The interrupt latency can be divided into two parts:
- Kernel latency: the time from the occurrence of the interrupt until the OS calls the interrupt handler. This is OS dependent.
- The transfer from the kernel mode to the user mode: the time from the call to the kernel until the release of the user-mode thread.
This latency depends on specific factors of your machine and software. You need to experiment with your software to test its speed.
WinDriver’s Kernel PlugIn feature (available from version 4.0) enables you to implement the crucial parts of your code (such as the interrupt handlers) in the kernel mode, thereby improving the overall performance of your driver. For the list of operating systems for which the Kernel PlugIn feature is available, refer to Technical Document #29.
Using the Kernel PlugIn you can expect to handle over 100,000 interrupts per second (see also www.jungo.com/st/support/windriver/windriver_faqs/#HMI).
For more information about the Kernel PlugIn feature, refer to the WinDriver PCI User’s Manual.