CODECOMPOSER: Adding a data watchpoint in CCS's debugger and changing compilation optimization

Hello Ki,

thank you for your response. As suggested, I did follow the manuals instructions but I am unable to set activate the watchpoint set at my variable of interest. I keep getting a warning: ' "numClusterEvents" does not resolve to an address'. Please take a look: 

Concerning the optimization, the provided link discusses the overall principle and tradeoffs - that I do know. What I would like to get is guidance how to change the optimization for the whole project in CCS.

Thank you for your help.
Best regards,
Mark

Hello,

Mark Passia said:

I keep getting a warning: ' "numClusterEvents" does not resolve to an address'

You will need to enter an expression that resolves to an address. Or simply specify the exact address.

Mark Passia said:

how to change the optimization for the whole project in CCS.

Check the compiler build properties:

https://software-dl.ti.com/ccs/esd/documents/users_guide_12.8.0/ccs_project-management.html#build-properties

Hello Ki,

thanks for your swift answer!

I managed to resolve the optimization level. I have set it to O0.

Concerning the watchpoint - it would be great if I could get some more assistance with this.

I know that inst->params.allocationParams.maxVelThre is written to upon initialization within mmw_load_params.c, so to my understanding if I set a watchpoint on it inside gtrack_module.c, which is used on a per frame basis. Then the program should stop when the parameter is initialized within mmw_load_params.c, is that correct? 

I did do what I described above, but the program did not halt it's execution at all - why is that?

You also write that I can simple specify the address that I want to watch, how can I do that?

Thanks for your help!
Best regards,
Mark

Mark Passia said:

You also write that I can simple specify the address that I want to watch, how can I do that?

In the Breakpoints view:

https://software-dl.ti.com/ccs/esd/documents/users_guide_12.8.0/ccs_debug-main.html#breakpoints-view

You can use the option to add a new hardware watchpoint:

A dialog will appear where you can explicitly specify an absolute address:

Please set on the absolute address and see if that helps. Open the address in the Memory view to check the contents.

Thanks

ki

Hi,

thanks for your answer. Unfortuneatly I still have some issue, here is what I do step by step:

1.) Set a breakpoint within a function (, in my case gtrack_step.c).

2.) When the program halts I use the "Expressions" view to see the address of interest, in my case it is a member of a struct:

inst = (GtrackModuleInstance *)handle;
inst->wasAllocateCalled = 0;

3.) I right click on its field in the watchpoint tab and select hardware watchpoint, which creates in the "Breakpoints" tab a new breakpoint. In my case at the address 0x102230D0. This corresponds to the address of the variable of interest as seen in the "Expressions" tab.

4.) I click step return or resume to continue the programs execution so that the code flows as during normal operations.

...But my watchpoint never halts the execution of the program despite me seeing that it's watched value gets toggled.

So my question is what am I doing wrong?

Can you show me the properties of the watchpoint you set? Please expand all the nodes so that they are visible:

https://software-dl.ti.com/ccs/esd/documents/users_guide_12.8.0/ccs_debug-main.html#breakpoint-properties

Hello Ki,

The action is "Remain Halted", as shown here in the breakpoints properties:

Moreover to make sure that I am not doing some stupid mistake, please take a look at this short screen capture. Note that gtrack_step.c is called before gtrack_module.c. In gtrack_step.c the value of inst->wasAllocateCalled = 0; is set to 0 and inside of gtrack_module.c it is set to 1, which happens every frame. My current frame rate is 250ms.

Thank you for the video. I don't see anything wrong with the watchpoint setup. I was testing watchpoints with various targets I have and it worked. However I did not try with the AWR2944. I will need to request a board and test it on this specific target.

Hello Ki,

Ki said:

However I did not try with the AWR2944. I will need to request a board and test it on this specific target

please note that the variable that I tested in this video is a variable added by me to the source code files. However, this one inst->params.allocationParams.maxVelThre is used by the original source code. Also note that there is a bug in TIs HECR demo as resolved here:  RE: AWR2944EVM: HECR demo tracker issues (, incl. documentation discrepancy)  

Can you get the watchpoint to trigger at all with any variable/address?

Can you also provide your programs? I have the toolbox and SDK installed but I am having issues building the examples. Instead of troubleshooting the build errors (I am a tools experts but not a device/SDK expert) it may be easier to use the program you are using. You can share privately via private E2E message.

Hello Ki,

I did try setting a breakpoint to other address locations, I cannot get this to work at all. For example I tried to set a watchpoint on numObjOut inside the function MmwDemo_handleObjectDetResult - this will change every frame. Moreover it is in the demo's mms_main.c file by default.

Concerning the program - yes I can. But I don't think that it changes much when it comes to the debugger. I was also unable to set the watchpoint in a straight out of the SDK HECR demo. 

For now maybe try it out on the default HECR demo? I am using radar_toolbox_3_20_00_04. As mentioned the tracker is not initialized properly. But this doesn't matter for setting breakpoints/watchpoints within the scope of the mss_main.c file. 

Best regards,

Mark

Mark Passia said:

For now maybe try it out on the default HECR demo? I am using radar_toolbox_3_20_00_04

Is that the awr2944_corner_radar projects? These seem to build ok. I will try with these. It is a local holiday today so I will give this a try tomorrow when I am back in the office.

Thanks

ki

Yes exactly, I am working with this demo: 

C:\ti\radar_toolbox_3_20_00_04\source\ti\examples\Automotive_ADAS_and_Parking\high_end_corner_radar\src

They do build out of the box, but the tracker doesn't initialize properly.

Enjoy your holiday!

Best,

Mark

Thanks, I did build and run those example and applied your fixes from the other thread. For some reason it doesn't appear to run correctly however. I'll need to investigate this further.

However I did a basic test of watchpoints on the R5 and it worked for me. Can you try this extremely simple test:

Create a global integer variable. Then in the source, access it somehow. Then build/load the program and set a watchpoint that will trigger a halt on ANY access. See if it triggers a halt.

An example of this is:

When line 3113 is executed, the watchpoint would detect the read access at &gVar and trigger a halt.

Hello Ki,

I did as instructed, the watchpoint worked:

Basic watchpoint functionality seems to work. I tried with both Memory for Write and Any and both seemed to work as expected. The question is why it wasn't working for the variables in question for you. Can you try "Any" for memory and see if at least that will trigger for anything?

Hello Ki, 

did you try to do it loc a function scope variable? I tried to do it as in this shown example but it doesn't work.

Maybe I don't understand something correctly, but to my knowledge as this variable has been initialized it should have the same in memory location.

Regards,

Mark