Perform for loop and carry out calculation separately

Dear community,

I´m once again asking for your help!

As visible in the code below, I´m trying to perform a for loop in steps of 1:1:823´543 (which corresponds to the length of vector a_1_7a. What I ultimately want to achieve, is that these c- and D-factors are calculated for every single value of the given range (which means that the related value is to be extracted from a previously defined vector, in this case depth_7a_p). After calculating 823´543 values for c and D a related number (823543) of matrices V_7a_p and vectors S_7a_p shall be created. Finally, the same amount of gamma factor vector is to be calculated and extracted from the for loop.

For some reason I don´t understand, this doesn´t work. Hopefully you´re able to help me.

Greetings

David

for n=1:1:length(a_1_7a)
    c_1_2=width*G./depth_7a_p(n,1).*10^3;
    c_2_3=width*G./depth_7a_p(n,2).*10^3;
    c_3_4=width*G./depth_7a_p(n,3).*10^3;
    c_4_5=width*G./depth_7a_p(n,4).*10^3;
    
    D_1=pi^2*E_0*width.*depth_7a_p(n,1)./(span).*10^-3;
    D_2=pi^2*E_0*width.*depth_7a_p(n,2)./(span).*10^-3;
    D_3=pi^2*E_0*width.*depth_7a_p(n,3)./(span).*10^-3;
    D_4=pi^2*E_0*width.*depth_7a_p(n,4)./(span).*10^-3;
    D_5=pi^2*E_0*width.*depth_7a_p(n,5)./(span).*10^-3;
        
    a_3_7a(n)=0;
      
    V_7a_p=[(c_1_2+D_1)*a_1_7a(n) -c_1_2*a_2_7a(n) 0 0 0; -c_1_2*a_1_7a(n) (c_1_2+c_2_3+D_2)*a_2_7a(n) -c_2_3*a_3_7a(n) 0 0; 0 -c_2_3*a_2_7a(n) (c_2_3+c_3_4+D_3)*a_3_7a(n) -c_3_4*a_4_7a(n) 0; 0 0 -c_3_4*a_3_7a(n) (c_3_4+c_4_5+D_4)*a_4_7a(n) -c_4_5*a_5_7a(n); 0 0 0 -c_4_5*a_4_7a(n) (c_4_5+D_5)*a_5_7a(n)];
    
    S_7a_p=[-c_1_2*(a_2_7a(n)-a_1_7a(n));-c_2_3*(a_3_7a(n)-a_2_7a(n))+c_1_2*(a_2_7a(n)-a_1_7a(n));-c_3_4*(a_4_7a(n)-a_3_7a(n))+c_2_3*(a_3_7a(n)-a_2_7a(n));-c_4_5*(a_5_7a(n)-a_4_7a(n))+c_3_4*(a_4_7a(n)-a_3_7a(n));c_4_5*(a_5_7a(n)-a_4_7a(n))];
    
    gamma_7a_1=(V_7a_p).^-1.*S_7a_p;
end
    

[Rik]

You are overwriting most variables in your loop, instead of indexing them.

The variable names look like you are using many numbered variables, which is generally a bad idea. You might be tempted to try to generate the variable names at runtime, instead of using indexing. You are also hiding a lot of information because you don't use descriptive variable names or comments.

Why don't you first try to get your code working and optimize it before you throw almost a million iterations at it. Can you try to find a way to vectorize your code?

[Davide Bertoldi]

Hello Rik,

firstly thanks for your answer which made to apply some modifications and comments to my code, which I now am showing as a whole. Obviously, If you have some concrete ideas on how to rationalize and improve it, I would be really thankful.

Still, even tough the code is working rather fast (25 sec), but this i set to change as soon as the calculation is repeated for more spans (see row 60-61). How would a preallocation look like? Could you give me some hints? I could not yet figure it out, because I´m a newbie to programming...

My 2nd question is about the loop. Even though I set the number of loops to a certain value, the extracted matrix produces some 600 ca. more. The way to extract vectors from a loop is obtained from: https://de.mathworks.com/videos/how-to-store-a-series-of-vectors-from-a-for-loop-97220.html.

