Toward GPGPU accelerated human electromechanical cardiac simulations.

Area of Science:

• Computational Science
• Biomedical Engineering
• High-Performance Computing

Background:

• Electromechanical modeling is crucial for understanding cardiac function.
• Existing CPU-based finite element codes face computational challenges for complex simulations.
• Graphics Processing Units (GPUs) offer potential for accelerating these demanding computations.

Purpose of the Study:

• To investigate the acceleration of weakly coupled electromechanics simulations using GPU computing.
• To port key components of the CHeart finite element code to a GPU architecture.
• To evaluate the performance improvements achieved through GPU acceleration.

Main Methods:

• Porting ordinary differential equation (ODE) and partial differential equation (PDE) solvers for electrophysiology to the GPU.
• Implementing Jacobian and residual evaluation for the mechanics problem on the GPU.
• Comparing GPU implementation performance against single-core (SC) and multi-core (MC) CPU executions.

Main Results:

• GPU acceleration yielded significant speedups for electrophysiology: 164x vs. SC and 5.5x vs. MC for ODEs; up to 72x vs. SC and 2.6x vs. MC for PDEs.
• GPU implementation of mechanics computations showed speedups of up to 44x vs. SC and 2.0x vs. MC.
• Performance gains were observed on a human-scale left ventricle mesh.

Conclusions:

• GPU acceleration is a viable strategy for significantly enhancing the speed of electromechanical simulations.
• The ported CHeart components demonstrate the effectiveness of GPUs in tackling multi-physics problems.
• This approach holds promise for more efficient and faster cardiac modeling and analysis.