CCS/AM5728: System trace - Memory throughput analysis - Uninitialized error text

Hello Rafael,

My board is TI AM572x Evaluation Module (www.ti.com/.../tmdsevm572x ). I have removed the screen on top of it. The serial console shows the following output at Uboot.

U-Boot SPL 2016.05-g6c5519b6fc (Dec 14 2016 - 20:19:24)
DRA752-GP ES2.0
Trying to boot from MMC1
reading args
spl_load_image_fat_os: error reading image args, err - -1
reading u-boot.img
reading u-boot.img
reading u-boot.img
reading u-boot.img


U-Boot 2016.05-g6c5519b6fc (Dec 14 2016 - 20:19:24 -0500)

CPU  : DRA752-GP ES2.0
Model: TI AM5728 BeagleBoard-X15
Board: AM572x EVM REV A.30
DRAM:  2 GiB
MMC:   no pinctrl for sdr104
no pinctrl for ddr50
no pinctrl for sdr50
no pinctrl for sdr25
no pinctrl for sdr12
OMAP SD/MMC: 0, OMAP SD/MMC: 1
reading uboot.env

** Unable to read "uboot.env" from mmc0:1 **
Using default environment

SCSI:  SATA link 0 timeout.
AHCI 0001.0300 32 slots 1 ports 3 Gbps 0x1 impl SATA mode
flags: 64bit ncq stag pm led clo only pmp pio slum part ccc apst 
scanning bus for devices...
Found 0 device(s).
Net:   
Warning: ethernet@48484000 using MAC address from ROM
eth0: ethernet@48484000
Hit any key to stop autoboot:  0

I get the same "Uninitialized error text message" error when I use GPEVM_AM572X instead.

I am using TI Emulators   7.0.100.1.

Thank you.

solid repellent said:

The video show exactly what and how I am doing.

As Rafael has already noted the BeagleBoard X15 has only wired EM0 and EM1 to the 20-pin compact TI JTAG header. Can you try setting the Number Of Pins to "1 pin" to see if that stops the errors (of the two EM0 and EM1 pins one is clock and one is STM trace data).

[I don't have a USB560 V2 system trace to try this]

Thank you Chester for the reply. I tried with "1 pin" (please check the attached video). It gives the same errors again. It gives the same errors when I select "2 pins" as well. Any other ideas?

From the TI AM572xevm_rev_a3a schematic diagram found here (http://www.ti.com/tool/tmdsevm572x) I see that EMU0 and EMU1 pins are connected. Whereas EMU2, EMU3, EMU4 are not connected.

desouza Were you able to test this out on your end with a Linux distro with the information I provided? Could you please share your insights? Thank you so much. 

Hi,

They tend to be quite nimble to release updates to their software stack. However since they weren't yet able to find the root cause, any guess as to when such release occurs is still unknown. This week is Spring Break in the US, which is traditionally a slow week and therefore I suspect to get an update next week.

As for other probes that allow System Trace on pins, I would also look at the TMDSEMU560V2STM-UE or the less costly variant XDS560v2 STM Traveler.

Regards,

Rafael

desouza said:

Apparently System Trace exported to pins is not functional with AM57x and Blackhawk Debug Probes.

I tried using a Spectrum Digital TMDSEMU560V2STM-UE with a BeagleBoard X-15. When attempted to setup Memory Throughput system trace in 1-pin mode with CCS 8.0.0.00016, TI Emulators 7.0.188.0 and Spectrum Digital Emulators 5.2.0.14 the following error was reported on the CCS GUI:

failed to calibrate channel : transmitter frequency measured at 0 Hz

And the information from the Trace Logging was:

M     15:22:46:314 | Trace Server: Channel event [15] Trace Channel Error: failed to calibrate channel : transmitter frequency measured at 0 Hz fired.
M     15:22:46:314 | Trace Operation 560V2STM: failed to configure channel
    E 15:22:46:314 | Trace Operation: Error Set:  failed to calibrate channel : transmitter frequency measured at 0 Hz

I confirmed that the TMDSEMU560V2STM-UE trace is functional, since was able to collect Memory Throughput 1-pin System Trace from a PandaBoard, where the Trace Logging reported that the trace was recording at 164.4 Mbits/second.

