add more comments for flag

microsoft · Dec 6, 2024 · 17d2c4d · 17d2c4d
1 parent 123a4f0
commit 17d2c4d
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 2 deletions.
diff --git a/include/onnxruntime/core/providers/coreml/coreml_provider_factory.h b/include/onnxruntime/core/providers/coreml/coreml_provider_factory.h
@@ -55,6 +55,8 @@ static const char* const kCoremlProviderOption_EnableOnSubgraphs = "EnableOnSubg
 // This process can affect the model loading time and the prediction latency.
 // Use this option to tailor the specialization strategy for your model.
 static const char* const kCoremlProviderOption_SpecializationStrategy = "SpecializationStrategy";
+// this flag is used to profile coreml computeplan after model compile.
+// The profile result includes what hardwares of each ops dispatched to and the estimate excution time.
 static const char* const kCoremlProviderOption_ProfileComputePlan = "ProfileComputePlan";
 // please refer to https://developer.apple.com/documentation/coreml/mlmodelconfiguration/allowlowprecisionaccumulationongpu
 static const char* const kCoremlProviderOption_AllowLowPrecisionAccumulationOnGPU = "AllowLowPrecisionAccumulationOnGPU";

diff --git a/onnxruntime/core/providers/coreml/builders/impl/squeeze_op_builder.cc b/onnxruntime/core/providers/coreml/builders/impl/squeeze_op_builder.cc
@@ -99,9 +99,8 @@ Status SqueezeOpBuilder::AddToModelBuilderImpl(ModelBuilder& model_builder,
     std::unique_ptr<Operation> op = model_builder.CreateOperation(node, coreml_op_type);
     AddOperationInput(*op, "x", input_defs[0]->Name());
 
-    // it's impossible to have empty exes input for unsqueeze
     if (!axes.empty()) {
-      // coreml squeeze op does support negative axes
+      // coreml supports negative axes
       AddOperationInput(*op, "axes", model_builder.AddConstant(op->type(), "axes", AsSpan(axes)));
     }
     AddOperationOutput(*op, *node.OutputDefs()[0]);

diff --git a/onnxruntime/core/providers/coreml/model/model.mm b/onnxruntime/core/providers/coreml/model/model.mm
@@ -301,6 +301,8 @@ Status GetMLMultiArrayCopyInfo(const MLMultiArray* _Nonnull array,
   return Status::OK();
 }
 
+// since __clang_major__ >= 15, MLComputePlan is introduced in <CoreML/CoreML.h>
+// we define __clang_analyzer__ here is for bypass static analysis
 void ProfileComputePlan(NSURL* compileUrl, MLModelConfiguration* config) {
 #if defined(__APPLE__) && defined(__clang__) && __clang_major__ >= 15 && !defined(__clang_analyzer__)
   if (@available(macOS 14.4, iOS 17.4, *)) {
@@ -454,6 +456,8 @@ Status Predict(const std::unordered_map<std::string, OnnxTensorData>& inputs,
       }
 
 // Set the specialization strategy to FastPrediction  for macOS 10.15+
+// since __clang_major__ >= 15, optimizationHints is introduced in <CoreML/CoreML.h>
+// we define __clang_analyzer__ here is for bypass static analysis
 #if defined(__APPLE__) && defined(__clang__) && __clang_major__ >= 15 && !defined(__clang_analyzer__)
       if (HAS_COREML8_OR_LATER) {
         MLOptimizationHints* optimizationHints = [[MLOptimizationHints alloc] init];