You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

115 lines
3.9 KiB

/*
* Copyright (C) 2019 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.
*/
#define LOG_TAG "Operations"
#include "OperationResolver.h"
#include "OperationsUtils.h"
namespace android {
namespace nn {
namespace fill_op {
constexpr uint32_t kNumInputs = 2;
constexpr uint32_t kDimsTensor = 0;
constexpr uint32_t kValueScalar = 1;
constexpr uint32_t kNumOutputs = 1;
constexpr uint32_t kOutputTensor = 0;
namespace {
template <typename T>
bool executeTyped(IOperationExecutionContext* context) {
T* output = context->getOutputBuffer<T>(kOutputTensor);
const int numElements = getNumberOfElements(context->getOutputShape(kOutputTensor));
const T value = context->getInputValue<T>(kValueScalar);
for (int i = 0; i < numElements; ++i) {
output[i] = value;
}
return true;
}
bool getValueType(OperandType outputType, OperandType* valueType) {
switch (outputType) {
case OperandType::TENSOR_FLOAT16:
*valueType = OperandType::FLOAT16;
return true;
case OperandType::TENSOR_FLOAT32:
*valueType = OperandType::FLOAT32;
return true;
case OperandType::TENSOR_INT32:
*valueType = OperandType::INT32;
return true;
default:
NN_RET_CHECK_FAIL() << "Unsupported value type for fill op: " << outputType;
}
}
} // namespace
Result<Version> validate(const IOperationValidationContext* context) {
NN_RET_CHECK_EQ(context->getNumInputs(), kNumInputs);
NN_RET_CHECK_EQ(context->getNumOutputs(), kNumOutputs);
// Check output type first because input value type is dependent on the
// output type.
OperandType outputType = context->getOutputType(kOutputTensor);
NN_RET_CHECK(outputType == OperandType::TENSOR_FLOAT16 ||
outputType == OperandType::TENSOR_FLOAT32 ||
outputType == OperandType::TENSOR_INT32)
<< "Unsupported output type for fill op: " << outputType;
NN_RET_CHECK(validateOutputTypes(context, {outputType}));
OperandType valueType;
NN_RET_CHECK(getValueType(outputType, &valueType));
NN_RET_CHECK(validateInputTypes(context, {OperandType::TENSOR_INT32, valueType}));
return Version::ANDROID_R;
}
bool prepare(IOperationExecutionContext* context) {
Shape dimsShape = context->getInputShape(kDimsTensor);
NN_RET_CHECK_EQ(getNumberOfDimensions(dimsShape), 1);
Shape outputShape = context->getOutputShape(kOutputTensor);
outputShape.dimensions.resize(dimsShape.dimensions[0]);
const int32_t* dims = context->getInputBuffer<int32_t>(kDimsTensor);
for (int i = 0; i < dimsShape.dimensions[0]; ++i) {
outputShape.dimensions[i] = dims[i];
}
return context->setOutputShape(kOutputTensor, outputShape);
}
bool execute(IOperationExecutionContext* context) {
switch (context->getInputType(kValueScalar)) {
case OperandType::FLOAT16:
return executeTyped<_Float16>(context);
case OperandType::FLOAT32:
return executeTyped<float>(context);
case OperandType::INT32:
return executeTyped<int32_t>(context);
default:
NN_RET_CHECK_FAIL() << "Unsupported value type for fill op.";
}
}
} // namespace fill_op
NN_REGISTER_OPERATION(FILL, "FILL", fill_op::validate, fill_op::prepare, fill_op::execute);
} // namespace nn
} // namespace android