BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111115T183000Z
DTEND:20111115T190000Z
LOCATION:TCC 305
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: GPUs are an excellent accelerator for data-parallel applications with regular data access patterns. It is challenging, however, to optimize computations with irregular data access patterns on GPUs.=0AOne such computation is the Symmetric Matrix Vector product (SYMV)=0Afor dense linear algebra. Optimizing the SYMV kernel is important be-=0Acause it forms the basis of fundamental algorithms such as linear solvers=0Aand eigenvalue solvers on symmetric matrices. In this work, we present=0Aa new algorithm for optimizing the SYMV kernel on GPUs. Our optimized SYMV in single precision brings up to 7X speed up compared to=0Athe (latest) CUBLAS 3.2 NVIDIA library on the GTX 280 GPU. Our=0ASYMV kernel tuned for Fermi C2050 is 4.5X faster than CUBLAS 3.2 in=0Asingle precision.
SUMMARY:Optimizing Symmetric Dense Matrix-Vector Multiplication on GPUs
PRIORITY:3
END:VEVENT
END:VCALENDAR