iPhoneAlle anzeigen
iPhone 11iPhone 11 ProiPhone 11 Pro MaxiPhone 12iPhone 12 ProiPhone 12 Pro MaxiPhone 12 miniiPhone 13iPhone 13 ProiPhone 13 Pro MaxiPhone 13 miniiPhone 14iPhone 14 PlusiPhone 14 ProiPhone 14 Pro MaxiPhone 15iPhone 15 PlusiPhone 15 ProiPhone 15 Pro MaxiPhone 16iPhone 16 PlusiPhone 16 ProiPhone 16 Pro MaxiPhone 16eiPhone 8iPhone 8 PlusiPhone SE 2020iPhone SE 2022iPhone XiPhone XRiPhone XSiPhone XS Max
Galaxy A-SerieAlle anzeigen
Galaxy A12Galaxy A13Galaxy A13 5GGalaxy A14Galaxy A15Galaxy A16Galaxy A17 5GGalaxy A20eGalaxy A21Galaxy A22Galaxy A23Galaxy A24Galaxy A25Galaxy A26 5GGalaxy A32Galaxy A33Galaxy A34Galaxy A35Galaxy A36 5GGalaxy A40Galaxy A41Galaxy A42Galaxy A50Galaxy A51Galaxy A52Galaxy A53Galaxy A54Galaxy A55Galaxy A56 5GGalaxy A70Galaxy A71Galaxy A72
Redmi SeriesAlle anzeigen
Redmi 10Redmi 13CRedmi 15 5GRedmi Note 10Redmi Note 10 ProRedmi Note 11Redmi Note 11 ProRedmi Note 11 Pro Plus 5GRedmi Note 11sRedmi Note 12Redmi Note 12 5GRedmi Note 12 ProRedmi Note 12 Pro 5GRedmi Note 12 Pro PlusRedmi Note 13Redmi Note 13 5GRedmi Note 13 ProRedmi Note 13 Pro PlusRedmi Note 14Redmi Note 14 Pro PlusRedmi Note 8Redmi Note 8 ProRedmi Note 9Redmi Note 9 Pro
- Startseite
Bücher Wissen & Bildung Naturwissenschaft, Medizin, Informatik, Technik Technik Elektronik, Elektro- und Nachrichtentechnik Sparse Matrix-Vector Multiplication Based on Network-on-Chip
Derzeit nicht verfügbar
Handgeprüfte Gebrauchtware
Bis zu 50 % günstiger als neu
Der Umwelt zuliebe
Sparse Matrix-Vector Multiplication Based on Network-on-Chip
Ahmad Mansour (Taschenbuch, Deutsch)
★★★★★
☆☆☆☆☆ Keine Bewertungen vorhanden
Optischer Zustand
Beschreibung
Sparse matrix-vector multiplication (SMVM) arises in a wide range of applications, and it often dominates the computational cost when solving sparse systems of linear equations or sparse eigenvalue problems iteratively.
The performance of this kernel on general purpose processors is much less than the peak performance due to the high number of required memory operations compared to the number of… floating-point operations and poor cache utilization caused by irregular memory access patterns and indirect memory references. Extensive research has been carried out in order to enhance the performance of this operation by using efficient data structures, matrix reordering techniques, parallel implementations, and accelerators.
With the increasing number of cores inside a single chip, providing efficient communication infrastructure became a critical challenge. Therefore, Network-on-Chip (NoC) is used in order to replace shared-bus connections by network-based communication. The extensibility and the flexibility of NoCs make them a promising solution for parallel implementations of algorithms with irregular data structures. For matrices with arbitrary sparsity structures, SMVM is a task which may exhibit very irregular data structures.
In order to achieve an efficient implementation of SMVM on NoC, investigations in different directions and combination of various results are required. First, SMVM data mapping methods are discussed in order to achieve balanced load distribution and minimal communication cost in the NoC. Second, as SMVM is mostly done in iterative algorithms, a combined design of algorithm and architecture is of high importance. A reduction mechanism is proposed to reduce the computational effort of SMVM on NoC benefiting from the robustness of iterative solvers. This mechanism reduces the number of matrix nonzero elements successively during the run of the iterative solver, without affecting the correctness of the solution. Third, an NoC simulation tool is built in order to facilitate investigations on various design factors. The proposed OMNeT++ based NoC simulator provides flexibility and extensibility not provided in hardware.
Finally, an FPGA prototype is proposed for performing SMVM on a data mapping-based NoC architecture. Dieses Produkt haben wir gerade leider nicht auf Lager.
Handgeprüfte Gebrauchtware
Bis zu 50 % günstiger als neu
Der Umwelt zuliebe
Technische Daten
Erscheinungsdatum
29.07.2015
Sprache
Deutsch
EAN
9783843922135, 9783843922135
Herausgeber
Dr. Hut
Serien- oder Bandtitel
Informationstechnik
Sonderedition
Nein
Autor
Ahmad Mansour
Seitenanzahl
155
Einbandart
Taschenbuch
Schlagwörter
Sparse Matrix, Network-on-Chip, Vector Multiplication
Thema-Inhalt
K - Volkswirtschaft, Finanzen, Betriebswirtschaft und Management
KNT - Medien, Kommunikation, Unterhaltung und Informationsindustrie
KNTX - Informationstechnologie
KN - Industrien und Branchen
Höhe
210 mm
Breite
14.8 cm
-.-
★★★★★
☆☆☆☆☆ Leider noch keine Bewertungen
Leider noch keine Bewertungen
Sicher bei rebuy kaufen
Schreib die erste Bewertung für dieses Produkt!
Wenn du eine Bewertung für dieses Produkt schreibst, hilfst du allen Kund:innen, die noch überlegen, ob sie das Produkt kaufen wollen. Vielen Dank, dass du mitmachst!
Sicher bei rebuy kaufen