TPTP often hangs after first execution... [message #56869] |
Thu, 09 March 2006 07:21  |
Eclipse User |
|
|
|
I observe, that very often profiling is unable to restart after
completion or
termination of a prior run. Is this a known effect/bug? Any workaround
(other than killing and restarting eclipse...).
This is on Eclipse 3.2M5 and TPTP-4.2.0-200602281515.
Michael
|
|
|
Re: TPTP often hangs after first execution... [message #57067 is a reply to message #56869] |
Thu, 09 March 2006 13:18   |
Eclipse User |
|
|
|
I don't know about the guys over on the profiling side, but it's not a
"known bug" to me. I'd file a bugzilla, and document what you can about
the scenario. In particular, are you going IAC (local connection ) or
RAC (agent controller in separate process)? Simple case (like the
tutorial CarModel) or something more complex? Make sure to indicate
JVM... and make sure you put the "a" on the 3.2M5 so they don't
immediately assume pilot error! :-)
I'm over on the (new technology) AC side, so when I see such things (and
I've seen things like this during the M5 development cycle) I assume
it's my stuff... One workaround I've developed for such situations
that's a half step above "killing and restarting eclipse" is to have
another workspace with very little in it (so startup is simple, e.g.,
have it in Resource mode) and just "Switch Workspace..." on the File
menu... then switch back. That seems to set things so that I can delete
the old run and run again.
But, again, directly to your question... I've actually run the CarModel
profiling multiple times on that stable release without needing to
restart... so your bugzilla could be very helpful.
--
RDS
Michael Moser wrote:
> I observe, that very often profiling is unable to restart after
> completion or
> termination of a prior run. Is this a known effect/bug? Any workaround
> (other than killing and restarting eclipse...).
>
> This is on Eclipse 3.2M5 and TPTP-4.2.0-200602281515.
>
> Michael
>
>
|
|
|
|
Re: TPTP often hangs after first execution... [message #57636 is a reply to message #57586] |
Sun, 12 March 2006 17:54  |
Eclipse User |
|
|
|
"Samson Wai" <samwai@ca.ibm.com> wrote in message
news:duvn4c$u77$1@utils.eclipse.org...
> Hi,
>
> Are you using the standalone Agent Controller (localhost:10002) or the
> Integrated Agent Controller (Local Direct Connection) for profiling?
> It seems that you are using the Integrated Agent Controller from you
> message.
yes, the integrated AC.
> We have found some performance and stability issue with the Integrated
> Agent Controller (IAC) in TPTP 4.1 and work has been done on improving
> it in 4.2. Are you collecting a large amount of data? Since IAC is
> residing in the same process as Eclipse, the heap and memory space
> allocated for running Eclipse will be used by handling the profiling
> data as well. This may be a problem if large amount of data is to be
> collected, especially if you don't have a profiling filter set. A
> workaround is to use a more restrictive profiling filter or use the
> standalone Agent Controller if you definitely need to collect that
> amount of data.
The amount of data is certainly non-negligible. Noticing that the
default settings would collect enourmous amounts of data (and slowed
things down such that I always had to abort the program at first) I then
quickly added a filter that filters out a single class only, but one
callback method of that class is called VERY often, so I added another
rule to exclude that particular method and moved that filter to the top
position hoping that this would eliminate the problem. After that the
recording appeared to work fine, but I noticed this restart issue.
I'll give the standalone AC a try and see if this resolves the issue.
Michael
|
|
|
Powered by
FUDForum. Page generated in 0.04940 seconds