Experimental Results

The experimental results presented in the following are described in detail here.

Performance Evaluation

Speedup of MDH (higher is better) over state-of-the-art approaches on GPUs and CPUs.

Linear Algebra

MDH Performance Experiments for Linear Algebra

Stencils

MDH Performance Experiments for Stencils

Quantum Chemistry

MDH Performance Experiments for Quantum Chemistry

Data Mining

MDH Performance Experiments for Data Mining

Deep Learning

MDH Performance Experiments for Deep Learning

Portability Evaluation

Our portability evaluation relies on the Pennycook Metric (higher is better). A value close to 1 indicates high portability, and a value close to 0 indicates low portability.

Portability Evaluation on GPUs+CPUs

MDH Portability Experiments on GPUs and CPUs

Portability Evaluation on either GPUs or CPUs

MDH Portability Experiments on GPUs or CPUs for Linear Algebra

MDH Portability Experiments on GPUs or CPUs for Stencils

MDH Portability Experiments on GPUs or CPUs for Quantum Chemistry

MDH Portability Experiments on GPUs or CPUs for Probabilistic Record Linkage