Vector LLVA: A virtual vector instruction set for media processing

Robert L. Bocchino, Vikram Sadanand Adve

Research output: Contribution to conferencePaper

Abstract

We present Vector LLVA, a virtual instruction set architecture (V-ISA) that exposes extensive static information about vector parallelism while avoiding the use of hardware-specific parameters. We provide both arbitrary-length vectors (for targets that allow vectors of arbitrary length, or where the target length is not known) and fixed-length vectors (for targets that have a fixed vector length, such as subword SIMD extensions), together with a rich set of operations on both vector types. We have implemented translators that compile (1) Vector LLVA written with arbitrary-length vectors to the Motorola RSVP architecture and (2) Vector LLVA written with fixed-length vectors to both AltiVec and Intel SSE2. Our translatorgenerated code achieves speedups competitive with handwritten native code versions of several benchmarks on all three architectures. These experiments show that our V-ISA design captures vector parallelism for two quite different classes of architectures and provides virtual object code portability within the class of subword SIMD architectures.

Original languageEnglish (US)
Pages46-56
Number of pages11
StatePublished - Jul 20 2006
EventVEE 2006 - 2nd International Conference on Virtual Execution Environments - Ottawa, ON, Canada
Duration: Jun 14 2006Jun 16 2006

Other

OtherVEE 2006 - 2nd International Conference on Virtual Execution Environments
CountryCanada
CityOttawa, ON
Period6/14/066/16/06

Fingerprint

Processing
Hardware
Experiments

Keywords

  • Multimedia
  • SIMD
  • Vector
  • Virtual Instruction Sets

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Bocchino, R. L., & Adve, V. S. (2006). Vector LLVA: A virtual vector instruction set for media processing. 46-56. Paper presented at VEE 2006 - 2nd International Conference on Virtual Execution Environments, Ottawa, ON, Canada.

Vector LLVA : A virtual vector instruction set for media processing. / Bocchino, Robert L.; Adve, Vikram Sadanand.

2006. 46-56 Paper presented at VEE 2006 - 2nd International Conference on Virtual Execution Environments, Ottawa, ON, Canada.

Research output: Contribution to conferencePaper

Bocchino, RL & Adve, VS 2006, 'Vector LLVA: A virtual vector instruction set for media processing', Paper presented at VEE 2006 - 2nd International Conference on Virtual Execution Environments, Ottawa, ON, Canada, 6/14/06 - 6/16/06 pp. 46-56.
Bocchino RL, Adve VS. Vector LLVA: A virtual vector instruction set for media processing. 2006. Paper presented at VEE 2006 - 2nd International Conference on Virtual Execution Environments, Ottawa, ON, Canada.
Bocchino, Robert L. ; Adve, Vikram Sadanand. / Vector LLVA : A virtual vector instruction set for media processing. Paper presented at VEE 2006 - 2nd International Conference on Virtual Execution Environments, Ottawa, ON, Canada.11 p.
@conference{772113f360664441a11771471026bd47,
title = "Vector LLVA: A virtual vector instruction set for media processing",
abstract = "We present Vector LLVA, a virtual instruction set architecture (V-ISA) that exposes extensive static information about vector parallelism while avoiding the use of hardware-specific parameters. We provide both arbitrary-length vectors (for targets that allow vectors of arbitrary length, or where the target length is not known) and fixed-length vectors (for targets that have a fixed vector length, such as subword SIMD extensions), together with a rich set of operations on both vector types. We have implemented translators that compile (1) Vector LLVA written with arbitrary-length vectors to the Motorola RSVP architecture and (2) Vector LLVA written with fixed-length vectors to both AltiVec and Intel SSE2. Our translatorgenerated code achieves speedups competitive with handwritten native code versions of several benchmarks on all three architectures. These experiments show that our V-ISA design captures vector parallelism for two quite different classes of architectures and provides virtual object code portability within the class of subword SIMD architectures.",
keywords = "Multimedia, SIMD, Vector, Virtual Instruction Sets",
author = "Bocchino, {Robert L.} and Adve, {Vikram Sadanand}",
year = "2006",
month = "7",
day = "20",
language = "English (US)",
pages = "46--56",
note = "VEE 2006 - 2nd International Conference on Virtual Execution Environments ; Conference date: 14-06-2006 Through 16-06-2006",

}

TY - CONF

T1 - Vector LLVA

T2 - A virtual vector instruction set for media processing

AU - Bocchino, Robert L.

AU - Adve, Vikram Sadanand

PY - 2006/7/20

Y1 - 2006/7/20

N2 - We present Vector LLVA, a virtual instruction set architecture (V-ISA) that exposes extensive static information about vector parallelism while avoiding the use of hardware-specific parameters. We provide both arbitrary-length vectors (for targets that allow vectors of arbitrary length, or where the target length is not known) and fixed-length vectors (for targets that have a fixed vector length, such as subword SIMD extensions), together with a rich set of operations on both vector types. We have implemented translators that compile (1) Vector LLVA written with arbitrary-length vectors to the Motorola RSVP architecture and (2) Vector LLVA written with fixed-length vectors to both AltiVec and Intel SSE2. Our translatorgenerated code achieves speedups competitive with handwritten native code versions of several benchmarks on all three architectures. These experiments show that our V-ISA design captures vector parallelism for two quite different classes of architectures and provides virtual object code portability within the class of subword SIMD architectures.

AB - We present Vector LLVA, a virtual instruction set architecture (V-ISA) that exposes extensive static information about vector parallelism while avoiding the use of hardware-specific parameters. We provide both arbitrary-length vectors (for targets that allow vectors of arbitrary length, or where the target length is not known) and fixed-length vectors (for targets that have a fixed vector length, such as subword SIMD extensions), together with a rich set of operations on both vector types. We have implemented translators that compile (1) Vector LLVA written with arbitrary-length vectors to the Motorola RSVP architecture and (2) Vector LLVA written with fixed-length vectors to both AltiVec and Intel SSE2. Our translatorgenerated code achieves speedups competitive with handwritten native code versions of several benchmarks on all three architectures. These experiments show that our V-ISA design captures vector parallelism for two quite different classes of architectures and provides virtual object code portability within the class of subword SIMD architectures.

KW - Multimedia

KW - SIMD

KW - Vector

KW - Virtual Instruction Sets

UR - http://www.scopus.com/inward/record.url?scp=33745944181&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33745944181&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:33745944181

SP - 46

EP - 56

ER -