Make Error in Simulink – Division by Zero?

I looked up Make Errors here, and most of the issues were solved by calling 'mex -setup'. I gave that a shot, and everything looked fine. I'm using MATLAB R2011a Student Version along with its built-in lcc C Compiler.

My Simulink model consists of two MATLAB functions, some constants and 'From Workspace' sources. One of the MATLAB functions is 154 lines long and includes four extrinsic functions, all of which are declared as such at the beginning of the function. The other is basically a placeholder and passes the outputs as the same as inputs.

Here are the errors, buried in a long error report:

Error c2_cogen2_simulink.c: 5380  Division by zero 
Error c2_cogen2_simulink.c: 5383  Division by zero 
Error c2_cogen2_simulink.c: 5411  Division by zero   

Here's the full error report, in case it sheds more light on the issue:

    Making simulation target "cogen2_simulink_sfun", ... 

C:\Users\Zack\Documents\Italy Research\cogen1\slprj\_sfprj\cogen2_simulink\_self\sfun\src>"C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\bin\lccmake" -f cogen2_simulink_sfun.lmk  
Warning c1_cogen2_simulink.c: 482  static `void function(pointer to void,pointer to const incomplete struct mxArray_tag defined at C:\Program Files (x86)\MATLAB\R2011a Student\extern\include\matrix.h 298,pointer to const char,pointer to void) c1_c_sf_marshallIn' is not referenced 
Warning c1_cogen2_simulink.c: 455  static `pointer to const incomplete struct mxArray_tag defined at C:\Program Files (x86)\MATLAB\R2011a Student\extern\include\matrix.h 298 function(pointer to void,pointer to void) c1_c_sf_marshallOut' is not referenced 
0 errors, 2 warnings 
Error c2_cogen2_simulink.c: 5380  Division by zero 
Error c2_cogen2_simulink.c: 5383  Division by zero 
Error c2_cogen2_simulink.c: 5411  Division by zero 
Warning c2_cogen2_simulink.c: 6780  static `void function(pointer to void,pointer to const incomplete struct mxArray_tag defined at C:\Program Files (x86)\MATLAB\R2011a Student\extern\include\matrix.h 298,pointer to const char,pointer to void) c2_u_sf_marshallIn' is not referenced 
Warning c2_cogen2_simulink.c: 6753  static `pointer to const incomplete struct mxArray_tag defined at C:\Program Files (x86)\MATLAB\R2011a Student\extern\include\matrix.h 298 function(pointer to void,pointer to void) c2_u_sf_marshallOut' is not referenced 
Warning c2_cogen2_simulink.c: 6671  static `double function(pointer to SFc2_cogen2_simulinkInstanceStruct,double) c2_sign' is not referenced 
Warning c2_cogen2_simulink.c: 6570  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,int,double,int,int,pointer to double,int,int,pointer to double,pointer to int,int,int,pointer to double,pointer to int) c2_ceval_xger' is not referenced 
Warning c2_cogen2_simulink.c: 6524  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,int,double,pointer to double,pointer to int,int,int,pointer to double,pointer to int,int,int,double,pointer to double,int,int,pointer to double) c2_ceval_xgemv' is not referenced 
Warning c2_cogen2_simulink.c: 6480  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,int,pointer to double,pointer to int,int,int,pointer to double,pointer to int,int,pointer to double,pointer to double) c2_eml_blas_xgemv' is not referenced 
Warning c2_cogen2_simulink.c: 6452  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,int,int,double,pointer to double,pointer to int,int,int,pointer to double,pointer to double,pointer to int,pointer to double) c2_eml_matlab_zlarf' is not referenced 
Warning c2_cogen2_simulink.c: 6443  static `double function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,double,double,int,int) c2_b_ceval_xscal' is not referenced 
Warning c2_cogen2_simulink.c: 6434  static `double function(pointer to SFc2_cogen2_simulinkInstanceStruct,double,double) c2_b_eml_xscal' is not referenced 
Warning c2_cogen2_simulink.c: 6425  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,double,double,pointer to double,pointer to double,pointer to double) c2_b_eml_matlab_zlarfg' is not referenced 
Warning c2_cogen2_simulink.c: 6403  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,double,pointer to double,pointer to int,int,int,pointer to double,pointer to int) c2_ceval_xscal' is not referenced 
Warning c2_cogen2_simulink.c: 6383  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,double,pointer to double,pointer to int,int,pointer to double,pointer to int) c2_eml_xscal' is not referenced 
Warning c2_cogen2_simulink.c: 6331  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,double,pointer to double,pointer to int,int,pointer to double,pointer to double,pointer to int,pointer to double) c2_eml_matlab_zlarfg' is not referenced 
Warning c2_cogen2_simulink.c: 6309  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,pointer to double,pointer to int,int,int,int,int,pointer to double,pointer to int) c2_ceval_xswap' is not referenced 
Warning c2_cogen2_simulink.c: 6287  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,pointer to double,pointer to int,int,int,int,int,pointer to double,pointer to int) c2_eml_refblas_xswap' is not referenced 
Warning c2_cogen2_simulink.c: 6265  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,pointer to double,pointer to int,int,int,int,int,pointer to double,pointer to int) c2_eml_blas_xswap' is not referenced 
Warning c2_cogen2_simulink.c: 6245  static `void function(pointer to SFc2_cogen2_simulinkInstanceStruct,int,pointer to double,pointer to int,int,int,pointer to double,pointer to int) c2_eml_xswap' is not referenced 
3 errors, 17 warnings 
    "C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\bin\lcc.exe" -c -Zp8 -DMATLAB_MEX_FILE -noregistrylookup -I"C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\extern\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\simulink\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\mex\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\debugger\include"  "cogen2_simulink_sfun.c" 
    "C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\bin\lcc.exe" -c -Zp8 -DMATLAB_MEX_FILE -noregistrylookup -I"C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\extern\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\simulink\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\mex\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\debugger\include"  "cogen2_simulink_sfun_registry.c" 
    "C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\bin\lcc.exe" -c -Zp8 -DMATLAB_MEX_FILE -noregistrylookup -I"C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\extern\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\simulink\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\mex\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\debugger\include"  "c1_cogen2_simulink.c" 
    "C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\bin\lcc.exe" -c -Zp8 -DMATLAB_MEX_FILE -noregistrylookup -I"C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\extern\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\simulink\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\mex\include"  -I"C:\Program Files (x86)\MATLAB\R2011a Student\stateflow\c\debugger\include"  "c2_cogen2_simulink.c" 
C:\Program Files (x86)\MATLAB\R2011a Student\sys\lcc\bin\lccmake: Error code 1

