This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

What does the "Others" value in CCS profiling data actually count ?

Hi,

I am trying to get the profile data for the entire code I am developing for C64x+ DSP. This code contains TI-VLIB API calls and hand-made C functions for image processing. Following the procedures for simulator profiling in ATK v1.3 guide (focus.ti.com/lit/ug/spru623d/spru623d.pdf ), I got profiling data. 

In the profiling summary (excel spreadsheet), I found that most of the cycle.CPU is taken by the "Others" (80~90 %). According to the ATK v1.3 guide, "Others" shows the counts incurred in the code with no debug information. I tried both Debug and Release modes when I build the code but I got similar results. Does anybody know what the "Others" acually count ? Can we estimate which part of the code contribute most highly to the "Others" value ?

[I am using CCS3.3.38.2. I selected C64x+ CPU Cycle Accurate Simulator in CCS set-up, checked both 'Collect app. level profile for total cycles and code size' and 'Collect code coverage and excl. profile data', enabled TI Lib in profile set-up.]

Any help would be very appreciated.

John

 

 

  • As you suggest, others encompasses the cycles that do not contain debug information, thus the profiler could not collect any more specific information. To be more specific, any time you are running code that was not built with debugging enabled you will end up with others, this will typically include libraries you are linking in which were built for release distribution and had debug information removed. If you had the source than you could get much more granular information from the profiler, however with just a prebuilt library that has no debug information you are limited like this.

  • Bernie,

     

    Thank you for your reply !  As I really need to achieve real-time (at least 15 frames/sec) performance for my current code, I have to identify which module is bottleneck (in the view of CPU cycles). Does TI always provide release mode VLIB ? Is there any way I could get pre-built TI-VLIB library that has debug information ?

    Thanks.

     

    John

  • Unfortunately I have not had a chance to work with VLIB myself, and I have not seen the source to it. Your best bet for this would probably be your local TI contact.

  •  

    OK. I will contact TI regarding debug-mode VLIB. Thank you for your help !!