BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121114T233000Z
DTEND:20121115T000000Z
LOCATION:255-EF
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: While high-efficiency machines are increasingly embracing heterogeneous=0Aarchitectures and massive multithreading, contemporary mainstream=0Aprogramming languages reflect a mental model in which processing=0Aelements are homogeneous, concurrency is limited, and memory is a flat=0Aundifferentiated pool of storage.  Moreover, the current state of the=0Aart in programming heterogeneous machines tends towards using separate=0Aprogramming models, such as OpenMP and CUDA, for different portions of=0Athe machine.  Both of these factors make programming emerging=0Aheterogeneous machines unnecessarily difficult.=0A=0AWe describe the design of the Phalanx programming model, which seeks to=0Aprovide a unified programming model for heterogeneous machines.  It=0Aprovides constructs for bulk parallelism, synchronization, and data=0Aplacement which operate across the entire machine.  Our prototype=0Aimplementation is able to launch and coordinate work on both CPU and GPU=0Aprocessors within a single node, and by leveraging the GASNet runtime,=0Ais able to run across all the nodes of a distributed-memory machine.
SUMMARY:Designing a Unified Programming Model for Heterogeneous Machines
PRIORITY:3
END:VEVENT
END:VCALENDAR
BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121114T233000Z
DTEND:20121115T000000Z
LOCATION:255-EF
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: While high-efficiency machines are increasingly embracing heterogeneous=0Aarchitectures and massive multithreading, contemporary mainstream=0Aprogramming languages reflect a mental model in which processing=0Aelements are homogeneous, concurrency is limited, and memory is a flat=0Aundifferentiated pool of storage.  Moreover, the current state of the=0Aart in programming heterogeneous machines tends towards using separate=0Aprogramming models, such as OpenMP and CUDA, for different portions of=0Athe machine.  Both of these factors make programming emerging=0Aheterogeneous machines unnecessarily difficult.=0A=0AWe describe the design of the Phalanx programming model, which seeks to=0Aprovide a unified programming model for heterogeneous machines.  It=0Aprovides constructs for bulk parallelism, synchronization, and data=0Aplacement which operate across the entire machine.  Our prototype=0Aimplementation is able to launch and coordinate work on both CPU and GPU=0Aprocessors within a single node, and by leveraging the GASNet runtime,=0Ais able to run across all the nodes of a distributed-memory machine.
SUMMARY:Designing a Unified Programming Model for Heterogeneous Machines
PRIORITY:3
END:VEVENT
END:VCALENDAR