mirror of https://github.com/alibaba/MNN.git
42 lines
1.4 KiB
C++
42 lines
1.4 KiB
C++
//
|
|
// ShapePadding.cpp
|
|
// MNN
|
|
//
|
|
// Created by MNN on 2019/6/24.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include "shape/SizeComputer.hpp"
|
|
#include "core/TensorUtils.hpp"
|
|
|
|
namespace MNN {
|
|
class PaddingComputer : public SizeComputer {
|
|
virtual bool onComputeSize(const MNN::Op* op, const std::vector<Tensor*>& inputs,
|
|
const std::vector<Tensor*>& outputs) const override {
|
|
if ((2 != inputs.size() && 3 != inputs.size()) || 1 != outputs.size()) {
|
|
MNN_ERROR("Padding inputs or outputs number error: %d -> %d\n", (int)inputs.size(), (int)outputs.size());
|
|
return false;
|
|
}
|
|
auto input = inputs[0];
|
|
auto paddings = inputs[1];
|
|
auto output = outputs[0];
|
|
output->buffer().type = input->buffer().type;
|
|
TensorUtils::copyShape(input, output, true);
|
|
|
|
auto size = paddings->elementSize();
|
|
if (size < output->dimensions() * 2) {
|
|
MNN_ERROR("Padding blob size not match output's dimension\n");
|
|
return false;
|
|
}
|
|
auto paddingPtr = paddings->host<int32_t>();
|
|
auto dimensions = input->dimensions();
|
|
for (int i = 0; i < dimensions; ++i) {
|
|
output->setLength(i, input->length(i) + paddingPtr[2 * i] + paddingPtr[2 * i + 1]);
|
|
}
|
|
return true;
|
|
}
|
|
};
|
|
|
|
REGISTER_SHAPE_INPUTS(PaddingComputer, OpType_Padding, {1});
|
|
} // namespace MNN
|