Merge "Add TENSOR_INT32 support for SUB"

6 files changed, 128 insertions, 11 deletions

diff --git a/nn/common/operations/Broadcast.cpp b/nn/common/operations/Broadcast.cpp
index 99bb18495..56069a852 100644
--- a/nn/common/operations/Broadcast.cpp
+++ b/nn/common/operations/Broadcast.cpp
@@ -205,8 +205,9 @@ bool addQuant8(const T* in1, const Shape& shape1, const T* in2, const Shape& sha
     return true;
 }
 
-bool addInt32(const int32_t* aData, const Shape& aShape, const int32_t* bData, const Shape& bShape,
-              int32_t activation, int32_t* outputData, const Shape& outputShape) {
+bool executeInt32(const int32_t* aData, const Shape& aShape, const int32_t* bData,
+                  const Shape& bShape, int32_t activation, int32_t* outputData,
+                  const Shape& outputShape, int32_t func(int32_t, int32_t)) {
     NN_RET_CHECK_EQ(activation, ANEURALNETWORKS_FUSED_NONE);
     IndexedShapeWrapper aShapeIndexed(aShape);
     IndexedShapeWrapper bShapeIndexed(bShape);
@@ -221,7 +222,7 @@ bool addInt32(const int32_t* aData, const Shape& aShape, const int32_t* bData, c
         uint32_t bFlatIndex;
         NN_RET_CHECK(bShapeIndexed.broadcastedIndexToFlatIndex(curIndex, &bFlatIndex));
 
-        outputData[outputFlatIndex] = aData[aFlatIndex] + bData[bFlatIndex];
+        outputData[outputFlatIndex] = func(aData[aFlatIndex], bData[bFlatIndex]);
 
