Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
13
0

Year Published

2019
2019
2021
2021

Publication Types

Select...
5

Relationship

0
5

Authors

Journals

citations
Cited by 11 publications
(13 citation statements)
references
References 4 publications
0
13
0
Order By: Relevance
“…This copying is only necessary for the part of the track data needed by the algorithm. The vectorization technology we chose is based on VecCore [3], an abstraction header library on top of a number of vectorization back-ends. We need to emphasize that most of the work required for migrating an algorithm from a pure scalar version to the basketized one is concentrated on refactoring the algorithm to work with vector types.…”
Section: Vectorizing On Track Datamentioning
confidence: 99%
See 2 more Smart Citations
“…This copying is only necessary for the part of the track data needed by the algorithm. The vectorization technology we chose is based on VecCore [3], an abstraction header library on top of a number of vectorization back-ends. We need to emphasize that most of the work required for migrating an algorithm from a pure scalar version to the basketized one is concentrated on refactoring the algorithm to work with vector types.…”
Section: Vectorizing On Track Datamentioning
confidence: 99%
“…The process allows to preserve the scalar version and even in most cases coalesce it with the vector one by templating the implementations on the data type. The GeantV scheduler actually makes use of both scalar and vector versions to optimize its workflow, as described in a previous paper [4].…”
Section: Vectorizing On Track Datamentioning
confidence: 99%
See 1 more Smart Citation
“…On the other hand, the use of intrinsics is complex and significantly decreases the readability and maintainability of the code, leading to platform-specific implementations, to the detriment of code portability. GeantV project implements an innovative approach to vectorization, based on the combined use of VecCore [8] and the concept of templated backends [9,10]. VecCore is a wrapper library that provides an abstraction layer on top of existing SIMD libraries, such as Vc [11] and UME::SIMD [12], that facilitate users to write generic vectorized code.…”
Section: Vectorization Is Good When It Happensmentioning
confidence: 99%
“…Data-level parallelism is achieved in ROOT by exploiting SIMD array operations (vectorization). For this, we rely in the VecCore [2] vectorization library, which we integrated in ROOT's mathematical libraries.…”
Section: Data-level Parallelismmentioning
confidence: 99%