The non-working trace on the BeagleBoard X15 used the MIPI to 20-pin Compact TI adapter.

For the working trace on the PandaBoard used the same MIPI to 20-pin Compact TI adapter plus an additional 20-pin Compact TI to 14-pin adapter. I.e. that shows that the MIPI to 20-pin Compact TI adapter which failed to collect the 1-pin trace from the BeagleBoard X15 is working.

I don't know if the failure to collect System Trace is a problem with the TMDSEMU560V2STM-UE drivers, or the GEL files used by the GPEVM_AM572X_SiRevA board not configuring the pin mapping correctly.

desouza said:

- If your board is the BeagleBoard X15, its schematics indicate it does not have the EMU2~4 pins connected, thus allowing only 1-pin STM Trace
elinux.org/Beagleboard:BeagleBoard-X15

Looking at the .TI-trace\data\TraceCntrl-Spectrum Digital XDS560V2 STM LAN Emulator_0-CT_STM_Config-560V2STM.cfg file created shows pins 2 and 0 are attempting to be used to receive the STM trace data (on the pinMap line I think the first number is the trace clock pin and the remainder are the trace data pins):

[Global]
dtc=sd560v2e-1062731773,sd560v2e.out

[Channel 0]
xmtrType=
targetSync=true
dtcBufMode=stop
numPins=1
pinMap=2,0
dtcBufSize=1048576
startPort=0
drmXmtr=tpiu

The AM572x TRM SPRUHZ6J shows that only emu2 can be used as the trace clock:

Therefore, I think to be able to collect STM 1-pin trace from an AM572X pins emu2 and emu0 must be connected, and since the BeagleBoard-X15 only connects emu1 and emu0 to the JTAG connector it is therefore not possible to collect STM trace data.

Whereas with an OMAP4430 I was able to collect STM 1-pin trace over emu1 and emu0 and the OMAP430 TRM shows that for an OMAP4430 1-pin STM trace can be used over emu1 and emu0:

desouza said:

I was chatting with the dev team and found out that what Chester is talking about is sound information.

In ccs800/ccsv8/ccs_base/emulation/analysis/xmldb/trace_config/devices/device_vayu.xml changed from:

				<map program="15" pinouts="tpiu=2,0" />

To:

				<map program="15" pinouts="tpiu=0,1" />

And in ccs800/ccsv8/ccs_base/emulation/analysis/xmldb/trace_config/devices/device_DRA7xx.xml changed from:

			<map program="14" pinouts="tpiu=2,0" />

To:

			<map program="14" pinouts="tpiu=0,1" />

The idea was to try and get the trace software to configure EMU0 for the clock and EMU1 for the data. The reason for changing device_vayu.xml and device_DRA7xx.xml was that both have a device ID which matches an AM5728. With the above changes made to CCS8 and a BeagleBoard-X15 with AM5728 rev 1.1 silicon connected to a Spectrum Digital XDS560V2 STM:

1) When a project which used a C66 core was able to collect 1-pin STM Trace, and the Memory Throughout Analysis looked valid.

2) When a project which used a A15 still failed to start a STM trace, but the error had changed to indicate a failure to detect the sample offset. Looking at the files in the .TI-trace directory shows that the trace clock frequency was detected on EMU0, but there was no data (all zeros) on EMU1.

In the case of the C66 core project which collected 1-pin STM trace on EMU0 and EMU1 the .cfg was set to use drmXmtr=pti, whereas in the case of the A15 core project was set to use drmXmtr=tpiu. Contrary to the AM5728 TRM it may be able to allow the CCS hardware trace analyser to allow configuration of 1-pin STM trace using EMU0 and EMU1 if have an option of selecting the pti to be used as the trace transmitter in the device. 

NOTE: I created this post from memory of the experiment, since the files are on a different computer, and don't have the actual modified files to attach.

desouza said:

At any rate, it is unfortunate the board was designed with the cTI 20-pin connector but the EMU2~4 pins were not routed - this would give maximum flexibility without requiring changes to the files that may have other unintended consequences - thus falling into the realm of "not supported" / "untested" from our point of view.

Could the CCS "Receiver/Transport Settings" dialogue be enhanced to list which EMU pins are used to clarify to the user which EMU pins need to be connected on the specific device being used for STM trace?