Use the graphics-rendering value to launch the GPU Rendering analysis (preview) and estimate your code performance based on the GPU usage per engine and GPU hardware metrics.
It focuses on the following usage models:
System-wide profiling on all virtual domains (Dom0, DomUs) running under the Xen* hypervisor to identify domains that take too many resources and introduce a bottleneck for the whole platform. Use the -target-system option to specify a remote machine connected to your host via SSH.
Profiling of OpenGL-ES applications running on Linux* systems to detect performance-critical API calls. For this mode, specify the application to analyze or a process to attach to, using the -target-process or -target-pid options.
For successful analysis, make sure to configure your system as follows:
For Xen virtualization platforms:
Virtualize CPU performance counters on a Xen platform to enable full-scale event-based sampling.
Establish a password-less SSH connection to the remote target system with the Xen platform installed.
To analyze Intel® HD and Intel® Iris® Graphics hardware events on a GPU, make sure to install the Intel Media Server Studio (starting with version 2015 R5) and build the kernel driver as described in the Getting Started Guide.
Knobs: gpu-sampling-interval, gpu-counters-mode=render-basic.
For the most current information on available knobs (configuration options) for the GPU Rendering, enter:
$ amplxe-cl -help collect graphics-rendering
This example runs system-wide GPU Rendering analysis for a remote Xen target:
host>./amplxe-cl --target-system=ssh:firstname.lastname@example.org –-collect graphics-rendering --duration 0
This example profiles an OpenGL-ES app running the GPU Rendering analysis:
host>./amplxe-cl –-collect graphics-rendering --target-process process1
When the data collection is complete, do one of the following to view the result:
Use the -report action to view the data from command line.
Use the -report-output action to write report to a .txt or .csv file