Optimized Data Transfers Based on the OpenCL Event Management Mechanism

Table 1

System specifications.

System	Masamune	Cichlid	RICC
CPU	Intel Xeon E5-2670	Intel Core i7 930	Intel Xeon 5570
GPU	GeForce GTX TITAN	Tesla C2070	Tesla C1060
NIC	GbE 1000BASE-T	GbE 1000BASE-T	InfiniBand DDR
OS	CentOS 6.4	CentOS 6.0	RHEL 5.3
Compiler	GCC-4.4.7	GCC-4.4.4	Intel Compiler 11.1
GPU Driver	319.37	290.10	295.41
OpenCL	OpenCL1.1 (CUDA5.5)	OpenCL1.1 (CUDA4.1.1)	OpenCL1.1 (CUDA 4.2.9)
MPI	Open MPI 1.5.4	Open MPI 1.6.0	Open MPI 1.6.1
Storage	SSD (Intel 910 400 GB)	NFS	NFS