Performance engineering for real and complex tall & skinny matrix multiplication kernels on GPUs

General matrix-matrix multiplications with double-precision real and complex entries (DGEMM and ZGEMM) in vendor-supplied BLAS libraries are best optimized for square matrices but often show bad performance for tall & skinny matrices, which are much taller than wide. NVIDIA’s current CUBLAS impl...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	The international journal of high performance computing applications 2021-01, Vol.35 (1), p.5-19
Hauptverfasser:	Ernst, Dominik, Hager, Georg, Thies, Jonas, Wellein, Gerhard
Format:	Artikel
Sprache:	eng
Schlagworte:	Multiplication Performance engineering Performance evaluation
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!