Merge pull request #40371 from tg-at-google:sign-compare-warning-fixes-batch-1

PiperOrigin-RevId: 316929432
Change-Id: Iea886c60e7aee2308157f1e4a0a06a4ff01ca6ad

tensorflow/compiler/mlir/lite/quantization/import_quant_stats_pass.cc

@@ -76,7 +76,8 @@ class ImportQuantStatsPass
   // If the index is out of range, this method returns false. Otherwise it
   // returns true if the value is a float tensor.
   bool IsQuantizableResult(Operation *op, int index) {
-    if (index < 0 || index >= op->getNumResults()) return false;
+    if (index < 0 || index >= static_cast<int>(op->getNumResults()))
+      return false;
     Value res = op->getResult(index);
     return res.getType().isa<ShapedType>() &&
            res.getType().cast<ShapedType>().getElementType().isa<FloatType>();
@@ -158,7 +159,7 @@ void ImportQuantStatsPass::ImportAsStatsOps(OpBuilder b, Operation *op,
     InsertStatsOpAtResult(b, op->getResult(index), layer_stats, axis_stats,
                           axis);
   } else {
-    for (int i = 0; i < op->getNumResults(); ++i) {
+    for (int i = 0, e = op->getNumResults(); i < e; ++i) {
       if (IsQuantizableResult(op, i)) {
         InsertStatsOpAtResult(b, op->getResult(i), layer_stats, axis_stats,
                               axis);

tensorflow/compiler/mlir/lite/quantization/quantization_config.cc

@@ -48,7 +48,7 @@ bool ParseInputNodeQuantSpecs(absl::string_view node_names,
   std::vector<llvm::Optional<double>> node_mins;
   if (!min_values.empty()) {
     std::vector<std::string> node_mins_str = absl::StrSplit(min_values, ',');
-    for (int i = 0; i < node_mins_str.size(); i++) {
+    for (int i = 0, e = node_mins_str.size(); i < e; i++) {
       double value;
       if (!absl::SimpleAtod(node_mins_str[i], &value)) {
         return true;
@@ -60,7 +60,7 @@ bool ParseInputNodeQuantSpecs(absl::string_view node_names,
   std::vector<llvm::Optional<double>> node_maxs;
   if (!max_values.empty()) {
     std::vector<std::string> node_maxs_str = absl::StrSplit(max_values, ',');
-    for (int i = 0; i < node_maxs_str.size(); i++) {
+    for (int i = 0, e = node_maxs_str.size(); i < e; i++) {
       double value;
       if (!absl::SimpleAtod(node_maxs_str[i], &value)) {
         llvm::errs() << "Unexpected mins: " << node_maxs_str[i] << "\n";

tensorflow/compiler/mlir/lite/quantization/quantization_driver.cc

@@ -294,7 +294,7 @@ class QuantizationDriver {
         return;
       if (current_op == op) llvm::errs() << "===>>>";
       llvm::errs() << op->getName() << " : (";
-      for (auto i = 0; i < op->getNumOperands(); ++i) {
+      for (int i = 0, e = op->getNumOperands(); i < e; ++i) {
         if (auto params = GetOperandQuantState(op, i).params)
           params.print(llvm::errs());
         else
@@ -303,7 +303,7 @@ class QuantizationDriver {
         llvm::errs() << ",";
       }
       llvm::errs() << ") -> (";
-      for (auto i = 0; i < op->getNumResults(); ++i) {
+      for (int i = 0, e = op->getNumResults(); i < e; ++i) {
         if (auto params = GetResultQuantState(op, i).params)
           params.print(llvm::errs());
         else

