How to speed up calculations of integral and summation

Question

Yuriy Yerin 2018-6-29

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/408080-how-to-speed-up-calculations-of-integral-and-summation

评论： Jeff Miller 2018-7-1

采纳的回答： Jeff Miller

在 MATLAB Online 中打开

Hello everybody

Regarding to my previous question I'd like to speed up calculations

function z=self_energy_summation
tic
ww=linspace(0,10,100);
for l=1:100
S(l)=integral(@(q)Gamma_0(q,ww(l)),0,10,'AbsTol',1e-6,'RelTol',1e-3,'ArrayValued',true);
end
plot(ww,S)
function y1=Gamma_0(q,w)
z2=load('Tc_Tmatrix_ordered.txt');
tt=z2(741,1);
T_c=z2(741,2);
mu=z2(741,3);
k=1;
N=-1000:1000;
nn=10^4;
fun1=@(a,q) a.*tanh((a.^2-mu)./(2*T_c)).*log((2*a.^2+2*a.*q+q.^2-2*mu-1i*2*pi*N*T_c)./(2*a.^2-2*a.*q+q.^2-2*mu-1i*2*pi*N*T_c))./q-2;
R= T_c*q./k.*log((-k.^2+2*k.*q-q.^2+mu+1i.*(2*pi*N*T_c-w))./(-k.^2-2*k.*q-q.^2+mu+1i.*(2*pi*N*T_c-w))).*1./((tt+integral(@(a)fun1(a,q),0,nn,'AbsTol',1e-6,'RelTol',1e-3,'ArrayValued',true)));
y1=sum(R);
end
Time = toc
end

On the first step I reduced the relative tolerance of integral calculations. But even on my computer with 24 Gb RAM and Intel I7 processor I can't finish these calculations during the 24 hours. My question: is it possible to improve the performance of this code and reduced the time?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jeff Miller 2018-6-30

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/408080-how-to-speed-up-calculations-of-integral-and-summation#answer_326937

在 MATLAB Online 中打开

It seems wasteful to do this

z2=load('Tc_Tmatrix_ordered.txt');

inside the function that you are integrating. Can't you load z2 just once inside self_energy_summation? The nest functions will have access to it.

2 个评论
显示无隐藏无

Yuriy Yerin 2018-7-1

A lot of thanks. Now it works much faster. I apologize, could you explain briefly the origin of the slowdown of my code?

Jeff Miller 2018-7-1

Good that it is faster. The reason is this: MATLAB's integral function calls your Gamma_0 function many times (look for an explanation of how numerical integration works). Loading the file inside Gamma_0 means that your function has to load the file again every time it is called, which wastes a lot of time because loading a file is a relatively slow operation.

There is also the fact that you are doing the integration 100 times, and that multiplies by 100 the number of times that the file is loaded. Loading the file before starting the for loop could easily save 999 repetitions of the file loading step, maybe more depending on your function.

请先登录，再进行评论。