build: update llvm tag to 3a020527

include/torch-mlir/Conversion/Utils/Utils.h

@@ -84,7 +84,7 @@ SmallVector<Value> getTypeConvertedValues(OpBuilder &b, Location loc,
 // should be converted builtin types.
 Value convertScalarToDtype(
     OpBuilder &b, Location loc, Value scalar, Type dtype,
-    llvm::Optional<Type> srcOriginalDtype = llvm::None);
+    llvm::Optional<Type> srcOriginalDtype = std::nullopt);
 
 } // namespace Torch
 } // namespace torch

include/torch-mlir/Dialect/Torch/IR/TorchOps.h

@@ -190,7 +190,7 @@ struct torch_list_of_optional_constant_ints_op_binder {
       if (matchPattern(value, m_TorchConstantInt(&num)))
         bind_values.push_back(num);
       else if (value.getType().isa<Torch::NoneType>())
-        bind_values.push_back(llvm::None);
+        bind_values.push_back(std::nullopt);
       else
         return false;
     }

lib/CAPI/TorchTypes.cpp

@@ -198,7 +198,7 @@ MlirType torchMlirTorchNonValueTensorTypeGet(MlirContext context,
                                              intptr_t numSizes,
                                              const int64_t *optionalSizes,
                                              MlirType optionalDtype) {
-  Optional<ArrayRef<int64_t>> optionalSizesArrayRef = None;
+  Optional<ArrayRef<int64_t>> optionalSizesArrayRef = std::nullopt;
   // if numSizes == -1, then it is unranked.
   if (numSizes > -1)
     optionalSizesArrayRef = llvm::makeArrayRef(optionalSizes, numSizes);
@@ -232,7 +232,7 @@ MlirType torchMlirTorchValueTensorTypeGet(MlirContext context,
                                           intptr_t numSizes,
                                           const int64_t *optionalSizes,
                                           MlirType optionalDtype) {
-  Optional<ArrayRef<int64_t>> optionalSizesArrayRef = None;
+  Optional<ArrayRef<int64_t>> optionalSizesArrayRef = std::nullopt;
   // if numSizes == -1, then it is unranked.
   if (numSizes > -1)
     optionalSizesArrayRef = llvm::makeArrayRef(optionalSizes, numSizes);

lib/Conversion/TorchToTosa/TosaLegalizeCommon.cpp

@@ -37,7 +37,7 @@ llvm::Optional<Value> convertReduceOpCommon(
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   ArrayRef<int64_t> input_shape = input_type.getShape();
   ArrayRef<int64_t> output_shape = output_type.getShape();
@@ -101,7 +101,7 @@ convertReduceAllOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   return convertReduceOpCommon<tosa::ReduceAllOp>(
       rewriter, op, output_type, input_value, axes_elems, keep_dims,
@@ -116,7 +116,7 @@ convertReduceAnyOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   return convertReduceOpCommon<tosa::ReduceAnyOp>(
       rewriter, op, output_type, input_value, axes_elems, keep_dims,
@@ -131,7 +131,7 @@ convertReduceMinOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   return convertReduceOpCommon<tosa::ReduceMinOp>(
       rewriter, op, output_type, input_value, axes_elems, keep_dims,
@@ -146,7 +146,7 @@ convertReduceMaxOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   return convertReduceOpCommon<tosa::ReduceMaxOp>(
       rewriter, op, output_type, input_value, axes_elems, keep_dims,
@@ -161,7 +161,7 @@ convertReduceProdOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   bool input_is_qtype =
       input_type.getElementType().isa<mlir::quant::UniformQuantizedType>();
@@ -171,7 +171,7 @@ convertReduceProdOp(PatternRewriter &rewriter, Operation *op,
   if (input_is_qtype || output_is_qtype) {
     op->emitOpError("ConvertReduceProdOp: input/output tensor should "
                     "be all floating-point.");
-    return llvm::None;
+    return std::nullopt;
   }
 
   return convertReduceOpCommon<tosa::ReduceProdOp>(
@@ -187,7 +187,7 @@ convertReduceSumOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   bool input_is_qtype =
       input_type.getElementType().isa<mlir::quant::UniformQuantizedType>();
@@ -197,7 +197,7 @@ convertReduceSumOp(PatternRewriter &rewriter, Operation *op,
   if (input_is_qtype != output_is_qtype) {
     op->emitOpError("ConvertReduceSumOp: input/output tensor should "
                     "be all quantized or all floating-point.");
-    return llvm::None;
+    return std::nullopt;
   }
 
