Search NASA⌕ Search

SEARCH · Search NASA

Results for “GEMM”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records
All sourcesNASA NTRSDOE OSTI
NASA NTRS2019

Gene Expression Measurement Module (GEMM) for space application: Design and validation

In order to facilitate studies on the impact of the space environment on biological systems, we have developed a prototype of GEMM (Gene Expression Measurement Module) - an automated, miniaturized, integrated fluidic system for in-situ measurements of gene expression in microbial samples. The GEMM instrument is capable of (1) lysing bacterial cell walls, (2) extracting and purifying RNA released from cells, (3) hybridizing the RNA to probes attached to a microarray and (4) providing electrochemical readout, all in a microfluidics cartridge. To function on small, uncrewed spacecraft, the conventional, laboratory protocols for both sample preparation and hybridization required significant modifications. Biological validation of the instrument was carried out on Synechococcus elongatus, a photosynthetic cyanobacterium known for its metabolic diversity and resilience to adverse conditions. It was demonstrated that GEMM yielded reliable, reproducible gene expression profiles. GEMM is the only high throughput instrument that can be deployed in near future on space platforms other than the ISS to advance biological research in space. It can also prove useful for numerous terrestrial applications in the field.

Kianoosh Peyvan↗
DOE OSTI2024

Understanding Mixed Precision GEMM with MPGemmFI: Insights into Fault Resilience

Emerging deep learning workloads urgently need fast general matrix multiplication (GEMM). Thus, one of the critical features of machine-learning-specific accelerators such as NVIDIA Tensor Cores, AMD Matrix Cores, and Google TPUs is the support of mixed-precision enabled GEMM. For DNN models, lower-precision FP data formats and computation offer acceptable correctness but significant performance, area, and memory footprint improvement. While promising, the mixed-precision computation on error resilience remains unexplored. To this end, we develop a fault injection framework that systematically injects fault into the mixed-precision computation results. We investigate how the faults affect the accuracy of machine learning applications. Based on the characteristics of error resilience, we offer lightweight error detection and correction solutions that significantly improve the overall model accuracy by 75% if the models experience hardware faults. The solutions can be efficiently integrated into the accelerator's pipelines.

Fang, Bo↗
DOE OSTI2021

Matrix Product (GEMM) Performance Data from GPUs

Timing data for mixed precision GEMM matrix product operations on several GPU models, including NVIDIA V100 and A100, AMD MI100 and Intel P580. Also data from machine learning model training on this data using Scikit-learn.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2012

Gene Expression Measurement Module (GEMM) - A Fully Automated, Miniaturized Instrument for Measuring Gene Expression in Space

The capability to measure gene expression on board spacecraft opens the door to a large number of high-value experiments on the influence of the space environment on biological systems. For example, measurements of gene expression will help us to understand adaptation of terrestrial life to conditions beyond the planet of origin, identify deleterious effects of the space environment on a wide range of organisms from microbes to humans, develop effective countermeasures against these effects, and determine the metabolic bases of microbial pathogenicity and drug resistance. These and other applications hold significant potential for discoveries in space biology, biotechnology, and medicine. Supported by funding from the NASA Astrobiology Science and Technology Instrument Development Program, we are developing a fully automated, miniaturized, integrated fluidic system for small spacecraft capable of in-situ measurement of expression of several hundreds of microbial genes from multiple samples. The instrument will be capable of (1) lysing cell walls of bacteria sampled from cultures grown in space, (2) extracting and purifying RNA released from cells, (3) hybridizing the RNA on a microarray and (4) providing readout of the microarray signal, all in a single microfluidics cartridge. The device is suitable for deployment on nanosatellite platforms developed by NASA Ames' Small Spacecraft Division. To meet space and other technical constraints imposed by these platforms, a number of technical innovations are being implemented. The integration and end-to-end technological and biological validation of the instrument are carried out using as a model the photosynthetic bacterium Synechococcus elongatus, known for its remarkable metabolic diversity and resilience to adverse conditions. Each step in the measurement process-lysis, nucleic acid extraction, purification, and hybridization to an array-is assessed through comparison of the results obtained using the instrument with those from standard laboratory protocols. Once developed, the system can be used with minor modifications for multiple experiments on different platforms in space, including extension to higher organisms and microbial monitoring. A proposed version of GEMM that is capable of handling both microbial and tissue samples on the International Space Station will be briefly summarized.

