COMPUTE OPTIMIZATION MECHANISM FOR DEEP NEURAL NETWORKS

The present disclosure provides a graphics processing unit comprising one or more multiprocessors, at least one of the one or more multiprocessors including a register file to store a plurality of different types of operands and a plurality of processing units, including a first set of execution uni...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Galoppo von Borries, Nicolas, Satish, Nadathur Rajagopalan, Ashbaugh, Ben J, Akhbari, Farshad, Ray, Joydeep, Srinivasa, Narayan, Maiyuran, Subramaniam, Schluessler, Travis T, Feit, John H, Gottschlich, Justin E, Boles, Jeffery S, Vaidyanathan, Karthik, Surti, Prasoonkumar, Nurvitadhi, Eriko, Burke, Devan, Hurd, Linda L, Appu, Abhishek R, Chen, Feng, Baghsorkhi, Sara S, Lake, Adam T, Lin, Tsung-Han, Fu, Wenyin, Koker, Altug, Kim, Dukhwan, Sinha, Kamal, Vembu, Balaji, Barik, Rajkishore, Mastronarde, Josh B
Format: Patent
Sprache:eng ; fre ; ger
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present disclosure provides a graphics processing unit comprising one or more multiprocessors, at least one of the one or more multiprocessors including a register file to store a plurality of different types of operands and a plurality of processing units, including a first set of execution units of a first type to process matrix instructions on a first set of operands stored in a first set of registers of the register file, wherein the first set of operands including one or more 64-bit operands and a second set of execution units of a second type, the second set of execution units being different from the first set of execution units, the second set of execution units to perform general purpose graphics processing unit, GPGPU, instructions on a second set of operands stored in a second set of registers of the register file.