Accelerating Particle Image Velocimetry Using Hybrid Architectures

High Performance Computing (HPC) applications are mapped to a cluster of multi-core processors communicating using high speed interconnects. More computational power is harnessed with the addition of hardware accelerators such as Graphics Processing Unit (GPU) cards and Field Programmable Gate Arrays (FPGAs). Particle Image Velocimetry (PIV) is an embarrassingly parallel application that can benefit from acceleration using hybrid architectures. The PIV application is mapped to a Nvidia GPU system, resulting in 3x speedup over a dual quad-core Intel processor implementation. The design methodology used to implement the PIV application on a specialized FPGA platform under development is described in brief and the resulting performance benefit is analyzed. I. PARTICLE IMAGE VELOCIMETRY ALGORITHM Cardiovascular Disease (CVD) is the leading cause of death in the United States and accounts for more than 36.3 % of all fatalities for 2006 [1]. Early and accurate detection of CVD would increase the median human life by a significant percentage. To facilitate this detection, the Advanced Experimental Thermofluids Engineering (AEThER) Lab at Virginia Tech is involved in the area of cardiovascular fluid dynamics. Their research focuses on the effects of coronary stents on blood flow, and the role of pressure gradients in the development of diastolic dysfunction and heart failure. The experimental method [2] involves using the data intensive Particle Image Velocimetry (PIV) technique for flow measurement and visualization. The PIV technique utilizes a high intensity pulsed laser sheet to illuminate a flow field seeded with reflective or fluorescent particles. The motion of the particles are captured using high speed digital cameras that produce images of the particle patterns. These images are correlated using the PIV algorithm to measure the fluid velocity, which influences the stress exerted on the arterial walls. The experimental setup produces large amount of data (each case results in 1250 image pairs × 5MB = 6.25 GB). On an average, there are 100 cases generated for each stent configuration from a total of 20 stent configurations (about 500 GB of data for each case). The PIV algorithm shown in Figure 1 consists of interrogation windows to capture the flow field. The window sizes are 16×16, 32×32 or 64×64 pixels. Each image is broken down into overlapping zones and the corresponding zones from both images are correlated. The peak detection is done FFT FFT Multiplication IFFT motion vector