Pohorille, Andrew↗
DOE OSTI2022

Threaded Multi-Core GEMM with MoA and Cache-Blocking: Preprint

A threaded multi-core implementation of the high performance dense linear algebra matrix-matrix multiply GEMM kernel is described. This kernel is widely implemented by vendors in the basic linear algebra subroutine BLAS library. The mathematics of arrays (MoA) paradigm due to Mullin (1988) results in contiguous memory accesses by employing outer-product forms. Our performance studies demonstrate that the MoA implementation of double precision DGEMM combined with optimal cache-blocking strategies results in at least a 25% performance gain on the Intel Xeon Skylake processor over the vendor supplied Intel MKL basic linear algebra libraries. Results are presented for the NREL Eagle supercomputer. The multi-core DGEMM achieves over 100 GigaFlops/sec with eight openMP threads.

cache-blocking↗
DOE OSTI2022

Matrix Multiply Performance of GPUs on Exascale-class HPE/Cray Systems

The computation of dense matrix-matrix products (GEMMs) is central to many modeling and simulation workloads as well as AI/ML deep learning campaigns. In fact, millions of dollars are spent annually on computing GEMMs, and large model training demands are increasing exponentially. Specialized processors such as GPUs are designed to perform well for these operations. However, the performance of GEMMs on GPUs can exhibit complex behaviors depending on many factors, making it challenging to optimize the performance of GEMMs on these processors. In this study we undertake an examination of GEMM performance on several leading GPU models taken from product lines of GPUs to be deployed in forthcoming exascale computing systems. We show results to illustrate the many factors that can affect performance of GEMMs on GPUs. We then present data collected from a large number of test runs for an example GEMM operation to show the dependence behaviors of GEMM rate on matrix dimensions. Finally, we show results from machine learning-based performance models using novel feature engineering methods to fit the measured performance, providing a potential basis for GEMM performance tuning and autotuning methods for GPUs. Recommendations are also given for how to achieve high GEMM performance on modern GPUs.

Melesse Vergara, Veronica↗
DOE OSTI2024

Kernel fusion in atomistic spin dynamics simulations on Nvidia GPUs using tensor core

In atomistic spin dynamics simulations, the time cost of constructing the space- and time-displaced pair correlation function in real space increases quadratically as the number of spins N, leading to significant computational effort. The GEMM subroutine can be adopted to accelerate the calculation of the dynamical spin-spin correlation function, but the computational cost of simulating large spin systems (>40000 spins) on CPUs remains expensive. In this work, we perform the simulation on the graphics processing unit (GPU), a hardware solution widely used as an accelerator for scientific computing and deep learning. Here we show that GPUs can accelerate the simulation up to 25-fold compared to multi-core CPUs when using the GEMM subroutine on both. To hide memory latency, we fuse the element-wise operation into the GEMM kernel using CUTLASS that can improve the performance by 26% ~ 33% compared to implementation based on cuBLAS. Furthermore, we perform the on-the-fly calculation in the epilogue of the GEMM subroutine to avoid saving intermediate results on global memory, which makes the large-scale atomistic spin dynamics simulation feasible and affordable.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication.

There is a growing interest in custom spatial accelerators for machine learning applications. These accelerators employ a spatial array of processing elements (PEs) interacting via custom buffer hierarchies and networks-on-chip. The efficiency of these accelerators comes from employing optimized dataflow (i.e., spatial/temporal partitioning of data across the PEs and fine-grained scheduling) strategies to optimize data reuse. The focus of this work is to evaluate these accelerator architectures using a tiled general matrix-matrix multiplication (GEMM) kernel. To do so, we develop a framework that finds optimized mappings (dataflow and tile sizes) for a tiled GEMM for a given spatial accelerator and workload combination, leveraging an analytical cost model for runtime and energy. Finally, our evaluations over five spatial accelerators demonstrate that the tiled GEMM mappings systematically generated by our framework achieve high performance on various GEMM workloads and accelerators.

