Custom metrics API
Lustre meta data NEW
CPU hardware counters NEW
Energy performance NEW
|Custom metrics API||Yes|
|Lustre meta data NEW||Yes|
|CPU hardware counters NEW||Yes|
|Energy performance NEW||Yes|
How well do your applications exploit your hardware?
Allinea Performance Reports are the most effective way to characterize and understand the performance of HPC application runs. One single-page HTML report elegantly answers a range of vital questions for any HPC site:
- Is this application well-optimized for the system and the processors it is running on?
- Does it benefit from running at this scale?
- Are there I/O, networking or threading bottlenecks affecting performance?
- Which hardware, software or configuration changes can we make to improve performance further.
- How much energy did this application use?
How do I use Allinea Performance Reports?
To invoke Allinea Performance Reports, you don't need to recompile or relink your application on most systems. Simply prefix the command that you use to start your application with the command "perf-report" - and let it do the rest. The results are a HTML, plain text, CSV or JSON file.
For example, if your application is started by
mpirun -np 4 a.out
Then the new command is:
What Systems Can I use it on?
Allinea Performance Reports is available for most flavors of Linux on Intel x86_64 (Xeon) hardware and supports standalone servers and supercomputers and clusters. Full details are available on the supported platforms page.
What Information Is In a Performance Report
A report measures overall information about the computation, communication and I/O - and provides detail for each of these areas.
- The time spent in various categories of instruction: memory access, numeric operations, floating point operations.
- I/O - time and the effective performance (transfers rate) of read and write operations to storage.
- Memory - the mean and peak usage of memory per node.
- Communication - MPI time and performance for collective and point-to-point operations.
- Threads - the time spent in computation and synchronization, the physical core utilization and system load.
- GPUs - the utilization and memory use of NVIDIA CUDA GPUs.
- Energy - the energy pack add-on reports energy usage and peak power - for system, CPU and any NVIDIA GPUs.
See Allinea Performance Reports in action
Watch the short video or browse through our selection of example reports
Build a culture of performance awareness
Scientists, system support staff and administrators all create, understand and use Allinea Performance Reports to make sure every production run counts.
- Transparent application performance monitoring
- Single-page HTML reports produced automatically
- Plain-text, JSON or CSV for logging and comparison and automated performance reporting integrations
The perfect addition to your HPC workflow
- Easily assess whether applications are well-suited to the system they’re about to run on
- Give applications a regular health check without adding downtime
- Set up queues to automatically generate and upload reports to your intranet
- Feed the contents of the output into custom scripts to track application behavior and development over time