Fast seismic modeling and reverse time migration on a graphics processing unit cluster

Abdelkhalek, Rached; Calandra, Henri; Coulaud, Olivier; Latu, Guillaume; Roman, Jean

doi:10.1002/cpe.1875

Cited by 19 publications

(23 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[7][8][9]11,19,21 This happens because, although using row-major order enables some coalesced memory accesses for block warps, all blocks in the grid will compete for space in the cache. However, as each thread within a block will need its own data plus its six neighbors, this may lead to data access redundancy.…”

Section: Figurementioning

confidence: 99%

“…This means that the chance of a cache miss increases as the grid size grows. [7][8][9]11,21 In CUDA, threads of the same block can communicate with each other through shared memory, which is a low-latency memory, but limited in total storage. According to Equation (5), at the time thread k asks for data in position (x + 1, y, z), thread k + 1 had requested this same data already, so it should be in cache memory.…”

Section: Figurementioning

confidence: 99%

“…Initial conditions are described by Equations (7) and (8). Initial conditions are described by Equations (7) and (8).…”

Section: Coherence Between Simulation Outputsmentioning

confidence: 99%

“…[7][8][9] With this in mind, we propose in this work a novel and smart data structure suitable for problems that access data from nearest neighbors, such as the discretization of the Laplacian operator, and we emphasize the GPU memory architecture as well as the CPU features. Discretization of the Laplacian requires some data memory accesses in locations that are physically close in space.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Accelerating simulations of cardiac electrical dynamics through a multi‐GPU platform and an optimized data structure

Vasconcellos

Clua

Fenton

et al. 2019

Concurrency and Computation

View full text Add to dashboard Cite

Simulations of cardiac electrophysiological models in tissue, particularly in 3D require the solutions of billions of differential equations even for just a couple of milliseconds, thus highly demanding in computational resources. In fact, even studies in small domains with very complex models may take several hours to reproduce seconds of electrical cardiac behavior.Today's Graphics Processor Units (GPUs) are becoming a way to accelerate such simulations, and give the added possibilities to run them locally without the need for supercomputers.Nevertheless, when using GPUs, bottlenecks related to global memory access caused by the spatial discretization of the large tissue domains being simulated, become a big challenge. For simulations in a single GPU, we propose a strategy to accelerate the computation of the diffusion term through a data-structure and memory access pattern designed to maximize coalescent memory transactions and minimize branch divergence, achieving results approximately 1.4 times faster than a standard GPU method. We also combine this data structure with a designed communication strategy to take advantage in the case of simulations in multi-GPU platforms.We demonstrate that, in the multi-GPU approach performs, simulations in 3D tissue can be just 4× slower than real time. KEYWORDScardiac electrophysiology models, GPU Computing, memory access optimization, parallel cardiac dynamics simulations INTRODUCTIONThe large increase of computational power over the last years shifted the bottleneck of different algorithms to the memory bandwidth and memory management. 1 One typical solution employed by hardware assemblers to minimize this issue is hardware hierarchical memory and memory locality optimization.Computational systems organize hierarchical memory system into levels. In the on-chip level, the registers are the fastest memory, with a high cost per byte and low capacity. Next, there are different cache levels according to the hardware architecture, typically called L1, L2, and so on.The main memory is the next level; here, the cost per byte is less than cache or registers, but latency is high. The last level is the secondary memory that has the highest latency with the lowest cost per byte. Overall, the cost per byte of each level determines the capacity and latency, which directly impact in performance.As each level of the hierarchical memory system has a different storage capacity and data is usually kept at the lowest memory level, computational systems must choose for each level which data will be prioritized to stay in memory and which will be removed when that memory level fills up. To do so, the computer memory system employs two fundamental principles, ie, temporal and spatial locality. 2 In general, these strategies aim to keep the most recently used data in the same memory level, since having to access higher memory levels drastically increases the time of the search.Based on the memory hierarchical principles, some researchers have tried to minimize memory system bottlenecks through s...

