Say hello to some of the latest features for the Intel® Graphics Performance Analyzers (Intel® GPA) tool suite.
Intel® GPA Framework
API Call Logging
We have enhanced verbose mode of the logging layer (--pre-layer logging:verbose=1) In addition to the API calls, you will see the values of variables in structs, the human readable names of enums, as well as arrays and other data structures.
You can see exactly what the application did while it was executing.
Using logging during playback, in addition to showing you the API calls and parameter values, we print out the unique IDs as they appear in the stream, so that if someone is doing verification or validation they can use the ID's to compare the same structure, class or function between runs.
gpa-metrics-collector.exe
gpa metrics collector provides query-based frame-level metrics. This new functionality allows you to automate performance work, monitoring changes. Previously the ability to collect query-based metrics from the GPU was only accessible if a developer created their own tool using our sample code as a starting point. Now you can use gpa-metrics-collector.exe..
Layer Generation Template - Jinja
For those of you who use GPA Framework’s ability to write your own layers we have changed our template backend from pybars to Jinja. You should update your layer templates to use Jinja syntax.
It's importatant to be aware that Intel® GPA Framework updates to stay in step with changes in Graphics APIs. When those API’s remove or deprecate certain functionality those changes are reflected in Intel® GPA Framework’s layer generation feature. So remember that you will need to adapt to these changes for any layers you have created.