42 ENGINEERING↗
DOE OSTI2021

Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication

There is a growing interest in custom spatial accelerators for machine learning applications. These accelerators employ a spatial array of processing elements (PEs) interacting via custom buffer hierarchies and networks-on-chip. The efficiency of these accelerators comes from employing optimized dataflow (i.e., spatial/temporal partitioning of data across the PEs and fine-grained scheduling) strategies to optimize data reuse. The focus of this work is to evaluate these accelerator architectures using a tiled general matrix-matrix multiplication (GEMM) kernel. To do so, we develop a framework that finds optimized mappings (dataflow and tile sizes) for a tiled GEMM for a given spatial accelerator and workload combination, leveraging an analytical cost model for runtime and energy. Our evaluations over five spatial accelerators demonstrate that the tiled GEMM mappings systematically generated by our framework achieve high performance on various GEMM workloads and accelerators.

43 PARTICLE ACCELERATORS↗
DOE OSTI2023

A high-performance implementation of atomistic spin dynamics simulations on x86 CPUs

Atomistic spin dynamics simulations provide valuable information about the energy spectrum of magnetic materials in different phases, allowing one to identify instabilities and the nature of their excitations. However, the time cost of evaluating the dynamical correlation function S(q,t) increases quadratically as the number of spins N, leading to significant computational effort, making the simulation of large spin systems very challenging. In this work, here, we propose to use a highly optimized general matrix multiply (GEMM) subroutine to calculate the dynamical spin-spin correlation function that can achieve near-optimal hardware utilization. Furthermore, we fuse the element-wise operations in the calculation of S(q,t) into the in-house GEMM kernel, which results in further performance improvements of 44\% - 71\% on several relatively large lattice sizes when compared to the implementation that uses the GEMM subroutine in OpenBLAS, which is the state-of-the-art open source library for Basic Linear Algebra Subroutine (BLAS).

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Genetically engineered mouse models of esophageal cancer

Esophageal cancer is the most common cause of cancer-related death worldwide with a diverse geographical distribution, poor prognosis, and diagnosis in advanced stages of the disease. Identification of the mechanisms involved in esophageal cancer development is evaluative to improve outcomes for patients. Genetically engineered mouse models (GEMMs) of cancer provide the physiologic, molecular, and histologic features of the human tumors to determine the pathogenesis and treatments for cancer, hence exhibiting a source of tremendous potential for oncology research. The advancement of cancer modeling in mice has improved to the extent that researchers can observe and manipulate the disease process in a specific manner. Despite the significant differences between mice and humans, mice can be great models for human oncology researches due to similarities between them at the molecular and physiological levels. Due to most of the existing esophageal cancer GEMMs do not propose an ideal system for pathogenesis of the disease, genetic risks, and microenvironment exposure, so identification of challenges in GEM modeling and well-developed technologies are required to obtain the most value for patients. In this review, we describe the biology of human and mouse, followed by the exciting esophageal cancer mouse models with a discussion of applicability and challenges of these models for generating new GEMMs in future studies.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2015

Lineage-related cytotoxicity and clonogenic profile of 1,4-benzoquinone-exposed hematopoietic stem and progenitor cells