   double input_scale = 1.0f;
@@ -242,7 +242,7 @@ convertReduceMeanOp(PatternRewriter &rewriter, Operation *op,
   RankedTensorType input_type =
       input_value.getType().dyn_cast<RankedTensorType>();
   if (!input_type)
-    return llvm::None;
+    return std::nullopt;
 
   bool input_is_qtype =
       input_type.getElementType().isa<mlir::quant::UniformQuantizedType>();
@@ -252,15 +252,15 @@ convertReduceMeanOp(PatternRewriter &rewriter, Operation *op,
   if (input_is_qtype != output_is_qtype) {
     op->emitOpError("ConvertReduceSumOp: input/output tensor should "
                     "be all quantized or all floating-point.");
-    return llvm::None;
+    return std::nullopt;
   }
 
   // Only supports float type mean() if it's non-quantized
   if (!input_is_qtype && !output_type.getElementType().isa<mlir::FloatType>()) {
     op->emitWarning(
         "Failed convertReduceMean: input unquantized type but output element "
         "not FloatType!");
-    return llvm::None;
+    return std::nullopt;
   }
 
   int64_t input_rank = input_type.getRank();
@@ -303,7 +303,7 @@ convertReduceMeanOp(PatternRewriter &rewriter, Operation *op,
       output_zp);
 
   if (!val.has_value())
-    return llvm::None;
+    return std::nullopt;
 
   if (!input_is_qtype) {
     Value div_const = getTosaConstTensorSingleF32(rewriter, op, div_scale);

lib/Conversion/TorchToTosa/TosaLegalizeUtils.cpp

@@ -162,7 +162,7 @@ llvm::Optional<Value> getConstTensor(PatternRewriter &rewriter, Operation *op,
 
   if (vec.size() != num_total_elements) {
     op->emitOpError("getConstTensor(): number of elements mismatch.");
-    return llvm::None;
+    return std::nullopt;
   }
 
   auto const_type =
@@ -186,7 +186,7 @@ llvm::Optional<Value> getConstTensor<APInt>(PatternRewriter &rewriter,
 
   if (vec.size() != num_total_elements) {
     op->emitOpError("getConstTensor(): number of elements mismatch.");
-    return llvm::None;
+    return std::nullopt;
   }
 
   auto const_type = RankedTensorType::get(
@@ -210,7 +210,7 @@ llvm::Optional<Value> getConstTensor<float>(PatternRewriter &rewriter,
 
   if (vec.size() != num_total_elements) {
     op->emitOpError("getConstTensor(): number of elements mismatch.");
-    return llvm::None;
+    return std::nullopt;
   }
 
   auto const_type = RankedTensorType::get(shape, rewriter.getF32Type());

lib/Dialect/Torch/IR/TorchOps.cpp

@@ -1028,7 +1028,7 @@ traceKnownSizeTensorType(Value value, llvm::Optional<int64_t> dim) {
     if (!tensorType.hasSizes())
       return false;
 
-    if (dim == llvm::None)
+    if (dim == std::nullopt)
       return tensorType.areAllSizesKnown();
 
     // If the dimension value is negative, then convert it to a positive value.
@@ -1062,7 +1062,7 @@ traceKnownSizeTensorType(Value value, llvm::Optional<int64_t> dim) {
 void AtenSizeOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                              MLIRContext *context) {
   patterns.add(+[](AtenSizeOp op, PatternRewriter &rewriter) {
-    auto type = traceKnownSizeTensorType(op.getOperand(), llvm::None);
+    auto type = traceKnownSizeTensorType(op.getOperand(), std::nullopt);
     if (failed(type))
       return rewriter.notifyMatchFailure(op, "all sizes not known");
     SmallVector<Value> listElements;

lib/Dialect/Torch/IR/TorchTypes.cpp

@@ -89,19 +89,19 @@ bool Torch::isValidSubtype(Type subtype, Type type) {
 static Optional<SmallVector<Type>>
 parseMultipleContainedTypes(AsmParser &parser) {
   if (parser.parseLess())
-    return None;
+    return std::nullopt;
 
