PerfView and TraceEvent 3.0-preview1

[brianrob]

Announcing the first preview release of PerfView and TraceEvent 3.0!

Action Requested: Please try this release and provide your feedback as issues in this repo, so that we can address any bugs and take your feedback into account on changes to the defaults. Please be sure start your issue title with "[PerfView 3.0]".

The goal of this and future previews is to allow for testing and feedback prior to an official stable release. There is not a set number of preview releases - this will be determined by feedback and bugs filed.

The main reason for incrementing the major version is that there are some changes that will break backwards compatibility:

• By default, all ETLX files are 64-bit and no longer have a limit of 20 million events.
• PerfView.exe is now 64-bit by default and PerfView64.exe has been removed from the build.

In addition, there are changes to the defaults for some of the views in PerfView. For example:

• The default grouping for all views is now [group module entries] instead of [Just My App] to help eliminate confusion when interpreting stacks.
• The default fold percentage of 1% has been removed to avoid situations where CPU time is so heavily spread that most CPU samples show up in the Process node when looking at stacks.
• Thread time views from nettrace files will now automatically fold UNMANAGED_CODE_TIME and CPU by default to help eliminate confusion when these values don’t properly represent the workload.

And of course, there are many other small fixes and improvements. Also notice that going forward releases of PerfView and TraceEvent will be merged together, rather than being posted as separate releases.

If you're interested in contributing to Perfview 3.0, please target your PR to the feature/perfview-3.0 branch.

---

This discussion was created from the release PerfView and TraceEvent 3.0-preview1.