Tracedata

RE: Tracedata

ovrebekk — Tue, 02 Oct 2018 06:43:06 GMT

Hi again

According to Segger the DWT can periodically sample certain memory locations, but apparently the solution is not very flexible. They needed some more time to figure out what the sampling period is, and whether or not this can be connected to the trace.

Best regards
Torbjørn

RE: Tracedata

ovrebekk — Mon, 01 Oct 2018 14:49:49 GMT

Sorry for the late reply. Segger confirmed that the ETM can not do memory tracing, only instruction trace.

The DWT sounds more promising, but the documentation is a bit unclear regarding it's capabilities. You can count the number of accesses to a certain memory address, but I am not sure you can also trace the data content as it is changed. I will have to confirm this with Segger also.

RE: Tracedata

Örjan — Mon, 01 Oct 2018 08:35:54 GMT

Found this entry at Segger Forum as per 24'th august 2018. Seems data tracing not yet supported in J-Trace but may be in the pipe. As the 52840 contains a DWT unit, I guess it's capable of producing CPU memory access traces. Is this correct?

RE: Tracedata

ovrebekk — Tue, 25 Sep 2018 11:02:44 GMT

I will have to check this with Segger directly. As soon as I hear from them I will update the case.

Best regards
Torbjørn

RE: Tracedata

Örjan — Mon, 24 Sep 2018 14:01:48 GMT

Breaking in with a related question: We are doing instruction trace debugging on nrf52840 with the Segger j-trace module / ozon using the four tracepins + clk. From what we are able to see, only the cpu instructions are traced. We would need also to see traces for memory accesses, i.e address and data info. Are we doing something wrong/looking at the wrong place or is this simply not supported?

RE: Tracedata

ovrebekk — Mon, 24 Sep 2018 11:45:17 GMT

Hi Peter

For full ETM trace you need all the pins unfortunately.

With a single data pin you only get ITM, which is not sufficient if you want to do live instruction trace.

Best regards
Torbjørn

RE: Tracedata

Pema — Fri, 21 Sep 2018 07:14:06 GMT

Hi,

We have dedicated trace hardware for this. But for trace functionality, do you need all five signals or does it work with a subset (like clk plus one or two of the tracedata pins)? I want to fulfill the needs of our sw guys, while not letting them occupy too many IO:s... If a subset works, what are the limitations compared with using all tracedata? Speed I guess?

RE: Tracedata

ovrebekk — Fri, 21 Sep 2018 07:08:13 GMT

Hi Peter

For regular debugging you only need the SWD pins (SWDIO and SWDCLK), and you are free to use the trace pins for other purposes.

The Tracedata pins are only required if you want to do a full instruction trace during debugging, but this requires dedicated trace hardware to use (the Segger chip on the development kits are not set up for this).

Best regards
Torbjørn