Functionally Arranged Data for Algorithms with Space-Time Wavefront

Perepelkina, Anastasia; Levchenko, V. D.

doi:10.1007/978-3-030-81691-9_10

Cited by 7 publications

(11 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The previous LRnLA LBM code on GPU has shown performance that is superior to other known approaches [62]. Its GPU acceleration is efficient due to the uniformity of the computations.…”

Section: Lrnla Subdivision On a Heterogenous Systemmentioning

confidence: 98%

“…The sequentially accessed lattice cells are along the direction of the cube diagonal. That is why the recently developed FArSh (Functionally Arranged Shadow) data structure [62] is used here.…”

Section: Data Structurementioning

confidence: 99%

“…Then the time coordinate t is added. Two dimensions of the FArSh i 1 , i 2 that span a hexagon are implemented with a Z-curve in the 3 continuous areas of the hexagon [62]. Finally, tiles is a 3D Z-curve array of groups (group is a cube of 8 cells), and FArSh is a 2D Z-order array of lines of cells.…”

Section: Data Structurementioning

confidence: 99%

“…The positioning of each pattern on the dependency graph is discussed in [50]. The compact elementary update is proposed in [50,51,62]. We choose it for the implementation since it provides optimal use of the memory throughput.…”

Section: Lrnla Cell For Lbmmentioning

confidence: 99%

“…Its GPU acceleration is efficient due to the uniformity of the computations. The presence of any inclusions introduces thread diversions, and some tiers of the ConeTorres have to be replaced with conditional statements or masks, and a notable performance drop is expected, so no boundary condition was used in [62]. That is why here it was decided to use GPU for the main part of computation and CPU for processing the boundary.…”

Section: Lrnla Subdivision On a Heterogenous Systemmentioning

confidence: 99%

See 4 more Smart Citations

Heterogeneous LBM Simulation Code with LRnLA Algorithms

null¹,

Perepelkina²

2023

CICP

View full text Add to dashboard Cite

A design of a new heterogeneous code for LBM simulations is proposed. By heterogeneous computing we mean a collaborative computation on CPU and GPU, which is characterized by the following features: the data is distributed between CPU and GPU memory spaces taking advantage of both parallel hierarchies; the capabilities of both SIMT GPU and SIMD GPU parallelization are used for calculations; the algorithms in use efficiently conceal the CPU-GPU data exchange; the subdivision of the computing task is performed with an account for the strong points of both processing units: high performance of GPU, low latency, and advanced memory hierarchy of CPU. This code is a continuation of our work in the development of LRnLA codes for LBM. Previous LRnLA codes had good efficiency both for CPU and GPU computing, and allowed GPU simulation performed on data stored in CPU RAM without performance loss on CPU-GPU data transfer. In the new code, we use methods and instruments that can be flexibly adapted to GPU and CPU instruction sets. We present the theoretical study of the performance of the proposed code and suggest implementation techniques. The bottlenecks are identified. As a result, we conclude that larger problems can be simulated with higher efficiency in the heterogeneous system.

show abstract

“…The previous LRnLA LBM code on GPU has shown performance that is superior to other known approaches [62]. Its GPU acceleration is efficient due to the uniformity of the computations.…”

Section: Lrnla Subdivision On a Heterogenous Systemmentioning

confidence: 98%

Section: Data Structurementioning

confidence: 99%

Section: Data Structurementioning

confidence: 99%