SkelCL
SkelCL is a high level multi GPU skeleton library developed at the university of Münster, Germany.

▼Algorithmic Skeletons  An overview of all algorithmic skeletons offered by SkelCL 
AllPairs Skeleton  The AllPairs skeleton describes a calculation of pairs of row vectors and column vectors with a user defined function which can be performed on one or more devices 
Map Skeletons  This group of classes implements the Map skeleton, which describes calculations performed on one or more devices. It invokes a unary useddefined function on a container in a parallel fashion 
MapOverlap Skeleton  The MapOverlap skeleton describes calculations performed on one or more devices. It invokes a unary userdefined function on a container in a parallel fashion. Unlike the Map skeleton the userdefined function can access more than just a single element of the container 
Reduce Skeleton  The Reduce skeleton describes a calculation on a Vector performed in parallel on a device. Given a binary userdefined function the input Vector is reduced to a scalar value 
Scan Skeleton  The Scan skeleton describes a calculation on a Vector performed in parallel on a device. Given a binary userdefined function the input Vector is transformed into an out Vector by performing a scan (a.k.a. prefix sum) operation 
Zip Skeleton  The Zip skeleton describes a calculation on two containers performed in parallel on one or more devices. It invokes a binary userdefined function on its two input containers in a parallel fashion 
▼Containers  An overview of all container data types offered by SkelCL 
Matrix  Two dimensional container data structures 
Vector  One dimensional container data structures 
Distributions  An overview of all distributions offered by SkelCL 