CCS/SIMPLELINK-CC2640R2-SDK: Faster Debug Stepping?

Markel Robregado said:

 I would like to know what is the requirement to achieve faster debug stepping at CCS.

Therefore, I tried my own measurements using CCS 7.1.0.00016 and a CC2538F53, under Windows 7. The test involved using a GEL script to step the target a number of instructions, where the script issued the next GEL_AsmStepInto() as soon as the target had halted after the previous instruction. The elapsed time was reported, which allowed the average time to step one instruction to be reported - which a combination of the emulator and CCS debugger performance.

The GEL code used was:

menuitem "Stepping Test";

unsigned int remaining_instructions = 0;

OnHalt ()
{
    if (remaining_instructions > 0)
    {
        remaining_instructions--;
        if (remaining_instructions > 0)
        {
            GEL_AsmStepInto ();
        }
        else
        {
            GEL_System ("echo %TIME%");
            GEL_TextOut ("Step complete\n");
        }
    }
}

start_stepping ()
{
    GEL_System ("echo %TIME%");
    GEL_TextOut ("Started stepping %d instructions\n",,,,,remaining_instructions);
    GEL_AsmStepInto ();
}

hotmenu STEP_10_instructions()
{
    remaining_instructions = 10;
    start_stepping ();
}

hotmenu STEP_100_instructions()
{
    remaining_instructions = 100;
    start_stepping ();
}

hotmenu STEP_1000_instructions()
{
    remaining_instructions = 1000;
    start_stepping ();
}

hotmenu STEP_10000_instructions()
{
    remaining_instructions = 10000;
    start_stepping ();
}

The results of the test are:

Notes:

a) The thread CCS/TMDSEMU110-U: Confirm the specification for TCLK speed says the maximum XDS110 TCLK frequency is 2.5 MHz, which was the frequency used in the above tests.

b) The Segger J-Link supports JTAG or SWD, and the CC2538 supports JTAG or cJTAG so JTAG was the only target interface which could be tested.

c) I also tried using a Blackhawk USB560-M but couldn't get the test connection to work, and so couldn't take any performance measurements.

The conclusions of the tests are:

1) When using a XDS110 the step time can be up to twice as fast when using 4-wire JTAG or cJTAG compared to 2-wire cJTAG. This is probably because 2-wire cJTAG requires 3 times more TCLKs per operation than 4-wire modes.

2) A Segger J-Link using 4-wire JTAG can be up to 60 times faster than a XDS110 using 4-wire mode, or up to 120 times faster than a XDS110 using 2-wire cJTAG mode.

I can't guarantee you would see the same speed-up in your environment, but it does show a wide range of step times for different debug probes / configurations.