaLegalizeUtils.h"
 
 #include "../PassDetail.h"
+#include "mlir/Dialect/Arithmetic/IR/Arithmetic.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
 #include "mlir/Dialect/Traits.h"
 #include "mlir/IR/Matchers.h"
@@ -404,6 +406,93 @@ class ConvertAtenAllDimsReductionOp
   }
 };
 
+template <>
+LogicalResult ConvertAtenOp<AtenArgmaxOp>::matchAndRewrite(
+    AtenArgmaxOp op, OpAdaptor adaptor,
+    ConversionPatternRewriter &rewriter) const {
+
+  Value self = adaptor.self();
+  auto selfTy = self.getType().template cast<RankedTensorType>();
+
+  if (!selfTy)
+    return op.emitError("Only ranked tensor types supported in TOSA argmax");
+
+  int64_t reduceDim;
+  if (!matchPattern(op.dim(), m_TorchConstantInt(&reduceDim))) {
+    // NoneType indicates reduce on all dims
+    reduceDim = -1;
+  }
+
+  bool keepDim = false;
+  if (!matchPattern(op.keepdim(), m_TorchConstantBool(&keepDim)))
+    return rewriter.notifyMatchFailure(
+        op, "non-const keepdim parameter unsupported");
+
+  auto resultTy = getTypeConverter()
+                      ->convertType(op.getResult().getType())
+                      .cast<RankedTensorType>();
+  auto outputETy = resultTy.getElementType();
+
+  // Create a single instance of tosa.argmax.
+  // Multiple dims require chained construct.
+  auto buildArgmax = [&](int64_t reduceDim, Value input) -> Value {
+    auto inputTy = input.getType().cast<RankedTensorType>();
+    auto inputShape = inputTy.getShape();
+    SmallVector<int64_t> outputShapeArr = {};
+    int32_t i = 0;
+
+    for (auto &dim : inputShape) {
+      if (i++ != reduceDim) {
+        outputShapeArr.push_back(dim);
+      } else {
+        if (keepDim)
+          outputShapeArr.push_back(1);
+      }
+    }
+
+    // Tosa argmax output is i32, while Torch backend mandates i64.
+    auto outputReduceTy = RankedTensorType::get(
+        ArrayRef<int64_t>(outputShapeArr), rewriter.getI32Type());
+    auto reduceDimAttr =
+        rewriter.getIntegerAttr(rewriter.getI64Type(), reduceDim);
+    return rewriter
+        .create<tosa::ArgMaxOp>(op->getLoc(),
+                                getTypeConverter()->convertType(outputReduceTy),
+                                input, reduceDimAttr)
+        .getResult();
+  };
+
+  // Convert the final index to i64 for backend finalization, However, i64
+  // is not a defined type for tosa.cast, so using arith.extsi instead.
+  auto castToInt64 = [&](Value result) -> LogicalResult {
+    auto resTy = result.getType().cast<ShapedType>();
+    if (!resTy)
+      return op.emitError("Argmax: Result is not a shaped type");
+
+    auto resShape = resTy.getShape();
+    auto outTy =
+        RankedTensorType::get(resShape, outputETy); // rewriter.getI64Type());
+
+    rewriter.replaceOpWithNewOp<arith::ExtSIOp>(
+        op, getTypeConverter()->convertType(outTy), result);
+
+    return success();
+  };
+
+  if (reduceDim == -1) { // reducing on all dims
+    Value input = self;
+    for (int dim = 0; dim < selfTy.getRank(); dim++) {
+      // progressively reduce each 0-th dim
+      input = buildArgmax(0, input);
+    }
+    return castToInt64(input);
+  } else {
+    return castToInt64(buildArgmax(reduceDim, self));
+  }
+
+  return success();
+}
+
 } // namespace
 
 // -----------------------------------------------------------------------------
@@ -415,13 +504,16 @@ class ConvertTorchToTosa : public ConvertTorchToTosaBase<ConvertTorchToTosa> {
 public:
   void getDependentDialects(DialectRegistry &registry) const override {
     registry.insert<tosa::TosaDialect>();
+    registry.insert<tensor::TensorDialect>();
+    registry.insert<arith::ArithmeticDialect>();
     TorchConversion::getBackendTypeConversionDependentDialects(registry);
   }
 
   void runOnOperation() override {
     MLIRContext *context = &getContext();
     ConversionTarget target(*context);
-    target.addLegalDialect<tosa::TosaDialect>();
+    target.addLegalDialect<tosa::TosaDialect, tensor::TensorDialect,
+                           arith::ArithmeticDialect>();
 
     TypeConverter typeConverter;
     typeConverter.addConversion([](Type type) { return type; });
@@ -491,6 +583,7 @@ class ConvertTorchToTosa : public ConvertTorchToTosaBase<ConvertTorchToTosa> {
     INSERT_ATENOP_PATTERN(AtenReluOp);
     INSERT_ATENOP_PATTERN(AtenMulTensorOp);
     INSERT_ATENOP_PATTERN(AtenDivTensorOp);
+    INSERT_ATENOP_PATTERN(AtenArgmaxOp);
 #undef INSERT_ATENOP_PATTERN
 
     if (failed(applyPartialConversion(getOperation(), target,

lib/Dialect/TorchConversion/Transforms/VerifyTosaBackendContract.cpp

@@ -42,6 +42,8 @@ class VerifyTosaBackendContractPass
     target.addDynamicallyLegalOp<ModuleOp, FuncOp, ReturnOp>(opHasLegalTypes);
     // Basic scalar operations.
     target.addLegalDialect<tosa::TosaDialect>();
+    target.addDynamicallyLegalOp<tensor::CastOp>(opHasLegalTypes);
+    target.addDynamicallyLegalOp<arith::ExtSIOp>(opHasLegalTypes);
 
     RewritePatternSet patterns(context);
     if (failed(applyFullConversion(module, target, std::move(patterns)))) {