   SmallVector<Type> containedTypes;
   if (!parser.parseOptionalGreater())
     return containedTypes;
   do {
     Type containedType = parseTorchDialectType(parser);
     if (!containedType)
-      return None;
+      return std::nullopt;
     containedTypes.push_back(containedType);
   } while (!parser.parseOptionalComma());
   if (parser.parseGreater())
-    return None;
+    return std::nullopt;
   return containedTypes;
 }
 
@@ -222,7 +222,8 @@ Type parseTensorType(MLIRContext *context, AsmParser &parser,
   llvm::SMLoc startLoc = parser.getCurrentLocation();
   if (parser.parseOptionalLess())
     return getTensorType(context,
-                         /*optionalSizes=*/None, /*optionalDtype=*/Type());
+                         /*optionalSizes=*/std::nullopt,
+                         /*optionalDtype=*/Type());
   bool hasSizes;
   SmallVector<int64_t> sizes;
   if (succeeded(parser.parseOptionalStar())) {
@@ -320,7 +321,7 @@ ValueTensorType NonValueTensorType::getWithValueSemantics() const {
 NonValueTensorType
 NonValueTensorType::getWithLeastStaticInformation(MLIRContext *context) {
   return NonValueTensorType::get(context,
-                                 /*optionalSizes=*/None,
+                                 /*optionalSizes=*/std::nullopt,
                                  /*optionalDtype=*/Type());
 }
 
@@ -357,7 +358,7 @@ NonValueTensorType ValueTensorType::getWithoutValueSemantics() const {
 ValueTensorType
 ValueTensorType::getWithLeastStaticInformation(MLIRContext *context) {
   return ValueTensorType::get(context,
-                              /*optionalSizes=*/None,
+                              /*optionalSizes=*/std::nullopt,
                               /*optionalDtype=*/Type());
 }
 
@@ -428,8 +429,8 @@ Type Torch::meetTensorTypes(BaseTensorType lhs, BaseTensorType rhs) {
 
   // If neither has sizes, we have nothing left to do.
   if (!lhs.hasSizes() && !rhs.hasSizes()) {
-    return ValueTensorType::get(lhs.getContext(), /*optionalSizes=*/None,
-                                dtype);
+    return ValueTensorType::get(lhs.getContext(),
+                                /*optionalSizes=*/std::nullopt, dtype);
   }
 
   // If the number of sizes is different, the two types are contradictory.

lib/Dialect/Torch/Transforms/GlobalizeObjectGraph.cpp

@@ -85,7 +85,7 @@ class ObjectGraphInfo {
                                            func::FuncOp methodFunc) {
     auto it = funcLinkageInfo.find({instance, methodFunc});
     if (it == funcLinkageInfo.end())
-      return None;
+      return std::nullopt;
     return it->second;
   }
 
@@ -638,7 +638,7 @@ static LogicalResult globalizeObjectGraph(ModuleOp module) {
   for (auto &monomorphization : tracker.getMonomorphizations()) {
     auto newFunc = cast<func::FuncOp>(monomorphization.func->clone());
     newFuncs[monomorphization] = newFunc;
-    Optional<LinkageInfo> linkageInfo = None;
+    Optional<LinkageInfo> linkageInfo = std::nullopt;
     // If it is potentially a method, check its linkage info.
     if (monomorphization.argInstances.size() != 0 &&
         monomorphization.argInstances[0].argIndex == 0) {

lib/Dialect/Torch/Transforms/RefineTypes.cpp

@@ -112,9 +112,9 @@ static torch_upstream::TypeKind getTypeKind(Type type) {
 }
 
 /// Returns the dtype that assumes information from both `lhs` and `rhs`.
-/// Returns `None` if the types are contradictory. Note this can only be used
-/// on the `dtype` from tensors and can't be used on other types like scalar
-/// types.
+/// Returns `std::nullopt` if the types are contradictory. Note this can only
+/// be used on the `dtype` from tensors and can't be used on other types like
+/// scalar types.
 static Optional<Type> meetElementTypes(Type lhs, Type rhs) {
   auto isNullOrBuiltIn = [](Type type) { return !type || isBuiltInType(type); };
   (void)isNullOrBuiltIn;
@@ -127,7 +127,7 @@ static Optional<Type> meetElementTypes(Type lhs, Type rhs) {
     return lhs;
   if (lhs == rhs)
     return lhs;
-  return None;
+  return std::nullopt;
 }
 
 enum class OptionalKnowledge {
@@ -137,7 +137,7 @@ enum class OptionalKnowledge {
 };
 
