Comments
Sort by recent activity
Even being taken to declaring source file would be already of great help. But I guess that the profiler has enough information available to even offer a selection of a) where the object in question was actually created, b) which ctors where used to do so, and then take me to any of those.
Taking it a step further it could even apply the same scheme used by VisualAssist and offer pseudo-source access throught the available meta data for assemblies without source code, for system types and imported assemblies. / comments
Even being taken to declaring source file would be already of great help. But I guess that the profiler has enough information available to even offer a selection of a) where the object in question...
No, no, I am really taken to the wrong source file. The file shows does not even create a single instance of the questionable object.
Rebuilding does not fix the problem. / comments
No, no, I am really taken to the wrong source file. The file shows does not even create a single instance of the questionable object.
Rebuilding does not fix the problem.
No, is a windows forms desktop application, heavily database dependant, C# 2.0, .Net 2.0, using VS2005 SP2. / comments
No, is a windows forms desktop application, heavily database dependant, C# 2.0, .Net 2.0, using VS2005 SP2.
Will verify, but i guess so, since usually we do a rebuild all before testing. I keep you informed as we go. / comments
Will verify, but i guess so, since usually we do a rebuild all before testing. I keep you informed as we go.
Right click in any source code window inside the IDE, brings up the editor context menu. Among other entries it contains 3 entries from the ANTS Profiler - Profile Method, Profile Class, Profile Namespace.
Those entries make not much sense during a debug session. / comments
Right click in any source code window inside the IDE, brings up the editor context menu. Among other entries it contains 3 entries from the ANTS Profiler - Profile Method, Profile Class, Profile Na...
How odd. That should be grayed out when unavailable but not removed. Thats irritating. / comments
How odd. That should be grayed out when unavailable but not removed. Thats irritating.
I dont know which version you are using, but mine definitely does not have a call stack view. I am using 3.1.0.20.
In the View menu I have
Back
Forward
Summary
All Methods
Start Page
Source Code
Hierarchy
Results
Thats it. Not finding anywhere a Call Stack view. / comments
I dont know which version you are using, but mine definitely does not have a call stack view. I am using 3.1.0.20.
In the View menu I have
Back
Forward
Summary
All Methods
Start Page
Source Code
Hi...
Definitely no low-memory situation, the application uses about 8mb at max during 20-30 seconds.
Perfmon shows a peak of 7.6 for %time spent in GC.
If it helps, I could send the app over for testing to you.
I am probably too spoiled from C++ Profilers where the whole call stack of the allocation location is shown. In my case I see only that most of the objects are allocated in System.Collections...whatever...Resize(). thats absolutely not very helpful. / comments
Definitely no low-memory situation, the application uses about 8mb at max during 20-30 seconds.
Perfmon shows a peak of 7.6 for %time spent in GC.
If it helps, I could send the app over for testing...
In the All Methods view I would suggest a slight change in the column sorting:
Almost always when using the profiler its with the intention to find bottlenecks and optimize code, so I want to see what uses most time, whats slowest.
When I open the all methods view and click on a column it will sort by this column, but always ascending at first. This is great for alphabetic columns but bad for number columns, as they show what I dont want to see. So I have to click a second time to sort descending.
I feel that the columns "Time", "Time with children" and "Hitcount" should by default sort descending when first clicked. It would simplify this view. / comments
In the All Methods view I would suggest a slight change in the column sorting:
Almost always when using the profiler its with the intention to find bottlenecks and optimize code, so I want to see w...
Most importantly - remember the docking location (thats true for all other tabs too)
I would like to be able to hide not-hit methods from the source window, to minimize information clutter. What good does a method that is not even hit?
Another great information to show in the source window is the total time and hitcount for every method. While this information is of course available in the All Methods tab, its tedious to find it there while you are browsing the source.
Optional hiding of the Line column would be also very welcome.
Entry times showing only 0,0000 might be better not shown at all I think, it does not provide any information (except that the time used was very few) and it clutters the display. Other entries might be overseen (like 0,0800 among some 0,0000)
Would be also nice if Select All would not jump to the end of the source, so one can copy and then continue where left off.
When selecting some lines, it would be useful to show the sum of time used of those lines either as a tooltip or in a statusbar field (a la excel). This would allow to get a timing of only a part of a method.
Another thing that would make browsing easier are clickable method calls. When there is timing information available on a method call, I want to drill down on that method by simply clicking on it in the source window. / comments
Most importantly - remember the docking location (thats true for all other tabs too)
I would like to be able to hide not-hit methods from the source window, to minimize information clutter. What go...