Scholarlink

With Artificial Intelligence applications becoming ubiquitous, both industrial and consumer applications require more computing availability. The increasing need for power-constrained Artificial Intelligence computing asks for new and innovative solutions, on both hardware design as well as on software toolchains that leverage them. At the same time, the challenge of EU technological sovereignty encourages an industry trend oriented to developing its own solutions for power efficient High Performance Computing (HPC). In this context, as partners in the ISOLDE project started in 2023, we propose the development of new power efficient embedded HPC accelerators, HPC design platforms, as well as High Level Synthesis (HLS) approaches and toolchain technology designs for RISC-V based applications. Specifically, we envisage applications such as Fast Fourier Transform, Number Theoretic Transforms, Post Quantum Computing and artificial intelligence, as well as designing ONNX targeted toolchains based on the LLVM compiler, open solutions for Neural Networks (NN) accelerators and new OpenMP extensions. Within the ISOLDE framework, the envisaged applications to prove the readiness of these technologies are the Automotive and the Smart Home demonstrators.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

RISC-V Accelerators, Enablement and Applications for Automotive and Smart Home in the ISOLDE Project

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Cătălin Bogdan Ciobanu,
  • Honorius Gâlmeanu,
  • Alexandru Puşcaşu,
  • Mihai Gologanu,
  • Octavian Buiu,
  • Mihai Antonescu,
  • Vlad-Gabriel Serbu,
  • Vasile-Mădălin Moise,
  • Cristian-Tiberius Axinte,
  • Alexandru-Tudor Popovici,
  • George-Iulian Uleru,
  • Andrei Stan,
  • Mihai Munteanu,
  • Alexandru Drîmbărean,
  • Csaba Nemeti,
  • Dănuţ Rotar,
  • Daniel Grosu,
  • Cosmin Moisă,
  • Bogdan Ditu,
  • Petre Cristian Trusca,
  • Marius Antache,
  • Simona Costinescu,
  • Mari-Anais Sachian,
  • George Suciu,
  • Cristian Gheorghe,
  • Cristina Tudor,
  • Kejsi Koci

摘要

With Artificial Intelligence applications becoming ubiquitous, both industrial and consumer applications require more computing availability. The increasing need for power-constrained Artificial Intelligence computing asks for new and innovative solutions, on both hardware design as well as on software toolchains that leverage them. At the same time, the challenge of EU technological sovereignty encourages an industry trend oriented to developing its own solutions for power efficient High Performance Computing (HPC). In this context, as partners in the ISOLDE project started in 2023, we propose the development of new power efficient embedded HPC accelerators, HPC design platforms, as well as High Level Synthesis (HLS) approaches and toolchain technology designs for RISC-V based applications. Specifically, we envisage applications such as Fast Fourier Transform, Number Theoretic Transforms, Post Quantum Computing and artificial intelligence, as well as designing ONNX targeted toolchains based on the LLVM compiler, open solutions for Neural Networks (NN) accelerators and new OpenMP extensions. Within the ISOLDE framework, the envisaged applications to prove the readiness of these technologies are the Automotive and the Smart Home demonstrators.