Search

Stencil computation patterns are the backbone of many scientific and engineering simulations. The stencil computation is known to be constrained by its high demand of memory bandwidth, which limits performance on accelerators such as GPUs. Prior GPU-based approaches concentrated on stencils with only axis-aligned grid points with 2D caching schemes. However, for stencils with non-axis grid points, prior approaches use 3D caching or multi-pass 2D caching schemes. These methods suffer from either large number of global memory accesses or required large size of the shared memory. In this work, we present an efficient GPU implementation scheme “Scatter Without Write Conflict” (SWiC) for large advanced 3D stencil patterns involving non-axis-aligned grid points. Unlike other 3D caching schemes, SWiC only needs 2D caching and a single pass over the stencil per iteration. SWiC achieves, significant reductions in the global memory accesses, without increasing the size of shared memory. SWiC can also be applied to simple axis-aligned stencils without any performance loss. Moreover, we propose a scalable implementation for the halo region exchange of 3D stencils on multi-GPU nodes. For evaluation, we test SWiC on three Nvidia GPU generations and show that our approach significantly outperforms existing state-of-the-art GPU implementations with a speedup ranging from 1.6x to 5.75x. We also provide a detailed scaling analysis in multi-node and multi-GPU environments.

Spinal muscular atrophy was recently added to the Wisconsin newborn screening panel. Here we report our screening methods, algorithm, and outcomes. Methods: A multiplex real-time PCR assay was used to identify newborns with homozygous SMN1 exon 7 deletion, and those newborns’ specimens further underwent a droplet digital PCR assay for SMN2 copy number assessment. An independent dried blood spot specimen was collected and tested to confirm the initial screening results for SMN1 and SMN2. Results: From October 15, 2019 to October 14, 2020, a total of 60,984 newborns were screened for spinal muscular atrophy. Six newborns screened positive for and were confirmed to have spinal muscular atrophy, making the Wisconsin spinal muscular atrophy birth prevalence 1 in 10,164. Of these six infants, two have two copies of SMN2, two have three copies of SMN2, and two have four copies of SMN2. Five newborns received Zolgensma therapy, and one newborn received Spinraza therapy. Conclusions: Our screening method'spositive predictive value is 100%. This comprehensive approach, providing both timely SMN2 information and SMN1 and SMN2 confirmation as parts of the algorithm for spinal muscular atrophy newborn screening, facilitated timely clinical follow-up, family counseling, and treatment planning.

BACKGROUND: Newborn screening identifies infants with congenital hypothyroidism (CH) for whom levothyroxine (L-T(4)) prevents cognitive impairment but also can identify infants with transient hypothyroidism. For some, transient hypothyroidism can be ruled out by thyroid gland imaging; otherwise, it is confirmed when thyroid stimulating hormone (TSH) concentrations remain normal after a supervised trial off L-T(4), typically after age 3 years. OBJECTIVES: To measure the rates of thyroid gland imaging and L-T(4) discontinuation and to assess whether discontinuation was monitored with TSH testing. METHODS: This is a retrospective analysis of claims data from the IBM® MarketScan® Databases for children born during 2010–2016 and continuously enrolled in a non-capitated employer-sponsored private health insurance plan or in Medicaid for ≥36 months from the date of the first filled L-T(4) prescription. RESULTS: 263 privately-insured and 241 Medicaid-enrolled children met the inclusion criteria. More privately-insured than Medicaid-enrolled children had imaging between the first filled prescription and 180 days after the last filled prescription (24.3% vs. 12.9%; P=0.001). By 36 months, 35.7% discontinued L-T(4), with no difference by insurance status (P=0.48). Among those who discontinued, 29.1% of privately-insured children and 47.7% of Medicaid-enrolled children had no claims for TSH testing within the next 180 days (P=0.01). CONCLUSIONS: Nearly one-third of children with suspected CH discontinued L-T(4) by 3 years and fewer Medicaid-enrolled than privately-insured children received timely follow-up TSH testing. Future studies are indicated to understand the quality of care and developmental outcomes for children with CH and barriers to guideline adherence in evaluating for transient CH.

