Optimizing OpenCLTM on CPUs

• OpenCL is a Platform API which supports a uniform programming environment across devices – Enables heterogeneous parallel computations – Unique in its ability to coordinate CPUs, GPUs, etc • The goal of using OpenCL should be to make the best use of all the available resources (CPU’s, GPU’s) from within a single program: – One program that runs well (i.e. reasonably close to “hand-tuned ” performance) on a heterogeneous mixture of processors. – Intel new generation of Processors: a new level of integration between CPU & GPU 3Writing OpenCL for the CPU • The challenge of unleashing the performance of modern CPU’s – Multi-Core / SMT – Vector Units (SSE ISA) • OpenCL is a great framework to harness Intel CPUs – Intuitive, easy & maintainable – Unleash the performance of Intel CPUs – Multi Core – Utilize the vector units (SSE ISA) – Close to hand-tuned code! – Performance-portable code – Forward compatibility between CPU generations – Aspire compatibility between devices 4OpenCL view of CoreTM i7 CoreTM i7 97