Search

The slow-down of technology scaling combined with the exponential growth of modern machine learning and artificial intelligence models has created a demand for specialized accelerators, such as GPUs, ASICs, and field-programmable gate arrays (FPGAs). FPGAs can be reconfigured and have the potential to outperform other accelerators, while also being more energy-efficient, but are cumbersome to use with today's fractured landscape of tool flows. We propose the concept of an operation set architecture to overcome the current incompatibilities and hurdles in using DNN-to-FPGA compilers by combining existing specialized frameworks into one organic compiler that also allows the efficient and automatic re-use of existing community tools. Furthermore, we demonstrate that mixing different existing frameworks can increase the efficiency by more than an order of magnitude.

Nowadays, a new parallel paradigm for energy-efficient heterogeneous hardware infrastructures is required to achieve better performance at a reasonable cost on high-performance computing applications. Under this new paradigm, some application parts are offloaded to specialized accelerators that run faster or are more energy-efficient than CPUs. Field-Programmable Gate Arrays (FPGA) are one of those types of accelerators that are becoming widely available in data centers. This paper proposes OmpSs@cloudFPGA, which includes novel extensions to parallel task-based programming models that enable easy and efficient programming of heterogeneous clusters with FPGAs. The programmer only needs to annotate, with OpenMP-like pragmas, the tasks of the application that should be accelerated in the cluster of FPGAs. Next, the proposed programming model framework automatically extracts parts annotated with High-Level Synthesis (HLS) pragmas and synthesizes them into hardware accelerator cores for FPGAs. Additionally, our extensions include and support two novel features: 1) FPGA-to-FPGA direct communication using a Message Passing Interface (MPI) similar Application Programming Interface (API) with one-to-one and collective communications to alleviate host communication channel bottleneck, and 2) creating and spawning work from inside the FPGAs to their own accelerator cores based on an MPI rank-like identification. These features break the classical host-accelerator model, where the host (typically the CPU) generates all the work and distributes it to each accelerator. We also present an evaluation of OmpSs@cloudFPGA for different parallel strategies of the N-Body application on the IBM cloudFPGA research platform. Results show that for cluster sizes up to 56 FPGAs, the performance scales linearly. To the best of our knowledge, this is the best performance obtained for N-body over FPGA platforms, reaching 344 Gpairs/s with 56 FPGAs. Finally, we compare the performance and power consumption of the proposed approach with the ones obtained by a classical execution on the MareNostrum 4 supercomputer, demonstrating that our FPGA approach reduces power consumption by an order of magnitude. This work has been done in the context of the IBM/BSC Deep Learning Center initiative. This work has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 754337 (EuroEXA), from Spanish Government (PID2019-107255GBC21/AEI/10.13039/501100011033), and from Generalitat de Catalunya (2017-SGR-1414 and 2017-SGR-1328).

The evolution of cloud applications into loosely-coupled microservices opens new opportunities for hardware accelerators to improve workload performance. Existing accelerator techniques for cloud sacrifice the consolidation benefits of microservices. This paper presents CloudiFi, a framework to deploy and compare accelerators as a cloud service. We evaluate our framework in the context of a financial workload and present early results indicating up to 485x gains in microservice response time., 3 pages, 6 figures

