Skip to main content

Computer Systems

2026/2027
Programme:
Computer Science and Mathematics, Second Cycle
Year:
1 ali 2 year
Semester:
second
Kind:
optional
ECTS:
6
Language:
slovenian
Course director:

Branko Šter

Hours per week – 2. semester:
Lectures
3
Seminar
0
Tutorial
2
Lab
0
Prerequisites

At least basic programming skills. Knowing basics of computer architecture is beneficial.

Content (Syllabus outline)

1. Advanced instruction-level parallelism: dynamic instruction scheduling, branch  prediction, speculative execution. 

2. Evaluation of parallel systems: architectural simulators (gem5), performance analysis and modeling. 

3. Thread-level parallelism: shared-memory architectures, symmetric (shared-memory) multiprocessors (SMP, UMA), snooping protocols (MSI, MESI, MOESI), multicores, distributed shared-memory (DSM, NUMA), directory protocols, cache coherence, atomicity, synchronization, memory consistency.

4. Data-level parallelism: vector architectures, SIMD instruction set extensions, roofline model, graphics processing units (GPU), GPU architectures, GPU memory structures, loop-level parallelism. 

5. Request-level parallelism: distributed memory systems, clusters, supercomputers. 

6. Interconnects: hierarchy, types, topologies, performance metrics, routing, accelerator interconnects (PCIe, NVLink, NVSwitch, Infinity Fabric, GPUDirect RDMA, GPUDirect Storage), supercomputers, trends.

7. Domain-specific architectures accelerator architectures, deep neural networks (DNN), tensor processing units (TPU), FPGA accelerators, heterogeneous systems.

Readings
  1. J.L. Hennessy, D.A. Patterson: Computer Architecture: A Quantitative Approach, 6th ed., Morgan Kaufmann, 2017. 

  2. M. Dubois, M. Annavaram, P. Stenström: Parallel Computer Organization and Design, 2nd ed., Cambridge University Press, 2026. 

  3. V. Eijkhout, E. Chow, R. van de Geijn: The Science of Computing, The Art of High Performance Computing, volume 1. 3rd edition 2022, https://theartofhpc.com  

  4. T.M. Aamodt, W.W.L. Fung, T.G. Rogers: General-Purpose Graphics Processor Architectures (Synthesis Lectures on Computer Architecture), ‎Morgan & Claypool Publishers, 2018. 

  5. C. Wang: Domain-Specific Computer Architectures for Emerging Applications: Machine Learning and Neural Networks, CRC Press, 2024. 

  6. T. Sterling, M. Brodowicz, M. Anderson: High Performance Computing: Modern Systems and Practices, Morgan Kaufmann, 2017. 

Objectives and competences

The course aims to present to graduate students modern parallel high performance computer systems.

Competences: Developing skills in critical, analytical and synthetic thinking.

The ability to define, understand and solve creative professional challenges in computer and information science.

The ability of professional communication in the native and in a foreign language.

The ability to apply acquired knowledge in independent work for solving technical and scientific problems in computer and information science; the ability to upgrade acquired knowledge.

Basic skills in computer and information science.

Practical knowledge and skills necessary for successful professional work in computer and information science.

The ability to independently perform engineering and organisational tasks in certain narrow areas and independently solve specific well-defined tasks in computer and information science.

Intended learning outcomes

Knowledge and understanding:

Knowledge of the principles of advanced state-of-the-art computer architectures.

Understanding of implications of different ways of using hardware parallelism (processors, interconnection networks and accelerators). Architectural design issues in parallel computers.

Application:

Programming high performance computer systems on various fields of science and engineering. Analyzing and solving architectural design problems.

Reflection:

Understanding theoretical methods for solving practical problems, and their limitations.

Understanding trends in parallel computer architectures. How can high performance systems be used in practice to solve complex problems.

Transferable skills:

Students will be able to combine the knowledge from other courses that cover computer architecture and organization, programming, parallel and distributed systems. The course will require students to acquire skills in literature search, search and application of existing parallel software, and engineering skills for solving real-life complex problems.

Learning and teaching methods

Lectures, calculation exercises, laboratory exercises, homeworks.

Assessment

Midterm work: laboratory exercises, homeworks.

Final exam: written and theoretical exam.
5 - 10, a student passes the exam if he is graded from 6 to 10

Lecturer's references
  1. Branko Šter: Selective recurrent neural network. Neural processing letters, 38(1): 1-15, 2013.
  2. Dominik Olszewski, Branko Šter: Asymmetric clustering using the alpha–beta divergence. Pattern Recognition, 47(5): 2031-2041, 2013.
  3. Rok Gaber, Tina Lebar, Andreja Majerle, Branko Šter, Andrej Dobnikar, Mojca Benčina, Roman Jerala: Designable DNA-binding domains enable construction of logic circuits in mammalian cells. Nature Chemical Biology, 10(3): 203-208, 2014.
  4. Jernej Zupanc, Damjana Drobne, Branko Šter: Markov random field model for segmenting large populations of lipid vesicles from micrographs. Journal of liposome research, 21(4): 315-323, 2011.
  5. Tom Vodopivec, Spyridon Samothrakis, Branko Šter: On Monte Carlo tree search and reinforcement learning. The journal of artificial intelligence research. Sep. 2017, vol. 60, str. 881-936.