Use of GPUs to boost the performance of a lattice-free tumour growth model

5th Young Researcher Meeting, Trieste 2014

Stella, Sabrina^1 ; Chignola, Roberto^2 ; Milotti, Edoardo^1,3

Department of Physics, University of Trieste, Italy^1

Department of Biotechnology, University of Verona, Italy^2

Istituto Nazionale di Fisica Nucleare, Sezione di Trieste, Italy^3

关键词: Computational model;    Design and implementations;    Different sizes;    Nearest neighbour search;    Simulation code;    Simulation program;    Simulation time;    Tumour growth models;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/566/1/012019/pdf
DOI  :  10.1088/1742-6596/566/1/012019

来源: IOP

【 摘 要 】

We recently developed a computational model of tumour growth. It is a cell- based model that can simulate the growth of multicellular tumour spheroids up to more than one million cells. The simulation program is very demanding and simulation time severely limits the integration of additional biological details, and indeed, at the moment, a typical simulation run requires tens of days to be completed. A new version of the code that exploits Graphics Processing Units (GPUs) to boost performance is being developed. In this paper we describe the design and implementation of a nearest-neighbour search (NNS) algorithm suitable to run on GPU. The algorithm will be integrated in the original code to manage the geometrical calculation in the simulation of the spheroid. Initially the stand alone NNS algorithm was tested for spheroids of different size: better efficency was obtained for bigger spheroids. Eventually the code was integrated in the whole simulation code and preliminary runs gave a speed up of about 5 for spheroids of relatively small size (15000 cells).