Background and ObjectivesThe decline of respiratory function in Duchenne muscular dystrophy (DMD) is associated with sleep disordered breathing (SDB) and alteration of nocturnal gas exchange, first manifesting as nocturnal hypoventilation (NH). However, the correlation between the pulmonary function measured by spirometry (PFT) and the onset of SDB with or without NH is unclear. The aims of this study are to identify the prevalence and features of SDB and to investigate the relationship between lung function determined by forced vital capacity (FVC) and sleep abnormalities in a large pediatric DMD population.MethodsThis was a retrospective, single-center cohort study. FVC% predicted (FVC%) was calculated using predicted equations from the Global Lung Function Initiative. NH was defined by transcutaneous (tc) CO2 >50 mm Hg for >25% of total sleep time (TST), borderline NH by a mean tcCO2 between 45 and 50 mm Hg or tcCO2>50 mm Hg for ≤25% of TST, and clinically meaningful obstructive sleep apnea (OSA) by obstructive apnea-hypopnea index >5. The sensitivity, specificity, and positive and negative predictive values of FVC < 50% to indicate the presence of nocturnal hypoventilation were calculated.ResultsOne hundred thirty-four patients underwent 284 sleep studies and 1222 PFT. The mean (SD) age at the first and the last sleep study was 12.9 (2.7) and 14.3 (2.6) years, respectively. Borderline NH (n = 31) was detected in both ambulant and early-nonambulant participants, while 100% of NH cases (n = 14) were nonambulant. NH was detected in 4 of the 14 patients despite an FVC >50%. Seventeen of the 26 patients with OSA presented with concomitant NH or borderline NH. FVC DiscussionFVC% 50%. These results are relevant for the clinical management of SDB.

Treatment of infants with craniofacial malformations, e.g. Robin sequence, is characterized by considerable heterogeneity and a lack of randomized trials to identify an optimal approach. We propose to establish an international register using a common minimal dataset that will better allow for a comparison between key determinants and outcomes in these patients. In infants, this should include an assessment of mandibular micrognathia, glossoptosis, upper airway obstruction, weight gain and mode of feeding. Later on, neurocognition, speech development, hearing and quality of life should also be included. Together, these data will help better to advice parents on which treatment to choose for their baby with a craniofacial malformation.

Aim The primary aim is to systematically review the specific airway and/or breathing problems which have been reported in Pierre Robin Sequence (PRS). Secondary aims are to understand the prevalence and severity of these airway and breathing problems, and options for screening, assessment, and monitoring. Method A search strategy will be designed to search the following databases: MEDLINE, Embase, Web of Science, and the Cochrane Central Register of Controlled Trials(CENTRAL) including a grey literature search, to identify publications on airway and/or breathing problems in Pierre Robin Sequence patients. Airway or breathing problems to be included are defined upper airway conditions (e.g., sleep apnoea) or clinically or objectively defined airway/breathing problems (e.g., increased respiratory rate, polysomnography). Screening will exclude non-English articles, abstracts, letters, editorials, expert opinions and breathing problems not anatomically related to craniofacial underdevelopment. Results This systematic review has been successfully registered on the PROSPERO International Prospective Register of Systematic Reviews (CRD42020210572) Conclusions PRS is a rare craniofacial condition which presents at birth, consisting of micrognathia, glossoptosis and airway problems, and is usually, but not always, associated with a cleft palate. There is no consensus on the type of airway problems that affect patients with PRS. The findings of this systematic review will help to develop a consensus on the airway and breathing problems in PRS and options for assessment and monitoring of the airway and breathing problem. This data may also contribute to development of a standardized guideline for management of airway and breathing problems in PRS.

Chiari malformation incorporate numerous forms of congenital or acquired cerebellar herniation through the foramen magnum. This may lead to brain stem, high spinal cord and cranial nerve compression resulting in obstructive and central apneas. This review highlights he high prevalence of sleep-disordered breathing in this population and the importance of refering these patients for sleep studies as part of their workup. A review of the literature was performed through a PubMed and EMBASE search of original articles and reviews using the key words “chiari” “chiari malformation” “hindbrain herniation” “sleep disordered breathing” “obstructive sleep apnea” “central sleep apnea" “sleep study” and “foramen magnum decompression”. We highlight the pathophysiology of sleep disordered breathing in patients with Chiari malformation, how it can be diagnosed and what the treatment options are. Sleep-disordered breathing is highly prevalent in patients with CM1. Clinicians caring for these patients should be aware of this and prioritise sleep diagnostic testing to allow for early diagnosis and management particularly in the presence of neurological symptoms and specific brain MRI pointers.

