xSCOPE offline capture issue

[XenoPhoenix]

Hi, I currently adding xSCOPE instrumentation to my project and I'm running into the following issue on the 13.0.2 tools release.

XScope runs and produces the xscope.xmt file, however the first entry in this file always looks like this:

Code: Select all

<XtEvents>
<XtEventTypes>
<XtEventType Id="0" Type="1" Name="Calculation" Units="ps" UserType="0" UserUnits="Time"/>
<XtEventType Id="1" Type="3" Name="State" Units="ps" UserType="1" UserUnits="States"/>
<XtEventType Id="2" Type="2" Name="Emit" Units="ps" UserType="1" UserUnits="Output"/>
<XtEventType Id="255" Type="1" Name="Missing Data" Units="ps" UserType="0" UserUnits="None"/>
</XtEventTypes>
<XtEventColors>
<XtEventColor Id="0" R="148" G="0" B="211" KeyText="None"/>
<XtEventColor Id="1" R="255" G="69" B="0" KeyText="None"/>
<XtEventColor Id="2" R="218" G="165" B="32" KeyText="None"/>
<XtEventColor Id="255" R="255" G="0" B="0" KeyText="None"/>
</XtEventColors>
<XtGlobalStart MilliSecs="1391164697222"/>
<TotalXtEvents Number="0"/>
<XtEventRecords>
<Record Type="0" Color="0" Start="18446744073629581616" End="1091410000" PC="67584" Tile="tile[8003]" LogicalCore="1" User="0"/>
<Record Type="1" Color="0" Start="1115570000" End="1116470000" PC="66524" Tile="tile[8003]" LogicalCore="2" User="0"/>
<Record Type="1" Color="0" Start="1116470000" End="1117370000" PC="66524" Tile="tile[8003]" LogicalCore="2" User="0"/>
<Record Type="0" Color="0" Start="1140710000" End="2312090000" PC="67584" Tile="tile[8003]" LogicalCore="1" User="0"/>
<Record Type="1" Color="0" Start="1117370000" End="3139040000" PC="66524" Tile="tile[8003]" LogicalCore="2" User="0"/>

The "Start" attribute on the first record has a number which is obviously far too big (indeed looks like two number concatenated together). This leads to an error when you attempt to load it in the XDE due to a failure to parse that number.

If I manually edit to file to make that number sane, it loads fine, however it's getting reasonably irritating have to do this every time after capturing a trace (even though it can be scripted).

Is there anything I could be doing wrong which would lead to this output or it a bug in the tools?

I am registering xScope as normal:

Code: Select all

void xscope_user_init(void) {
    // Register XSCOPE with the system.
   xscope_register(3,
       XSCOPE_STARTSTOP,    "Calculation", XSCOPE_NONE, "Time",
       XSCOPE_STATEMACHINE, "State",       XSCOPE_UINT, "States",
       XSCOPE_CONTINUOUS,   "Emit",        XSCOPE_UINT, "Output");
   xscope_config_io(XSCOPE_IO_BASIC);
}

And capturing data:

Code: Select all

xscope_core_start(0);
...
xscope_core_stop(0);

Code: Select all

xscope_core_int(1, state);

Code: Select all

xscope_core_int(2, output);

Interestingly, this issue did not occur when I only had a single probe (the first STARTSTOP) one.

Any help / confirmation if this is a bug would be useful. Thanks.

[XMatt]

I will have a look at trying to reproduce this, it looks like a bug as the timestamp of the first record should be 0.

Matt

[XenoPhoenix]

Thanks Matt, let me know if there is anything else you need.

[sprajagopal]

I have this exact same problem. Except the values are recurring. Is this bug not fixed? If so, what triggers this and how do I avoid it?

I have the scope data attached to this forum post:
http://www.xcore.com/questions/4152/why ... scope-data