How can I accelerate my Matlab program

Question

0 个投票

Hi everybody, the working of the my matlab program durate very long, and i want to short the operation time. I write the outline of the my program below.

f=(400:1000);

q=5;

m=5;

rk=4096;

f1=length(f);

for a=1:f1

for b=1:q

for c=1:m

for d=1:rk

end

Also, how can be shorted the operation time of the any matlab program?

2 个评论
显示无隐藏无

Voss 2022-5-14

在 MATLAB Online 中打开

Here's an outline of a solution:

f=(400:1000);
q=5;
m=5;
rk=4096;
f1=length(f);
% use vectorized operations instead of for loops

It's possible that the operations inside the for loops can be vectorized, which will very likely speed up the program. However, without knowing what those operations are, it's not possible to say the extent to which vectorization is possible or how to do it. Please share the full code.

Erkan 2022-5-17

thanks for your advise. I tried to vectorize the codes that being one part of my main program, but i cannot do because i know a few on the vectorization. I written those codes below. Also, in those codes, there is the coupled differential equations (4th Runge-Kutta method).

clc, clear all, close all

q=1.6e-19;

Beta=1e-4;

Gama=0.025;

Alfa_gs=4.6e-14;

Alfa_es=9.2e-14;

h=6.62e-34;

Lamda=1.55e-4;

Epsilon=0;

R1=0.95;

R2=0.05;

Kb=(1.38e-23)/q;

T=300;

Aw=1.35e10;

Cw=5e-9;

Ae=1.5e10;

Ce=9e-8;

ue=4;

uw=10;

ug=2;

Ees=0.84;

Egs=0.792;

Ewl=1.05;

twr=500e-12;

ter=500e-12;

tgr=1.2e-9;

tp=8.92e-12;

Length=2.45e-1;

Width=12e-4;

Thickness=2e-7;

Va=(Length*Width*Thickness*3);

N0=6e16;

Nr=3.27;

c=3e10;

Vg=c/Nr;

Loss_m=((log(1/(R1*R2)))/(2*Length*Nr));

Loss_i=6;

Loss=Loss_m+Loss_i;

Pcon=((Vg*h*Va*Loss_m*c)/(Lamda*Gama));

f=1000e6;

I=(0:200)*1e-3;

loop1=length(I);

t=linspace(0,1/f,3000);

step=(1/f)/3000;

loop2=length(t);

power1=zeros(1,loop1);

power2=zeros(1,loop1);

power3=zeros(1,loop1);

power4=zeros(1,loop1);

power5=zeros(1,loop1);

power6=zeros(1,loop1);

Gain_es=zeros(1,loop1);

Gain_gs=zeros(1,loop1);

Gain_total=zeros(1,loop1);

Nw=zeros(1,loop2);

Ng=zeros(1,loop2);

Ne=zeros(1,loop2);

Se=zeros(1,loop2);

Sg=zeros(1,loop2);

for j=1:loop1

% starting the 4th Runge-Kutta method at this for loop

for i=1:loop2

fes(i)=1-(Ne(i)/(4*N0));

fgs(i)=1-(Ng(i)/(2*N0));

twe(i)=1/(Aw+Cw*Nw(i));

teg(i)=1/(Ae+Ce*Nw(i));

twg(i)=twe(i);

tge(i)=(ug/ue)*teg(i)*exp((Ees-Egs)/(Kb*T));

tew(i)=(ue/uw)*twe(i)*exp((Ewl-Ees)/(Kb*T));

fNw=@(t,Nw,Ne) I(j)/(q*Va)-fes(i)*(Nw/twe(i))-fgs(i)*(Nw/twg(i))-(Nw/twr)+(Ne/tew(i));

fNe=@(t,Nw,Ne,Ng,Se) fes(i)*(Nw/twe(i))+fes(i)*(Ng/tge(i))-fgs(i)*(Ne/teg(i))-(Ne/tew(i))-(Ne/ter)-Vg*Gama*Alfa_es*((Ne/2)-N0)*(Se/(1+(Epsilon*Se)));

fNg=@(t,Nw,Ne,Ng,Sg) fgs(i)*(Nw/twg(i))+fgs(i)*(Ne/teg(i))-fes(i)*(Ng/tge(i))-(Ng/tgr)-Vg*Gama*Alfa_gs*(Ng-N0)*(Sg/(1+(Epsilon*Sg)));

