benchmarking profiling code on C6424 hardware

Hi farhad farahani,

All of SYS/BIOS time benchmarks are done using timing APIs.

e.g.

Bench_calib();

Bench_begin();

<code to benchmark>

Bench_end();

farhad farahani said:

I am not interested in manually putting timers around all my functions so please do not suggest that. 

These methods would certainly require you to update your code.  I'm not sure if this is what you are trying to avoid in the above statement or not, but this is how it is done for BIOS.  If you are interested in the same method, I can provide you the details of the above macros so that you could use them in your application.

Steve

OK. Here is the situation: My application code consists of thousands of functions. There are system stuff, audio coding, audio decoding, audio post processing, etc going on and I am sitting at 90% CPU usage. I am interested to figure out which functions are eating the MIPS.  People call this different names : statistical profiling, call graph profiling, etc. I have always done this easily on other platforms like linux/intel, VDK/Sharc or VXworks/ARM etc. Somehow I can not find out how to do this on TI/SYSBIOS. 

If I put Bench_begin() and Bench_end() around my main() is it gonna tell me how much each function ( using the symbols ) inside my program takes up ? I am guessing not ? Is there a way to automatically instrument all the function entry and exit points and post process every thing ? 

To clarify, I am looking for a result like this :

  %   cumulative   self              self     total           
 time   seconds   seconds    calls  ms/call  ms/call  name    
 19.48    251.46   251.46   203160     1.24     1.24  Radix2FFT::FFT(double*, double*)
 16.91    469.79   218.33    25395     8.60    19.34  Upmixer::CalculateBand()
 13.00    637.65   167.86    25395     6.61     6.61  PeakLimiter::Process(double**, int, int)
 11.57    787.08   149.43   152370     0.98     2.31  Synthesis::Transform(Complex*, double*)
  6.54    871.47    84.39  1168170     0.07     0.12  BandCalculator::ComputeFrontStats(int, Complex**)
  4.21    925.88    54.41    25395     2.14     2.14  Upmixer::ApplyHSFilters()
  4.15    979.44    53.56    50790     1.05     2.29  Analysis::Transform(double*, Complex*)
  3.79   1028.43    48.99  7009020     0.01     0.01  RSmoothMatrix(double**, double**, double, int, int)
  2.94   1066.40    37.97  1168170     0.03     0.06  BandCalculator::ComputeAllLRStats(int, Complex**)
  2.91   1104.03    37.63                             main