         NN_RET_CHECK(outputShapeIndexed.nextIndexInplace(&curIndex, &lastIndex));
     } while (!lastIndex);
@@ -461,7 +462,8 @@ bool validate(OperationType opType, const IOperationValidationContext* context)
         }
     } else if (inputType == OperandType::TENSOR_QUANT8_ASYMM_SIGNED) {
         NN_RET_CHECK(validateHalVersion(context, std::max(HalVersion::V1_3, opIntroducedAt)));
-    } else if (inputType == OperandType::TENSOR_INT32 && opType == OperationType::ADD) {
+    } else if (inputType == OperandType::TENSOR_INT32 &&
+               (opType == OperationType::ADD || opType == OperationType::SUB)) {
         NN_RET_CHECK(validateHalVersion(context, HalVersion::V1_3));
     } else {
         NN_RET_CHECK_FAIL() << "Unsupported tensor type for operation " << getOperationName(opType);
@@ -517,13 +519,14 @@ bool executeAdd(IOperationExecutionContext* context) {
                              context->getOutputBuffer<int8_t>(kOutputTensor),
                              context->getOutputShape(kOutputTensor));
         case OperandType::TENSOR_INT32:
-            return addInt32(context->getInputBuffer<int32_t>(kInputTensor1),
-                            context->getInputShape(kInputTensor1),
-                            context->getInputBuffer<int32_t>(kInputTensor2),
-                            context->getInputShape(kInputTensor2),
-                            context->getInputValue<int32_t>(kActivationScalar),
-                            context->getOutputBuffer<int32_t>(kOutputTensor),
-                            context->getOutputShape(kOutputTensor));
+            return executeInt32(context->getInputBuffer<int32_t>(kInputTensor1),
+                                context->getInputShape(kInputTensor1),
+                                context->getInputBuffer<int32_t>(kInputTensor2),
+                                context->getInputShape(kInputTensor2),
+                                context->getInputValue<int32_t>(kActivationScalar),
+                                context->getOutputBuffer<int32_t>(kOutputTensor),
+                                context->getOutputShape(kOutputTensor),
+                                [](int32_t a, int32_t b) { return a + b; });
         default:
             NN_RET_CHECK_FAIL() << "Unsupported tensor type for operation ADD";
     }
@@ -606,6 +609,15 @@ bool executeSub(IOperationExecutionContext* context) {
                              context->getInputValue<int32_t>(kActivationScalar),
                              context->getOutputBuffer<int8_t>(kOutputTensor),
                              context->getOutputShape(kOutputTensor));
+        case OperandType::TENSOR_INT32:
+            return executeInt32(context->getInputBuffer<int32_t>(kInputTensor1),
+                                context->getInputShape(kInputTensor1),
+                                context->getInputBuffer<int32_t>(kInputTensor2),
+                                context->getInputShape(kInputTensor2),
+                                context->getInputValue<int32_t>(kActivationScalar),
+                                context->getOutputBuffer<int32_t>(kOutputTensor),
+                                context->getOutputShape(kOutputTensor),
+                                [](int32_t a, int32_t b) { return a - b; });
         default:
             NN_RET_CHECK_FAIL() << "Unsupported tensor type for operation SUB";
     }
diff --git a/nn/runtime/include/NeuralNetworks.h b/nn/runtime/include/NeuralNetworks.h
index b2f72c47d..505a19bfb 100644
--- a/nn/runtime/include/NeuralNetworks.h
+++ b/nn/runtime/include/NeuralNetworks.h
@@ -2379,6 +2379,7 @@ typedef enum {
      * * {@link ANEURALNETWORKS_TENSOR_FLOAT32}
      * * {@link ANEURALNETWORKS_TENSOR_QUANT8_ASYMM} (since API level 29)
      * * {@link ANEURALNETWORKS_TENSOR_QUANT8_ASYMM_SIGNED} (since API level 30)
+     * * {@link ANEURALNETWORKS_TENSOR_INT32} (since API level 30)
      *
      * Supported tensor rank: up to 4
      *
@@ -2389,6 +2390,8 @@ typedef enum {
      * * 2: An {@link ANEURALNETWORKS_INT32} scalar, and has to be one of the
      *      {@link FuseCode} values. Specifies the activation to
      *      invoke on the result.
+     *      For a {@link ANEURALNETWORKS_TENSOR_INT32} tensor,
+     *      the {@link FuseCode} must be "NONE".
      *
      * Outputs:
      * * 0: A tensor of the same {@link OperandCode} as input0.
diff --git a/nn/runtime/test/TestValidateOperations.cpp b/nn/runtime/test/TestValidateOperations.cpp
index 4f17c3d31..d42323c94 100644
--- a/nn/runtime/test/TestValidateOperations.cpp
+++ b/nn/runtime/test/TestValidateOperations.cpp
@@ -682,6 +682,10 @@ TEST(OperationValidationTest, SUB_quant8_signed) {
     simpleMathOpTest(ANEURALNETWORKS_SUB, ANEURALNETWORKS_TENSOR_QUANT8_ASYMM_SIGNED);
 }
 
