For Inspector XE and VTune XE :
User interface is not very user friendly lots of fudging around to set up the file to process when in standalone mode. Lots of repeats to tell it where to find things. Does not remember previous choices need to restart at root of system.
Not possible to define subset as to what is to be inspected on large projects previous versions had a configuration utility which enabled a subset of the entire application to be selected. The only response to this issue is in Inspector (but not in VTune) Modules to exclude but there is no list such as in the previous version they must be expressly typed in.
Results on using VTune XE
Event Based Sampling previous version had a neat configuration window which organized events on a ratio basis. Chose ratios (ex. CPI and others) and the utility choses the events. Results showed both ratios and events and enabled navigation down from modules all the way to source. New version seems to have predefined everything, user selects one of the predefined types of analyses and pushes the button. Some events are displayed but no ratios except CPI on one or two of the analyses types. Help displays nothing on ratios. There are custom types but these seem very kludgy when compared to the configuration mode of the VTune previous version.
Call Graph analysis and critical path seems missing.
Performance counters seems to have been removed.
Thread Profiler functionality could not find time line and statistics such as are exhibited in Thread Profiler Old Thread Profiler works.
Call Graph table which shows statistics on callers and callees around pivot functions seem missing.
Old Thread Profiler and VTune seem to run.
VTune is a different animal altogether It seems less flexible than the previous version, appears to be missing a lot of functionality and seems to not be configurable as to reducing analysis to a subset.
I know that the mind resists change, I consequently tried to keep an open objective mind when looking at these products hope to have succeeded.
Question : Did Intel indeed remove a lot of features from old Vtune with the new XE version ?
As I know that statistical call graph data with sampling resultwill be provided, the performance is better than former call graph feature. Yes, critical path is missed but the user can start to investigating performance issue from Top N hot functions.
New tool provides "predefined" Analysis Types which includes more "commonly use" events in different analysis type. Absolutely, the user candefine their analysis type by selecting other events - that is for senior user. I think that most of senior user can define theirnew analysis type, for example - L2 Misses / CPU CLK, forL2 miss rateper cycle;Branch Misprediction / Instruction Retired, for B-Misprediction per instruction, etc. Yes,senior users can define their event ratios if they like.
I don't know what feature was missedin new tool from Intel Thread Profiler, would you please describe them in detail?Thanks!
One thing I really liked about old VTune was configuration of event based sampling via predefined ratios. It enables the JUNIOR programmer such as I to progressively understand how to better use the CPU. If my application reports a poor CPI, I could quickly narrow down via ratios where the problem lies. This would not be the case with new VTune. I'd furthermore say that if one has to be a SENIOR programmer to achieve this, then I believe that Intel is defeating its alleged intention to make parallel programming accessible to the mainstream.
Concerning Thread Profiler :
The new version's presentatoin is missing the clear vertical bar summary of the selected time line. It does present thread concurrency but it is missing thread beavior (critical path, blocking, impact and overhead).
Concerning Call Graph :
The tabular presentation which enambes to chose a pivot function to analyse upstream and downstream statistics is a useful tool. Also, the graphical presentation gives a bird's eye view as to where the critical path lies. These are very useful and save time. Hot spot analysis is useful but one could be spending time on a function which consulmes CPU but which is not on the ciritcal path.
The presentation of the new version is indeed more consistent and well organized. But this seems to have been done at the expense of certain features present in the old version.
I completely agree with you that Event Ratios can help users to create a new Analysis Type - understand and selectuseful events from ratios quickly.
It seems that Thread Concurrency, CP, impact time, blocking time and overheads in each thread are not presented in new tool.
I have send above as new feature requests to development team. Thanks for your feedbacks!
Performance Counters measure some performance objects, which are from operation system. You can use PerfMon.exe instead, or see example from MSDNhttp://msdn.microsoft.com/en-us/library/aa645516(VS.71).aspx
Sorry to say, your response is not very satisfying.
Vtune's perf counter is infinetly better than Microsoft's. It is very user friendly, enables you to collect a run and to drill down into event based sampling all the way into source. You cannot possibly have run both tools to propose that MS's PerfMon.exe can replace Vtune's perf monitoring.
Yes. We can select multiple collectors (sampling collector and performance monitor) in VTune Analyzer. We can select interest of timerangebased on performance monitor result,then observerelated sampling resultin same timerange.
Now VTune Amplifier XE 2011 has same "Zoom-in / Filter on Selection" function, but not based on performance monitor result:-)