mirror of https://github.com/alibaba/MNN.git
56 lines
2.0 KiB
C++
56 lines
2.0 KiB
C++
//
|
|
// ShapeBroadcastTo.cpp
|
|
// MNN
|
|
//
|
|
// Created by MNN on 2019/12/2.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include "shape/SizeComputer.hpp"
|
|
#include "core/Macro.h"
|
|
#include "core/TensorUtils.hpp"
|
|
|
|
namespace MNN {
|
|
|
|
class ShapeBroadcastTo : public SizeComputer {
|
|
virtual bool onComputeSize(const MNN::Op* op, const std::vector<Tensor*>& inputs,
|
|
const std::vector<Tensor*>& outputs) const override {
|
|
MNN_ASSERT(inputs.size() == 2);
|
|
MNN_ASSERT(outputs.size() == 1);
|
|
auto input = inputs[0];
|
|
auto shape = inputs[1];
|
|
auto output = outputs[0];
|
|
int inputDims = input->dimensions();
|
|
int shapeDims = shape->elementSize();
|
|
output->buffer().dimensions = inputDims > shapeDims ? inputDims : shapeDims;
|
|
const int dimension = output->dimensions();
|
|
const int* shapeData = shape->host<int>();
|
|
for (int i = 1; i <= dimension; ++i) {
|
|
int inputDim = 1, shapeDim = 1;
|
|
if (i <= inputDims) {
|
|
inputDim = input->length(inputDims - i);
|
|
}
|
|
if (i <= shapeDims) {
|
|
shapeDim = shapeData[shapeDims - i];
|
|
}
|
|
if (shapeDim <= 1) {
|
|
// shapeDim is {-1,0,1}, keep inputDim
|
|
output->setLength(dimension - i, inputDim);
|
|
} else {
|
|
// broadcast inputDim to shapeDim, need shapDim % inputDim == 0
|
|
// inputDim == 0, need shapeDim <= 0 keep dim
|
|
MNN_ASSERT(inputDim != 0);
|
|
MNN_ASSERT(shapeDim % inputDim == 0);
|
|
output->setLength(dimension - i, shapeDim);
|
|
}
|
|
}
|
|
output->buffer().type = input->buffer().type;
|
|
TensorUtils::getDescribe(output)->dimensionFormat = TensorUtils::getDescribe(input)->dimensionFormat;
|
|
return true;
|
|
}
|
|
};
|
|
|
|
REGISTER_SHAPE_INPUTS(ShapeBroadcastTo, OpType_BroadcastTo, {1});
|
|
|
|
} // namespace MNN
|