I`m looking forward to your answer and help

%Strength values for CLT
G=0.69; %kN/mm²
E_0=11;
E_0_1_7=ones(1,7).*11;%kN/mm²
E_90=0.37; %kN/mm²

%Number of lamellas [-]
lam=3:2:7;

%Span length [m]
span=3.0:0.5:6.0;

%Analysed width [m]
width=1;

%Depth outer lamella parallel to span [mm]
lam_o_p=20:5:50;

%Depth inner lamella parallel to span [mm]
lam_i_p=20:5:50;

%Depth inner lamella cross to span [mm]
lam_i_c=20:5:50;

%Depth vector of CLT panels [mm] total number of possible combinations
depth_7a=combvec(lam_o_p,lam_i_p,lam_i_c,lam_i_p,lam_i_c,lam_i_p,lam_o_p)'; %2*lam_o_p+lam_i_c+lam_i_p+lam_i_c+2*lam_o_p

%Depth vector of CLT panels [mm] parallel to span [mm]
depth_7a_p=depth_7a(:,[1 2 4 6 7]);
depth_7a_p_1=depth_7a_p(:,1);
depth_7a_p_2=depth_7a_p(:,2);
depth_7a_p_3=depth_7a_p(:,3);
depth_7a_p_4=depth_7a_p(:,4);
depth_7a_p_5=depth_7a_p(:,5);

%Depth vector of CLT panels [mm] cross to span [mm]
depth_7a_c=depth_7a(:,[3 5]);

%Total depth of CLT panels [mm]
depth_7a_tot=sum(depth_7a,2); %sum(depth_x,2) where 2 stands for row-wise addition

%Determination of Ares CLT panel parallel to span [m²]
A_CLT_7a_p=10^-3.*depth_7a_p.*width;

%Determination of Ares CLT panel cross to span [m²]
A_CLT_7a_c=10^-3.*depth_7a_c.*width;

%Calculation position center of gravity for CLT_7a,CLT_7b [mm]
zs_CLT_7a=(A_CLT_7a_p(:,1).*(0.5.*depth_7a_p(:,1))+A_CLT_7a_p(:,2).*(depth_7a_p(:,1)+0.5.*depth_7a_p(:,2))+A_CLT_7a_c(:,1).*(depth_7a_p(:,1)+depth_7a_p(:,2)+0.5.*depth_7a_c(:,1))+A_CLT_7a_p(:,3).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+0.5.*depth_7a_p(:,3))+A_CLT_7a_c(:,2).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+depth_7a_p(:,3)+0.5.*depth_7a_c(:,2))+A_CLT_7a_p(:,4).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+depth_7a_p(:,3)+depth_7a_c(:,2)+0.5.*(depth_7a_p(:,4)))+A_CLT_7a_p(:,5).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+depth_7a_p(:,3)+depth_7a_c(:,2)+depth_7a_p(:,4)+0.5.*(depth_7a_p(:,5))))./(A_CLT_7a_p(:,1)+A_CLT_7a_p(:,2)+A_CLT_7a_p(:,3)+A_CLT_7a_p(:,4)+A_CLT_7a_p(:,5)+A_CLT_7a_c(:,1)+A_CLT_7a_c(:,2)); %[mm]

%Distances between center of gravity and lamellas [mm]
a_1_7a=zs_CLT_7a-(0.5.*depth_7a_p(:,1));
a_2_7a=zs_CLT_7a-(depth_7a_p(:,1)+0.5.*depth_7a_p(:,2));
a_2_3_7a=zs_CLT_7a-(depth_7a_p(:,1)+depth_7a_p(:,2)+0.5.*depth_7a_c(:,1));%cross layer in between
a_3_7a=zeros(823543,1);
a_3_4_7a=-zs_CLT_7a+(depth_7a_p(:,4)+depth_7a_p(:,5)+0.5.*depth_7a_c(:,2));%cross layer in between
a_4_7a=-zs_CLT_7a+(0.5.*depth_7a_p(:,4)+depth_7a_p(:,5));
a_5_7a=-zs_CLT_7a+(depth_7a_p(:,5));