Purpose We aimed to estimate the carrier frequency of Zellweger spectrum disorder (ZSD), a rare autosomal-recessive disease, and the associated disease incidence based on data from the Exome Aggregation Consortium (ExAC) of approximately 60,000 individuals. Methods We obtained variants from ExAC in 13 PEX genes associated with ZSD. Potentially pathogenic missense variants were identified with computational variant analysis tools according to three stringency levels. Using variants classified as potentially pathogenic, we estimated the carrier frequency and the associated incidence for the entire ExAC population and its sub-populations. We also evaluated variants based on pathogenicity criteria for sequence variant interpretation outlined by the American College of Medical Genetics and Genomics (ACMG) and calculated the carrier frequency and incidence based on those variants. Results The bioinformatically estimated incidence rate of ZSD in the ExAC population is 1 in 83,841 using our least stringent pathogenicity cutoff. Under clinical guidelines outlined by ACMG, the estimated incidence is 1 in 3,275,751 births. Conclusion We outlined a process for estimating the ZSD disease carrier frequency and incidence in a large consortium using bioinformatics tools. Our results are close to current newborn screening estimates in New York of 1 in 90,000 births, estimated from 1.08 million screenings.

Nowadays, computer researchers can face ever-complex scientific problems by using a hardware and software co-design. One successful approach is exploring novel massively-parallel Natural-inspired algorithms, such as the Ant Colony Optimization (ACO) algorithm, through the exploitation of high-throughput accelerators such as GPUs, which are designed to provide high levels of parallelism and low Energy per instruction (EP) cost through heavy vectorization. In this paper, we demonstrate how to take advantage of contemporary hardware-based CUDA vectorization to optimize the ACO algorithm when applied to the Traveling Salesman Problem (TSP). Several parallel designs are proposed and analyzed on two different CUDA architectures. Our results reveal that our vectorization approaches can obtain good performance on these architectures. Moreover, atomic operations are under study showing good benefits on latest generations of CUDA architectures. This work lays the groundwork for future developments of ACO algorithm on high-performance platforms.

With the society’s growing adoption of machine learning (ML) and deep learning (DL) for various intelligent solutions, it becomes increasingly imperative to standardize a common set of measures for ML/DL models with large scale open datasets under common development practices and resources so that people can benchmark and compare models quality and performance on a common ground. MLCommons has emerged recently as a driving force from both industry and academia to orchestrate such an effort. Despite its wide adoption as standardized benchmarks, MLCommons Inference has only included a limited number of ML/DL models (in fact seven models in total). This significantly limits the generality of MLCommons Inference’s benchmarking results because there are many more novel ML/DL models from the research community, solving a wide range of problems with different inputs and outputs modalities. To address such a limitation, we propose MLHarness, a scalable benchmarking harness system for MLCommons Inference with three distinctive features: (1) it codifies the standard benchmark process as defined by MLCommons Inference including the models, datasets, DL frameworks, and software and hardware systems; (2) it provides an easy and declarative approach for model developers to contribute their models and datasets to MLCommons Inference; and (3) it includes the support of a wide range of models with varying inputs/outputs modalities so that we can scalably benchmark these models across different datasets, frameworks, and hardware systems. This harness system is developed on top of the MLModelScope system, and will be open sourced to the community. Our experimental results demonstrate the superior flexibility and scalability of this harness system for MLCommons Inference benchmarking.

In the Plain Community, there is an increased frequency of genetic disorders including phenylalanine hydroxylase (PAH) deficiency. Common pathogenic variants have been observed due to founder effect and closed community. This study obtained genotypes of 12 Plain individuals with PAH deficiency, identified through newborn screen or diagnosed by symptomatic presentation, who are receiving medical care at the University of Wisconsin metabolic clinic. Genotype and phenotypic data were evaluated to characterize genotype-phenotype correlations. Results can inform the need for confirmatory testing for the disorder and provide a better understanding of the biochemical phenotype, which may help with management. Keywords: Neonatal screening, Phenylketonuria, Genotype