tensorflow/compiler/mlir/lite/quantization/quantization_utils.cc

@@ -55,7 +55,7 @@ static Type GetQuantizedType(Builder builder, Type input_type,
   } else if (min.size() == max.size()) {
     auto shape = input_type.dyn_cast<ShapedType>();
     if (!shape || shape.getRank() <= quant_dim ||
-        min.size() != shape.getDimSize(quant_dim)) {
+        static_cast<int64_t>(min.size()) != shape.getDimSize(quant_dim)) {
       return {};
     }
     // TODO(b/141508873): the quantization dim is set to the last dimension.
@@ -76,7 +76,8 @@ TypeAttr RescaleQuantizedType(Type input, Attribute factor) {
   if (auto qtype = ele_type.dyn_cast<quant::UniformQuantizedPerAxisType>()) {
     ArrayRef<double> scales = qtype.getScales();
     // Broadcasting hasn't been implemented yet.
-    if (scales.size() != factor_values.getNumElements()) return {};
+    if (static_cast<int64_t>(scales.size()) != factor_values.getNumElements())
+      return {};
     SmallVector<double, 4> new_scales;
     new_scales.reserve(scales.size());
     auto scales_iter = scales.begin();
@@ -270,7 +271,7 @@ Type GetUniformQuantizedPerAxisTypeForWeight(ElementsAttr attr, int quant_dim,
                                              bool narrow_range) {
   Builder builder(attr.getContext());
   auto shape = attr.getType().cast<ShapedType>().getShape();
-  if (shape.size() <= quant_dim) return {};
+  if (static_cast<int>(shape.size()) <= quant_dim) return {};
   // `symmetric` can only be used when it is `signed` and `narrow_range`.
   if (symmetric && (!is_signed || !narrow_range)) return {};
 
@@ -335,7 +336,7 @@ quant::QuantizedType GetUniformQuantizedTypeForBias(
     const std::vector<quant::QuantizedType>& op_types) {
   if (op_types.empty()) return {};
 
-  int axis_size = 1;
+  size_t axis_size = 1;
   int32_t quant_dim = -1;
   Type expressed_type;
   // Requires all the op types are valid UniformQuantizedTypes or
@@ -369,7 +370,7 @@ quant::QuantizedType GetUniformQuantizedTypeForBias(
         scales[index_scale.index()] *= index_scale.value();
       }
     } else if (auto type = op_type.dyn_cast<quant::UniformQuantizedType>()) {
-      for (int index = 0; index != axis_size; ++index) {
+      for (int index = 0, e = axis_size; index != e; ++index) {
         scales[index] *= type.getScale();
       }
     }

tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc

@@ -41,7 +41,7 @@ std::string MakeUniqueFilename(string name) {
   static NameCounts& instance = *new NameCounts;
 
   // Remove illegal characters from `name`.
-  for (int i = 0; i < name.size(); ++i) {
+  for (int i = 0, e = name.size(); i < e; ++i) {
     char ch = name[i];
     if (ch == '/' || ch == '[' || ch == ']' || ch == '*' || ch == '?' ||
         ch == '\\') {

tensorflow/compiler/mlir/xla/ir/chlo_ops.cc

@@ -49,7 +49,7 @@ static Type GetBroadcastType(Type x, Type y, Type element_type,
 
   if (shape_x.size() == shape_y.size()) {
     llvm::SmallVector<int64_t, 4> out_shape(shape_x.size());
-    for (int i = 0; i < shape_x.size(); i++) {
+    for (int i = 0, e = shape_x.size(); i < e; i++) {
       auto x_val = shape_x[i];
       auto y_val = shape_y[i];
       if (x_val == -1 || y_val == -1) {

tensorflow/compiler/mlir/xla/ir/hlo_ops.cc

@@ -143,7 +143,7 @@ DenseIntElementsAttr BuildConvPaddingAttrs(
 
   int rank = padding_low.size();
   SmallVector<int64_t, 8> padding;
-  for (unsigned i = 0; i < rank; ++i) {
+  for (unsigned i = 0, e = rank; i < e; ++i) {
     padding.push_back(GetPaddingValue(padding_attr, {i, 0}) + padding_low[i]);
     padding.push_back(GetPaddingValue(padding_attr, {i, 1}) + padding_high[i]);
   }
@@ -891,7 +891,7 @@ static Attribute foldConcatenateHelper(ConcatenateOp* op,
   auto shape = type.getShape();
 
   size_t top_size = 1;
-  for (int i = 0; i < axis; i++) {
+  for (int i = 0, e = axis; i < e; i++) {
     top_size = top_size * shape[i];
   }
 
@@ -1169,7 +1169,7 @@ static LogicalResult Verify(MapOp op) {
   // increasing.
   auto values = op.dimensions().getValues<int64_t>();
   auto dimensions = std::vector<int64_t>{values.begin(), values.end()};
-  for (int i = 0; i < dimensions.size(); ++i) {
+  for (int i = 0, e = dimensions.size(); i < e; ++i) {
     if (dimensions[i] != i)
       return op.emitOpError() << "requires monotonically increasing dimension "
                                  "numbers, but got: "