+TEST(OperationValidationTest, SUB_int32) {
+    simpleMathOpTest(ANEURALNETWORKS_SUB, ANEURALNETWORKS_TENSOR_INT32);
+}
+
 TEST(OperationValidationTest, DIV_float16) {
     simpleMathOpTest(ANEURALNETWORKS_DIV, ANEURALNETWORKS_TENSOR_FLOAT16);
 }
diff --git a/nn/runtime/test/generated/spec_V1_3/sub_int32.example.cpp b/nn/runtime/test/generated/spec_V1_3/sub_int32.example.cpp
new file mode 100644
index 000000000..baa8741be
--- /dev/null
+++ b/nn/runtime/test/generated/spec_V1_3/sub_int32.example.cpp
@@ -0,0 +1,70 @@
+// Generated from sub_int32.mod.py
+// DO NOT EDIT
+// clang-format off
+#include "TestHarness.h"
+using namespace test_helper;
+
+namespace generated_tests::sub_int32 {
+
+const TestModel& get_test_model() {
+    static TestModel model = {
+        .expectFailure = false,
+        .expectedMultinomialDistributionTolerance = 0,
+        .inputIndexes = {0, 1},
+        .isRelaxed = false,
+        .minSupportedVersion = TestHalVersion::V1_3,
+        .operands = {{
+                .channelQuant = {},
+                .data = TestBuffer::createFromVector<int32_t>({2, -4, 8, -16}),
+                .dimensions = {1, 2, 2},
+                .isIgnored = false,
+                .lifetime = TestOperandLifeTime::MODEL_INPUT,
+                .numberOfConsumers = 1,
+                .scale = 0.0f,
+                .type = TestOperandType::TENSOR_INT32,
+                .zeroPoint = 0
+            }, {
+                .channelQuant = {},
+                .data = TestBuffer::createFromVector<int32_t>({2, -2, -4, 4}),
+                .dimensions = {1, 2, 2},
+                .isIgnored = false,
+                .lifetime = TestOperandLifeTime::MODEL_INPUT,
+                .numberOfConsumers = 1,
+                .scale = 0.0f,
+                .type = TestOperandType::TENSOR_INT32,
+                .zeroPoint = 0
+            }, {
+                .channelQuant = {},
+                .data = TestBuffer::createFromVector<int32_t>({0}),
+                .dimensions = {},
+                .isIgnored = false,
+                .lifetime = TestOperandLifeTime::CONSTANT_COPY,
+                .numberOfConsumers = 1,
+                .scale = 0.0f,
+                .type = TestOperandType::INT32,
+                .zeroPoint = 0
+            }, {
+                .channelQuant = {},
+                .data = TestBuffer::createFromVector<int32_t>({0, -2, 12, -20}),
+                .dimensions = {1, 2, 2},
+                .isIgnored = false,
+                .lifetime = TestOperandLifeTime::MODEL_OUTPUT,
+                .numberOfConsumers = 0,
+                .scale = 0.0f,
+                .type = TestOperandType::TENSOR_INT32,
+                .zeroPoint = 0
+            }},
+        .operations = {{
+                .inputs = {0, 1, 2},
+                .outputs = {3},
+                .type = TestOperationType::SUB
+            }},
+        .outputIndexes = {3}
+    };
+    return model;
+}
+
+const auto dummy_test_model = TestModelManager::get().add("sub_int32", get_test_model());
+
+}  // namespace generated_tests::sub_int32
+
diff --git a/nn/runtime/test/specs/V1_3/sub_int32.mod.py b/nn/runtime/test/specs/V1_3/sub_int32.mod.py
new file mode 100644
index 000000000..82fe66f81
--- /dev/null
+++ b/nn/runtime/test/specs/V1_3/sub_int32.mod.py
@@ -0,0 +1,23 @@
+#
+# Copyright (C) 2020 The Android Open Source Project
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+model = Model()
+input0 = Input("input0", "TENSOR_INT32", "{1, 2, 2}")
+input1 = Input("input1", "TENSOR_INT32", "{1, 2, 2}")
+output = Output("output", "TENSOR_INT32", "{1, 2, 2}")
+model = model.Operation("SUB", input0, input1, 0).To(output)
+Example({input0: [2, -4, 8, -16],
+         input1: [2, -2, -4, 4],
+         output: [0, -2, 12, -20]})
diff --git a/nn/tools/api/types.spec b/nn/tools/api/types.spec
index c6650d959..095f491ed 100644
--- a/nn/tools/api/types.spec
+++ b/nn/tools/api/types.spec
@@ -2728,6 +2728,7 @@
 %/kind
 %kind ndk hal_1.3+
      * * {@link %{OperandTypeLinkPfx}TENSOR_QUANT8_ASYMM_SIGNED} (since %{APILevel30})
+     * * {@link %{OperandTypeLinkPfx}TENSOR_INT32} (since %{APILevel30})
 %/kind
      *
      * Supported tensor rank: up to 4
@@ -2739,6 +2740,10 @@
      * * 2: An {@link %{OperandTypeLinkPfx}INT32} scalar, and has to be one of the
      *      {@link %{FusedActivationFunc}} values. Specifies the activation to
      *      invoke on the result.
+%kind ndk hal_1.3+
+     *      For a {@link %{OperandTypeLinkPfx}TENSOR_INT32} tensor,
+     *      the {@link %{FusedActivationFunc}} must be "NONE".
+%/kind
      *
      * Outputs:
      * * 0: A tensor of the same {@link %{OperandType}} as input0.