• Course code:63509
  • Credits:6
  • Semester: summer
  • Contents
  1. Shared memory multiprocessors

- Multicores (architectures UMA and NUMA, cache coherence, hardware multithreading, atomic instructions).

- Differences between architectures Intel,ARM,RISC

- Vector units in modern processors

  1. Accelerator Architectures 

- GPU architectures. Schedulers on HW level, working with memory, latency hiding, data transfer between GPU. Nvidia, AMD, Intel, EUPAC. Tensor cores.

- Unified memory

- FPGA accelerators.

  1. Distributed memory systems

- cluster architectures

- supercomputers

- unified global memory

- interconnection networks (switching fabric, virtual channels, Infiniband, Ethernet)

- scalable systems

- latency issues (of memory accesses and interprocessor communication), precommunication, block data transfer, multithreading

- cluster cooling methods, energy consumption of processors and cooling system

  1. Evaluation of parallel systems
  2. Architectural issues in designing power-aware computing systems

- high-performance systems

- embedded systems

  • Study programmes
  • Distribution of hours per semester
laboratory work
  • Professor
Room:R2.48 - Kabinet
Teaching Assistant
Room:R2.41 - Laboratorij LASPP