Ookami: Deployment and Initial Experiences

Burford, Andrew; Calder, A. C.; Carlson, David E.; Chapman, Barbara; Coskun, Firat; Curtis, Tony; Feldman, Catherine; Harrison, Robert J.; Kang, Yan; Michalowicz, Benjamin; Raut, Eric; Siegmann, Eva; Wood, Daniel; DeLeon, Robert L.; Jones, Mathew; Simakov, Nikolay A.; White, John; Oryspayev, Dossay

doi:10.1145/3437359.3465578

Cited by 13 publications

(7 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Early study of the performance of FLASH on Ookami may be found in [8], and similar experiences are reported in [4,7]. The unoptimized performance on A64FX did not compare well to that found on traditional X86 architectures [5].…”

Section: Thermonuclear Supernovae With Flashmentioning

confidence: 73%

“…The nodes have 32 GB of high-bandwidth memory, where 5 GB are reserved for the OS, leaving 27 GB for the user. These processors use the ARMv8.2-A Scalable Vector Extension (SVE) SIMD instruction set with a 512 bit vector implementation, allowing for vector lengths anywhere from 128-2048 bits and enabling vector length agnostic programming [5].…”

Section: Introduction 1ookami and A64fxmentioning

confidence: 99%

See 1 more Smart Citation

A Further Study of Linux Kernel Hugepages on A64FX with FLASH, an Astrophysical Simulation Code

Feldman,

Chheda,

Calder

et al. 2023

Practice and Experience in Advanced Research Computing

View full text Add to dashboard Cite

We present an expanded study of the performance of FLASH when using Linux Kernel Hugepages on Ookami, an HPE Apollo 80 A64FX platform. FLASH is a multi-scale, multi-physics simulation code written principally in modern Fortran and makes use of the PARAMESH library to manage a block-structured adaptive mesh. Our initial study used only the Fujitsu compiler to utilize standard hugepages (hp), but further investigation allowed us to utilize hp for multiple compilers by linking to the Fujitsu library libmpg and transparent hugepages (thp) by enabling it at the node level. By comparing the results of hardware counters and in-code timers, we found that hp and thp do not significantly impact the runtime performance of FLASH. Interestingly, there is a significant reduction in the TLB misses, differences in cache and memory access counters, and strange behavior is observed when using thp. CCS CONCEPTS• Computer systems organization → Parallel architectures; • Computing methodologies → Modeling and simulation; Massively parallel and high-performance simulations; • Applied computing → Physics; Astronomy.

show abstract

Section: Thermonuclear Supernovae With Flashmentioning

confidence: 73%

Section: Introduction 1ookami and A64fxmentioning

confidence: 99%

A Further Study of Linux Kernel Hugepages on A64FX with FLASH, an Astrophysical Simulation Code

Feldman,

Chheda,

Calder

et al. 2023

Practice and Experience in Advanced Research Computing

View full text Add to dashboard Cite

show abstract

“…Mutations were introduced into wild-type (WT) models with MOE and PyMOL (PyMOL molecular graphics system). All MD simulations were conducted using the 2021 version of Gromacs software packages with the GROMOS96 53a6 force field. The protonation states of the amino acid residues under physiological conditions were determined using the H++ server .…”

Section: Methodsmentioning

confidence: 99%

Computationally Guided Enzymatic Studies on Schizochytrium-Sourced Malonyl-CoA:ACP Transacylase

Chi

Cao

et al. 2022

J. Agric. Food Chem.

View full text Add to dashboard Cite

The malonyl-CoA:ACP transacylase (MAT) domain is responsible for the selection and incorporation of malonyl building blocks in the biosynthesis of polyunsaturated fatty acids (PUFAs) in eukaryotic microalgae (Schizochytrium) and marine bacteria (Moritella marina, Photobacterium profundum, and Shewanella). Elucidation of the structural basis underlying the substrate specificity and catalytic mechanism of the MAT will help to improve the yield and quality of PUFAs. Here, a methodology guided by molecular dynamics simulations was carried out to identify and mutate specificity-conferring residues within the MAT domain of Schizochytrium. Combining mutagenesis, cell-free protein synthesis, and in vitro biochemical assay, we dissected nearby interactions and molecular mechanisms relevant for binding and catalysis and found that the reorientation of the Ser154 Cβ–Oγ bond establishes distinctive proton-transfer chains (His153-Ser154 and Asn235-His153-Ser154) for catalysis. Gln66 can be replaced by tyrosine to shorten the distance between His153 (Nε2) and Ser154 (Oγ), which facilitates a faster proton-transfer rate, allowing better use of acyl substrates than the wild type. Furthermore, we screened a mutant that displayed an 18.4% increase in PUFA accumulation. These findings provide important insights into the study of MAT through protein engineering and will benefit dissecting the molecular mechanisms of other PUFA-related catalytic domains.

show abstract

“…As it was one of the marquee applications for the Ookami project, FLASH was ported as soon as Ookami was up and available. Sorting out the compilers and their options, versions of MPI, and requisite packages like the HDF5 library took some effort, but our initial experience with FLASH and other applications was overwhelmingly positive [3]. FLASH ran "right out of the box" with several compilers and MPI implementations, scaling reasonably well with no tuning.…”

Section: V2dmentioning

confidence: 99%

“…The project is run by Stony Brook University (SBU) in cooperation with the University at Buffalo [2] and provides open access along with training and resources to effectively use such hardware. Users have been able to port, analyze, and optimize the performance of many applications [3]. Below, we describe the hardware, our user base and education program, and provide highlights from previous performance studies of two astrophysical simulation codes, the astrophysical radiation hydrodynamics code V2D [4] and multi-application package FLASH [5,6], here applied to thermonuclear supernovae.…”

Section: Introductionmentioning

confidence: 99%

Ookami: An A64FX Computing Resource

Calder,

Siegmann,

Feldman

et al. 2024

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

We present a look at Ookami, a project providing community access to a testbed supercomputer with the ARM-based A64FX processors developed by a collaboration between RIKEN and Fujitsu and deployed in the Japanese supercomputer Fugaku. We provide an overview of the project and details of the hardware, and describe the user base and education/training program. We present highlights from previous performance studies of two astrophysical simulation codes and present a strong scaling study of a full 3D supernova simulation as an example of the the machine’s capability.

show abstract

Ookami: Deployment and Initial Experiences

Cited by 13 publications

References 12 publications

A Further Study of Linux Kernel Hugepages on A64FX with FLASH, an Astrophysical Simulation Code

A Further Study of Linux Kernel Hugepages on A64FX with FLASH, an Astrophysical Simulation Code

Computationally Guided Enzymatic Studies on Schizochytrium-Sourced Malonyl-CoA:ACP Transacylase

Ookami: An A64FX Computing Resource

Contact Info

Product

Resources

About