Implement NNAPI canonical interfaces

This CL implements the canonical IDevice, IPreparedModel, and IBuffer
interfaces for the 1.0, 1.1, 1.2, and 1.3 NN HIDL HAL interfaces.
Further, it introduces "Resilient" adapter interfaces to automatically
retrieve a handle to a recovered interface object after it has died and
rebooted.

This CL also updates the conversion code from returning nn::Result to
nn::GeneralResult, which includes a ErrorStatus code in the case of an
error.

Finally, this CL introduces a new static library
neuralnetworks_utils_hal_service which consists of a single function
::android::nn::hal::getDevices which can be used by the NNAPI runtime to
retrieve the HIDL services without knowing the underlying HIDL types.

Bug: 160668438
Test: mma
Test: NeuralNetworksTest_static
Change-Id: Iec6ae739df196b4034ffb35ea76781fd541ffec3
Merged-In: Iec6ae739df196b4034ffb35ea76781fd541ffec3
(cherry picked from commit 3670c385c4b12aef975ab67e5d2b0f5fe79134c2)
diff --git a/neuralnetworks/1.1/utils/src/Conversions.cpp b/neuralnetworks/1.1/utils/src/Conversions.cpp
index 7fee16b..ffe0752 100644
--- a/neuralnetworks/1.1/utils/src/Conversions.cpp
+++ b/neuralnetworks/1.1/utils/src/Conversions.cpp
@@ -42,7 +42,7 @@
 using convertOutput = std::decay_t<decltype(convert(std::declval<Input>()).value())>;
 
 template <typename Type>
-Result<std::vector<convertOutput<Type>>> convert(const hidl_vec<Type>& arguments) {
+GeneralResult<std::vector<convertOutput<Type>>> convert(const hidl_vec<Type>& arguments) {
     std::vector<convertOutput<Type>> canonical;
     canonical.reserve(arguments.size());
     for (const auto& argument : arguments) {
@@ -53,11 +53,11 @@
 
 }  // anonymous namespace
 
-Result<OperationType> convert(const hal::V1_1::OperationType& operationType) {
+GeneralResult<OperationType> convert(const hal::V1_1::OperationType& operationType) {
     return static_cast<OperationType>(operationType);
 }
 
-Result<Capabilities> convert(const hal::V1_1::Capabilities& capabilities) {
+GeneralResult<Capabilities> convert(const hal::V1_1::Capabilities& capabilities) {
     const auto quantized8Performance = NN_TRY(convert(capabilities.quantized8Performance));
     const auto float32Performance = NN_TRY(convert(capabilities.float32Performance));
     const auto relaxedFloat32toFloat16Performance =
@@ -73,7 +73,7 @@
     };
 }
 
-Result<Operation> convert(const hal::V1_1::Operation& operation) {
+GeneralResult<Operation> convert(const hal::V1_1::Operation& operation) {
     return Operation{
             .type = NN_TRY(convert(operation.type)),
             .inputs = operation.inputs,
@@ -81,7 +81,7 @@
     };
 }
 
-Result<Model> convert(const hal::V1_1::Model& model) {
+GeneralResult<Model> convert(const hal::V1_1::Model& model) {
     auto operations = NN_TRY(convert(model.operations));
 
     // Verify number of consumers.
@@ -90,9 +90,9 @@
     CHECK(model.operands.size() == numberOfConsumers.size());
     for (size_t i = 0; i < model.operands.size(); ++i) {
         if (model.operands[i].numberOfConsumers != numberOfConsumers[i]) {
-            return NN_ERROR() << "Invalid numberOfConsumers for operand " << i << ", expected "
-                              << numberOfConsumers[i] << " but found "
-                              << model.operands[i].numberOfConsumers;
+            return NN_ERROR(nn::ErrorStatus::GENERAL_FAILURE)
+                   << "Invalid numberOfConsumers for operand " << i << ", expected "
+                   << numberOfConsumers[i] << " but found " << model.operands[i].numberOfConsumers;
         }
     }
 
