2019
DOI: 10.1007/978-3-030-27562-4_29
|View full text |Cite
|
Sign up to set email alerts
|

Low Precision Processing for High Order Stencil Computations

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
8
0

Year Published

2021
2021
2023
2023

Publication Types

Select...
4
2
1

Relationship

5
2

Authors

Journals

citations
Cited by 9 publications
(8 citation statements)
references
References 14 publications
0
8
0
Order By: Relevance
“…Accuracy impact with FP16 precision -Although traditional stencil applications use higher precision such as FP64 and FP32, there are increasing research works demonstrate the success of using lower precision such as FP16 in stencil application [9,12,35,36].…”
Section: Discussionmentioning
confidence: 99%
“…Accuracy impact with FP16 precision -Although traditional stencil applications use higher precision such as FP64 and FP32, there are increasing research works demonstrate the success of using lower precision such as FP16 in stencil application [9,12,35,36].…”
Section: Discussionmentioning
confidence: 99%
“…Unlike stencils found in the literature [51,55,56,135,154,168,169], real-world compound stencils consist of a collection of stencils that perform a sequence of element-wise computations with complex interdependencies. Such compound kernels have complex memory access patterns and low arithmetic intensity because they have limited operations per loaded value.…”
Section: Related Workmentioning
confidence: 99%
“…Szustak et al accelerate the MPDATA advection scheme on multi-core CPU [159] and computational luid dynamics kernels on FPGA [133]. Singh et al [154] explore the applicability of diferent number formats and exhaustively search for the appropriate bit-width for memory-bound stencil kernels to improve performance and energy eiciency with minimal loss in the accuracy. Bianco et al [41] optimize the COSMO weather prediction model for GPUs.…”
Section: Related Workmentioning
confidence: 99%
“…Szustak et al accelerate the MPDATA advection scheme on multi-core CPU [134] and computational fluid dynamics kernels on FPGA [116]. Singh et al [130] explore the applicability of different number formats and exhaustively search for the appropriate bit-width for memory-bound stencil kernels to improve performance and energy-efficiency with minimal loss in the accuracy. Bianco et al [29] optimize the COSMO weather prediction model for GPUs while Thaler et al [136] port COSMO to a many-core system.…”
Section: Related Workmentioning
confidence: 99%