 /// Returns the OptionalKnowledge that assumes information from both `lhs` and
-/// `rhs`. Returns `None` if the knowledges are contradictory.
+/// `rhs`. Returns `std::nullopt` if the knowledges are contradictory.
 static Optional<OptionalKnowledge>
 meetOptionalKnowledge(OptionalKnowledge lhs, OptionalKnowledge rhs) {
   if (lhs == OptionalKnowledge::unKnown)
@@ -146,7 +146,7 @@ meetOptionalKnowledge(OptionalKnowledge lhs, OptionalKnowledge rhs) {
     return lhs;
   if (lhs == rhs)
     return lhs;
-  return None;
+  return std::nullopt;
 }
 
 static OptionalKnowledge joinOptionalKnowledge(OptionalKnowledge lhs,
@@ -327,7 +327,7 @@ struct ValueKnowledge {
 
   // Given two pieces of static knowledge, calculate new knowledge that assumes
   // the facts from both.
-  // If the two pieces of knowledge are contradictory, None is returned.
+  // If the two pieces of knowledge are contradictory, std::nullopt is returned.
   static Optional<ValueKnowledge> meet(const ValueKnowledge &lhs,
                                        const ValueKnowledge &rhs) {
     if (!lhs.isInitialized)
@@ -338,13 +338,13 @@ struct ValueKnowledge {
     Optional<ValueKnowledge> knowledge = meetTypes(lhs, rhs);
 
     if (!knowledge.has_value())
-      return None;
+      return std::nullopt;
     ValueKnowledge result = knowledge.value();
 
     Optional<OptionalKnowledge> optional =
         meetOptionalKnowledge(lhs.optional, rhs.optional);
     if (!optional.has_value())
-      return None;
+      return std::nullopt;
     result.optional = optional.value();
     return result;
   }
@@ -362,7 +362,7 @@ struct ValueKnowledge {
       return rhs;
     if (lhs == rhs)
       return lhs;
-    return None;
+    return std::nullopt;
   }
 
   // We start in the uninitialized state by default.
@@ -559,7 +559,7 @@ static Type getPromotedResultDType(ValueKnowledge *tensor, Type scalarType) {
   torch_upstream::ResultTypeState state = {};
   // No need to check if rank is zero for tensor because scalar uses
   // wrappedResult which is a lower priority than both dimResult and zeroResult.
-  state = updateResultTypeState(tensor, /*rankIsNonZero=*/None, state,
+  state = updateResultTypeState(tensor, /*rankIsNonZero=*/std::nullopt, state,
                                 /*skipRankCheck=*/true);
   state =
       updateResultTypeState(getDefaultDtypeForTorchScalar(scalarType), state);
@@ -573,7 +573,7 @@ static SmallVector<Optional<bool>> getRankIsNonZeroArray(ValueRange values) {
       if (tensorType.hasSizes()) {
         rankIsNonZero.push_back(tensorType.getSizes().size() != 0);
       } else {
-        rankIsNonZero.push_back(None);
+        rankIsNonZero.push_back(std::nullopt);
       }
     }
   }

lib/Dialect/Torch/Utils/Utils.cpp

@@ -27,10 +27,10 @@ llvm::Optional<int64_t>
 Torch::matchLegalConstantIndexIntoListOfSize(Value v, int64_t length) {
   int64_t dim;
   if (!matchPattern(v, m_TorchConstantInt(&dim)))
-    return llvm::None;
+    return std::nullopt;
   dim = toPositiveDim(dim, length);
   if (!isValidDim(dim, length))
-    return llvm::None;
+    return std::nullopt;
   return dim;
 }
 
@@ -169,7 +169,7 @@ bool Torch::isBuiltInType(Type type) {
 Optional<unsigned> Torch::getTensorRank(Value tensor) {
   BaseTensorType tensorType = tensor.getType().cast<BaseTensorType>();
   if (!tensorType.hasSizes())
-    return llvm::None;
+    return std::nullopt;
   return tensorType.getSizes().size();
 }