CCS loads .bss section generated by GCC

I hearded from somebody that "GEL_EnableZeroFill(1)" can be used to enable the CCS to load the .bss section, but when I tried to add it to my GEL file and execute it, following error was reported by CCS:

C66xx_1: GEL: Error while executing OnTargetConnect(): 'GEL_EnableZeroFill()' not found

Thank you!

Hi Brighton,

Could you provide the OnTargetConnect() call that is calling GEL_EnableZeroFill(1)? You can just cut & paste the whole function here.

Thanks

ki

Ki,

GEL_EnableZeroFill(1) is the only call in the OnTargetConnect().

Thank you!

Brighton Feng said:

GEL_EnableZeroFill(1) is the only call in the OnTargetConnect().

hmmm... that should work.

But I am confused about another thing. You mentioned:

Brighton Feng said:

I am using ARM CortexA15 GCC compiler to generate .out file, and using CCS5.3 to debug it.

Brighton Feng said:

following error was reported by CCS:

C66xx_1: GEL: Error while executing OnTargetConnect(): 'GEL_EnableZeroFill()' not found

You have an executable for ARM Cortex A15 but you are calling GEL_EnableZeroFill on the C66x

ki

Ki,

I tried the GEL_EnableZeroFill() in the  OnTargetConnect() of GEL for both C66x and ARM CortexA15. It did not work in either case.

Thank you!

Brighton - after starting a debug session and connecting to the target, can you open the Scripting Console and type in:

js:> activeDS.expression.evaluate("GEL_EnableZeroFill(1)")

And let me know what it returns?

Thanks

ki

Brighton,

Which exact version of CCS are you using? This enhancement was included in CCS 5.3.242. Here is the CQ tracking this SDSCM00045176.

Mandeep

Ki this should be in 5.3 GA.

Mandeep

Ki and Mandeep,

Sorry, I made a typo. the CCS I used for this test in the desktop PC is CCS5.2.

CCS5.3.0.00090 is the one in my laptop.

Is there other way to workaround it in CCS5.2, most of our desktop PCs in lab are still using CCS5.2 or CCS5.1.

Thank you very much!

Brighton,

Sorry there are no work around for this in particular in earlier versions of CCS. Zero initialization should only apply to uninitialized global variables. If you initialize the data explicitly you should not need to depend on the loader to zero fill anything.

Mandeep

Hi, Mandeep,

That is true for TI's compiler.

When I use GCC, the problem I encoutered was, I defined a global variable with zero intialized, such as,

int a=0;

The GCC compiler puts this varialbe in .bss section, because the compiler assumes the .bss section will be zero filled and will be loaded.

However, the CCS did not load the .bss section, and the variable was not initialized with 0. So, after I saw many strange behaviors, then I realized that the variable was not initalized, and finanlly I found that is becuase the CCS does not load the .bss section.

Thank you very much!

Brighton,

You are correct about the behavior regarding TI's tool chain. The tool chain support 2 options for dealing with these types of variables ram and rom ( exposed via the build options ). When using the ram model TI's tool gen will do the initialization of a global in your example at run time ( c_init ); when using rom model you will see similar behavior to gcc. You can look in to seeing if GCC has a simlar switch/flag/option.

Mandeep