Compare Profiler Results

Comments

• StephenC Redgate › Posts: 314

  November 25, 2009 11:38AM

  This is something we get asked for on a reasonably regular basis although there are very good reasons for it not being implemented - results can change based on what the CPU was also doing at that point in time, different stack traces can't be compared easily if they have changed a great deal, comparing various time regions is difficult. There are many other reasons too which i'll not go into.
  
  You can export the All Methods grid for each run and and compare the results that way. We will be improving the exporting in the next version but for now if you:
  
  - Select a row in the methods grid
  - Ctrl + A
  - Ctrl + C
  
  Then paste into the Excel file. You can compare various methods from this grid. Alternating between CPU time and Wall-Clock time can also help.
  
  There is one other possibility by implementing our UserEvents.dll
  You can place this in your code to mark/annotate events on the timeline giving you start and end times that you can compare.
  
  http://www.red-gate.com/supportcenter/C ... wledgebase\ANTS_Profiler\KB200811000319.htm
  
  Perhaps you could tell me what would be most useful for you?
  
  Stephen

  0 · Share on Twitter
• ianormy Posts: 3

  November 25, 2009 12:03PM

  Thanks Stephen for your prompt reply. Lots of people ask for it because its a really useful feature
  
  The main use for it is that its saves a lot of time to be able to just load up 2 files and have up and down arrows showing improvements. The more of your own time you have to spend to do the comparison, the less profitable it becomes. I realise that the comparison may be difficult to implement, but even something that gives a rough idea really helps.
  
  Ah well! Thanks again,
  
  Ian

  0 · Share on Twitter
• StephenC Redgate › Posts: 314

  November 25, 2009 12:18PM

  On the face of it - it does look like it would be useful but the problems with it generally outweigh any positives (or at least that's the view at the moment but we are an open bunch so i wouldn't rule it out).
  
  I'd be really interested to know - what sort of time comparison granularity are you talking about?
  
  Going from 10mins to 2mins for example is a pretty obvious change. Shaving off a couple of seconds though probably isn't generally particularly useful to know.
  
  You can open a second instance of ANTS and load the other file in and check the timelines for both although again i have to hold up my hands and admit it's not how we would implement it if we ever decided to.
  
  Stephen

  0 · Share on Twitter
• Brian Donahue Posts: 6,590 Bronze 1

  November 25, 2009 12:31PM

  Hi Ian,
  
  You may also be interested in the Performance Profiler results comparison tool.
  
  ftp://support.red-gate.com/patches/ants ... ompare.zip
  
  If you copy the all methods grid from each profiling session to the clipboard, it will create XML files with the contents of the grid and show the deltas between the same method in each session. You could, for instance, sort on the method with the greatest time improvement, etc.

  0 · Share on Twitter