Profiling using VxWorks and AdaMULTI.

Profiling using VxWorks and AdaMULTI.

[Petter Fryklund]

Does anybody have any experience with profiling in VxWorks. We are trying to
do some profiling. We're currently using the AdaMULTI debugger to collect
data, but when we try to analys, we get an error message:

protrans: fatal: TickTime not implemented.

We can't find TickTime in any documentation.

Re: Profiling using VxWorks and AdaMULTI.

[Ted Dennison]

In article <9f2jl6$l89$1@newstoo.ericsson.se>, Petter Fryklund says...
>
>Does anybody have any experience with profiling in VxWorks. We are trying to
>do some profiling. We're currently using the AdaMULTI debugger to collect
>data, but when we try to analys, we get an error message:

I've done it using ScopeProfile (WindRiver, ex RTI), but not using AdaMULTI.
You'd probably be better off contacting GreenHills support (assuming you have
support).

---
T.E.D.    homepage   - http://www.telepath.com/dennison/Ted/TED.html
          home email - mailto:dennison@telepath.com

Re: Profiling using VxWorks and AdaMULTI.

[Petter Fryklund]

We've already been in contact with GHS, they haven't yet found out what
TickTime is. I've also tried to get in contact with WRS. I've just hoping
that somebody out there knew.

Ted Dennison wrote in message <2E6R6.5179$rn5.256863@www.newsranger.com>...
>In article <9f2jl6$l89$1@newstoo.ericsson.se>, Petter Fryklund says...
>>
>>Does anybody have any experience with profiling in VxWorks. We are trying
to
>>do some profiling. We're currently using the AdaMULTI debugger to collect
>>data, but when we try to analys, we get an error message:
>
>I've done it using ScopeProfile (WindRiver, ex RTI), but not using
AdaMULTI.
>You'd probably be better off contacting GreenHills support (assuming you
have
>support).
>
>---
>T.E.D.    homepage   - http://www.telepath.com/dennison/Ted/TED.html
>          home email - mailto:dennison@telepath.com

Re: Profiling using VxWorks and AdaMULTI.

[Ted Dennison]

In article <9f4oae$2tc$1@newstoo.ericsson.se>, Petter Fryklund says...
>
>We've already been in contact with GHS, they haven't yet found out what
>TickTime is. I've also tried to get in contact with WRS. I've just hoping
>that somebody out there knew.

Well, it doesn't appear to be anything in vxWorks (which shouldn't be shocking,
since you apparently *don't* have it). I can't find any reference to it in
vxWorks 5.3 or 5.4, and it is not a symbol in our 5.4 kernel.

The only explanations I can think of are that it either:

a)  is a symbol that the GreenHills profiler uses. 

That seems rather unlikely, as it then wouldn't work for anyone.

b)  is a symbol used by some code of yours (most likely custom C code linked
into the kernel) that happens to have the same name as a kernel subroutine.
Everything in vxWorks exists in the same symbol space, so its quite possible for
this kind of thing to happen (another good reason to not use C). When linking a
kernel, I don't think there are any checks for undefined or redefined
references.

To check for b), first see if you have any (non-WindRiver) code being linked
into your kernel. If you do, look through it for the offending symbol. If its
found in there, you have 2 problems: 1 - there's an undefined symbol you have to
do something with; 2 - something that is calling that has the same name as a
kernel (or profiler) routine, and *must* be renamed.

We had something similar to b) happen to us. There was a C routine which we were
reusing from another project which happened to have a routine named "checksum"
in it. When we took our first kernel out to a system to try it out, we couldn't
get networking to work. We couldn't even ping the loopback address. It turns out
that "checksum" is also the name of a routine in the TCP/IP stack, and we were
inadvertantly replacing it with our C program's "checksum" routine. Whoops.



---
T.E.D.    homepage   - http://www.telepath.com/dennison/Ted/TED.html
          home email - mailto:dennison@telepath.com

Re: Profiling using VxWorks and AdaMULTI.

[Martin Dowie]

> The only explanations I can think of are that it either:
>
> a)  is a symbol that the GreenHills profiler uses.
>
> That seems rather unlikely, as it then wouldn't work for anyone.

well, after doing a winNT 'find...' for any file containing TickTime the
only
occurance we found was in the protrans application too (it is found in
some VxWorks headers but only in the middle of #define, e.g.
#define WRS_TICKTIMEEVENT)

so it looks to me like something is missing from the GHS distribution...

...maybe no one has ever tried this tool ever?!?! :-)