Hematopoietic stem cells (HSCs) and hematopoietic progenitor cells (HPCs) are sensitive targets for benzene-induced hematotoxicity and leukemogenesis. The impact of benzene exposure on the complex microenvironment of HSCs and HPCs remains elusive. This study aims to investigate the mechanism linking benzene exposure to targeting HSCs and HPCs using phenotypic and clonogenic analyses. Mouse bone marrow (BM) cells were exposed ex vivo to the benzene metabolite, 1,4-benzoquinone (1,4-BQ), for 24 h. Expression of cellular surface antigens for HSC (Sca-1), myeloid (Gr-1, CD11b), and lymphoid (CD45, CD3e) populations were confirmed by flow cytometry. The clonogenicity of cells was studied using the colony-forming unit (CFU) assay for multilineage (CFU-GM and CFU-GEMM) and single-lineage (CFU-E, BFU-E, CFU-G, and CFU-M) progenitors. 1,4-BQ demonstrated concentration-dependent cytotoxicity in mouse BM cells. The percentage of apoptotic cells increased (p < 0.05) following 1,4-BQ exposure. Exposure to 1,4-BQ showed no significant effect on CD3e{sup +} cells but reduced the total counts of Sca-1{sup +}, CD11b{sup +}, Gr-1{sup +}, and CD45{sup +} cells at 7 and 12 μM (p < 0.05). Furthermore, the CFU assay showed reduced (p < 0.05) clonogenicity in 1,4-BQ-treated cells. 1,4-BQ induced CFU-dependent cytotoxicity by significantly inhibiting colony growth for CFU-E, BFU-E, CFU-G, and CFU-M starting at a low concentration of exposure (5 μM); whereas for the CFU-GM and CFU-GEMM, the inhibition of colony growth was remarkable only at 7 and 12 μM of 1,4-BQ, respectively. Taken together, 1,4-BQ caused lineage-related cytotoxicity in mouse HPCs, demonstrating greater toxicity in single-lineage progenitors than in those of multi-lineage. - Highlights: • We examine 1,4-BQ toxicity targeting mouse hematopoietic cell lineages. • 1,4-BQ induces concentration-dependent cytotoxicity in bone marrow (BM) cells. • 1,4-BQ shows lineage-related toxicity on hematopoietic stem and progenitors. • 1,4-BQ toxicity is greater in single- than multilineage committed progenitors.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2018

A Flexible-blocking Based Approach for Performance Tuning of Matrix Multiplication Routines for Large Matrices with Edge Cases

Efficient and scalable matrix operations are being highly demanding in the recent era of Machine Learning, Deep Learning, and Big Data Analytics. The two commonly used matrix-matrix operations in the Basic Linear Algebra Subprograms (BLAS) specification are General Matrix-Matrix multiplication (GEMM) and Symmetric Rank-k update (SYRK). The SYRK routine is a specialization of the GEMM routine, where half of the multiplications are skipped as the resultant matrix is known to be symmetric. Fortunately, several linear algebra libraries implement these BLAS routines quite efficiently. The libraries usually partition the input matrices into blocks and place them in processor caches, thus improving performance by leveraging the caches. However, the contemporary libraries are highly optimized for squarish matrices, but the performance degrades significantly for the matrices with edge case (strictly thin or strictly fat shapes) in the multicore machine. The primary reason is that the current state-of-the-art libraries make fixed block shapes based on a processor architecture, and do not consider the shape of the input matrices. In this paper, we propose a new blocking approach, we name it Flexible-blocking, to mitigate the scalability issues. In contrast to the contemporary libraries, our approach formulates the blocks of the input matrices based on the shapes of the matrices as well as the number of threads used in the implementation. Our proposed technique shows noticeable performance improvement on multicore shared-memory machines for the edge case matrices.

Hossain, Md Mosharaf↗
DOE OSTI2022

Towards Precision-Aware Fault Tolerance Approaches for Mixed-Precision Applications

Graphics Processing Units (GPUs), the dominantly adopted accelerators in HPC systems, are susceptible to transient hardware fault. New generation of GPUs feature mixed-precision architectures such as NVIDIA Tensor Cores to accelerate matrix multiplications. While being widely adapted, how would they behave under transient hardware faults remain unclear. In this study, we conduct a large-scale fault injection experiments on GEMM kernels implemented with different floating-point data types on the V100 and A100 Tensor Cores, and show distinct error resilience characteristics for the GEMMS with different formats. In the future, we plan to explore this space by building precision-aware floating-point fault tolerance techniques for applications such as DNNs that exercise low-precision computations.

