How to pass data divided to two GPU for parallel calculations?

1 次查看（过去 30 天）

Mantas Vaitonis 2019-3-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/452691-how-to-pass-data-divided-to-two-gpu-for-parallel-calculations

Hello,

I have a group of large *.mat files that are arrays of size Nx130. I would like to load this data and sent it to two GPUs (workers) divided like Nx1:75 to 1GPU and Nx76:130 to 2GPU. I need to do both calculations in parallel, thus I use spmd. Firstly, I would like to ask if this method in the code below is correct. If yes, then if this is the optimal one?

clear;
gpuDevice(1);
nGPUs = gpuDeviceCount(); %start pool of 2 GPUS
parpool('local', nGPUs );
labdata=load ('07_04.mat');%load data to workspace
  lab1=labdata.c;
  [~,zz]=size(lab1);
 zz=zz/2;
zz = fix(zz)*2;
spmd 
   er=zz-zz/labindex+1;
   er1=zz/2*labindex;
   lab=lab1(:,er:er1); %creat array depending of worker (Nx1:75 or Nx76:130)
   dydis=5050;
   [z,~]=size(lab);
z=z/dydis;
z = fix(z);
[a,b]=parallel (lab,z);
end