Thanks for any help you can give.

Zack

Answer 1

[Zachary Santer]

Turns out, trying to use 'load' and 'save' as extrinsic functions in a MATLAB Function Block in Simulink doesn't work.

That situation has been replaced by global variables, Data Store Memory Blocks, and an If Action Subsystem. The If Action Subsystem has variables being loaded from the workspace through Constant Blocks and loaded into Data Store Write Blocks.

Everything works fine, now.

Thanks anyway,

Zack

[Kaustubha Govind]

Thanks for posting the solution!

Answer 2

[Fangjun Jiang]

I assume you understand what is division by zero error. Every time you do a division, in C or in MATLAB, the safest way is to check whether the denominator is zero.

if abs(b)<eps
% do something, such as assign b or c to a fixed number
else
c=a/b;
end

So check the code in your MATLAB function to see if you have such operation and if you have the safeguard.

In Simulink, you can turn on the diagnositcs. Under menu Simulation->Configuration Parameters...->Diagnostics-> Data validity, turn on the check for "Signal"->"Inf or NaN block output". So when you run simulation, hopefully it can help you find out these division by zero errors early on so you don't have to deal with it until generating code and compiling it.

[Kaustubha Govind]

You beat me to it, Fangjun. :)

@Zachary: Please also follow Fangjun's advice to verify if the problem isn't indeed with your code.

[Zachary Santer]

I added the little if abs(var)<eps in one spot where the code could have hypothetically had this type of issue. I also turned on the diagnostics you mentioned.

I still have the same issue.

I actually opened the .c file that had the issues, in Notepad, and found the following on lines 5387 to 5421.

if (c2_xi > 0.0) {

c2_n = 5;

c2_r = c2_xi / 0.0;

} else if (c2_xi < 0.0) {

c2_n = 1;

c2_r = c2_xi / 0.0;

} else {

c2_b_xi = c2_xi;

c2_low_i = 1U;

c2_low_ip1 = 1U;

c2_high_i = 6U;

while (c2_high_i > c2_low_ip1 + 1U) {

c2_b_low_i = c2_low_i;

c2_b_high_i = c2_high_i;

c2_mid_i = (c2_b_low_i + c2_b_high_i) >> 1U;

(uint32_T)_SFD_EML_ARRAY_BOUNDS_CHECK("", (int32_T)c2_mid_i, 1, 6, 1, 0);

if (c2_b_xi >= 0.0) {

c2_low_i = c2_mid_i;

c2_low_ip1 = c2_low_i;

} else {

c2_high_i = c2_mid_i;

}

}

c2_u1 = c2_low_i;

if (c2_u1 > 2147483647U) {

c2_u1 = 2147483647U;

}

c2_n = (int32_T)c2_u1;

c2_a = c2_n;

c2_c = c2_a;

_SFD_EML_ARRAY_BOUNDS_CHECK("", c2_c + 1, 1, 6, 1, 0);

c2_r = c2_xi / 0.0;

}

I don't know much about C. Nor do I know anything about what Simulink does to create C code. Nothing in there looks anything like my MATLAB code, though.

