Texas Instruments SPRAA56 appendix Programmatic Measurement of Total CPU Load

SPRAA56

last30frame.current = CLK_getltime();

// check to see if we dropped any frames

benchVid.framesDropped.current = last30frame.current - last30frame.previous; benchVid.framesDropped.current -= 1000*(frameCnt / DISPLAYRATE); benchVid.framesDropped.current /= DISPLAYRATE;

last30frame.previous = last30frame.current;

if (benchVid.framesDropped.current > 0 && frameRateTarget == DISPLAYRATE ) { LOG_error("Dropped %d frames", benchVid.framesDropped.current); UTL_logDebug2("Dropped %d frames, after %d frameCount",

benchVid.framesDropped.current, frameProcessCnt);

benchVid.framesDropped.previous = benchVid.framesDropped.current; if (benchVid.framesDropped.current > benchVid.framesDropped.max) {

benchVid.framesDropped.max = benchVid.framesDropped.current;

}

} // end of dropped frame detection

A UTL_logDebug API call is made during the benchmarking routine every 30 or 25 frames, to report any dropped frames during the last group. Additionally, a call to LOG_error is made, which will insert a red mark in the DSP/BIOS execution graph and insert the text string specified in the API into the execution graph details, which are visible from the Message Log RTA tool.

4.5Simulating High CPU Load Stress Conditions with Dummy NOP Loads

The H.263 encoder algorithm in this example has a relatively moderate CPU load benchmark of about 50%. Other applications may require encoders with higher CPU loading or additional post- processing stages that add to the load. Before integrating such functions into the system, you may want to estimate their effects on real-time performance.

One way to estimate the effects of an additional load is with a dummy load of NOP instructions. Such a dummy load function is provided in the dummyLoad.c file of this example. It can be controlled from the h263rateControl.gel file, which manipulates the controlVideoProc.dummyProcessLoad variable containing the number of NOP instructions the function will execute. The dummy load function can also be used to test a system beyond typical stress conditions to ensure that it performs correctly, and drops frames gracefully if necessary.

4.6Programmatic Measurement of Total CPU Load

The DSP/BIOS Real-Time Analysis (RTA) tools already provide a CPU Load Graph tool within CCStudio on the host PC. However, some applications could benefit from an awareness of the CPU load within the program itself. A program with awareness of the current CPU load, for instance, could decide not to instantiate new processing routines when the load is high, or could turn on additional post-processing when the load is lower.

Programmatic calculation of the CPU load could also be useful if RTA is disabled, causing the CPU Load Graph to be inoperative. The application could periodically report the current CPU load using the UTL_logDebug APIs, and that data could be viewed at a breakpoint or halt.

This application note introduces a LOAD module that allows you to check the CPU load via an API call. That data is reported during benchmark output in the processing task. Figure 4 shows how the LOAD module works.

14DSP/BIOS Real-Time Analysis (RTA) and Debugging Applied to a Video Application