通用图形处理器缓存子系统性能优化方法综述

随着工艺和制程技术的不断发展以及体系架构的日趋完善,通用图形处理器(general purpose graphics processing units, GPGPU)的并行计算能力得到了很大的提升,其在高性能、高吞吐量等通用计算应用场景的使用越来越广泛.GPGPU通过支持大量线程的并发执行,可以较好地隐藏长延时访存操作,从而获得高并行计算能力.然而,GPGPU在处理计算和访存不规则的应用时,其存储子系统的效率受到很大影响,尤其是片上缓存的争用情况尤为突出,难以及时提供计算操作所需的数据,使得GPGPU的高并行计算能力不能得到充分发挥.解决片上缓存的争用问题、优化缓存子系统的性能,是优化GPGPU性能的主要解决方案之一,也是目前研究GPGPU性能优化的主要热点之一.目前,针对GPGPU缓存子系统的性能优化研究主要集中在线程级并行度(thread level parallelism, TLP)调节、访存顺序调节、数据通量增强、最后一级缓存(last level cache, LLC)优化和基于非易失性存储(non-volatile memory, NVM)的GPGPU缓存新架构设计等5个方面.也从这5个方面重点分析讨论了目前主要的GPGPU缓存子系统性能优化方法,并在最后指出了未来GPGPU缓存子系统优化需要进一步探讨的问题,对GPGPU缓存子系统性能优化的研究有重要意义.     

Shrunken Pore Syndrome Is Frequently Occurring in Severe COVID-19

A selective decrease in the renal filtration of larger molecules is attributed to the shrinkage of glomerular pores, a condition termed Shrunken Pore Syndrome (SPS). SPS is associated with poor long-term prognosis. We studied SPS as a risk marker in a cohort of patients with COVID-19 treated in an intensive care unit. SPS was defined as a ratio < 0.7 when the estimated glomerular filtration rate (eGFR), determined by cystatin C, calculated by the Cystatin C Caucasian-Asian-Pediatric-Adult equation (CAPA), was divided by the eGFR determined by creatinine, calculated by the revised Lund–Malmö creatinine equation (LMR). Clinical data were prospectively collected. In total, SPS was present in 86 (24%) of 352 patients with COVID-19 on ICU admission. Patients with SPS had a higher BMI, Simplified Physiology Score (SAPS3), and had diabetes and/or hypertension more frequently than patients without SPS. Ninety-nine patients in the total cohort were women, 50 of whom had SPS. In dexamethasone-naïve patients, C-reactive protein (CRP ), TNF-alpha, and interleukin-6 did not differ between SPS and non-SPS patients. Demographic factors (gender, BMI) and illness severity (SAPS3) were independent predictors of SPS. Age and dexamethasone treatment did not affect the frequency of SPS after adjustments for age, sex, BMI, and acute severity. SPS is frequent in severely ill COVID-19 patients. Female gender was associated with a higher proportion of SPS. Demographic factors and illness severity were independent predictors of SPS.     

基于联合块稀疏模型的随机调频步进ISAR成像方法

在回波数据稀疏、低信噪比等不利条件下,利用随机调频步进信号进行ISAR成像时,成像性能将会严重下降。针对上述问题,该文在充分分析随机调频步进信号回波特性的基础上,提出利用目标距离向具有的联合块稀疏特征来获得高质量ISAR图像的新方法。首先,推导了在随机调频步进信号发射波形条件下目标回波信号的联合块稀疏成像模型并分析了该模型特征;其次,提出了联合块稀疏正交匹配追踪稀疏重构算法(JBOMP)实现对模型的求解。该算法利用ISAR回波信号具有的块稀疏以及联合稀疏等先验信息,因此在低量测值、低信噪比条件下的ISAR成像性能得到了增强。所提算法还可以实现对多维信号的联合处理,且具有较快的运算速度。理论分析与仿真实验均验证了所提方法的有效性。     

Eine hydrodynamische Theorie der Flüssigkeits-reibung in Rollenlagern

Ohne Zusammenfassung VDI