I have the OMAP-l137 running at 300mHZ and suppose that I have a clock that is giving 300.000.000instructions/sec, if I have a sample rate at 20KHZ, equals 15.000 cycle/sample to complete the sample and the analyse. Now I have build in a buffer to store samples to buffer what I can't process within the 50us, I have added several optimization to the compiler and changed the software in several ways to improve the performance, but I have problems to get the performance where expected. My program is analysing for 10 sec and shall during that period make the DSP analyse. If I look at the program in ASM in CCS5.3 I can count what looks like 3700 steps/analyse, counted with single step. If I run the analyse an measure what time the counted 3700 steps takes, it looks like that 20.000 samples takes 1sec. I can't see that I have 15.000steps witch should finish the 20.000 samples in 1/4 sec. I have checked the clock and I get it to 300Mhz where is the missing 75% in performance or is my calculation wrong?
How is the 3848MIPS found (from the tms320c6747.pdf ) ?
Claus.