MAXQ Competitive Performance

Analysis of the ATmega8

1. The MODEL used for each benchmark was SMALL.
2. The 8-bit and 16-bit matrix benchmarks would not run correctly under the fastest configuration (simulation would not terminate) unless the medium optimization level was used (instead of the highest).
3. FIR filter on the smallest configuration would not simulate (simulation would not terminate) even at the lowest optimization settings. Therefore, size and performance numbers from the fastest configuration are shown here.
4. The Matrix Multiplication benchmark would not simulate correctly (simulation would not terminate) on the ATmega8, ATmega16, or ATmega32 targets. The ATmega64 target was used for this benchmark.