- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I swallowed hard and downgraded my workstation back to RedHat 7.3. I am now able to run VTune 1.1.
To profile my application I gave the command:
% vtl activity my_act -c sampling -d 999 -app my_app run
followed by:
% vtl view a1::r1 -ha -mn lib1so,lib2.so,...
and VTune spits out a hot-spot report. So far so good.
Now for my questions:
1) How do I get this report in a human readable format (I got
something which has over 600 characters per line)? I have
been unable to find any formatting options for the report
generation.
2) Data is given in exclusive time (time spent in the function)
but I would like the possibility of seeing (and ordering by)
inclusive time (function plus descendants). Can I ask for
this info?
3) Can I get finer resolution than function? I would like to
know which *line* I am spending my time in. Is it possible
to get the hotspot line? Is it possible to get source files
annotated with cycle count?
Should I perhaps not be using sampling+hotspot to get the info I am looking for in 2 and 3?
Please note that I have absolutely no problem with text output. This is how I have *preferred* to use the profiling tools I have used in the past, e.g., pixie on SGI.
Also note that I do not have access to computers running Microsoft Windows nor do I believe I could persuade my company to ante up for another VTune license to process the collected data.
Regards.
Gisli
PS: This forum software allows a user which is not logged in to create a new message. Only when "Submit Message" is selected does the software inform the user that he is not logged in. Needless to say, the act of logging in clears this window and the message must be retyped. Ouch!
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
OUCH you bet, gisli. Happens to moderators too, and we have it bug reported. HATE IT WHEN THAT HAPPENS.
Now, in order:
1) most users either import the data into their favorite database (oracle, informix, whatever) and manipulate the data that way, or just awk sed and grep there way around through txt files. Both work real well.
2) not in this release of the software, but will be in the GUI version of same, currently planned to be in beta Q2 2004. You should definitely get that beta software when it's available!
3) that is possible with vtune 2.0, which will be generally available sometime during the first three weeks in JAN 04.
I'll be posting when vtune 2.0 is available, as well as when the next release, the GUI version will be starting beta. Keep checking here, Gisli!
cheers
jdg
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page