The slowdown of Moore's law and the end of Dennard scaling created a demand for specialized accelerators, including Field Programmable Gate Arrays (FPGAs), in cloud data centers. At the same time, compute resources are increasingly consumed via public and private clouds and traditional applications are modernized using scalable microservices and Function-as-a-Service (FaaS) offerings. Nonetheless, true FaaS based on FPGAs or other accelerators is virtually absent from the offering catalogs of all major cloud providers. In addition, FPGA applications are typically coded in a monolithic fashion, due to device and vendor specific dependencies, which reduces the portability and usability of FPGA cloud offerings further. However, FPGA-based FaaS can improve execution efficiency and minimize (tail-) latencies while decreasing costs. We propose a novel system architecture, called Mantle, that uses disaggregated FPGAs to enable scalable, usable, portable and efficient FaaS offerings for FPGAs. Our experimental results demonstrate a significant reduction of end-to-end service provisioning time to below 7 seconds and an increase in execution efficiency by a factor of 4 with negligible overhead.

With the slowdown of Moore's law and the stop of Dennard scaling, energy efficiency of compute hardware translates to compute power. Therefore, High-Performance Computing (HPC) systems tend to rely more and more on accelerators such as Field-Programmable Gate Arrays (FPGAs) to fuel high demanding workloads, like Big Data applications or Deep Neuronal Networks. These FPGAs are reconfigurable and sometimes no longer bus-attached to a CPU but directly connected to the data center network fabric as standalone nodes. This mix of CPUs and FPGAs leads to the creation of Reconfigurable Heterogeneous HPC (ReH2PC) clusters for which no established programming model exists, despite many proposals in the past. In contrast to this, the Message Passing Interface (MPI) has evolved as the de-facto standard to program classical HPC clusters, due to its high-re-usability and fast development of applications. This paper revisits the programming model of ReH2PC clusters and argues that MPI is suitable for program-ming heterogeneous clusters of FPGAs and CPUs. Our experiments with 31 FPGAs show an average speedup of 4 and a 90% reduction of power consumption compared to a cluster of CPUs. We demonstrate a one-click solution for compiling and deploying a standard MPI application on ReH2PC clusters. Our framework implements a High-Level Synthesis (HLS) library, a specific run-time environment for FPGAs and CPUs, and a transpiler that closes the semantic gap between the MPI API and FPGA designs. Our experiments with 31 FPGAs show an average speedup of 4 and a 90% reduction of power consumption compared to a cluster of CPUs.

Nusinersen (NUS), the first treatment approved for Spinal Muscular Atrophy type 1 (SMA1), was made available in the UK for SMA1 through the Expanded Access Program (EAP) in 2017. The Great Ormond Street Respiratory (GSR) score was developed as an objective respiratory assessment for children with SMA1 during their treatment. Aims: Track respiratory status of SMA1 children over the course of Nusinersen treatment and compare GSR scores amongst SMA1 sub-types. Single centre study on SMA1 patients using the GSR score at set time points: prior to first NUS dose; 2 weeks post end of loading doses; 2 weeks post-subsequent doses. GSR score ranges 1–28, being 1–9 = Stable minimal support, thorough to 23–28 = Poor reserve with maximum support. 20 SMA1 children underwent NUS treatment between January 2017 - November 2018. Median age of diagnosis was 5.0 months. NUS started at median of 9.57 months. From 5th dose onwards, GSR scores were significantly lower for Type 1C patients compared to Type 1B By month 18, irrespective of subtypes, the whole cohort appears to stabilise GSR Scores. As treatment duration increases, an overall stabilisation of respiratory status across the cohort was observed. Further longitudinal studies are needed to validate the GSR.