Activities such as the suppression of coal dust would benefit from an improved understanding about how the presence of surfactant alters the charge distribution of spray drops. However, there is no standard method to measure charge and size distributions of nozzle-generated spray drops. To fill this gap, a new spray drop charge measurement system was developed to measure a broad range of sizes and electrostatic charges for spray drops based on the principle that the free-falling drops can be separated according to their electrostatic mobility in an electrical field with known strength. High and low concentration (1 × 10−4 and 1 × 10−6 M, respectively) anionic, cationic, and nonionic surfactant sprays and water spray were tested. Nonionic and cationic surfactant-containing sprays and water spray carried net positive charge on average, while anionic surfactant-containing sprays tended to carry net negative charge on average. Increasing surfactant concentration did not increase the magnitude of drop charge among the tested surfactants. Drop charge level was significantly higher among larger size drops, with the average charge per particle on 320 μm drops 1000 to 10,000 times higher than on 20 μm drops. The spray drop charge measurement system developed in this study can be used to measure spray drops with an electrostatic charge ranging from 0 to 960,000 charges per drop and diameter ranging from 20 to 270 μm. Potential design improvements described in this study and additional computing power could improve measurement accuracy and precision, as well as the capacity of the system.

Orotic aciduria in the presence of hyperammonemia is a key indicator for a defect in the urea cycle, specifically ornithine transcarbamylase (OTC) deficiency. Current newborn screening (NBS) protocols can detect several defects of the urea cycle, but screening for OTC deficiency remains a challenge due to the lack of a suitable assay. The purpose of this study was to develop a high-throughput assay to measure orotic acid in dried blood spot (DBS) specimens as an indicator for urea cycle dysfunction, which can be readily incorporated into routine NBS.Orotic acid was extracted from DBS punches and analyzed using flow-injection analysis tandem mass spectrometry (FIA-MS/MS) with negative-mode ionization, requiring2 min/sample run time. This method was then multiplexed into a conventional newborn screening assay for analysis of amino acids, acylcarnitines, and orotic acid.We describe 2 assays which can quantify orotic acid in DBS: a stand-alone method and a combined method for analysis of orotic acid, amino acids, and acylcarnitines. Both methods demonstrated orotic acid recovery of 75-85% at multiple levels of enrichment. Precision was also comparable to traditional FIA-MS/MS methods. Analysis of residual presumptively normal NBS specimens demonstrated a 5:1 signal to noise ratio and the average concentration of orotic acid was approximately 1.2 μmol/l. The concentration of amino acids and acylcarnitines as measured by the combined method showed no significant differences when compared to the conventional newborn screening assay. In addition, retrospective analysis of confirmed patients and presumptively normal newborn screening specimens suggests potential for the methods to identify patients with OTC deficiency, as well as other urea cycle defects.The assays described here quantify orotic acid in DBS using a simple extraction and FIA-MS/MS analysis procedures that can be implemented into current NBS protocols.

With the industry-wide switch to multicore and manycore architectures, parallel computing has become the only venue in sight for continued growth in application performance. In order for the performance of an application to grow with future generations of hardware, a significant portion of its computation must be done with scalable parallel algorithms. It is therefore important to develop and deploy as many scalable parallel algorithms as possible. This paper takes a critical look at the major challenges involved in the development of scalable parallel algorithms and points to needs for compiler tool innovations to help address these challenges.

BackgroundThere has been great variation and uncertainty about how many and what CFTR mutations to include in cystic fibrosis (CF) newborn screening algorithms, and very little research on this topic using large populations of newborns.MethodsWe reviewed Wisconsin screening results for 1994–2008 to identify an ideal panel.ResultsUpon analyzing approximately 1 million screening results, we found it optimal to use a 23 CFTR mutation panel as a second tier when an immunoreactive trypsinogen (IRT)/DNA algorithm was applied for CF screening. This panel in association with a 96th percentile IRT cutoff gave a sensitivity of 97.3%, but restricting the DNA tier to F508del was associated with 90% (P