[Fangjun Jiang]

Wow, the generated C code actually contains "c2_r = c2_xi / 0.0;"! I am speechless! Without seeing your MATLAB function, I don't think there is much to help. You could post your M code here, or contact TWM tech support for help.

[Walter Roberson]

If I read the filename correctly, it seems more likely that it is the simulink portion that is generating the division by 0. But only more likely, not a certainty.

[Fangjun Jiang]

Yes. The C code is generated but the source my be in the M code, or could be the problem of the code generation.

[Zachary Santer]

Okay. I examined my M code carefully, found a couple of issues and fixed them. Running the simulation still gives the same error. I built the same simulation in another window, copying and pasting my M code, thinking that could do some good. I still got the same error.

Here's my M code. Sorry that it's going to look cruddy here.

function [command_el_MCI,MCI_grid_out] = ...

MCI(MCI_th_set,MCI_el_set,simulation_time,water_temp_in,water_temp_out,...

water_M_dot,Pel,m_dot_fuel)

%#codegen

persistent MCI_grid i predictor_time intercept_time corrector_count ...

corrector_wait corrector_coeff time_vector Pth_vector saved_time ...

Pel_vector m_dot_fuel_vector water_data max_index step_time stop_time

coder.extrinsic('load','save','now','medfilt1')

world_time = 0;

world_time = now;

%initialize everything

if isempty(MCI_grid)

%doing this before loading the variables from a file insures they are

%in a Simulink-compatible format

MCI_grid = zeros(16,5,26);

corrector_coeff = 1;

water_data = zeros(6,3);

step_time = 1;

stop_time = 30;

load('MCI_data.mat')

% load('MCI_grid.mat')

i = 0;

predictor_time = 180; %s

intercept_time = 90; %s

corrector_count = 0;

corrector_wait = 90; %s must be less than predictor time

% load('corrector_coeff.mat')

max_index = ceil(max([predictor_time intercept_time corrector_wait])/step_time);%/step_time

% load('water_data.mat')

time_vector = zeros(1,max_index);

Pth_vector = time_vector;

Pel_vector = time_vector;

m_dot_fuel_vector = time_vector;

saved_time = world_time;

end

hour_index = floor(rem(world_time,1)*24) + 1;

hour_part = rem(world_time*24,1);

if hour_part < 0.5

early_weight = 0.5 - hour_part;

center_weight = 0.5 + hour_part;

late_weight = 0;

% elseif hour_part == 0.5 redundant

% early_weight = 0;

% center_weight = 1;

% late_weight = 0;

else

early_weight = 0;

center_weight = 1.5 - hour_part;

late_weight = -0.5 + hour_part;

end

command_el_MCI_early = interp1(MCI_grid(:,4,hour_index-1),MCI_grid(:,3,hour_index-1),MCI_th_set,'linear','extrap');%point

command_el_MCI_center = interp1(MCI_grid(:,4,hour_index),MCI_grid(:,3,hour_index),MCI_th_set,'linear','extrap');

command_el_MCI_late = interp1(MCI_grid(:,4,hour_index+1),MCI_grid(:,3,hour_index+1),MCI_th_set,'linear','extrap');

command_el_MCI = command_el_MCI_early*early_weight + command_el_MCI_center*center_weight + command_el_MCI_late*late_weight;

%assumed temperature inputs are in C, water volumetric flow input is in L/sec

%water pressure passing through the engine is 2 Bar. The data used here is

%for 1 Bar. Fix this.

H_in = interp1(water_data(:,1),water_data(:,2),water_temp_in,'linear','extrap'); %kJ/kg

H_out = interp1(water_data(:,1),water_data(:,2),water_temp_out,'linear','extrap'); %kJ/kg

water_temp_ave = (water_temp_in + water_temp_out)/2; %degC

water_density = interp1(water_data(:,1),water_data(:,3),water_temp_ave,'linear','extrap');%kg/m^3

Pth = water_M_dot*water_density*0.001*(H_out - H_in);%kW

if i < max_index

i = i + 1;

time_vector(i) = simulation_time;

Pth_vector(i) = Pth;

Pel_vector(i) = Pel;

m_dot_fuel_vector(i) = m_dot_fuel;

data_length = i;

else

time_vector(1:max_index-1) = time_vector(2:max_index);

time_vector(max_index) = simulation_time;