Over the past two decades, the Message Passing Interface (MPI) has evolved as the de-facto standard for programming High-Performance Computing (HPC) clusters. Its widespread utilization led to the rapid development of applications and high reusability. Meanwhile, energy- and compute-efficient devices such as Field-Programmable Gate Arrays (FPGAs) are stepping into modern data centers and HPC clusters to address the nearing end of technology scaling. This combination of traditional CPU servers and FPGA nodes leads to Reconfigurable Heterogeneous HPC (ReH 2 PC) systems that are particularly cumbersome to program because of the absence of a standard programming model. This work advocates the use of MPI to program such ReH 2 PC clusters and presents a proof of concept based on a cross-compiler, a High-Level Synthesis library, C++ library, an FPGA- and a CPU-runtime environment. The result is a one-click solution, which compiles a standard MPI application for a ReH 2 PC cluster.

Spinal muscular atrophy (SMA) is a degenerative motor neurone disorder causing progressive muscular weakness. Without assisted ventilation or novel therapies, most children with SMA type 1 die before the second year of life due to respiratory failure as the respiratory muscles and bulbar function are severely affected. Active respiratory treatment (mechanically assisted cough, invasive or non-invasive ventilation) has improved survival significantly in recent decades, but often at the cost of becoming ventilator dependent. The advent of a new oligonucleotide based therapy (Nusinersen) has created new optimism for improving motor function. However, the long-term effect on respiratory function is unclear and non-invasive respiratory support will remain an important part of medical management in patients with SMA. This review summarises the existing knowledge about sleep-disordered breathing and respiratory failure in patients with SMA, especially type 1, as well as the evidence of improved outcome and survival in patients treated with non-invasive or invasive ventilation. Practical considerations and ethical concerns are delineated with discussion on how these may be affected by the advent of new therapies such as Nusinersen.

ObjectiveTo test the hypothesis that children with Prader-Willi syndrome (PWS) and obstructive sleep apnoea syndrome (OSAS) have hypercapnia for higher proportion of total sleep time (TST) than non-syndromic children with similar obstructive apnoea–hypopnoea index (OAHI).DesignCross-sectional study.SettingTwo tertiary care hospitals.PatientsPatients with PWS and non-syndromic children with snoring who underwent polygraphy and were of similar age, body mass index (BMI) z-score and OAHI.Main outcome measureThe two groups were compared regarding %TST with transcutaneous CO2 (PtcCO2) >50 mm Hg. The interaction between PWS diagnosis and OSAS severity (OAHI 5 episodes/h) regarding %TST with PtcCO2 >50 mm Hg was tested using multiple linear regression.Results48 children with PWS and 92 controls were included (median age 2.3 (range 0.2–14.1) years vs 2.2 (0.3–15.1) years; BMI z-score 0.7±1.9 vs 0.8±1.7; median OAHI 0.5 (0–29.5) episodes/h vs 0.5 (0–33.9) episodes/h; p>0.05). The two groups did not differ in %TST with PtcCO2 >50 mm Hg (median 0% (0–100%) vs 0% (0–81.3%), respectively; p>0.05). However, the interaction between PWS and OSAS severity with respect to duration of hypoventilation was significant (p2 >50 mm Hg between children with PWS and controls for OAHI 5 episodes/h were +0.2%, +1% and +33%, respectively.ConclusionIncreasing severity of upper airway obstruction during sleep in children with PWS is accompanied by disproportionately longer periods of hypoventilation when compared with non-syndromic children with similar frequency of obstructive events.

There are a wide range of surgical and maxillofacial options to alleviate upper airway obstruction in children with craniofacial disorders. The nasopharyngeal prong (NPP) is a simple idea where the airway obstruction arising from the posteriorly placed tongue secondary to a small mandible can be overcome quickly and without resorting to more invasive surgical procedures. The role of the NPP is of particular interest in Robin sequence (RS). RS describes a congenital anomaly with retrognathia (often with associated U-shaped cleft palate) where in some children the tongue and small jaw can significantly impact on airway patency with upper airway obstruction. The NPP is a modified endotracheal tube, of ideal diameter and cut to a desired length that can bypass the airway obstruction and regain patency to the upper airway. RS has a natural history of mandibular growth and resolution of the symptoms with time and thus the NPP presents a medium term solution precluding the need in selected children of more permanent and potentially unnecessary surgical procedures. The role of the NPP in other craniofacial disorders (either as a temporary solution or longer term option) requires further research.

