Part Number: F29H850TU
I have been trying to read data from the F29x processor while it is running. I have a F29x controlCard with a USB wire connected to it from my PC. It has a XDS110 debug chip.
I use memory.read() JS scripting API available in CCS.
What I have observed with it is it takes a minimum of around 1.5ms to run and provide me data, regardless of the datatype of the variable. I profiled it using performance.now() scripts available as JS module.
Note that I am trying to read data while the processor is running code, without halting it using breakpoints. It does read it properly, but there is a clear bottleneck here with XDS110 probe chip, in that the speed of data reads is relatively slow, compared to the ARM CoreSight tech for ARM chips, wherein J-Link segger with SWO has about 1.5Mb/s throughput of the processor reads without halting it.
I improved memory.read() timing marginally to about 0.9ms by increasing the JTAG clock frequency to a maximum of 14MHz in the ccxml in CCS 2050 version, however, it is not significant.
Is there anything I am missing that can improve speed of data reads from the processor?
Here is an use case: if I want to read ADC data, at say 10kHz fro the AdcResultRegs, the debugSession is simply not capable enough to do this with the current speeds I see. I could read it in batches yes, but it will affect the memory of the target.
I want to make this memory.read() as non-intrusive to the actual code as possible, without affecting the processor's memory, and in real-time without halting the processor.
Is this a limitation of XDS110 debug probe with TI processors, or the drivers in CCS and the firmware in the XDS110 probe chip that memory.read() uses for JTAG?
I know there are advanced debuggers in the XDS560, but they are costly.