%first round of calculation for span_1=3.0m
span_1=span(1);

%Calculation of c-factors for matrix [kN/mm]
c_1_2=width*G./depth_7a_p_1.*10^3;
c_2_3=width*G./depth_7a_p_2.*10^3;
c_3_4=width*G./depth_7a_p_3.*10^3;
c_4_5=width*G./depth_7a_p_4.*10^3;
   
%Calculation of D-factors for matrix [kN/mm]
D_1=pi^2*E_0*width.*depth_7a_p_1./(span_1).*10^-3;
D_2=pi^2*E_0*width.*depth_7a_p_2./(span_1).*10^-3;
D_3=pi^2*E_0*width.*depth_7a_p_3./(span_1).*10^-3;
D_4=pi^2*E_0*width.*depth_7a_p_4./(span_1).*10^-3;
D_5=pi^2*E_0*width.*depth_7a_p_5./(span_1).*10^-3;

n=1:1:823543;

%Matrix [5x5] where every element is a (823543x1) vector
V_7a_p=[(c_1_2(n)+D_1(n)).*a_1_7a(n) -c_1_2(n).*a_2_7a(n) zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2(n).*a_1_7a(n) (c_1_2(n)+c_2_3(n)+D_2(n)).*a_2_7a(n) -c_2_3(n).*a_3_7a(n) zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3(n).*a_2_7a(n) (c_2_3(n)+c_3_4(n)+D_3(n)).*a_3_7a(n) -c_3_4(n).*a_4_7a(n) zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4(n).*a_3_7a(n) (c_3_4(n)+c_4_5(n)+D_4(n)).*a_4_7a(n) -c_4_5(n).*a_5_7a(n); zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5(n).*a_4_7a(n) (c_4_5(n)+D_5(n)).*a_5_7a(n)];
    
%S-Vector [(5x1] where every element is a (823543x1) vector
S_7a_p=[-c_1_2(n).*(a_2_7a(n)-a_1_7a(n));-c_2_3(n).*(a_3_7a(n)-a_2_7a(n))+c_1_2(n).*(a_2_7a(n)-a_1_7a(n));-c_3_4(n).*(a_4_7a(n)-a_3_7a(n))+c_2_3(n).*(a_3_7a(n)-a_2_7a(n));-c_4_5(n).*(a_5_7a(n)-a_4_7a(n))+c_3_4(n).*(a_4_7a(n)-a_3_7a(n));c_4_5(n).*(a_5_7a(n)-a_4_7a(n))];

%Start of for loop with the idea to calculate 823543 5x5 matrices with related number of s-vectors
for n=1:1:823543
V_7a_p_n=V_7a_p([n,823543+n,1647086+n,2470629+n,3294172+n],:);
S_7a_p_n=S_7a_p([n,823543+n,1647086+n,2470629+n,3294172+n],:);

gamma_7a_n=pinv(V_7a_p_n)*S_7a_p_n;

mat_gamma_7a(n,:)=gamma_7a_n;
end

%Extracting wanted results for further calculation from the 1st to the 823543th row. For some reason the
%code gives 824516 rows. 
mat_gamma_7a_p=mat_gamma_7a(1:1:823543,:);

Answer 1

[Shadaab Siddiqie]

From my understanding you want to accelerate your code. Please vectorize your code. This reference might help you.

[Davide Bertoldi]

Hello, so I tried to vectorize the code. Below you can see the whole vectorized code instead of the for loop. What I couldn´t achieve until now is that the matrix V_7a_p and vector S_7a_p are calculated separately for every value of n_1. I need to perform such task because I need separated values for gamma_7a_n_span_1. Currently I cannot think of any other way than a for loop as shown above. How would you handle it?

