C Code taking too much time

Usman,

Can you share more details about the environment in which you are running the code before and after? Are you using an operating system, etc?

yp I am using DSP-BIOS and using software, hardware interrupts and tasks in my code. And the specific code section is managed by task. 

Usman,

What was your previous environment, where you were getting better performance? How much time does your algorithm takes now, and before? Give as much details as you can, then it will be easier to understand. 

I am not finding much difference in both the environments except that the new one is more complex in terms of code size and functionality. Initially my code was taking about 0.6msec but now it is taking almost 2.3msec. I am also using dsplib64plus and its function DSP_fir_gen multiple times and have observed that it is also taking considerable time from the previous environment.

More specifically what can be the causes of different execution time of calling same library function in two different environments.

Usman,

First you've to check the clock frequency, DDR clock frequency etc. Since you are not giving an specific details about the setup I can't give any specific step to try out. 

In both the environments Clock frequency and DDR clock frequency is same.

Clock frequency is 459MHz.

DDR clock frequency is 126MHz.

Hi Usman,

Have you checked your build configuration? Do you see any difference like Debug Vs Release, optimization level etc?

I assume no other program is running other than your C code in both the environments, is that correct?

Optimization settings in both the environments are same. Yes no other program is running beside the change that i have mentioned that timing is increased after putting my module in big framework and built in library functions are also taking unusually larger time from the previous environment.

Hi Usman,

when you say bigger framework and more complex, what does that mean? 

The reason behind I asked for other threads running because, it may so happen that the newer framework might have threads having more access to DDR causing stalls in your program causing increase in access time and eventually affecting your program performance. I am just guessing based on whatever inputs you have provided.

Bigger framework contains other code sections and threads linked to hardware interrupts but during calculation of cycles consumed in my code section using DSP-BIOS function CLK_gethtime i disable all the hardware interrupts and hence all the other threads. I too think it is some kind of memory related problem but where do i start ?

Thanks

Radhesh , Ranjith thanks

Thanks for your time. Some of my variables were declared on-chip which was not the case in the 2nd environment and that's why i was paying timing penalty. Now the problem is solved  :)