show abstract

Section: Figurementioning

confidence: 99%

Section: Figurementioning

confidence: 99%

“…Initial conditions are described by Equations (7) and (8). Initial conditions are described by Equations (7) and (8).…”

Section: Coherence Between Simulation Outputsmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Accelerating simulations of cardiac electrical dynamics through a multi‐GPU platform and an optimized data structure

Vasconcellos

Clua

Fenton

et al. 2019

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…Using graphical processors to speed up computations of 3D finite difference algorithms is a problem that has recently been well addressed in many articles [14][15][16][17][18][19]. Methods described in the previous sections fall within complex 3D stencil computations.…”

Section: Related Workmentioning

confidence: 99%

Optimizing the computation of a parallel 3D finite difference algorithm for graphics processing units

Porter-Sobieraj

Cygert²,

Kikoła

et al. 2014

Concurrency and Computation

View full text Add to dashboard Cite

This paper explores the possibilities of using a graphics processing unit for complex 3D finite difference computation via MUSTA-FORCE and WENO algorithms. We propose a novel algorithm based on the new properties of CUDA surface memory optimized for 2D spatial locality and compare it with 3D stencil computations carried out via shared memory, which is currently considered to be the best approach. A case study was performed for the extensive generation of a time series of 3D grids of arbitrary size used in the computation of collisions between heavy nuclei in terms of relativistic hydrodynamics. It proved that implementation based on surface memory is as much as 23% faster than an equivalent implementation using shared memory. 4 4 1592 J. PORTER-SOBIERAJ ET AL.heavy ion collisions create a highly dynamic system, hydrodynamic models can be used to describe an intermediate stage in the collisions, where the assumption is valid [1].Hydrodynamic simulations of heavy ion collisions include three major components:1. Initial conditions for the simulations; 2. Hydrodynamic evolution of the system; 3. Freeze-out, where a hydrodynamic picture is no longer valid and particles have to be treated as individual entities.At the beginning, nuclear matter is in the pre-equilibrium stage, and a hydro approach can be only used when the system reaches the local thermal equilibrium. For the initial conditions, another model has to be used. The Glauber model [2] is often used for this purpose, although there are other choices available. Freeze-out is another stage where a hydro-like description is not valid because the distance between particles no longer allows for treating them as a continuous medium. For this stage, a transport theory is more appropriate. Freeze-out usually occurs as a sudden freeze-out, where a particle decouples from a bulk when the energy density is below a certain threshold. Then, after the hydro stage, further propagation of particles is simulated using a transport model, UrQMD [3] for instance.To exploit the full potential of realistic hydrodynamics, fully (3 + 1)-dimensional simulations (the three spatial dimensions + time) are necessary to study the system's evolution without any assumptions regarding its symmetries. It is a compelling idea; however, such simulations are extremely expensive in terms of computing power. Initial conditions for the hydro models are also important. The fluctuations in the pre-equilibrium stage phase might have an impact on the dynamics of the nuclear medium; therefore, event-by-event studies (where an event is a single collision between heavy ions) with fluctuating initial conditions and with a large amount of statistics are of particular interest and require a fast and efficient computer code. Another important topic, from the physics perspective, is studies of jets. Jets are narrow beams of particles with a high momentum, and their propagation through the hot nuclear medium can provide information about the fundamental properties of the Quark-Gluon Plasma (transport coefficien...

show abstract

Special Issue on Advances in High Performance Computing and Simulation

Smari

Fiore

Nygaard

2012

Concurrency and Computation

View full text Add to dashboard Cite

Fast seismic modeling and reverse time migration on a graphics processing unit cluster

Cited by 19 publications

References 15 publications

Accelerating simulations of cardiac electrical dynamics through a multi‐GPU platform and an optimized data structure

Accelerating simulations of cardiac electrical dynamics through a multi‐GPU platform and an optimized data structure

Optimizing the computation of a parallel 3D finite difference algorithm for graphics processing units

Special Issue on Advances in High Performance Computing and Simulation

Contact Info

Product

Resources

About