Fang, Bo↗
DOE OSTI2024

Large language model evaluation for high–performance computing software development

We apply AI-assisted large language model (LLM) capabilities of GPT-3 targeting high-performance computing (HPC) kernels for (i) code generation, and (ii) auto-parallelization of serial code in C ++, Fortran, Python and Julia. Our scope includes the following fundamental numerical kernels: AXPY, GEMV, GEMM, SpMV, Jacobi Stencil, and CG, and language/programming models: (1) C++ (e.g., OpenMP [including offload], OpenACC, Kokkos, SyCL, CUDA, and HIP), (2) Fortran (e.g., OpenMP [including offload] and OpenACC), (3) Python (e.g., numpy, Numba, cuPy, and pyCUDA), and (4) Julia (e.g., Threads, CUDA.jl, AMDGPU.jl, and KernelAbstractions.jl). Kernel implementations are generated using GitHub Copilot capabilities powered by the GPT-based OpenAI Codex available in Visual Studio Code given simple + + prompt variants. To quantify and compare the generated results, we propose a proficiency metric around the initial 10 suggestions given for each prompt. For auto-parallelization, we use ChatGPT interactively giving simple prompts as in a dialogue with another human including simple “prompt engineering” follow ups. Results suggest that correct outputs for C++ correlate with the adoption and maturity of programming models. For example, OpenMP and CUDA score really high, whereas HIP is still lacking. We found that prompts from either a targeted language such as Fortran or the more general-purpose Python can benefit from adding language keywords, while Julia prompts perform acceptably well for its Threads and CUDA.jl programming models. Finally, we expect to provide an initial quantifiable point of reference for code generation in each programming model using a state-of-the-art LLM. Overall, understanding the convergence of LLMs, AI, and HPC is crucial due to its rapidly evolving nature and how it is redefining human-computer interactions.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Potential reductions in fine particulate matter and premature mortality following implementation of air pollution controls on coal-fired power plants in India

Coal-fired power plants (CFPPs) account for > 70% of electricity generation in India, but < 5% of facilities have installed technologies for sulfur dioxide (SO 2 ) and nitrogen oxide (NO X ) removal. Emissions of these pollutants lead to the formation of fine particulate matter (PM 2.5 ) and an increased risk of premature mortality for exposed populations. Here, we use a nested version of the GEOS-Chem global chemical transport model (0.5° × 0.625° resolution) for India to estimate reductions in PM 2.5 concentrations that could have been achieved by implementing existing emission control technologies like flue-gas desulfurization (FGD) and/or selective catalytic reduction (SCR). We quantify the associated burden of disease using the integrated exposure response (IER) and global exposure mortality model (GEMM) functions and compare the costs of premature mortality to those for FGD installation. Model simulations for 2010 suggest installation of FGD would have reduced mean annual PM 2.5 concentrations across India by 8%, compared to 3% with SCR installation, and 11% with both FGD and SCR. A 7–28% reduction in PM 2.5 was simulated for local communities closest to CFPPs (same model grid cell), leading to up to 17% reduction in annual premature mortality. Overall, more than 0.21–0.48 million premature deaths would have been avoided over a 10-year period if FGD had been implemented on all CFPPs, compared to 0.09–0.21 million with SCR and 0.22–0.72 million with both FGD and SCR. Benefits associated with such actions are approximately $\$18.1$–$\$604$ billion USD per year, which is equivalent to ~ 0.44 to 10% of India’s GDP. These results suggest that monetary benefits from avoided premature mortality far outweigh the capital and operational costs of FGD and/or SCR installation of $\$19.5$ billion and/or $\$32.8$ billion per year, respectively. This information is essential because the high costs of installation and operation are often given as reasons for delaying installation and commissioning. Finally, we conclude that policy actions to control air pollution from CFPPs are economically justifiable.

63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTAN↗