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.

176 lines
6.8 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 "ModelArgumentInfo"
#include "ModelArgumentInfo.h"
#include <LegacyUtils.h>
#include <algorithm>
#include <utility>
#include <vector>
#include "NeuralNetworks.h"
#include "TypeManager.h"
namespace android {
namespace nn {
static const std::pair<int, ModelArgumentInfo> kBadDataModelArgumentInfo{ANEURALNETWORKS_BAD_DATA,
{}};
std::pair<int, ModelArgumentInfo> ModelArgumentInfo::createFromPointer(
const Operand& operand, const ANeuralNetworksOperandType* type, void* data, uint32_t length,
bool paddingEnabled) {
if ((data == nullptr) != (length == 0)) {
const char* dataPtrMsg = data ? "NOT_NULLPTR" : "NULLPTR";
LOG(ERROR) << "Data pointer must be nullptr if and only if length is zero (data = "
<< dataPtrMsg << ", length = " << length << ")";
return kBadDataModelArgumentInfo;
}
ModelArgumentInfo ret;
uint32_t neededLength = 0;
if (data == nullptr) {
ret.mState = ModelArgumentInfo::HAS_NO_VALUE;
} else {
if (int n = ret.updateDimensionInfo(operand, type)) {
return {n, ModelArgumentInfo()};
}
if (operand.type != OperandType::OEM) {
neededLength = TypeManager::get()->getSizeOfData(operand.type, ret.mDimensions);
if (neededLength > length) {
LOG(ERROR) << "Setting argument with invalid length: " << length
<< ", minimum length expected: " << neededLength;
return kBadDataModelArgumentInfo;
}
}
ret.mState = ModelArgumentInfo::POINTER;
}
const uint32_t rawLength = neededLength == 0 ? length : neededLength;
const uint32_t padding = length - rawLength;
if (!paddingEnabled && padding > 0) {
LOG(ERROR) << "Setting argument with padded length without enabling input and output "
"padding -- length: "
<< length << ", expected length: " << neededLength;
return kBadDataModelArgumentInfo;
}
ret.mBuffer = data;
ret.mLocationAndLength = {.poolIndex = 0, .offset = 0, .length = rawLength, .padding = padding};
return {ANEURALNETWORKS_NO_ERROR, ret};
}
std::pair<int, ModelArgumentInfo> ModelArgumentInfo::createFromMemory(
const Operand& operand, const ANeuralNetworksOperandType* type, uint32_t poolIndex,
uint32_t offset, uint32_t length, bool paddingEnabled) {
ModelArgumentInfo ret;
if (int n = ret.updateDimensionInfo(operand, type)) {
return {n, ModelArgumentInfo()};
}
const bool isMemorySizeKnown = offset != 0 || length != 0;
uint32_t neededLength = 0;
if (isMemorySizeKnown && operand.type != OperandType::OEM) {
neededLength = TypeManager::get()->getSizeOfData(operand.type, ret.mDimensions);
if (neededLength > length) {
LOG(ERROR) << "Setting argument with invalid length: " << length
<< " (offset: " << offset << "), minimum length expected: " << neededLength;
return kBadDataModelArgumentInfo;
}
}
const uint32_t rawLength = neededLength == 0 ? length : neededLength;
const uint32_t padding = length - rawLength;
if (!paddingEnabled && padding > 0) {
LOG(ERROR) << "Setting argument with padded length without enabling input and output "
"padding -- length: "
<< length << ", offset: " << offset << ", expected length: " << neededLength;
return kBadDataModelArgumentInfo;
}
ret.mState = ModelArgumentInfo::MEMORY;
ret.mLocationAndLength = {
.poolIndex = poolIndex, .offset = offset, .length = rawLength, .padding = padding};
ret.mBuffer = nullptr;
return {ANEURALNETWORKS_NO_ERROR, ret};
}
int ModelArgumentInfo::updateDimensionInfo(const Operand& operand,
const ANeuralNetworksOperandType* newType) {
if (newType == nullptr) {
mInitialDimensions = operand.dimensions;
} else {
const uint32_t count = newType->dimensionCount;
mInitialDimensions = std::vector<uint32_t>(count);
std::copy(&newType->dimensions[0], &newType->dimensions[count], mInitialDimensions.begin());
}
mDimensions = mInitialDimensions;
return ANEURALNETWORKS_NO_ERROR;
}
Request::Argument ModelArgumentInfo::createRequestArgument() const {
switch (mState) {
case ModelArgumentInfo::POINTER: {
Request::Argument arg = {.lifetime = Request::Argument::LifeTime::POINTER,
.location = mLocationAndLength,
.dimensions = mDimensions};
arg.location.pointer = mBuffer;
return arg;
}
case ModelArgumentInfo::MEMORY:
return {.lifetime = Request::Argument::LifeTime::POOL,
.location = mLocationAndLength,
.dimensions = mDimensions};
case ModelArgumentInfo::HAS_NO_VALUE:
return {.lifetime = Request::Argument::LifeTime::NO_VALUE};
case ModelArgumentInfo::UNSPECIFIED:
LOG(FATAL) << "Invalid state: UNSPECIFIED";
return {};
};
LOG(FATAL) << "Invalid state: " << mState;
return {};
}
std::vector<Request::Argument> createRequestArguments(
const std::vector<ModelArgumentInfo>& argumentInfos,
const std::vector<DataLocation>& ptrArgsLocations) {
const size_t count = argumentInfos.size();
std::vector<Request::Argument> ioInfos(count);
uint32_t ptrArgsIndex = 0;
for (size_t i = 0; i < count; i++) {
const auto& info = argumentInfos[i];
switch (info.state()) {
case ModelArgumentInfo::POINTER:
ioInfos[i] = {.lifetime = Request::Argument::LifeTime::POOL,
.location = ptrArgsLocations[ptrArgsIndex++],
.dimensions = info.dimensions()};
break;
case ModelArgumentInfo::MEMORY:
case ModelArgumentInfo::HAS_NO_VALUE:
ioInfos[i] = info.createRequestArgument();
break;
default:
CHECK(false);
};
}
return ioInfos;
}
} // namespace nn
} // namespace android