Performance Tuning and Analysis for Stencil-Based Applications on POWER8 Processor

Xu, Jia; Fu, Haohuan; Shi, Wen; Gan, Lin; Li, Yuxuan; Luk, Wayne; Yang, Guangwen

doi:10.1145/3264422

Cited by 11 publications

(12 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This poses a fundamental challenge to acceleration. CPU implementations of these kernels [175] sufer from limited data locality and ineicient memory usage, as our rooline analysis in Figure 1 exposes. In Figure 3 we implement a copy stencil from the COSMO weather model to evaluate the performance potential of our HBM-based FPGA platform for the weather prediction application.…”

Section: Flux Resultsmentioning

confidence: 99%

“…Unlike the conventional stencil kernels, vertical advection has dependencies in the vertical direction, which leads to limited available parallelism and irregular memory access patterns. For example, when the input grid is stored by row, accessing data elements in the depth dimension typically results in many cache misses [175].…”

Section: Flux Resultsmentioning

confidence: 99%

“…We co-design a hardware-software framework and provide an optimized API to interface eiciently with the rest of the COSMO model, which runs on the CPU. Our FPGA-based solution for hdiff and vadvc leads to performance improvements of 5.3× and 12.7× and energy reductions of 12× and 35×, respectively, with respect to optimized CPU implementations [175].…”

Section: At T Ainable Perform Ance Is Const Rained By M Em Ory Bandwi...mentioning

confidence: 97%

“…Figure 1 shows the rooline plot [173] for an IBM 16-core POWER9 CPU (IC922). 1 After optimizing the vadvc and hdiff kernels for the POWER architecture 2 by following the approach in [175], they achieve 29.1 GFLOP/s and 58.5 GFLOP/s, respectively, for 64 threads. Our rooline analysis indicates that these kernels are constrained by the host DRAM bandwidth.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Accelerating Weather Prediction Using Near-Memory Reconfigurable Fabric

Singh

Diamantopoulos

Gómez-Luna

et al. 2022

ACM Trans. Reconfigurable Technol. Syst.

View full text Add to dashboard Cite

Ongoing climate change calls for fast and accurate weather and climate modeling. However, when solving large-scale weather prediction simulations, state-of-the-art CPU and GPU implementations suffer from limited performance and high energy consumption. These implementations are dominated by complex irregular memory access patterns and low arithmetic intensity that pose fundamental challenges to acceleration. To overcome these challenges, we propose and evaluate the use of near-memory acceleration using a reconfigurable fabric with high-bandwidth memory (HBM). We focus on compound stencils that are fundamental kernels in weather prediction models. By using high-level synthesis techniques, we develop NERO, an FPGA+HBM-based accelerator connected through OCAPI (Open Coherent Accelerator Processor Interface) to an IBM POWER9 host system. Our experimental results show that NERO outperforms a 16-core POWER9 system by 5.3 × and 12.7 × when running two different compound stencil kernels. NERO reduces the energy consumption by 12 × and 35 × for the same two kernels over the POWER9 system with an energy efficiency of 1.61 GFLOPS/Watt and 21.01 GFLOPS/Watt. We conclude that employing near-memory acceleration solutions for weather prediction modeling is promising as a means to achieve both high performance and high energy efficiency.

show abstract

Section: Flux Resultsmentioning

confidence: 99%

Section: Flux Resultsmentioning

confidence: 99%

Section: At T Ainable Perform Ance Is Const Rained By M Em Ory Bandwi...mentioning

confidence: 97%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Accelerating Weather Prediction Using Near-Memory Reconfigurable Fabric

Singh

Diamantopoulos

Gómez-Luna

et al. 2022

ACM Trans. Reconfigurable Technol. Syst.

View full text Add to dashboard Cite

show abstract

“…Figure 1 shows the roo ine plot [104] for an IBM 16core POWER9 CPU (IC922). 1 After optimizing the vadvc and hdiff kernels for the POWER architecture by following the approach in [105], they achieve 29.1 GFLOP/s and 58.5 GFLOP/s, respectively, for 64 threads. Our roo ine analysis indicates that these kernels are constrained by the host DRAM bandwidth.…”

Section: Introductionmentioning

confidence: 99%

NERO: A Near High-Bandwidth Memory Stencil Accelerator for Weather Prediction Modeling

Singh¹,

Diamantopoulos²,

Hagleitner³

et al. 2020

Preprint

View full text Add to dashboard Cite

Ongoing climate change calls for fast and accurate weather and climate modeling. However, when solving large-scale weather prediction simulations, state-of-the-art CPU and GPU implementations su er from limited performance and high energy consumption. These implementations are dominated by complex irregular memory access patterns and low arithmetic intensity that pose fundamental challenges to acceleration. To overcome these challenges, we propose and evaluate the use of near-memory acceleration using a recon gurable fabric with high-bandwidth memory (HBM). We focus on compound stencils that are fundamental kernels in weather prediction models. By using high-level synthesis techniques, we develop NERO, an FPGA+HBM-based accelerator connected through IBM CAPI2 (Coherent Accelerator Processor Interface) to an IBM POWER9 host system. Our experimental results show that NERO outperforms a 16-core POWER9 system by 4.2× and 8.3× when running two di erent compound stencil kernels. NERO reduces the energy consumption by 22× and 29× for the same two kernels over the POWER9 system with an energy e ciency of 1.5 GFLOPS/Watt and 17.3 GFLOPS/Watt. We conclude that employing near-memory acceleration solutions for weather prediction modeling is promising as a means to achieve both high performance and high energy e ciency.

show abstract

Highly scalable parallel genetic algorithm on Sunway many-core processors

Xiao

et al. 2021

Future Generation Computer Systems

View full text Add to dashboard Cite

Performance Tuning and Analysis for Stencil-Based Applications on POWER8 Processor

Cited by 11 publications

References 32 publications

Accelerating Weather Prediction Using Near-Memory Reconfigurable Fabric

Accelerating Weather Prediction Using Near-Memory Reconfigurable Fabric

NERO: A Near High-Bandwidth Memory Stencil Accelerator for Weather Prediction Modeling

Highly scalable parallel genetic algorithm on Sunway many-core processors

Contact Info

Product

Resources

About