- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
my question is not VTune-specific, it is generally related to program profiling under Linux. Is there a tool to conduct instruction level profiling for a program and gain results summarizing the instruction mix of the program execution? (e.g. 30% integer add, 15% integer mul, 13% fp div, etc.)
Thanks in advance
my question is not VTune-specific, it is generally related to program profiling under Linux. Is there a tool to conduct instruction level profiling for a program and gain results summarizing the instruction mix of the program execution? (e.g. 30% integer add, 15% integer mul, 13% fp div, etc.)
Thanks in advance
Link Copied
4 Replies
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Do you need the static mix or the dynamic mix ?
For dynamic mix of the programs here is an interesting thread.
http://compilers.iecc.com/comparch/article/95-02-068
Cheers
Sandeep
For dynamic mix of the programs here is an interesting thread.
http://compilers.iecc.com/comparch/article/95-02-068
Cheers
Sandeep
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I am interested in the dynamic instruction mix of a program. The motivation behind my question is basically an effort to discover at runtime the profile (e.g. "fp-add" oriented or "fp-mul" oriented) of certain threads executing on HT processors. This could assist for example the scheduling of threads on HT processors, provided that there is prior knowledge regarding the execution units on which resource conflicts are observed.
Currenty, the only related events supported on HT enabled processors, are the total number of instructions, total number of fp instructions, but there isn't any distinction at a finer granularity (e.g. total #fp-add, total #fp-mul, etc.). It would be interesting to see Intel integrating such kind of events in the future.
Currenty, the only related events supported on HT enabled processors, are the total number of instructions, total number of fp instructions, but there isn't any distinction at a finer granularity (e.g. total #fp-add, total #fp-mul, etc.). It would be interesting to see Intel integrating such kind of events in the future.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
OK,
it seems that Pin (program instrumentation tool - http://rogue.colorado.edu/Pin/index.html) does the work I want (specifically, the tool 'opcodemix' - check it at http://rogue.colorado.edu/Pin/pintools.html).
it seems that Pin (program instrumentation tool - http://rogue.colorado.edu/Pin/index.html) does the work I want (specifically, the tool 'opcodemix' - check it at http://rogue.colorado.edu/Pin/pintools.html).
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
hey nik80
you were randomly selected as a VTUNE book winner, but the email announcing your having won bounced. Aside from doing nothing and letting us pick another winner, you could:
1) amend the email in your profile
2) Email vtune.sales@intel.com an alternate email address by which we can reach you.
Cheers
jdg

Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page