Re: Why systemtap loses events and what can be done about it?

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -

On Mon, Sep 10, 2007 at 06:13:52PM -0700, Brad Peters wrote:

> [...]  From this discussion, I mentioned I would assess possible
> time savings gained via this approach.

Thanks.  If I read the data correctly, among the effects of this
change would be:

- a considerable fraction of probe handling time becomes interruptible;

- probe handlers run overall 3 us longer ("step 3" of "scenario 2");

- a new dynamically managed resource (word queue items containing input
  data snapshots) is necessary

- it would be necessary to find a way of classifying which
  script-level inputs have to be gathered in context ($target vars,
  contextinfo tapset stuff like cpu(), execname(), ...?) and which can
  be deferred (script-level globals, other tapset stuff?)
  

- FChE