Technical Report: Accessor Realization

Anzt, Hartwig; Cojean, Terry; Gruetzmacher, Thomas

doi:10.2172/1773264

Title: Technical Report: Accessor Realization

Full Record
Other Related Research

Abstract

As we have seen the arithmetic power of hardware architectures growing much faster than the memory bandwidth, there is an urgent need to rethink the way we operate, communicate, and store data. A promising strategy for overcoming the “memory wall” is the decoupling of memory precision from arithmetic precision as proposed in Anzt et al.. Such a decoupling of memory precision and arithmetic precision requires a “memory accessor” that handles the conversion between formats (resp. compression of the data) on the fly in registers. For such a memory accessor to be useful in practice, it has to combine multiple characteristics.

Authors:

Anzt, Hartwig ^[1]; Cojean, Terry ^[1]; Gruetzmacher, Thomas ^[1]

Karlsruhe Inst. of Technology (KIT) (Germany)

Publication Date:: 2021-03-26

Research Org.:: Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States); Karlsruhe Inst. of Technology (KIT) (Germany)

Sponsoring Org.:: USDOE National Nuclear Security Administration (NNSA)

OSTI Identifier:: 1773264

Report Number(s):: LLNL-SR-820918
1032697

DOE Contract Number:: AC52-07NA27344; NA0003525

Resource Type:: Technical Report

Country of Publication:: United States

Language:: English

Subject:: 97 MATHEMATICS AND COMPUTING

Citation Formats


                    Anzt, Hartwig, Cojean, Terry, and Gruetzmacher, Thomas. Technical Report: Accessor Realization.  United States: N. p., 2021. 
        Web.  doi:10.2172/1773264.

Copy to clipboard


                    Anzt, Hartwig, Cojean, Terry, & Gruetzmacher, Thomas. Technical Report: Accessor Realization.  United States.  https://doi.org/10.2172/1773264

Copy to clipboard


                    Anzt, Hartwig, Cojean, Terry, and Gruetzmacher, Thomas. 2021.  
        "Technical Report: Accessor Realization".  United States.  https://doi.org/10.2172/1773264.  https://www.osti.gov/servlets/purl/1773264.

Copy to clipboard


                    
@article{osti_1773264,

  title        = {Technical Report: Accessor Realization},

  author       = {Anzt, Hartwig and Cojean, Terry and Gruetzmacher, Thomas},

  abstractNote = {As we have seen the arithmetic power of hardware architectures growing much faster than the memory bandwidth, there is an urgent need to rethink the way we operate, communicate, and store data. A promising strategy for overcoming the “memory wall” is the decoupling of memory precision from arithmetic precision as proposed in Anzt et al.. Such a decoupling of memory precision and arithmetic precision requires a “memory accessor” that handles the conversion between formats (resp. compression of the data) on the fly in registers. For such a memory accessor to be useful in practice, it has to combine multiple characteristics.},

  doi          = {10.2172/1773264},

  url          = {https://www.osti.gov/biblio/1773264},
  journal      = {},
number       = ,

  volume       = ,

  place        = {United States},

  year         = {2021},

  month        = {3}

}

Copy to clipboard

Technical Report:

View Technical Report (0.44 MB)

https://doi.org/10.2172/1773264

Save / Share:

Export Metadata

Save to My Library

Similar records in OSTI.GOV collections:

Technical report: Design of the accessor

Technical Report Anzt, Hartwig; Cojean, Terry; Gruetzmacher, Thoma

As we have seen the arithmetic power of hardware architectures growing much faster than the memory bandwidth, there is an urgent need to rethink the way we operate, communicate, and store data. A promising strategy for overcoming the “memory wall” is the decoupling of memory precision from arithmetic precision as proposed in Anzt et al. Such a decoupling of memory precision and arithmetic precision requires an “accessor” that handles the conversion between formats (resp. compression of the data) on the fly in registers. Such an accessor has to be general in the sense that it can handle a wide rangemore »« less
https://doi.org/10.2172/1763198

Full Text Available
Parallel processing and pipelining usher DSP model into the future

Journal Article Kampen, T; Anders, P - Electron. Des.; (United States)

The course of digital signal processing is well plotted into the future. When standard microprocessors, constrained by their von Neumann architectures and weak arithmetic ability, proved inadequate for the task, the first specialized chips appeared. The devices were fortified with Harvard-like parallel architecture, multiplication-accumulation hardware, and instruction pipelines. In the next stage of their evolution, DSP chips will have to rely on faster IC technologies and even greater degrees of parallel operation. Two versions of such a DSP chip are planned, one with and one without data ROM and program memory, and the latter has been cast in silicon. Themore »« less
High-bandwidth/low-latency temporary storage for supercomputers

Thesis/Dissertation Swensen, J

The traditional use of memory and a symmetrical set of registers for storage of temporary results of scientific programs requires more execution time, hardware, and instruction-stream bandwidth than necessary. Novel register organizations that can be easily integrated into traditional supercomputer architectures can reduce all of these requirements. Execution speed can be more than doubled by storing temporary results in an asymmetrical set of general-purpose registers or an asymmetrical set of vector registers, instead of in memory and a small register-set. Faster access and a hardware cost one fourth that of traditional vector registers can be had by using a vectormore »« less
High-bandwidth/low-latency temporary storage for supercomputers

Book Swensen, J

The traditional use of memory and a symmetrical set of registers for storage of temporary results of scientific programs requires more execution time, hardware, and instruction-stream bandwidth than necessary. Novel register organizations that can be easily integrated into traditional supercomputer architectures can reduce all of these requirements. Execution speed can be more than doubled by storing temporary results in an asymmetrical set of general-purpose registers or an asymmetrical set of vector registers, instead of in memory and a small register-set. Faster access and a hardware cost one fourth that of traditional vector registers can be had by using a vectormore »« less
High-bandwidth/low latency temporary storage for supercomputers

Technical Report Swensen, J

The traditional use of memory and symmetrical set of registers for storage of temporary results of scientific programs requires more execution time, hardware, and instruction-stream bandwidth than necessary. Novel register organizations that can be easily integrated into traditional supercomputer architectures can reduce all of these requirements. Execution speed can be more than doubled by storing temporary results in an asymmetrical set of general-purpose registers or an asymmetrical set of vector registers, instead of in memory and a small register-set. Faster access and a hardware cost one fourth that of traditional vector registers can be had by using a vector registermore »« less

Similar Records