%Strength values for CLT
G=0.69; %kN/mm²
E_0=11;
E_0_1_7=ones(1,7).*11;%kN/mm²
E_90=0.37; %kN/mm²

%Number of lamellas [-]
lam=3:2:7;

%Span length [m]
span=3.0:0.5:6.0;

%Analysed width [m]
width=1;

%Depth outer lamella parallel to span [mm]
lam_o_p=20:5:50;

%Depth inner lamella parallel to span [mm]
lam_i_p=20:5:50;

%Depth inner lamella cross to span [mm]
lam_i_c=20:5:50;

%Depth vector of CLT panels [mm] total number of possible combinations
depth_7a=combvec(lam_o_p,lam_i_p,lam_i_c,lam_i_p,lam_i_c,lam_i_p,lam_o_p)'; %2*lam_o_p+lam_i_c+lam_i_p+lam_i_c+2*lam_o_p

%Depth vector of CLT panels [mm] parallel to span [mm]
depth_7a_p=depth_7a(:,[1 2 4 6 7]);
depth_7a_p_1=depth_7a_p(:,1);
depth_7a_p_2=depth_7a_p(:,2);
depth_7a_p_3=depth_7a_p(:,3);
depth_7a_p_4=depth_7a_p(:,4);
depth_7a_p_5=depth_7a_p(:,5);

%Depth vector of CLT panels [mm] cross to span [mm]
depth_7a_c=depth_7a(:,[3 5]);

%Total depth of CLT panels [mm]
depth_7a_tot=sum(depth_7a,2); %sum(depth_x,2) where 2 stands for row-wise addition

%Determination of Ares CLT panel parallel to span [m²]
A_CLT_7a_p=10^-3.*depth_7a_p.*width;

%Determination of Ares CLT panel cross to span [m²]
A_CLT_7a_c=10^-3.*depth_7a_c.*width;

%Calculation position center of gravity for CLT_7a,CLT_7b [mm]
zs_CLT_7a=(A_CLT_7a_p(:,1).*(0.5.*depth_7a_p(:,1))+A_CLT_7a_p(:,2).*(depth_7a_p(:,1)+0.5.*depth_7a_p(:,2))+A_CLT_7a_c(:,1).*(depth_7a_p(:,1)+depth_7a_p(:,2)+0.5.*depth_7a_c(:,1))+A_CLT_7a_p(:,3).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+0.5.*depth_7a_p(:,3))+A_CLT_7a_c(:,2).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+depth_7a_p(:,3)+0.5.*depth_7a_c(:,2))+A_CLT_7a_p(:,4).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+depth_7a_p(:,3)+depth_7a_c(:,2)+0.5.*(depth_7a_p(:,4)))+A_CLT_7a_p(:,5).*(depth_7a_p(:,1)+depth_7a_p(:,2)+depth_7a_c(:,1)+depth_7a_p(:,3)+depth_7a_c(:,2)+depth_7a_p(:,4)+0.5.*(depth_7a_p(:,5))))./(A_CLT_7a_p(:,1)+A_CLT_7a_p(:,2)+A_CLT_7a_p(:,3)+A_CLT_7a_p(:,4)+A_CLT_7a_p(:,5)+A_CLT_7a_c(:,1)+A_CLT_7a_c(:,2)); %[mm]

%Distances between center of gravity and lamellas [mm]
a_1_7a=zs_CLT_7a-(0.5.*depth_7a_p(:,1));
a_2_7a=zs_CLT_7a-(depth_7a_p(:,1)+0.5.*depth_7a_p(:,2));
a_2_3_7a=zs_CLT_7a-(depth_7a_p(:,1)+depth_7a_p(:,2)+0.5.*depth_7a_c(:,1));%cross layer in between
a_3_7a=zeros(823543,1);
a_3_4_7a=-zs_CLT_7a+(depth_7a_p(:,4)+depth_7a_p(:,5)+0.5.*depth_7a_c(:,2));%cross layer in between
a_4_7a=-zs_CLT_7a+(0.5.*depth_7a_p(:,4)+depth_7a_p(:,5));
a_5_7a=-zs_CLT_7a+(depth_7a_p(:,5));

%Calculation of c-factors for matrix [kN/mm]
c_1_2=width*G./depth_7a_p_1.*10^3;
c_2_3=width*G./depth_7a_p_2.*10^3;
c_3_4=width*G./depth_7a_p_3.*10^3;
c_4_5=width*G./depth_7a_p_4.*10^3;
   