@@ -111,7 +111,8 @@
     };
 }
 
-Result<ExecutionPreference> convert(const hal::V1_1::ExecutionPreference& executionPreference) {
+GeneralResult<ExecutionPreference> convert(
+        const hal::V1_1::ExecutionPreference& executionPreference) {
     return static_cast<ExecutionPreference>(executionPreference);
 }
 
@@ -122,20 +123,20 @@
 
 using utils::convert;
 
-nn::Result<V1_0::PerformanceInfo> convert(
+nn::GeneralResult<V1_0::PerformanceInfo> convert(
         const nn::Capabilities::PerformanceInfo& performanceInfo) {
     return V1_0::utils::convert(performanceInfo);
 }
 
-nn::Result<V1_0::Operand> convert(const nn::Operand& operand) {
+nn::GeneralResult<V1_0::Operand> convert(const nn::Operand& operand) {
     return V1_0::utils::convert(operand);
 }
 
-nn::Result<hidl_vec<uint8_t>> convert(const nn::Model::OperandValues& operandValues) {
+nn::GeneralResult<hidl_vec<uint8_t>> convert(const nn::Model::OperandValues& operandValues) {
     return V1_0::utils::convert(operandValues);
 }
 
-nn::Result<hidl_memory> convert(const nn::Memory& memory) {
+nn::GeneralResult<hidl_memory> convert(const nn::Memory& memory) {
     return V1_0::utils::convert(memory);
 }
 
@@ -143,7 +144,7 @@
 using convertOutput = std::decay_t<decltype(convert(std::declval<Input>()).value())>;
 
 template <typename Type>
-nn::Result<hidl_vec<convertOutput<Type>>> convert(const std::vector<Type>& arguments) {
+nn::GeneralResult<hidl_vec<convertOutput<Type>>> convert(const std::vector<Type>& arguments) {
     hidl_vec<convertOutput<Type>> halObject(arguments.size());
     for (size_t i = 0; i < arguments.size(); ++i) {
         halObject[i] = NN_TRY(convert(arguments[i]));
@@ -153,11 +154,11 @@
 
 }  // anonymous namespace
 
-nn::Result<OperationType> convert(const nn::OperationType& operationType) {
+nn::GeneralResult<OperationType> convert(const nn::OperationType& operationType) {
     return static_cast<OperationType>(operationType);
 }
 
-nn::Result<Capabilities> convert(const nn::Capabilities& capabilities) {
+nn::GeneralResult<Capabilities> convert(const nn::Capabilities& capabilities) {
     return Capabilities{
             .float32Performance = NN_TRY(convert(
                     capabilities.operandPerformance.lookup(nn::OperandType::TENSOR_FLOAT32))),
@@ -168,7 +169,7 @@
     };
 }
 
-nn::Result<Operation> convert(const nn::Operation& operation) {
+nn::GeneralResult<Operation> convert(const nn::Operation& operation) {
     return Operation{
             .type = NN_TRY(convert(operation.type)),
             .inputs = operation.inputs,
@@ -176,9 +177,10 @@
     };
 }
 
-nn::Result<Model> convert(const nn::Model& model) {
+nn::GeneralResult<Model> convert(const nn::Model& model) {
     if (!hal::utils::hasNoPointerData(model)) {
-        return NN_ERROR() << "Mdoel cannot be converted because it contains pointer-based memory";
+        return NN_ERROR(nn::ErrorStatus::INVALID_ARGUMENT)
+               << "Mdoel cannot be converted because it contains pointer-based memory";
     }
 
     auto operands = NN_TRY(convert(model.main.operands));
@@ -202,7 +204,7 @@
     };
 }
 
-nn::Result<ExecutionPreference> convert(const nn::ExecutionPreference& executionPreference) {
+nn::GeneralResult<ExecutionPreference> convert(const nn::ExecutionPreference& executionPreference) {
     return static_cast<ExecutionPreference>(executionPreference);
 }