Background Despite recent clinical guideline development, the best pathway of care for children with symptoms of obstructive sleep-disordered breathing (oSDB) is still debated. This systematic review aims to map the research in childhood oSDB that has been conducted so far to support further guideline development, identify evidence gaps, and guide future research. Methods A systematic search was performed in PubMed, EMBASE, and the Cochrane Library from inception to November 26, 2015. All publications on childhood oSDB were included, irrespective of type and language. The annual number of publications in the field of oSDB was counted over all years; for those published since January 1, 2011 (ie, the date of the latest literature search of the American Academy of Pediatrics guideline), total and annual numbers of publications across main research themes and methodologies were calculated. Results Of the 7,637 unique records retrieved, 5,871 publications were eligible for inclusion. There was an increase in annual publications since 2000, with 46% published since 2011. Most publications (61%) focused on individual treatment modalities, incidence, or prognosis. Few publications (2.7%) focused on health service delivery, outcomes, and health economics. Observational studies composed 78.5% of publications, 2.4% were randomized controlled trials, and 0.4% used a qualitative approach as their main methodology. Conclusions A recent surge in research activity into childhood oSDB has improved the knowledge base for this condition; however, the lack of health services, health economics, and outcomes research impacts the applicability of evidence informing current guidance and leaves important questions for future research. Registration PROSPERO: registration number CRD42015029291

Corticosteroids (CSs) have prolonged survival and respiratory function in boys with Duchenne muscular dystrophy (DMD) when compared with CSs-naïve boys.The differential impact of frequently used CSs and their regimens on long-term (5 years) cardiorespiratory progression in children with DMD is unknown.This was a retrospective longitudinal study including children with DMD followed at Dubowitz Neuromuscular Centre, Great Ormond Street Hospital London, England, from May 2000 to June 2017. Patients enrolled in any interventional clinical trials were excluded. We collected patients' anthropometrics and respiratory (FVC, FVC % predicted and absolute FVC, and noninvasive ventilation requirement [NIV]) and cardiac (left ventricular shortening function [LVFS%]) function. CSs-naïve patients had never received CSs. Patients who were treated with CSs took either deflazacort or prednisolone, daily or intermittently (10 days on/10 days off) for1 month. Average longitudinal models were fitted for yearly respiratory (FVC % predicted) and cardiac (LVFS%) progression. A time-to-event analysis to FVC % predicted 50%, NIV start, and cardiomyopathy (LVFS% 28%) was performed in CS-treated (daily and intermittent) vs CS-naïve patients.There were 270 patients, with a mean age at baseline of 6.2 ± 2.3 years. The median follow-up time was 5.6 ± 3.5 years. At baseline, 263 patients were ambulant. Sixty-six patients were treated with CSs daily, 182 patients underwent CSs intermittent60% treatment, and 22 were CS-naïve patients. Yearly FVC % predicted declined similarly from 9 years (5.9% and 6.9% per year, respectively; P = .27) in the CSs-daily and CSs-intermittent groups. The CSs-daily group declined from a higher FVC % predicted than the CSs-intermittent group (P .05), and both reached FVC % predicted 50% and NIV requirement at a similar age,2 years later than the CS-naïve group. LVFS% declined by 0.53% per year in the CSs-treated group irrespective of the CSs regimen, significantly slower (P .01) than the CSs-naïve group progressing by 1.17% per year. The age at cardiomyopathy was 16.6 years in the CSs-treated group (P .05) irrespective of regimen and 13.9 years in the CSs-naïve group.CSs irrespective of the regimen significantly improved respiratory function and delayed NIV requirement and cardiomyopathy.