In my application, I am trying to speed up the code execution in the CPU1 CLA.
Following the cla_iir2p2z_cpu01 example, I initially had all variables as floats.
However, there are some flag variables that do not need to be 32-bit wide floats.
I have changed those to "unsigned int" and the program stopped working.
It was a simple change in the shared struct.
====
typedef struct{
float cla_sin_theta;
float cpu2cla_d;
// unsigned int cpu2cla_SP; <-- does not work
// unsigned int cpu2cla_SN; <-- does not work
float cpu2cla_SP; <-- works OK
float cpu2cla_SN; <-- works OK
} CPU2CLA ;
extern CPU2CLA cpu2cla; // in the shared header file
====
in a CPU1 .c file:
#pragma DATA_SECTION(cpu2cla,"CpuToCla1MsgRAM")
CPU2CLA cpu2cla;
=====
The SP and SN flags are then copied into an internal structure:
====
in cla1.cla file:
conv_sim.d = cpu2cla.cpu2cla_d; // input
conv_sim.SN = cpu2cla.cpu2cla_SN; // input .... works when both FLOATs but not when both UNSIGNED INTs
conv_sim.SP = cpu2cla.cpu2cla_SP; // input .... works when both FLOATs but not when both UNSIGNED INTs
...
in the shared header file:
float Tsw;
float Ts_step_to_Tsw_ratio;
float theta_increment;
// unsigned int SN; <-- does not work
// unsigned int SP; <-- does not work
float SN; <-- works OK
float SP; <-- works OK
} CONV_SIM;
====
in a .c file:
#pragma DATA_SECTION(conv_sim,"CLADataLS0")
CONV_SIM conv_sim;
====
CPU1 writes the correct values to cpu2cla.cpu2cla_SN and cpu2cla.cpu2cla_SP but these values are not reflected in the CLA-only structure CONV_SIM
Is there any limitation on mixing ints and floats in shared structures between CPU1 and CLA?
