Modern and emerging architectures demand increasingly complex compiler analyses and transformations. As the emphasis on compiler infrastructure moves beyond support for peephole optimizations and the extraction of instruction-level parallelism, they should support custom tools designed to meet these demands with higher-level analysis-powered abstractions of wider program scope. This paper introduces NOELLE, a robust open-source domain-independent compilation layer built upon LLVM providing this support. NOELLE is modular and demand-driven, making it easy-to-extend and adaptable to custom-tool-specific needs without unduly wasting compile time and memory. This paper shows the power of NOELLE by presenting a diverse set of ten custom tools built upon it, with a 33.2% to 99.2% reduction in code size (LoC) compared to their counterparts without NOELLE.
Semantic segmentation and extraction based on remote sensing images has important theory and significance. Deep learning has become one of the mainstream methods to extract information from remote sensing images. In this paper, based on the improvement of U-Net network structure, we combine ASPP and skip connection. Improve the residual module to improve the information extraction method. The main improvements of this paper are: 1 Based on the U-Net network structure, we use the multi-scale feature detection capabilities of Pyramid to introduce. The ASPP module and the residual structure are improved, paying more attention to semantic and detail informatization, overcoming the limitations of U-Net in small target detection; 2 We have improved the U-Net network, using skip connections to get more layers of information. Experiments show that the model proposed in this paper has significantly higher MPA and MIOU than the U-Net model on both the VOC dataset and the Vaihingen dataset. It means that ARU-Net can extract information better.
In view of the low precision of DV-Hop positioning algorithm in wireless sensor network, an improved DV-Hop localization algorithm that fuses chaotic sequences and Grey Wolf Optimization (GWO) algorithms is proposed. Firstly, in order to reduce the average jump distance error, the algorithm introduces the correction factor, and secondly, for the cumulative calculation error brought about by the least squares method, the improved gray wolf optimization algorithm is used as an alternative, in order to further improve the positioning accuracy and convergence speed, the chaotic sequence is introduced to initialize the gray wolf population, adaptive adjustment strategy is applied to control parameter a, and the algorithm global search ability is improved by using random walking strategy. The simulation results show that compared with the traditional DV-Hop algorithm, GA-DVhop algorithm and PSO-DVhop algorithm, the average positioning accuracy is improved by 90%, 86% and 78%.
With the rapid development of automatic driving technology, semantic segmentation of road image has become a hot research object of semantic segmentation. In order to better solve the problems of long image prediction time and rough category edges in road image segmentation, this paper proposes a segmentation network model based on deep separation and attention mechanism based on the optimization design of the backbone network of DeepLabV3+model. By improving the adaptability of the BNeck structure composed of DSC backbone network, reducing the number of model parameters to improve the prediction speed of the model, through the combination of spatial attention mechanism and channel attention mechanism, the weight of the model is more distributed among the regions of interest and channels, and finally using the overall architecture of encoder decoder to design a more efficient segmentation model.
Since the outbreak of the epidemic, it is extremely important to wear a mask correctly in the natural environment. Therefore, this paper proposes an improved SSD detection method. This paper takes the original SSD algorithm as the overall framework, and uses the lightweight network MobieNet-V2 to replace the original VGG-16 as the backbone network;In order not to compensate for the loss, the attention mechanism senet is added to improve its accuracy;And K-Means algorithm is used to cluster and generate anchor boxes suitable for data sets;Finally, Focal Loss is used to replace the Softmax of the original network as a new loss function. The experimental results show that compared with the original network, The improved network has better mAP, 8.9% higher than the original model.
A compiler's intermediate representation (IR) defines a program's execution plan by encoding its instructions and their relative order. Compiler optimizations aim to replace a given execution plan with a semantically-equivalent one that increases the program's performance for the target architecture. Alternative representations of an IR, like the Program Dependence Graph (PDG), aid this process by capturing the minimum set of constraints that semantically-equivalent execution plans must satisfy. Parallel programming like OpenMP extends a sequential execution plan by adding the possibility of running instructions in parallel, creating a parallel execution plan. Recently introduced parallel IRs, like TAPIR, explicitly encode a parallel execution plan. These new IRs finally make it possible for compilers to change the parallel execution plan expressed by programmers to better fit the target parallel architecture. Unfortunately, parallel IRs do not help compilers in identifying the set of parallel execution plans that preserve the original semantics. In other words, we are still lacking an alternative representation of parallel IRs to capture the minimum set of constraints that parallel execution plans must satisfy to be semantically-equivalent. Unfortunately, the PDG is not an ideal candidate for this task as it was designed for sequential code. We propose the Parallel Semantics Program Dependence Graph (PS-PDG) to precisely capture the salient program constraints that all semantically-equivalent parallel execution plans must satisfy. This paper defines the PS-PDG, justifies the necessity of each extension to the PDG, and demonstrates the increased optimization power of the PS-PDG over an existing PDG-based automatic-parallelizing compiler. Compilers can now rely on the PS-PDG to select different parallel execution plans while maintaining the same original semantics.
Aiming at the poor accuracy and slow segmentation speed of the current road scene semantic segmentation, an improved DeepLabV3+neural network segmentation algorithm is proposed.Firstly, the encoder network is changed to a more lightweight MobileNetV3 network, which increases the segmentation speed without sacrificing segmentation accuracy.Secondly, the atrous spatial convolutional pyramid pooling (ASPP) is improved in the form of dense connection, which further expands the receptive field while maintaining the size of the feature spatial resolution, and increases the feature reuse efficiency. And the dual attention mechanism is added to the encoder and Dense-ASPP module, which improves the recognition effect of the edge area of the model and improves the model segmentation accuracy.Finally, the multi-scale low-level semantic features extracted at the encoder are fused at the decoder, thereby reducing the loss of spatial information due to subsampling.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)