A newer version of this document is available. Customers should click here to go to the newest version.
- Configure Yocto Project*/Wind River* Linux* and Intel® VTune™ Profiler with the Intel System Studio Integration Layer
Hardware Event Skid
Event skid is the recording of an event not exactly on the code line that caused the event.
Event skids may even result in a caller function event being recorded in the callee function.
Event skid is caused by a number of factors:
The delay in propagating the event out of the processor's microcode through the interrupt controller (APIC) and back into the processor.
The current instruction retirement cycle must be completed.
When the interrupt is received, the processor must serialize its instruction stream which causes a flushing of the execution pipeline.
Intel® processors support accurate event location for some events. These events are called precise events.
The event skid affects the accuracy of your analysis results. When the grouping level is very small (for example, instruction, source line, or basic block), the Intel® VTune™ Profiler attributes performance results incorrectly. For example, when row A induces a problem, row B shows up as a hotspot. If different CPU events in the formula of a hardware event-based metric have different skids, the VTune Profiler may attribute data to different blocks, which makes all metrics invalid. This type of issue typically does not show up at the function granularity.
Example: Interpreting Jump and Call Instructions
Events that happen in the execution time of the jmp or call instruction, may appear on an instruction that is one or two instructions away from original jmp/ call in the execution flow. In this example, the mov instruction at the top of the loop is not responsible for the 1.02% of the events because the mov instruction is the target of the branch at the bottom of the loop. The real source of the events is the jmp instruction at the bottom of the loop.
Did you find the information on this page useful?