Search

As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.

Through the development and calibration of a reference material which is 209.8 Ma old using a newly-developed Laser Ablation (LA) Multi-Collector Inductively Coupled Plasma Mass Spectrometry (MC-ICP-MS) technique, we successfully overcome the difficulty in sampling and dating ultra-low U-Pb ancient marine carbonates, which was previously untenable by isotope dilution (ID) methods. We developed the LA-MC-ICP-MS in situ U-Pb dating technique for ancient marine carbonates for the study of diagenesis-porosity evolution history in Sinian Dengying Formation, Sichuan Basin. By systematically dating of dolomitic cements from vugs, matrix pores and fractures, we found that the burial and diagenetic process of dolomite reservoirs in Sinian Dengying Formation was characterized by progressive filling-up of primary pores and epigenic dissolution vugs. The filling of vugs happened in three stages, early Caledonian, late Hercynian-Indosinian and Yanshanian-Himalayan, while the filling of matrix pores mainly took place in early Caledonian. The unfilled residual vugs, pores and fractures constitute the main reservoir sapce. Based on the above knowledge, we established the diagenesis-porosity evolution history of the dolomite reservoir in Sinian Dengying Formation, Sichuan Basin. These findings are highly consistent with the tectonic-burial and basin thermal histories of the study area. Our study confirmed the reliability of this in situ U-Pb dating technique, which provides an effective way for the investigation of diagenesis-porosity evolution history and evaluation of porosity in ancient marine carbonate reservoirs before hydrocarbon migration. Key words: laser ablation, in-situ U-Pb dating, Sichuan Basin, Sinian Dengying Formation, diagenesis-porosity evolution, carbonates

The problem of large area damage was analyzed when cleaning and dismantling gas-gas heater(GGH) corrugated plate in a factory. The causes of corrugated plate damage were mainly studied from the field soot and operating conditions. The results show that high pressure steam is not suitable for direct soot blowing, and the safe soot blowing method is decompressing high pressure steam into low pressure steam before soot blowing. High pressure steam with pressure higher than 0.8 MPa, and enthalpy value more than 2 978.4 kJ/kg can be used as the corrugated plate soot source. Research findings would play a positive role in maintaining the life of the corrugated plate and provide reference for safe blowing of air preheater and GGH.