Background Severe combined immunodeficiency (SCID) is characterized by the absence of functional T cells and B cells. Without early diagnosis and treatment, infants with SCID die from severe infections within the first year of life. Objective To determined the feasibility of detecting SCID in newborns by quantitating T-cell receptor excision circles (TRECs) from dried blood spots (DBSs) on newborn screening (NBS) cards. Methods DNA was extracted from DBSs on deidentified NBS cards, and real-time quantitative PCR (RT-qPCR) was used to determine the number of TRECs. Positive controls consisted of DBS from a 1-week-old T − B − NK + patient with SCID and whole blood specimens selectively depleted of naive T cells. Results The mean and median numbers of TRECs from 5766 deidentified DBSs were 827 and 708, respectively, per 3.2-mm punch (∼3 μL whole blood). Ten samples failed to amplify TRECs on initial analysis; all but 1 demonstrated normal TRECs and β-actin amplification on retesting. No TRECs were detected in either the SCID or naive T-cell–depleted samples, despite the presence of normal levels of β-actin. Conclusions The use of RT-qPCR to quantitate TRECs from DNA extracted from newborn DBSs is a highly sensitive and specific screening test for SCID. This assay is currently being used in Wisconsin for routine screening infants for SCID.

To efficiently accommodate standards changes and algorithmic improvements, functional reconfigurability is increasingly desired for media processing. Such adaptability, however, generally comes at significant power cost. This work suggests that another dimension of adaptation can be beneficial -power adaptation. Through a unique compiler-hardware approach, we (1) demonstrate an extension to the state-of-the-art in data analyzability, yielding better control over scratchpad data management, and (2) combine this knowledge with an SRAM having variable latency and access properties, yielding adjustable power savings. Building upon the compiler techniques presented by [1], we evaluate the severity of the current on-chip storage power problem and detail how SRAM structures can be built to enable data power savings for media applications. We show how the implemented compiler techniques can be applied to other problems in the embedded/media processing domain, and present net data power savings results for a suite of media and telecommunication applications, including MPEG-2, MPEG-4, H.263, and JPEG-2000.

Contemporary many-core processors such as the GeForce 8800 GTX enable application developers to utilize various levels of parallelism to enhance the performance of their applications. However, iterative optimization for such a system may lead to a local performance maximum, due to the complexity of the system. We propose program optimization carving, a technique that begins with a complete optimization space and prunes it down to a set of configurations that is likely to contain the global maximum. The remaining configurations can then be evaluated to determine the one with the best performance. The technique can reduce the number of configurations to be evaluated by as much as 98% and is successful at finding a near-best configuration. For some applications, we show that this approach is significantly superior to random sampling of the search space.

This study investigates the potential of using transmission power control in wireless packet networks with differing number of hops between source and destination nodes. Here we exploit the benefits of power control in the context of multi-hop wireless ad hoc type networks with a distributed media access control. For our investigations we choose several general ad hoc network topologies and study the effects of power control with respect to energy consumption and network capacity, We show that power control largely improves the network capacity and energy savings in all investigated scenarios, and that utilizing a greater number of intermediate hops between source and destination nodes improves the energy savings though may cause a tradeoff in capacity, depending on the network topology considered.

This paper studies the performance implications of architectural synchronization support for automatically parallelized numerical programs. As the basis for this work, we analyze the needs for synchronization in automatically parallelized numerical programs. The needs are due to task scheduling, iteration scheduling, barriers, and data dependence handling. We present synchronization algorithms for efficient execution of programs with nested parallel loops. Next, we identify how various hardware synchronization support can be used to satisfy these software synchronization needs. The synchronization primitives studied are test&set , fetch&add , and exchange-byte operations. In addition to these, synchronization bus implementation of lock/unlock and fetch&add operations are also considered. Finally, we ran experiments to quantify the impact of various architectural support on the performance of a bus-based shared memory multiprocessor running automatically parallelized numerical programs. We found that supporting an atomic fetch&add primitive in shared memory is as effective as supporting lock/unlock operations with a synchronization bus. Both achieve substantial performance improvement over the cases where atomic test&set and exchange-byte operations are supported in shared memory.