Opened 13 years ago

Last modified 13 years ago

#1574 closed defect (fixed)

Parameter matrix used in equations generate wrong code!

Reported by: adrpo Owned by: adrpo
Priority: high Milestone:
Component: Simulation Code Generation Version:
Keywords: Cc: adrpo, sjoelund.se

Description

This model:

model petrou
 parameter Real A[14,14] = zeros(14,14);
 parameter Real F[14,4] = ones(14,4) * 10.5;
 // make it 12 to be sure we don't run out of bounds
 parameter Real u[12,4] = ones(12,4) * 1.5;

 Real x[14];
 // Real v[4];
 Integer i(start = 0);
equation
 i = integer(floor(time + 1));
 der(x) = A*x + F* u[i]; // if you use v here and un-comment below and above is OK
 // for j in 1:4 loop v[j] = u[i,j]; end for;
end petrou;

generates code such as:

#define $Pu localData->parameters[252]
...
*real_array_element_addr(&$Pu, 2, (modelica_integer)$Pi, (modelica_integer) 4)

which is wrong as localData->parameters[252] is not a real_array!

Change History (3)

comment:1 Changed 13 years ago by adrpo

No idea on how to solve this, maybe we could create an array from the parameter array
before using it.

comment:2 Changed 13 years ago by sjoelund.se

I will clean up the code generation for rhs crefs a bit... It looks like a mess right now :(

comment:3 Changed 13 years ago by sjoelund.se

Ok, too much of a mess to clean it up. I made it more complex instead :)
We now address constant dimensions (i.e. simulation context) and expression subscripts directly on the (double*) for performance reasons (yes, for performance despite this being the only testcase for this)...

Note: See TracTickets for help on using tickets.