GP-SIMD processing-in-memory

Amir Morad, Leonid Yavits, Ran Ginosar

Research output: Contribution to journalArticlepeer-review

19 Scopus citations


GP-SIMD, a novel hybrid general-purpose SIMD computer architecture, resolves the issue of data synchronization by in-memory computing through combining data storageand massively parallel processing. GP-SIMD employs a two-dimensional access memory with modified SRAM storage cells and a bit-serial processing unit per each memory row. An analytic performance model of the GP-SIMD architecture is presented, comparing it to associative processor and to conventional SIMD architectures. Cycle-accurate simulation of four workloads supports the analytical comparison. Assuming a moderate die area, GP-SIMD architecture outperforms both the associative processor and conventional SIMD coprocessor architectures by almost an order of magnitude while consuming less power.

Original languageEnglish
Article number53
JournalTransactions on Architecture and Code Optimization
Issue number4
StatePublished - 1 Dec 2015
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2014 ACM.


  • Associative processor
  • Multicore
  • PIM
  • Processing in memory
  • SIMD


Dive into the research topics of 'GP-SIMD processing-in-memory'. Together they form a unique fingerprint.

Cite this