mirror of https://github.com/alibaba/MNN.git
81 lines
2.7 KiB
C++
81 lines
2.7 KiB
C++
|
//
|
||
|
|
// CPUBatchMatMul.cpp
|
||
|
|
// MNN
|
||
|
|
//
|
||
|
|
// Created by MNN on 2019/03/25.
|
||
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
||
|
|
//
|
||
|
|
|
||
|
|
#include "CPUBatchMatMul.hpp"
|
||
|
|
#include "CPUBackend.hpp"
|
||
|
|
#include "Matrix.hpp"
|
||
|
|
|
||
|
|
namespace MNN {
|
||
|
|
|
||
|
|
CPUBatchMatMul::CPUBatchMatMul(const Op* op, Backend* backend) : Execution(backend) {
|
||
|
|
// nothing to do
|
||
|
|
}
|
||
|
|
|
||
|
|
ErrorCode CPUBatchMatMul::onResize(const std::vector<Tensor*>& inputs, const std::vector<Tensor*>& outputs) {
|
||
|
|
auto input0 = inputs[0];
|
||
|
|
auto input1 = inputs[1];
|
||
|
|
auto output = outputs[0];
|
||
|
|
const int dimensions = input0->dimensions();
|
||
|
|
|
||
|
|
int batch = 1;
|
||
|
|
for (int i = 0; i < dimensions - 2; ++i) {
|
||
|
|
batch *= input0->length(i);
|
||
|
|
}
|
||
|
|
mBatch = batch;
|
||
|
|
|
||
|
|
std::vector<int> dimSizes(2);
|
||
|
|
|
||
|
|
dimSizes[0] = input0->length(dimensions - 2);
|
||
|
|
dimSizes[1] = input0->length(dimensions - 1);
|
||
|
|
mMatrixA.reset(Tensor::createDevice<float>(dimSizes));
|
||
|
|
|
||
|
|
dimSizes[0] = input1->length(dimensions - 2);
|
||
|
|
dimSizes[1] = input1->length(dimensions - 1);
|
||
|
|
mMatrixB.reset(Tensor::createDevice<float>(dimSizes));
|
||
|
|
|
||
|
|
dimSizes[0] = output->length(dimensions - 2);
|
||
|
|
dimSizes[1] = output->length(dimensions - 1);
|
||
|
|
mMatrixC.reset(Tensor::createDevice<float>(dimSizes));
|
||
|
|
|
||
|
|
return NO_ERROR;
|
||
|
|
}
|
||
|
|
|
||
|
|
ErrorCode CPUBatchMatMul::onExecute(const std::vector<Tensor*>& inputs, const std::vector<Tensor*>& outputs) {
|
||
|
|
auto input0 = inputs[0];
|
||
|
|
auto input1 = inputs[1];
|
||
|
|
auto output = outputs[0];
|
||
|
|
const int dimensions = input0->dimensions();
|
||
|
|
MNN_ASSERT(dimensions >= 3);
|
||
|
|
const int input0Stride = input0->stride(dimensions - 3);
|
||
|
|
const int input1Stride = input1->stride(dimensions - 3);
|
||
|
|
const int outputStride = output->stride(dimensions - 3);
|
||
|
|
const auto input0Ptr = input0->host<float>();
|
||
|
|
const auto input1Ptr = input1->host<float>();
|
||
|
|
float* const outputPtr = output->host<float>();
|
||
|
|
|
||
|
|
for (int i = 0; i < mBatch; ++i) {
|
||
|
|
mMatrixA->buffer().host = reinterpret_cast<uint8_t*>(input0Ptr + i * input0Stride);
|
||
|
|
mMatrixB->buffer().host = reinterpret_cast<uint8_t*>(input1Ptr + i * input1Stride);
|
||
|
|
mMatrixC->buffer().host = reinterpret_cast<uint8_t*>(outputPtr + i * outputStride);
|
||
|
|
Math::Matrix::multi(mMatrixC.get(), mMatrixA.get(), mMatrixB.get());
|
||
|
|
}
|
||
|
|
return NO_ERROR;
|
||
|
|
}
|
||
|
|
|
||
|
|
class CPUBatchMatMulCreator : public CPUBackend::Creator {
|
||
|
|
public:
|
||
|
|
virtual Execution* onCreate(const std::vector<Tensor*>& inputs, const std::vector<Tensor*>& outputs,
|
||
|
|
const MNN::Op* op, Backend* backend) const override {
|
||
|
|
return new CPUBatchMatMul(op, backend);
|
||
|
|
}
|
||
|
|
};
|
||
|
|
|
||
|
|
REGISTER_CPU_OP_CREATOR(CPUBatchMatMulCreator, OpType_BatchMatMul);
|
||
|
|
|
||
|
|
} // namespace MNN
|