Pth_vector(1:max_index-1) = Pth_vector(2:max_index);

Pth_vector(max_index) = Pth;

Pel_vector(1:max_index-1) = Pel_vector(2:max_index);

Pel_vector(max_index) = Pel;

m_dot_fuel_vector(1:max_index-1) = m_dot_fuel_vector(2:max_index);%point

m_dot_fuel_vector(max_index) = m_dot_fuel;

data_length = max_index;

end

%This part finds whether or not the MCI_grid predicted the current response

%correctly. If it did not, a correction factor is found and multiplied into

%the result from MCI_grid.

if simulation_time > predictor_time

cutoff = simulation_time - predictor_time;

ii = i;

while time_vector(ii)>cutoff

ii = ii - 1;

end

th_predictor = polyfit(time_vector(ii:i),Pth_vector(ii:i),1);

th_predicted = th_predictor(1)*(simulation_time + intercept_time) + th_predictor(2); %check this

th_delta_predicted = th_predicted - MCI_th_set;

th_delta = Pth - MCI_th_set;

if sign(th_delta_predicted) == sign(th_delta)

corrector_count = corrector_count + 1;

corrector_index = data_length - corrector_count;

if (simulation_time - time_vector(corrector_index)) > corrector_wait

if abs(Pth) < eps

corrector_count = 0;

else

correction = MCI_th_set/Pth;

corrector_coeff = corrector_coeff*correction;

corrector_count = 0;

end

end

else

corrector_count = 0;

end

end

%make routine to replace elements in MCI_grid

%use medfilt1

if i >= max_index %consider putting this under the else statement of the if structure begun on line 34

filter_order = 5; %consider other n's than 5

Pth_vector_filtered = zeros(size(Pth_vector));%point 'double'

Pel_vector_filtered = Pth_vector_filtered;

m_dot_fuel_vector_filtered = Pth_vector_filtered;

Pth_vector_filtered = medfilt1(Pth_vector,filter_order);

Pel_vector_filtered = medfilt1(Pel_vector,filter_order);

m_dot_fuel_vector_filtered = medfilt1(m_dot_fuel_vector,filter_order);

for MCI_grid_step = 1:length(MCI_grid(:,1,hour_index))

for filtered_vector_step = floor(filter_order/2):(max_index-floor(filter_order/2))

if (Pth_vector_filtered(filtered_vector_step) > (MCI_grid(MCI_grid_step,2,hour_index) - 0.5)) && (Pth_vector_filtered(filtered_vector_step) < (MCI_grid(MCI_grid_step,2,hour_index) + 0.5)) && ((MCI_grid(MCI_grid_step,1,hour_index) + (23/24)) < world_time)

MCI_grid(MCI_grid_step,1,hour_index) = world_time - (time_vector(max_index) - time_vector(filtered_vector_step))/86400; %serial date number of the time data was taken

MCI_grid(MCI_grid_step,3,hour_index) = Pel_vector_filtered(filtered_vector_step); %kW electric

MCI_grid(MCI_grid_step,4,hour_index) = Pth_vector_filtered(filtered_vector_step); %kW thermal power

MCI_grid(MCI_grid_step,5,hour_index) = m_dot_fuel_vector_filtered(filtered_vector_step); %l/s fuel

if hour_index == 2

MCI_grid(:,:,26) = MCI_grid(:,:,2);

elseif hour_index == 25

MCI_grid(:,:,1) = MCI_grid(:,:,25);

end

end

end

end

end

MCI_grid_out = MCI_grid;

if (simulation_time == stop_time) || ((saved_time + 4/24) < world_time);

save('MCI_data.mat','MCI_grid','corrector_coeff','-append')

% save('MCI_grid.mat','MCI_grid')

% save('corrector_coeff.mat','corrector_coeff')

saved_time = world_time;

end

Answer 3

[Kaustubha Govind]

There seems to be have been a similar bug before. See Bug#534529. However, since it was fixed in R2009b, it is likely not the same issue.

You could however try the workaround suggested that Bug Report and see if it works for you:

1. In the Simulink Editor, select Simulation > Configuration Parameters.
2. In the Configuration Parameters dialog box, select Optimization.
3. Change the Loop unrolling threshold value to 1, or to any number smaller than the length of the vector passed to norm.

If not, I would recommend reporting this to MathWorks Tech Support to see if this really is another bug.

[Zachary Santer]

R2011a apparently doesn't have a 'Loop unrolling threshold' option in the Configuration Parameters dialog box. Additionally, I am not using 'norm.'

If I don't find a solution in the next day or two, I will go ahead and report this to Tech Support.

Thanks to both of you for your help.