fSe=@(t,Ne,Se) Vg*Gama*Alfa_es*((Ne/2)-N0)*(Se/(1+(Epsilon*Se)))+(Gama*Beta*(Ne/ter))-(Se/tp);

fSg=@(t,Ng,Sg) Vg*Gama*Alfa_gs*(Ng-N0)*(Sg/(1+(Epsilon*Sg)))+(Gama*Beta*(Ng/tgr))-(Sg/tp);

k1=fNw(t(i),Nw(i),Ne(i));

m1=fNe(t(i),Nw(i),Ne(i),Ng(i),Se(i));

n1=fNg(t(i),Nw(i),Ne(i),Ng(i),Sg(i));

p1=fSe(t(i),Ne(i),Se(i));

r1=fSg(t(i),Ng(i),Sg(i));

k2=fNw(t(i)+step/2,Nw(i)+step/2*k1,Ne(i)+step/2*m1);

m2=fNe(t(i)+step/2,Nw(i)+step/2*k1,Ne(i)+step/2*m1,Ng(i)+step/2*n1,Se(i)+step/2*p1);

n2=fNg(t(i)+step/2,Nw(i)+step/2*k1,Ne(i)+step/2*m1,Ng(i)+step/2*n1,Sg(i)+step/2*r1);

p2=fSe(t(i)+step/2,Ne(i)+step/2*m1,Se(i)+step/2*p1);

r2=fSg(t(i)+step/2,Ng(i)+step/2*n1,Sg(i)+step/2*r1);

k3=fNw(t(i)+step/2,Nw(i)+step/2*k2,Ne(i)+step/2*m2);

m3=fNe(t(i)+step/2,Nw(i)+step/2*k2,Ne(i)+step/2*m2,Ng(i)+step/2*n2,Se(i)+step/2*p2);

n3=fNg(t(i)+step/2,Nw(i)+step/2*k2,Ne(i)+step/2*m2,Ng(i)+step/2*n2,Sg(i)+step/2*r2);

p3=fSe(t(i)+step/2,Ne(i)+step/2*m2,Se(i)+step/2*p2);

r3=fSg(t(i)+step/2,Ng(i)+step/2*n2,Sg(i)+step/2*r2);

k4=fNw(t(i)+step,Nw(i)+step*k3,Ne(i)+step*m3);

m4=fNe(t(i)+step,Nw(i)+step*k3,Ne(i)+step*m3,Ng(i)+step*n3,Se(i)+step*p3);

n4=fNg(t(i)+step,Nw(i)+step*k3,Ne(i)+step*m3,Ng(i)+step*n3,Sg(i)+step*r3);

p4=fSe(t(i)+step,Ne(i)+step*m3,Se(i)+step*p3);

r4=fSg(t(i)+step,Ng(i)+step*n3,Sg(i)+step*r3);

Nw(i+1)=Nw(i)+step/6*(k1+2*k2+2*k3+k4);

Ne(i+1)=Ne(i)+step/6*(m1+2*m2+2*m3+m4);

Ng(i+1)=Ng(i)+step/6*(n1+2*n2+2*n3+n4);

Se(i+1)=Se(i)+step/6*(p1+2*p2+2*p3+p4);

Sg(i+1)=Sg(i)+step/6*(r1+2*r2+2*r3+r4);

end

Gain_e=Gama*Alfa_es*((Ne/2)-N0);

Gain_g=Gama*Alfa_gs*(Ng-N0);

Gain=(Gain_e+Gain_g);

Gain_es(j)=mean(Gain_e);

Gain_gs(j)=mean(Gain_g);

Gain_total(j)=mean(Gain);

power1(j)=mean(Nw);

power2(j)=mean(Ne);

power3(j)=mean(Ng);

power4(j)=mean(Se);

power5(j)=mean(Sg);

power6(j)=mean(Se+Sg);

end

请先登录，再进行评论。

请先登录，再回答此问题。

Follow Question

How can I accelerate my Matlab program

2 个评论
显示无隐藏无

回答（0 个）

类别

产品

版本

标签

Community Treasure Hunt

How can I accelerate my Matlab program

2 个评论 显示 无 隐藏 无

回答（0 个）

类别

产品

版本

标签

另请参阅

Community Treasure Hunt

2 个评论
显示无隐藏无