%Calculation of D-factors for matrix [kN/mm]
D_1=pi^2*E_0*width.*depth_7a_p_1./(span).*10^-3;
D_2=pi^2*E_0*width.*depth_7a_p_2./(span).*10^-3;
D_3=pi^2*E_0*width.*depth_7a_p_3./(span).*10^-3;
D_4=pi^2*E_0*width.*depth_7a_p_4./(span).*10^-3;
D_5=pi^2*E_0*width.*depth_7a_p_5./(span).*10^-3;

n=[(1:1:length(a_1_7a)),1]';
length_n=length(n);

%Matrix [5x5] where every element is a (823543x1) vector for span_1 to span_7
V_7a_p_span_1=[(c_1_2+D_1(1)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(1)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(1)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(1)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(1)).*a_5_7a];
V_7a_p_span_2=[(c_1_2+D_1(2)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(2)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(2)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(2)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(2)).*a_5_7a];
V_7a_p_span_3=[(c_1_2+D_1(3)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(3)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(3)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(3)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(3)).*a_5_7a];
V_7a_p_span_4=[(c_1_2+D_1(4)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(4)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(4)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(4)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(4)).*a_5_7a];
V_7a_p_span_5=[(c_1_2+D_1(5)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(5)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(5)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(5)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(5)).*a_5_7a];
V_7a_p_span_6=[(c_1_2+D_1(6)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(6)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(6)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(6)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(6)).*a_5_7a];
V_7a_p_span_7=[(c_1_2+D_1(7)).*a_1_7a -c_1_2.*a_2_7a zeros(823543,1) zeros(823543,1) zeros(823543,1); -c_1_2.*a_1_7a (c_1_2+c_2_3+D_2(7)).*a_2_7a -c_2_3.*a_3_7a zeros(823543,1) zeros(823543,1); zeros(823543,1) -c_2_3.*a_2_7a (c_2_3+c_3_4+D_3(7)).*a_3_7a -c_3_4.*a_4_7a zeros(823543,1); zeros(823543,1) zeros(823543,1) -c_3_4.*a_3_7a (c_3_4+c_4_5+D_4(7)).*a_4_7a -c_4_5.*a_5_7a; zeros(823543,1) zeros(823543,1) zeros(823543,1) -c_4_5.*a_4_7a (c_4_5+D_5(7)).*a_5_7a];

%S-Vector [(5x1] where every element is a (823543x1) vector
S_7a_p=[-c_1_2.*(a_2_7a-a_1_7a);-c_2_3.*(a_3_7a-a_2_7a)+c_1_2.*(a_2_7a-a_1_7a);-c_3_4.*(a_4_7a-a_3_7a)+c_2_3.*(a_3_7a-a_2_7a);-c_4_5.*(a_5_7a-a_4_7a)+c_3_4.*(a_4_7a-a_3_7a);c_4_5.*(a_5_7a-a_4_7a)];

n_1=(1:1:823543)';

V_7a_p_span_1_n=V_7a_p_span_1([n_1, length(n_1)+n_1, 2*length(n_1)+n_1,3*length(n_1)+n_1,4*length(n_1)+n_1],:);
S_7a_p_n=S_7a_p([n_1, length(n_1)+n_1, 2*length(n_1)+n_1, 3*length(n_1)+n_1, 4*length(n_1)+n_1],:);

gamma_7a_n_span_1=pinv(V_7a_p_span_1_n)*S_7a_p_n;                

[Shadaab Siddiqie]

You can still vectorize it. One example would be:

c_1_2=width*G./depth_7a_p_1.*10^3;
c_2_3=width*G./depth_7a_p_2.*10^3;
c_3_4=width*G./depth_7a_p_3.*10^3;
c_4_5=width*G./depth_7a_p_4.*10^3;
%insted of above code you can convert depth_7a_p_1,depth_7a_p_2,..,depth_7a_p_4 into a matrix
C = width*G./D.*10^3;
% Here D = [depth_7a_p_1 ; depth_7a_p_2 ;depth_7a_p_3 ;depth_7a_p_4]

Dont create depth_7a_p_1,depth_7a_p_2,..,depth_7a_p_4 at all create D directly from depth_7a_p.