University of Piraeus (UNIPI)
Προηγμένες μεθοδολογίες δοκιμής για αρχιτεκτονικές μικροεπεξεργαστών με αριθμητική υψηλών επιδόσεων
Abstract
dc:descriptionAdvances in the microelectronics science lead to manufacturing technologies with smaller feature sizes. The predictions of the International Technology Roadmap for Semiconductors - ITRS mentions feature sizes of 10 nm in 2022. As feature sizes are getting smaller, defects that hinder the correct operation of a digital circuit or affect its operating frequency, thus its performance, appear more frequently. The testing of digital circuits aims at detecting defective circuits and it is carried out after the circuit is manufactured (manufacturing testing). In addition, testing must be performed simultaneously with normal circuit operation (on-line testing), in order to face the new challenges that appear with the advent of nanotechnology, such as the continuously increasing soft error rate which is attributed to a-particles and cosmic ray (neutrons), and the ageing of materials which leads to lower system performance. Among all digital circuits, the testing of microprocessors presents the most challenges as the most aggressive manufacturing is applied first at them. Given the fact that nowadays most embedded systems contain at least one or more microprocessors ranging from simple 8-bit cores to 32/64-bit ones with modern architectures (superscalar, VLIW), the testing of microprocessors becomes even more important. This PhD thesis contributes to the field of microprocessor testing which integrates floating point units to achieve high performance arithmetic. In the first part of the thesis, we present the methodology to develop testability conditions for floating point units. Testability conditions consist of controllability and observability conditions and are used to extract test patterns for floating point units. They are scalable to every floating point precision (single, double) and they are also independent of the internal architecture of the subcomponents that comprise the floating point unit. In the second part of the thesis, we first present the criteria (small memory footprint, small execution time and low-power consumption) for the development of self-test software for on-line microprocessor testing. Furthermore, we present a new methodology for the development of self-test software for floating point units in modern 32/64-biy microprocessors for both single and double precision units. Finally, experimental results in the high performance floating point units which are used in modern microprocessors (MIPS, PowerPC, Alpha) show the efficacy of the proposed methodology.
Degree
thesis:*- Grantor dc:publisher
- University of Piraeus (UNIPI)
- Year dc:date
- 2009
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Ξενούλης, Γεώργιος
- Xenoulis, Georgios
Subjects
dc:subject × 14- Μικροεπεξεργαστές
- Μονάδες κινητής υποδιαστολής
- Δοκιμή μικροεπεξεργαστών
- Δοκιμή κατά τη λειτουργία
- Δοκιμή μονάδων κινητής υποδιαστολής
- Microprocessors
- Floating point units
- Microprocessor testing
- On-line testing
- Floating point unit testing
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/18060